Verb-Verb Complexes in Asian Languages 2021931021, 9780198759508

Citation preview

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Verb-Verb Complexes in Asian Languages

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages Edited by

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

TARO KAGEYAMA, PETER E. HOOK, AND PRASHANT PARDESHI

1

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Taro Kageyama, Peter E. Hook, and Prashant Pardeshi 2021 © the chapters their several authors 2021 The moral rights of the authors have been asserted First Edition published in 2021 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2021931021 ISBN 978–0–19–875950–8 DOI: 10.1093/oso/9780198759508.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

For Colin P. Masica

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Contents Acknowledgments List of figures List of tables List of abbreviations The contributors

1. Introduction Taro Kageyama, Peter E. Hook, and Prashant Pardeshi

ix xi xiii xv xxi

1

PART I. LANGUAGES OF EAST ASIA 2. Between lexical verbs and auxiliaries: The architecture of Japanese verb-verb complexes Taro Kageyama

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3. Verb verb complex predicates in Old and Middle Japanese Hirofumi Aoki and Bjarke Frellesvig 4. Grammaticalization and constructionalization in Japanese lexical compound verbs Taro Kageyama 5. Syntactic verb-verb compounds in Japanese Hideki Kishimoto 6. The semantic differentiation of verb-te verb complexes and verb-verb compounds in Japanese Yo Matsumoto

15 44

70 103

139

7. Verb-verb complexes in Irabu Ryukyuan Michinori Shimoji

165

8. Korean verb-verb sequences Hyun Kyung Hwang and John Whitman

193

PART II. LANGUAGES OF SOUTH ASIA 9. Classification of complex verbs and the evolution of the compound verb in Marathi Prashant Pardeshi 10. Development of verb-verb complexes in Indo-Aryan Benjamin Slade

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

223 249

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

viii



11. The Hindi-Urdu compound verb and its covert semantics: Births, earthquakes, meteors, and other autogenous expressions Peter E. Hook

275

12. The matrix of verb-verb sequences in Tamil E. Annamalai

300

13. Verb + verb sequences in Dravidian Sanford B. Steever

327

14. Semantically related verb verb combinations in Tibetan and Ladakhi: 1300 years of stable transition Bettina Zeisler

354

PART III. LANGUAGES OF CENTRAL AND NORTHWEST ASIA

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15. Verb–verb complexes in Turkic languages: Interaction of lexical and delexicalized verbs Andrey Shluinsky

397

16. Verb-verb complexes in Central and Eastern Turkic languages Noriko Ohsaki and Fuyuki Ebata

430

17. Turkish and Uyghur verb-verb complexes in contrast Yuu Kuribayashi

455

18. Verb-verb complexes in Avar Hisanari Yamada

471

PART IV. LANGUAGES OF EAST AND SOUTHEAST ASIA 19. Verbal complexes in Thai Kingkarn Thepkanjana and Satoshi Uehara

499

20. Verb-verb sequences in Mandarin and Hindi-Urdu Hsin-hsin Liang and Peter E. Hook

521

References Index of languages Index of subjects

545 573 575

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Acknowledgments This volume grew out of an international conference entitled “Mysteries of VerbVerb Complexes in Asian Languages,” held at the National Institute for Japanese Language and Linguistics (NINJAL), Tokyo, on December 14 and 15, 2013. The two-day conference, comprising sixteen invited talks and fifteen poster presentations, was a great success, with a total of about 250 people attending from diverse countries of the world. We thank all the participants in the conference for valuable feedback and the staff members of NINJAL for logistics support. We are particularly grateful to Miriam Butt for presenting a plenary talk which served as a bridge to link the researchers in South Asian linguistics, Japanese linguistics, and Turkic linguistics. The conference was supported by the NINJAL funding for two largescale collaborative research projects: “Syntactic, Semantic, and Morphological Characteristics of the Japanese Lexicon” (Project leader: Taro Kageyama) and “Universals and Cross-Linguistic Variations in the Semantic Structure of Predicates” (Project leader: Prashant Pardeshi). After the conference, specialists working on Ryukyuan, Indo-Aryan and Dravidian diachrony, Korean, Thai, and Mandarin Chinese were specifically invited for publication to make the language coverage complete. The chapters of this volume thus consist of selected papers from the conference and the newly invited contributions, all of which were read by internal and/or external reviewers. Special thanks are due to the anonymous reviewers for the volume and for the conference. Thanks are also due to Ms. Miyako Hara for her help and patience in the painstaking job of creating the map included in Chapter 1. Finally, we offer our profound thanks to John Haig (retired professor of Japanese linguistics at the University of Hawai‘i at Mānoa) for serving as copy-editor, to Doshisha University (Faculty of Culture and Information Science) for providing the first co-editor, Taro Kageyama, with the research funding necessary for the editorial work at the last stage (2017–18), and to the NINJAL projects “Development of and Linguistic Research with a Parsed Corpus of Japanese” and “Contrastive Studies of Japanese Prosody and Grammar” for providing the third co-author, Prashant Pardeshi, with the research funding necessary for the editorial work. Taro Kageyama Peter E. Hook Prashant Pardeshi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

List of figures 1.1. Delexicalized verb constructions in Asia

11

2.1. Semilexical categories in predicate agglutination

40

4.1. Paths to Type 2 aspectual compounds

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

9.1. Increase in frequency of use of the Marathi, Marwari, and Bangla CV

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

90 244

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

List of tables 2.1. Four types of V–V complexes in Japanese

17

2.2. Auxiliary verbs in Type 4 complex predicates

20

2.3. Phasal verbs in Type 3 syntactic compound verbs

31

2.4. Selected aktionsart meanings in Type 2 compound verbs

36

2.5. Form-meaning correspondence in V-V complexes

38

2.6. Different stages of grammaticalization in delexicalized verbs

42

4.1. The vicissitudes of auxiliary verbs in Old Japanese

82

4.2. Examples of the independent usage antedating the V2 usage

87

4.3. Examples of the V2 usage antedating the independent usage

88

4.4. Examples of the V2 usage and the independent usage appearing concurrently

89

4.5. The V2 usage evolving only in the compound structure

89

6.1. Subtypes of V-te V complexes and V-V compounds

151

6.2. V-te V complexes and three factors favorable to the choice of their form

161

7.1. Finite inflection in Irabu Ryukyuan

167

7.2. A partial paradigm of nonfinite inflections

168

7.3. Stem forms for V-V complexes

169

7.4. Irabu V-V complexes

170

7.5. V₂ of syntactic compounds

174

7.6. V-V complexes in Irabu and in Japanese

189

7.7. Irabu SVCs which are cognates with Japanese Types 1 and 2

192

8.1. Nonthematic V₂ selecting infinitive V1-e/a

199

8.2. Object-sharing V₁-e/a thematic V₂ complexes with V₁ mil-/nwul- ‘push’

206

8.3. V₁-e/a V₂ sequences taking pwuth- ‘attach (intransitive)’ as V₂

215

9.1. Damle’s classification of arthavān (meaningful) V₂s in V₁+V₂ complexes

228

9.2. Variation across Indo-Aryan languages in ratios of CVs to total verbs

242

9.3. Variation across time in ratios of compound verbs to total verbs in Marathi texts

243

9.4. Variation across time in ratios of compound verbs to total verbs in Bangla texts

243

9.5. Variation across time in ratios of compound verbs to total verbs in Marwari texts

244

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

xiv

  

10.1. Light verbs that occur in VVs from selected Dravidian languages

271

10.2. Light verbs that occur in VVs from selected Indo-Aryan languages

272

10.3. Selected Hindi and Nepali light verbs compared

273

11.1. Ratios of compound verbs to total verbs in Hindi-Urdu, Marathi, and Nepali

281

12.1. List of light verbs and the corresponding full verbs in Tamil

301

13.1. Analysis of Dravidian compound verbs

352

14.1. SVVCs in Tibetic languages, overview

358

14.2. Percentage of SVVCs in the LLV

364

14.3. Distribution of SVVCs

364

14.4. Percentage of SVVCs among motion and movement verbs

365

14.5. Locating the verb-verb constructions on the univerbation path

391

15.1. Meaning of the construction -p tur- in Karachay-Balkar depending on the actional class of the lexical verb

409

15.2. Meaning of the construction -sa il- in Anatri Chuvash depending on the actional class of the lexical verb

413

15.3. Restrictions on lexical verbs with V–V complexes in Tubalar Altai

428

16.1. Subclasses of VVCs in Central and Eastern Turkic

446

16.2. The list of V2s in AVCs in Kyrgyz and Sakha

449

17.1. Delexicalized V2 verbs in gerundive-type V-V complexes in Uyghur

456

17.2. Delexicalized V2 verbs in converb-type V-V complexes in Uyghur

458

17.3. Auxiliaries in gerundive-type V-V complexes in Turkish

462

17.4. Auxiliary verbs in converb-type V-V complexes in Turkish

463

17.5. Auxiliary verbs in Orkhon Turkic V-V complexes

467

18.1. i-second verbs in Avar

476

18.2. gp-second verbs

477

18.3. pc-second verbs

477

19.1. Grammaticalized verbs in Thai

508

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

List of abbreviations 1 2 3 Ø ØABS ABL ABS ACC ACV ADJVZR ADN ADV AES AF AGR AGT ALL AM AOR ASP ASS ASSO ATB AUX BCV BEN BSC BVB CAUS CC CED CF CL CND CNT COM COMP CONCESS COND

first person second person third person zero form Ø-marked absolutive ablative case absolutive accusative (case) auxiliary compound verb adjectivizer adnominal adverb; adverbial aesthetive (experiencer marking) affective voice agreement marker agentive (~ ergative) case allative agreement marker aorist aspect assertive associative case across-the-board auxiliary balance compound verb benefactive basic stem bare verb base causative clause chaining Condition on Extraction Domains conjunctive form classifier conditional continuative comitative complementizer concessive conditional

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

xvi

  

CONJ CONJCT CONV COP CP CS CT CTR CV CVB CVB.ANT CVB.IPFV D/A DAT DEC DEF DESID DF DIR DM DOM DST ECHO ECV EF EMJ EMPH EP ERG ES EXCL F FEM FOC FUT GEN GER GNL GYA H HAB HON HS HUM IA

conjectural conjunctive converb copula conjunctive participle change of state Classical Tibetan controllable compound verb converb anterior converb imperfective converb dative/accusative dative (case) declarative definitive aspect desiderative definiteness marker directional directive marker Domkhar (Shamskat) distance marker; distal deictic form echo word echo compound verb effective voice Early Middle Japanese emphatic entry into a process ergative (agent marking) entry into a state exclamatory feminine feminine focus future (tense) genitive (case) gerund; gerundive general Gya-Miru (Kenhat) high form habitual honorific hearsay humble; humilific Indo-Aryan

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

   IKS IMP IMPF_PTCP IND INDIC INF INFR INST INT INTR IPFV JUSS KHAL LCS LCV LEH LLV LMJ LOC LQ M MANNER MJ MOD MP MPST MSC MW NEG NEUT NH NJ NML NOM NPST NVIS OBJ OBL OIA OJ ONOM OPT OT PA PAS

Indus Kohistan Shina imperative imperfect participle indicative mood indicative infinitive inferential instrumental intentional intransitive; intransitivizer imperfective jussive Khalatse (Shamskat) lexical conceptual structure lexical compound verb Leh (Kenhat) Lower Ladakhi version of the Kesar epic Late Middle Japanese locative (case); locational limiting quantifier masculine subordinate clause of manner Middle Japanese modality multiplicative process modal past masculine measure word negative; negation neuter non-human Modern Japanese nominalizer nominative (case) non-past non-visual perceptive evidence object oblique (case) Old Indo-Aryan Old Japanese onomatopoetic word optative Old Tibetan past (stem) predicate argument structure

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

xvii

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

xviii

  

PASS PAT PBT PC PDJ PER PERF PERF_PTCP PERS PFT PFV PFX PL PNT POL POSS POTEN PP PRES PRESPRT PRF PRG PRM PROG PROV PROX PRS PRSMP PRT PRX PSN PST PST_PTCP PTCP PURP PV Q QM QOM QUOT RCV REFL REL REL_PRON RESP

passive patient possibility perfective converb Present-day Japanese perfect (present) perfect perfective participle person perfect perfective prefix plural present tense polite possessive potential postposition present tense present participle perfect progressive permissive progressive provisional proximate case suffix present (tense) presumptive particle proximal deictic form personal name past (tense) past participle participle purposive perfective question (particle/marker) interrogative quote marker quotative reduplicating compound verb reflexive relative verbal participle relative pronoun respect

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

   RM S SAS SBJ SBJNC SEQ SFP SG SIM SLOC STAT STM SV SVA SVC SVVC TA_PTCP THM TOP TP TPE TR TRNS TYA V VIS VN VOC VOL VOV Vv

remoteness marker state Saspol (Shamskat) subject subjunctive sequential (converb) sentence-final particle singular simultaneous converb superlocative stative stem simple verb subject-verb agreement Serial Verb Construction semantically related verb verb combination ta participle thematic vowel topic transition particle truncated personal ending transitive (verb) transferative Teya (Shamskat) verb visual evidence verbal noun vocative case volitional vocative main verb plus lexicalized verb

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

xix

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

The contributors E. Annamalai did his doctoral work on Tamil syntax at the University of Chicago. His research expanded to the study of multilingualism, language modernization, and politics of language at the Central Institute of Indian Languages, Mysore. He was a research professor for a year at the Institute of Languages and Cultures of Asia and Africa, Tokyo, where he carried out a research project on Tamil predicates. Currently, he is visiting professor of Tamil at the University of Chicago. His publications include Tamil Today: Questions and Answers (in Tamil), Social Dimensions of Modern Tamil, and Dynamics of Verbal Extension in Tamil.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Hirofumi Aoki (D.Litt., Kyushu University) is an associate professor of Japanese language and literature on the Graduate Faculty of Humanities at Kyushu University and an invited associate professor at the National Institute for Japanese Language and Linguistics (NINJAL). He works on the historical change of Japanese grammar and grammaticalization. His publications are exclusively written in Japanese, including two monographs A Historical Study of Word Formation: Its Role and Significance in Japanese Grammar (Hituzi Syobo, 2010) and An Introduction to the Historical Syntax of Japanese (Hituzi Syobo, 2016). Fuyuki Ebata (D.Litt., University of Tokyo) is Associate Professor at the Faculty of Humanities, Niigata University, Japan. His main research interests are in the morphosyntax of Sakha and Tuvan, and he has conducted field research on Sakha and Tuvan languages for almost twenty years. His articles are published in Gengo Kenkyu and Northern Language Studies. Bjarke Frellesvig is Professor of Japanese Linguistics at the University of Oxford. He received his Ph.D. from the University of Copenhagen in 1993. He is the author of, amongst others, A History of the Japanese Language (Cambridge University Press, 2010) and his main research interests fall within Japanese historical linguistics, in particular Old Japanese. He is director and editor-in-chief of the Oxford-NINJAL Corpus of Old Japanese. Peter E. Hook (Ph.D., University of Pennsylvania, 1973) is Professor Emeritus at the University of Michigan. His doctoral dissertation, published as The Compound Verb in Hindi (University of Michigan, 1974), is a pioneering work in the study of Indo-Aryan compound verbs. He has published numerous articles on the syntax, semantics, and grammaticalization of compound verbs and other phenomena in Hindi-Urdu and other languages which substantiate the idea of the “Indo-Turanian” linguistic area involving South, Central, and Northeast Asia. His major articles appeared in Linguistics, Language Sciences, International Journal of Dravidian Languages, Approaches to Grammaticalization (eds E. Traugott and B. Heine), Complex Predicates in South Asian Languages (ed. M. K. Verma), and The Yearbook of South Asian Languages and Linguistics.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

xxii

 

Hyun Kyung Hwang received her Ph.D. from Cornell University in 2011 and is now an associate professor at the University of Tsukuba. Her primary research interests lie in the interface between prosody and phonology/syntax/semantics in Korean and Japanese. Her recent journal papers appeared in Journal of East Asian Linguistics and Korean Linguistics. Taro Kageyama (Ph.D., University of Southern California, 1977) is Professor Emeritus at the National Institute for Japanese Language and Linguistics (Director-General, 2009–17). He is the author or (co-)editor of forty books in the fields of morphology, lexical semantics, and syntax. Among his books in Japanese, Grammar and Word Formation (1993) is a seminal monograph on Japanese compound and complex predicates. His co-edited books in English include Handbook of Japanese Lexicon and Word Formation (2016), Transitivity and Valency Alternations (2016), and Handbook of Japanese Contrastive Linguistics (2018), and his major articles have appeared in Language, Lingua, Journal of Japanese Linguistics, Yearbook of Morphology, The Oxford Handbook of Compounding, and Oxford Research Encyclopedia of Linguistics (editorial adviser).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Hideki Kishimoto (Ph.D., Kobe University, 1991) is Professor of Linguistics at Kobe University, Japan. His research interest includes lexical semantics and syntax, on which he has published numerous articles in such journals as Language, Linguistic Inquiry, Natural Language and Linguistic Theory, Journal of Linguistics, The Linguistic Review, Studies in Language, Journal of East Asian Linguistics, and Gengo Kenkyu. He authored Syntactic Structures and Grammatical Relations (in Japanese; Kurosio, 2005) and co-edited Handbook of Japanese Lexicon and Word Formation (De Gruyter Mouton, 2016). Yuu Kuribayashi (D.Litt., Kobe University, 2010) is a professor of linguistics at the Graduate School of Humanities and Social Sciences, Okayama University. His research interests are in the morphology and syntax of Turkish and Turkic languages. His publications include a monograph, Structure and Description of the Southwestern Group of Turkic Languages (in Japanese; Kurosio, 2010) and book chapters “Causative/Anti-Causative Alternations in Turkish, Old Turkic and Khalaj” (Ankara Papers in Turkish and Turkic Linguistics, 2015) and “Verb-Verb Compounding in Japanese and Turkish” (Handbook of Japanese Contrastive Linguistics, 2018). Hsin-hsin Liang (Ph.D. in linguistics, University of Michigan) is Professor of Chinese and former Director of the Chinese Language Program in the Department of East Asian Languages, Literatures and Cultures, University of Virginia. Her research interests are in the pedagogy of Chinese, teacher training, and curriculum design, as well as in Chinese linguistics. Her recent publications include Chinese as a Second/Foreign Language in the Study Abroad Context, and four co-authored volumes entitled Success in China: Mandarin for Living (all from Beijing University Press). Yo Matsumoto (Ph.D. in linguistics, Stanford University, 1992), is Professor of Linguistic Theory and Typology at the National Institute for Japanese Language and Linguistics (NINJAL), Tokyo. He has worked extensively on various topics of lexical semantics and its relation to syntax, morphology, pragmatics, and cognition in general. His publications include the monograph Complex Predicates in Japanese (CSLI Publications, 1996) and numerous journal papers in Linguistics and Philosophy, Linguistics, Journal of East Asian

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 

xxiii

Linguistics, Cognitive Linguistics, and Gengo Kenkyu. He has also co-authored or edited numerous books in Japanese. Noriko Ohsaki received her Ph.D. in linguistics from Kyoto University in 2006 and is currently a researcher at the Haneda Memorial Hall at Kyoto University. Her research interest is in Kyrgyz morphology and syntax with special focus on the morphosyntax of verbs. She is a co-author of The Berlin Chinese Text U 5335 Written in Uighur Script (Brepols, 2015). Prashant Pardeshi (Ph.D., Kobe University, 2000) is a professor in the Division of Linguistic Theory and Typology, the National Institute for Japanese Language and Linguistics. He is interested in functional linguistic typology and has published articles in such journals as Journal of Pragmatics, Linguistics, Journal of Japanese Linguistics, Annual Review of South Asian Languages and Linguistics, and Gengo Kenkyu, as well as in collections of articles published in the West and in the East. He is also the author of A Functional Account of Marathi’s Voice Phenomena (Brill, 2016) and a co-editor of Handbook of Japanese Contrastive Linguistics (De Gruyter Mouton, 2018).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Michinori Shimoji is Associate Professor of Linguistics at Kyushu University and Invited Associate Professor at the National Institute for Japanese Language and Linguistics (NINJAL). He received his Ph.D. in Linguistics from the Australian National University in 2009 and his dissertation was published as A Grammar of Irabu, a Southern Ryukyuan Language (Kyushu University Press, 2017). His major research interest is in linguistic typology based on fieldwork on Japonic languages, especially Irabu Ryukyuan and the Kyushu dialects of Japanese. He co-edited Handbook of Ryukyuan Languages (De Gruyter Mouton, 2015). Andrey Shluinsky is a deputy director of the Institute of Linguistics, Russian Academy of Sciences (Moscow), and a team member on a research project at Moscow State University. He received his MA (2002) and Ph.D. (2005) from Moscow State University for crosslinguistic studies of clause-chaining and pluractionality. His research projects are related to the documentation and description of West African (Kwa, Mande) and North Eurasian (Samoyedic, Turkic) languages, as well as to linguistic typology. Benjamin Slade is Associate Professor of Linguistics at the University of Utah. He received his Ph.D. from the University of Illinois at Urbana-Champaign in 2011. In addition to verbverb collocations in Indian languages, he currently works on the semantic and structural analysis of aspectual adverbials crosslinguistically, and computational approaches to focus detection and identification. His past work includes studies of quantifier particles and epistemic indefinites in Sinhala and other languages, morphological processes in Rastafari Dread Talk, and the rise and spread of morphological and orthographic innovations in the cyberpunk subculture. Sanford B. Steever received his Ph.D. in Linguistics from the University of Chicago in 1982. His research interests include syntax and comparative linguistics, particularly of Tamil and the Dravidian languages. He has published numerous articles and reviews, as well as five books, including Analysis to Synthesis (Oxford University Press, 1993), The Dravidian

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

xxiv

 

Languages (Routledge, 1998), The Tamil Auxiliary Verb System (Routledge, 2005), and Dravidian Syntactic Typology (Dravidian Linguistics Association, 2017). Kingkarn Thepkanjana (Ph.D. in linguistics, University of Michigan, 1986) is Professor of Linguistics at the Faculty of Arts, Chulalongkorn University, Bangkok, Thailand. Her research interests include cognitive and typological studies of Asian languages, especially Thai and Chinese. Her publications include an authored book written in Thai titled Grammaticalization (2016). Some of her research articles are published in international journals such as Linguistics, Language Sciences, Mon-Khmer Studies, Taiwan Journal of Linguistics, and Language and Linguistics. Satoshi Uehara is Professor of Japanese and Linguistics at Tohoku University. His Ph.D. dissertation (University of Michigan) was published as Syntactic Categories in Japanese: A Cognitive and Typological Introduction (Kurosio Publishers, 1998). His research interests include cognitive linguistics and linguistic typology, and his research focuses on Japanese and East and Southeast Asian languages.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

John Whitman (Ph.D. in linguistics, Harvard University, 1985) is professor and chair of the Cornell University Department of Linguistics. His specializations are East Asian linguistics, comparative syntax, language typology, and historical linguistics. Recent publications include Korean: A Linguistic Introduction (2019, Cambridge, with Sungdai Cho), “Topic Prominence” in the Blackwell Companion to Syntax (2017, with Waltraud Paul), and the co-edited volume Ryūkyū shogo to Kodai Nihongo: Nichiryū sogo no saiken ni mukete (Ryūkyūan and Premodern Japanese: Toward the Reconstruction of Proto-JapaneseRyūkyūan, 2016, Kurosio, with Tatsuya Hirako and Yukinori Takubo). Hisanari Yamada did his graduate work at the University of Tsukuba and is Professor at the Center for Language Studies, Otaru University of Commerce, Japan. His research focuses on Avar syntax and morphology from a crosslinguistic perspective. He has published papers on reciprocal and reflexive constructions, comitative constructions, periphrastic verb formation, and case-marking and agreement patterns in simplex and periphrastic verb constructions. Bettina Zeisler, who did her Ph.D. at the Freie Universität Berlin in 1999, has taken up several research positions at the Universität Tübingen, sponsored by the Deutsche Forschungsgemeinschaft. She was the principal investigator in research projects on crossclausal reference, on verb valency, and on evidentiality or speaker attitude in Tibetic languages, particularly Ladakhi. She has been conducting fieldwork in Ladakh for almost twenty years and has published on various topics concerning the Ladakhi language as well as on Old and Classical Tibetan.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

1 Introduction Taro Kageyama, Peter E. Hook, and Prashant Pardeshi

1.1 Scope and purpose of this volume While European languages generally employ affixes or particles to expand the meaning of a single verb into a more complex one narrating how an event unfolds, as in English misread [fail to execute an intended action] or Dutch door-zingen [through-sing] ‘keep on singing’ [continuation], Asian languages predominantly resort to verb-verb sequences to achieve the same purpose or to convey even more complex meanings involved in the unfolding of events. This volume presents a comprehensive survey of the systems of multiverb predicates that characteristically occur across languages of Asia. The languages discussed in the body of the volume are listed in (1) (see also the map in Figure 1.1 in Section 1.4).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

a. East Asia (Part I): Japonic languages (Japanese, Ryukyuan), Ainu,1 Korean b. South Asia (Part II): Indo-Aryan languages (Hindi-Urdu, Marathi, Sinhala, Gujarati, Nepali, Kashmiri, etc.), Dravidian languages (Tamil, Kannada, Telugu, Malto, Malayalam, Parji, Pengo, etc.), Tibeto-Burman (Limbu), Tibetic languages (Tibetan, Ladakhi), Austro-Asiatic (Munda), Kasunda (isolate) c. Central and Northwest Asia (Part III): Turkic languages (Bashkir, Tatar, Chuvash, Khakas, Uzbek, Tubalar Altai, Karachay-Balkar, Kyrgyz, Sakha, Tuvan, Ottoman Turkish, Uyghur, Kazakh), NE Caucasian languages (Avar, Tsez)2 d. East and Southeast Asia (Part IV): Thai, Chinese

Although diverse types of verb-verb sequences are observed in the languages of Asia as reported in the individual chapters that follow, the central concern of the volume resides in a special type of V-V complex consisting of a main verb that ¹ Spoken in northern Japan (previously in Sakhalin as well), Ainu is a critically endangered language unrelated to Japanese. It is not included in this volume. See Bugaeva (2018: 266), who discusses its main verb + auxiliary verb sequences, some of which seem to have been present since the thirteenth century. ² Technically, the Caucasus area where NE Caucasian languages are spoken is part of Europe.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

2

. , . . ,  . 

provides full lexical meaning and full argument structure and a “quasi-auxiliary” verb that retains its grammatical function as a verb in exhibiting tense and other inflections but is semantically bleached to express a variety of functional notions that would not be properly subsumed under the standard notions of tense, aspect, and modality. Such quasi-auxiliary verbs, termed “vectors” (Hook 1974), “explicators” (Masica 1976), or “light verbs” (Butt 2003), recur characteristically across Asian languages and are considered to have a unique status distinct from both full lexical verbs and genuine auxiliaries of aspect and modality. Other verb-verb sequences, such as “verb + modal auxiliary,” “verb + phasal verb,” and serial sequences “full verb + full verb,” which are outside the main scope of this volume, may also be taken up in some chapters for comparison. However, “verbal noun + light verb” sequences and prefixed or suffixed verbs, which are at times included in previous studies on complex predicates, are systematically excluded. The term “verb-verb complex” in the book title is a cover term for such multiverb predicates, which may or may not form morphological compounds. The present study is thus aligned with the recent surge of interest in compound and complex predicates within and across languages. This line of research, sparked by Alsina et al. (1997) and followed by Amberber et al. (2010), Nash and Samvelian (2016), Nolan and Diedrichsen (2017), and other works, is of special significance in deepening our understanding of how human languages conceptualize and express the complex processes of the unfolding of events. Compared with the above-mentioned works, which are oriented toward theoretical formalization of the mechanisms unifying the semantic and syntactic properties of multiple predicates in a single clause, however, the present volume is more restrictive and more specific in its scope and goal. In terms of scope, the principal focus centers on a special type of verb-verb complex called “vector verb construction” (to be explained in section 1.2) that is widely spread across languages of Asia. As to the goal, the volume aims to broaden our understanding of language universality and areal typology by testing the empirical feasibility of an overarching hypothesis advanced by Masica (1976). In his pioneering monograph Defining a Linguistic Area: South Asia, Masica (1976) puts forth the idea of identifying the gigantic landmass ranging from South Asia to Central and East Asia as the “Indo-Altaic linguistic area,” covering a large part of Asia including the languages mentioned in (1a, b, c) but excluding those in (1d). Masica’s hypothesis is built on the observation that OV word order, bound causative verbs, conjunctive participles, dative subject constructions, and verb-verb complexes which he calls the “explicator compound verb” cluster together in this area. Of these, we sort out the last feature as the most important in characterizing Asian languages, while leaving open its possible correlation with the other features. The twofold delimitation in function and geographic distribution as just explained will make it possible to resolve two independent but ultimately related questions that have long confronted researchers in Japanese linguistics, on the one

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi



3

hand, and in South Asian linguistics, on the other. In Japanese linguistics, the historical development of the V-V compound verbs that abound in Present-day Japanese has long been a puzzle since Kindaichi (1953) denied their existence in Old Japanese (see Chapters 3 (Aoki and Frellesvig) and 4 (Kageyama)). Likewise in South Asian linguistics, the origin and geographic distribution of innovative verb clusters composed of a main verb and a quasi-auxiliary verb (vector verb), which are not documented in Sanskrit, remain as open questions (see Chapters 9 (Pardeshi) and 10 (Slade)). A unified solution to these two questions has the potential of entailing far-reaching consequences that go beyond the boundaries of Asian linguistics, thus helping to demystify, for example, the multiverb constructions in such little-studied languages as Kichwa (Andean), Omotic (NE African), and Wardaman (NW Australian). Furthermore, the data and analyses presented in this volume are expected to contribute to sharpening the general understanding of universal issues in the grammaticalization of verb categories, complex predicate formation, aktionsart and event semantics, morphology-syntax-semantics interactions, areal linguistics, and linguistic typology.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

1.2 Vector verb constructions Many Asian languages feature a special type of V-V complex in which a secondary element is (i) the desemanticized form of a basic lexical verb or adverb³ and (ii) alternates with its own absence. The constructions are illustrated with Japanese (2a, b), Korean (3a, b), Hindi-Urdu (4a, b), and Marathi (5a, b) examples.⁴ In the English glosses of numbered examples here and elsewhere in the volume, the literal meaning of a desemanticized verb is represented in small capitals. Japanese (2)

a. subete o wasure-te simai tai , b. subete o wasure tai all  forget- . want all  forget want ‘(I) want to forget everything’. ‘(I) want to forget everything’.

³ Verbs optionally extended with desemanticized adverb paro ‘away’ are found in South Rajasthani dialects spoken in Raniwara and Tartoli: hathi maro (paro) ‘The elephant died ()’ (Hook 1988: 177) and in the Himachali dialect Kotgarhi: (poru) mŏru ‘Will I die ()?’ (Hendriksen 1979: 25). ⁴ The optional elements are identified by the meanings (in small caps) of the basic lexical items from which they derive. Sources of online examples: (1a) Carol Marinelli. 2018. kisu wa yami ni magirete (A kiss under cover of the dark). Toronto: Harlequin; (1b) Yamaguchi Yōko. 2007. mō ichido waratte yo (Laugh one more time). Tokyo: Skyviewplanning. p 25; (2a) https://brunch.co. kr/@parkdabin/527; (2b) https://allklyrics.com; (3a) Shakeel Badayuni. from film dūr kī āwāz (A voice from afar); (3b) https://pann.nate.com/talk/339748998; (4a) Vijaya Rajadhyaksh. 2017. akhertse parv (The final stage). Pune: Mehta Publishing House; (4b) https://misalpav.com/node/ 43829.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

4

. , . . ,  . 

Korean (3)

a. da ic-e peli-ko sip-ta , b. da ic-ko sip-ta all forget- - want- all forget- want- ‘(I) want to forget everything’. ‘(I) want to forget everything’.

Hindi-Urdu (4)

a. sab.kučh bhūl jā-nā čāh.tā hũũ , b. sab.kučh bhūl-nā čāh.tā hũũ all forget - want am all forget- want am ‘I want to forget everything’. ‘I want to forget everything’.

Marathi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(5)

a. sag:la visar-un dzā.y-tsa all forget- - ‘(I) must forget everything’.

,

b. sag:la visar.āy-tsa all forget- ‘(I) must forget everything’.

Compared with the simple verbs in (2b)–(5b) on the right, the predicates in (2a)–(5a) on the left additionally have delexicalized, quasi-auxiliary verbs, shown in boldface. These quasi-auxiliary verbs serve to add to the main verbs’ conceptual meanings some subtler emotional overtones like ‘frustration’ or ‘relief ’. The Japanese example in (2a), for example, implies that the speaker will feel relieved if he can forget all the bad things that worry him. As the translations of each (a–b) pair are intended to show, such a semantic distinction between simple verbs and their extended counterparts is not easily or systematically made in languages such as English, French, or Sanskrit, which lack vector verbs. Unlike the V-Vs in (2a)–(5a) in which presence or absence of the quasi-auxiliary verb concerns aspectual and pragmatic overtones, there are other verb-verb sequences, as exemplified in (6) and (7), in which deletion of the quasi-auxiliary verb causes greater differences in conceptual meaning. ,?

(6)

a. sooan-o kaki-nagut-ta. draft- write-- ‘(X) dashed off a draft’.

b. sooan-o kai-ta (Japanese) draft- write- ‘(X) wrote a draft’.

(7)

a. agar: am.bagar: am likh mār-ā ,? b. agar: am.bagar: am likh-ā (Hindi-Urdu) nonsense write - nonsense write- ‘(X) dashed off nonsense’. ‘(X) wrote nonsense’.

The quasi-auxiliary verbs in (6a) and (7a), originally meaning ‘to beat, strike’ when used as independent verbs, are desemanticized to express adverbial meanings like the manner (‘roughly’) or promptness (‘hastiness’) of action when they

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi



5

are combined with certain main verbs. Such adverbial meanings are not profiled in the corresponding simplex verbs (6b, 7b). In still other cases, as in Marathi (8a) and Japanese (9a), verb-verb complexes depend on the presence of each verbal element to express meanings that are “noncompositional”: (8)

a. šatru-var tut:-un enemy-on - ‘X attacked the enemy’. (www.tarunbharat.net)

pad.li : 

≠

b. šatru-var tut:.li ≠ enemy-on broke ‘X broke on the enemy?’

c. šatru-var pad.li : enemy-on fell ‘X affected the enemy’ (oldhistoricity.lbp.world) (9)

a. yuusikin-ga koge-tui-ta loan- -- ‘The loaned money became unrecoverable’

≠

b. yuusikin-ga koge-ta ≠ loan- burn- ‘The loaned money burned’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. yuusikin-ga tui-ta loan- attach- ‘The loaned money was secured’ (8a) and (9a) are examples of lexicalized idioms in which both members of a V-V complex are desemanticized, yielding a new meaning as a constellation of verbs. While a variety of V-V sequences are discussed in the chapters that follow, the major concern of the volume is verb-verb complexes of the first type (2–5) and second type (6–7) composed of a main verb with full lexical content and a desemanticized quasi-auxiliary verb. Setting aside exceptions like EAT, DIE, and KILL, since most of the secondary verbs that recur across languages originate from primary verbs of locomotion (e.g. COME, GO), positional change (e.g. SIT, RISE, FALL), displacement (e.g. PUT, THROW, SEND), and change in possession (e.g. GIVE, TAKE) and to a greater or lesser extent share the notion of directionality that is implicit in the word vector, we adopt vector verb⁵ as the term used to refer to the desemanticized secondary verbs seen in (2a–5a) and (6a–7a). The vector verb construction is now generalized as a formmeaning pair in (10).

⁵ While Pray (1970) may be credited as the originator of the term “vector,” this term, however, has not gained global recognition outside of its homeland, South Asian linguistics.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

6

. , . . ,  . 

(10)

The vector verb construction a. Form: main, lexical verb + secondary, delexicalized verb (vector) b. Meaning: The delexicalized verb (vector) adds a diversity of functional or pragmatic meanings/connotations to the lexical meaning of the main verb.

As a consequence of the partial or even total bleaching of the meanings of their lexical counterparts, vector verbs have evolved functions such as the expression of aspect, manner and intensity of action, benefaction, attitude, derogation, or politeness (see Chapter 2 (Kageyama) for examples of the last). Verb-verb complexes of the idiomatic or semi-idiomatic type in (8a) and (9a), common in East Asia, rarer in South Asia, are discussed in Chapters 4 (Kageyama on Japanese), 8 (Hwang and Whitman on Korean), and 9 (Pardeshi on Marathi). A fourth type of verb-verb complex in which the second element refers to the response or reaction to an action expressed by the first is examined in Chapters 19 (Thepkanjana and Uehara on Thai) and 20 (Liang and Hook on Chinese).

1.3 Descriptive and theoretical issues: a brief overview

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The delexicalized (vector) verb constructions raise a number of questions that pertain not only to Asian languages but also to general theories of morphology, syntax, semantics, pragmatics, and the lexicon. Some of the questions discussed in the volume are listed in the following section.

1.3.1 Questions about morphological and syntactic conditions • Order of two verbs Normally, the main verb precedes its vector in the languages discussed in this book. As Hook (1974: 55–64) pointed out for Hindi-Urdu, however, reversal of order is observed in rare cases where the vector shows up to the left of its main verb. Relevant discussion will be found in Chapter 3 for Old Japanese (Aoki and Fellesvig) and Chapter 12 for Tamil (Annamalai). • Nonfinite forms in V1 As a rule, a main verb takes a nonfinite form. Different languages utilize different nonfinite strategies such as infinitive, gerund/gerundive, unmarked bare stem, participial, converb, and conjunctive, and a single language may allow two or more options, depending on the type of extended verb (see individual chapters for concrete strategies). Regardless of the nonfinite strategies, it is a vector verb that selects the morphological shape of a main verb. In fact, the suffix (or its absence) on the main verb may best be considered as an integral, albeit discontinuous part of the vector verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi



7

• Biclausal or monoclausal Because of their semilexical status, vector verbs preserve the syntactic behavior of the original verbs to varying degrees. Because of this, some V-V sequences are analyzed as having a biclausal complementation structure in syntax. The clausality issue is addressed specifically in Chapter 5 (Kishimoto) for Japanese, Chapter 7 (Shimoji) for Ryukyuan, Chapter 8 (Hwang and Whitman) for Korean, and Chapter 14 (Zeisler) for Tibetan.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

• Word status of V-V complexes Languages differ as to whether the combinations of main and vector verbs in them are morphological compounds or freely formed syntactic concatenations, and even a single language, such as Japanese (Chapters 2 (Kageyama) and 6 (Matsumoto)), may have both compounds and syntactic concatenations. Morphological words typically exhibit syntactic non-interruptibility (so-called “lexical integrity”). In theory the tighter the link between them, the harder it becomes to separate them by insertion of particles or other syntactic items. The chapters that follow will clarify the morphological status of V-V sequences in individual languages. Variation in morphological status accounts for the notational variation in chapter titles: “verb-verb complexes” (with a short hyphen), “verb–verb complexes” (with a long hyphen (‘n’ dash); Chapter 15 (Shluinsky)), “verb + verb sequences” (with a plus sign; Chapter 13 (Steever)), and “verb verb complexes” (with a space; Chapters 3 (Aoki and Frellesvig) and 14 (Zeisler)). • Restrictions on collocability Depending on the degree of grammaticalization, some vectors combine with main verbs in V1 rather freely, while others are more selective in combinations with V1s. In general the more desemanticized (“bleached”) the vector is, the greater the range of V1s it is free to co-occur with (Chapter 11 (Hook)). On the other hand, vector verbs that retain much of their syntactic properties such as transitivity/ ergativity may participate in the determination of the transitivity of the whole clause or in the selection of subject or object nouns. Such syntactic restrictions are discussed in many of the chapters that follow.

1.3.2 Lexical and semantic issues • Semantics of vector verbs Despite the crosslinguistic differences in the size and content of the vector verb roster, some of them appear to be governed by possibly universal semantic concepts. In (11), representative meanings are grouped into five categories.⁶ ⁶ An alternative typology may be found in Indo-Aryan and Dravidian languages (Abbi and Gopalakrishnan 1991).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

8

. , . . ,  . 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(11) a. Lexical aspect: ‘perfective’, ‘completive’, ‘continuative’, ‘incompletion/failure’, ‘gradualness’, etc. b. Direction, manner, degree: ‘away from speaker or protagonist’, ‘toward speaker or protagonist’, ‘suddenly, abruptly’, ‘without volition, inadvertently’, ‘deliberately’, ‘with difficulty’, ‘with ease’, ‘casually, carelessly’, ‘violently, intensely’, ‘exhaustively’, etc. c. Attitude: ‘regret about an unwanted occurrence of irreversible action’, ‘preparatory action in anticipation of a future event’, etc. d. Benefaction, malefaction: ‘benefit or harm to the speaker’, ‘benefit / harm to the hearer/other’, etc. e. Speech style or register: ‘respect, politeness, contempt, condemnation, humiliation, exasperation’, etc. The examples given earlier in (2)–(5) represent the attitudinal class (11b), where the speaker expresses his subjective attitudes or feelings about the execution of a main verb. Viewed from a crosslinguistic perspective, the relation between the basic lexical meaning of a putative source verb and its extended meaning as vector is not fixed. A given lexical verb may give rise to different concepts in different languages. Vector GIVE, for example, may represent a benefaction (or malefaction) in many languages (Japanese, Hindi-Urdu), but suddenness or ease of action in another (Anatolian Turkish). Vector SEE in Japanese (Chapter 2 (Kageyama)), Korean (Chapter 8 (Hwang and Whitman)), and Avar (Chapter 18 (Yamada)) means a tentative trial to judge feasibility (e.g. ‘He tried opening the door’), while the same verb in Tubalar Altai and Sakha (Chapters 15 (Shluinsky) and 16 (Ohsaki and Ebata)) has a conative meaning like ‘He tried to open the door (but couldn’t)’. Conversely, one and the same concept, say, perfectivity or completeness of an action, may be expressed by GO/ TAKE/GIVE in Hindi, Panjabi, and Bengali, by TAKE/COME/PUT ON in Kashmiri, GO/LEAVE/PUT in Tamil, PUT/GO/THROW in Malayalam, and GO/RELEASE in Kannada (Abbi and Gopalakrishnan 1991). The precise evolution in semantics from the original physical meaning to the new functional meaning remains unclear. However, from examples like Hindi-Urdu’s SIT ‘regret’ and Korean’s EAT ‘disgust’ it is clear that at times what happens is not a linear bleaching, emptying, or loss of original meaning but may instead involve the rise of novel unexpected meaning. • Lexical correspondence and grammaticalization A hallmark of vector verbs is that each of them, in principle, derives from a basic lexical counterpart, a full verb. From the viewpoint of grammaticalization, where a

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi



9

full verb is thought to undergo change in its status and function to a bound auxiliary, the pairwise correspondence of a vector and its homophonous full-verb counterpart will be reasonably interpreted as suggesting that the desemanticized variant is at an intermediate stage on a grammaticalization cline, as schematically shown in (12): (12) full (lexical) verb

>

vector verb

> auxiliary > affix (cf. Hopper and Traugott 2003: 107)

An alternative model proposed by Butt (2010) and Butt and Lahiri (2013) looks like (13), where a light-verb (=vector) counterpart is paired with its main-verb counterpart in a single lexical entry. This representation is based on the hypothesis that the paired lexical and light verbs are always “form-identical” through time. (13)

main verb > auxiliary > clitic > affix

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

light verb

(Butt 2010: 68)

In this model, if the composite lexical entry as a whole is deleted from the grammar of a language in historical change, it is expected that both the main- and the light-verb use will cease to exist at the same time (Butt and Tantos 2004: 126). Which of the two models in (12) and (13) is more feasible is an important issue that has nontrivial implications for grammaticalization and language change and hence must be carefully examined on the basis of empirical data. A case in point will be the Korean vector malda and the Japanese pejorative vector yagaru, where vectors of obscure origin survive as orphans. The problem is addressed specifically in Chapters 2 (Kageyama), 3 (Aoki and Frellesvig), and 4 (Kageyama) for Japanese and in Chapters 9 (Pardeshi) and 10 (Slade) for IndoAryan languages.

1.4 Geographical distribution of vector verbs in Asia The notion of the vector verb plays a pivotal role in unifying the diverse languages sampled in (1a) (East Asian languages), (1b) (South Asian languages), and (1c) (Central and Northwest Asian languages) as forming a linguistic area, called the “Indo-Turanian linguistic area” (see Masica 1976 and Hook 1987 for detailed discussions on the linguistic criteria for bundling diverse languages of different genealogical origins together as a linguistic area). On the map in Figure 1.1, the forward slashes (///) mark the Indo-Turanian linguistic area, where vector verbs are used productively (the “explicator-type” in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

10

. , . . ,  . 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Masica’s original terminology). This map⁷ is created on the basis of Masica’s map 5 (1976: 151) but with some significant modifications. For example, the Tibetic languages of the Himalayas, which are left “blank” in Masica’s (1976) map, are indicated by forward slashes because they are shown to have vector verbs, albeit in modest numbers (see Chapter 14 (Zeisler) for discussion). Chapter 16 (Ohsaki and Ebata) shows that Sakha (also known as Yakut), a Turkic language spoken in the Sakha Republic of Russia, has vector verbs, although it does not appear on the map, as its location is far to the north, facing the Arctic Ocean. As Masica (1976) assumed, South Asian influence extended westward through contact with languages of Central and Northwest Asia but stopped at Turkish, which exhibits only an impoverished inventory of vector verbs (see Chapter 17 (Kuribayashi)). In the languages of East Asia, on the other hand, where there is no historical evidence for language contact with South Asian languages, their vector verbs seem to have developed internally (or possibly through language contact within the area). See Chapters 3 (Aoki and Frellesvig) and 4 (Kageyama) for Japanese, Chapter 7 (Shimoji) for Ryukyuan, and Bugaeva (2018) for Ainu. The backward slash banding (\\\), extending from Chinese to Southeast Asian languages including Thai and characterized as “Action-reaction type” on the map, covers the areas where V-Vs operate on an absolutive pivot: In them, while vectors modify the actions of intransitive subjects, in transitive clauses vectors refer to the reactions of direct objects. See Chapters 19 (Thepkanjana and Uehara) and 20 (Liang and Hook; especially page 436, footnote 33). Languages like Lahu and Burmese, spoken in the area on the map where forward and backward slash marks overlap, exhibit both nominative and absolutive pivots in forming their V-V sequences.

1.5 Prospect The chapters that follow present data on at least three major types of delexicalized verb constructions: vector verb constructions, idiomatic V-V constellations, and action-reaction constructions. The range of data adduced is so wide as to resist coherent generalization, and accordingly we do not attempt one. Rather, it is our hope that the descriptions of V-V complexes in Asian languages put forward here will inspire future study and publication. Although this volume is devoted exclusively to Asian languages, research on vector verbs and similar constructions should not be limited within this geographical area. Masica (1976: 148) has already pointed out analogous expressions in English colloquial speech like go and V expressing a “negative sentiment regarding the action” (e.g. See what you’ve gone and done.) and take and V representing ⁷ Special thanks to Ms. Miyako Hara and Ms. Yoko Kasai (both from NINJAL) for creating Figure 1.1 for the present volume.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Figure 1.1 Delexicalized verb constructions in Asia

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

12

. , . . ,  . 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

suddenness (e.g. He took and left.). Another examples is She up(ped) and married someone else. Besides English, some other northern European languages (Swedish, Russian) marginally exhibit analogous “double verb” constructions expressing suddenness and surprise: [Swedish] han gick och dog ‘He  and died’, [Russian] dožd’ vzjal da i perestal ‘The rain  and (suddenly) stopped’. See Wiklund (2009) for Swedish, Weiss (2012) for comparative discussion of “double verb” in some north European languages, and Stoynova (2007) for even broader comparisons. It is hoped that the present volume provides a window from which to view analogous constructions reportedly available in non-Asian languages on common ground.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

PART I

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

L A N G U A G E S OF E A S T A S I A

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

2 Between lexical verbs and auxiliaries The architecture of Japanese verb-verb complexes Taro Kageyama

2.1 Introduction

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Among Asian languages, Japanese has an especially rich system of verb-verb complexes.¹ In most cases, the second verb members are semantically bleached in varying degrees to represent a wide range of functional meanings while maintaining their verb category in the conjugational and inflectional paradigms. This chapter will explore the nature of the “semilexical” status of V2 verbs in the Japanese V-V complexes by dividing them into four major types according to three crosscutting criteria: (i) Morphological criterion. Japanese V-V complexes are characterized as a contiguous sequence of two verbs joined by one of two conjugated/ inflected forms of V1, and the choice of the form determines their status as morphological compound or syntactic concatenation. One is the infinitive, also known as ren’yōkei (lit. ‘verb-modifying form’) in traditional Japanese grammar. It is a conjugational ending manifested as the vowel -i appended to the bare stem of a verb if it ends in a consonant, as in kaer.ikakeru [go.home.-] ‘be about to go home’; otherwise, it is realized as “zero,” whereby the infinitive verb takes the same form as the bare stem, as in tabe-kakeru [eat.-] ‘be about to eat’. The infinitive verb may be used in clause combining, as in Kaze ga huki, ame ga hut-ta [wind  blow, rain  fall-] ‘It blew and rained’), or as a base of compound formation, as in compound nouns like tabe-mono [eat.-thing] ‘food’ and in the verb-verb compounds discussed in this chapter. The other conjunctive form is the gerundive, which is signaled by the ending -te (or its phonetic variant -de that occurs after verb stems ending in voiced consonants [g], [b], [m], [n]: see Section 2.3.2). This ending is considered ¹ A survey of studies on Japanese compound and complex predicates is found in Kageyama (2018), and an overview of Japanese morphology and predicate agglutination in Kageyama (2020).

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

16

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

a nonfinite inflection corresponding to the finite counterpart -ta (or its voiced variant -da) for past tense. Like the past tense inflection, the gerundive inflection attaches to the infinitive conjugation of V1, as in kas.i-te ageru [lend.- ] ‘lend to other’ and tabe-te simau [eat- .] ‘finish eating, eat up’.² The gerundive form characterizes Type 4 complexes, and the infinitive form all the other types. (ii) Syntactic criterion. There is a general rule that no syntactically motivated element can be embedded inside a compound word. While Types 1 and 2 conform to this regulation, Types 3 and 4 deviate from it in their capacity to house in their V1 position such syntactically motivated elements as passive, causative, and honorific verbs. The first two types are monoclausal, while the last two are considered to originate from biclausal syntactic structures. (iii) Semantic criterion. While the verb category is invariably determined by the V2 member, the locus of the semantic core that determines the conceptual meaning and argument information of the whole complex may vary between V1 or V2. Only in Type 1 compounds do the V2s qualify both as categorial head and as semantic core; in the other types, the V1 verbs are the semantic core, to which the V2 verbs supply a variety of functional meanings. The three criteria lead to postulating four similar but distinct types, as shown in Table 2.1.³ The morphological criterion separates “Type 4 V-te V complex predicates” (Section 2.3) from the other V-V types, which count as compound words. Within the compound group, the syntactic criterion differentiates the monoclausal, lexical compounds of Type 1 and Type 2 from “Type 3 syntactic compound verbs” (Section 2.5), which are biclausal like Type 4 complex predicates. Finally, within the lexical compound verbs (Section 2.6), “Type 1 lexical thematic compound verbs,” composed of two lexical verbs, are separated from “Type 2 lexical aspectual compound verbs,” whose second members (V2) are virtually empty of argument structure and are semantically bleached in much the same way as in Type 3 and Type 4. ² The view that -te as well as -ta attaches to the infinitive conjugation of a verb is based on the conjugational paradigm of traditional Japanese grammar and is substantiated by those verbs whose bare stems end in [s], as in kas.i-te [lend.-] and kas.i-ta [lend.-]. Bloch (1946), on the other hand, holds that -te and -ta attach directly to a verb’s bare stem, in which case, however, two allomorphs need be postulated for the bare stems that end in [s], as in kas- ~ kasi-. Regardless of which analysis one adopts, phonetic adjustments (called onbin ‘euphonic sound change’) have to be postulated for those verbs whose bare stems end in /r, w, k, g, b, m/, as in: (i) . . . r/w- !~ . . . t- (e.g. hor- ‘dig’ ! hot-te [dig-], kaw- ‘buy’ ! kat-te [buy-]), (ii) . . . k/g- ! . . . i- (e.g. kak- ‘write’ ! kai-te [write-], kag- ‘smell’ ! kai-de [smell-]), and (iii) . . . b/m- ! . . . n- (e.g. yob- ‘call’ ! yon-de [call-], yom- ‘read’ ! yon-de [read-]). ³ A fifth type consisting of auxiliary verbs of politeness and humiliation will be introduced in Section 2.7.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

17

Table 2.1 Four types of V–V complexes in Japanese T 1 (lexical thematic compounds)

T 3 T 2 (syntactic (lexical compounds) aspectual compounds)

T 4 (V-te V complex predicates)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(i) morphological criterion compound word syntactic cluster (ii) syntactic criterion monoclausal biclausal (iii) semantic criterion V2’s meaning: V2’s meaning: bleached (except for motion literal subtype in Type 4)

From the viewpoint of grammaticalization, Types 2, 3, and 4 stand out as a unique group in having their V2 verbs semantically bleached. Their bleached meanings involve a bewildering diversity of functional concepts ranging from temporal aspect and spatial aspect (direction of action) to intensity or manner of action/state to subjective attitudes toward V1’s event/action, social deixis (benefaction), and other discourse-related concepts. These functional meanings do not correspond straightforwardly to the familiar TAM (tense-aspect-modality) categories associated with auxiliaries across languages of the world. The V2s in Types 2–4 are thus neither full-fledged lexical verbs nor purely grammatical auxiliaries. Rather, they are intermediate between lexical categories (content words) and functional categories (grammatical morphemes) and hence fall under the rubric of “semi-lexical categories” (Corver and van Riemsdijk 2001). Importantly, however, the Japanese semilexical verbals do not form a unitary class but are themselves divided into a few different subtypes depending on their morphosyntactic behavior and phonetic reduction, as explained in Sections 2.2–2.6. The existence of disparate subclasses of semilexical verbals has interesting repercussions for typological studies of V-V complexes as well as for theoretical discussions on the verb-to-auxiliary grammaticalization along the lines of Heine (1993), Kuteva (2001), and Anderson (2006) vis-à-vis Butt’s (2010) theory of light verbs (see Section 2.7).

2.2 Compound verbs vs. complex predicates Drawing on the morphological criterion introduced in Section 2.1, Japanese V-V complexes are partitioned into morphological words and syntactic clusters. A hallmark of words is morphological integrity, which is typically manifested as noninterruptibility. The internal structure of a compound word cannot be disrupted by insertion of focus particles like sae ‘even’ and mo ‘also’, which take scope over syntactic phrases. As demonstrated in (1), this restriction is observed strictly by all three types of V-V complexes that have their first members marked in the infinitive. The same restriction, however, is relaxed, if not completely lifted, in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 

18

the case of Type 4 complexes with the -te inflection signaling a phrasal boundary. Compared with the outright ungrammaticality of (1), particle insertion in Type 4 complexes is marginally acceptable, as in (2). (1)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(2)

Compound verbs a. Kare wa zibun no gitaa o tataki-(*sae)-kowasi-ta. (Type 1) he  self  guitar  beat-(*even)-destroy- ‘He pounded his own guitar to pieces.’ b. Pan ga yaki-(*sae)-agat-ta. bread  bake-(*even)-- ‘The bread has finished baking.’

(Type 2)

c. Haha wa pan o yaki-(*sae)-kake-ta. mother  bread  bake-(*even)-- ‘Mother was about to bake bread.’

(Type 3)

Type 4 complex predicate Kare wa zibun no gitaa o kowasi-te he  self  guitar  destroy- simat-ta. .- ‘He even shattered his own guitar (to an irreversible extent).’

(?sae) (even)

The disparity between (1) and (2) shows that an infinitive-marked verb is tightly integrated with the following verb to form a morphological compound, whereas the gerundive inflection -te syntactically unites two verbs as a complex predicate. • Types 1–3: compound verbs of the form “V1INF-V2” • Type 4: complex predicates of the form “V1- V2” Apart from syntactic noninterruptibility, morphological evidence is available to endorse the word status of the three types of compounds. The evidence concerns nominalization, which comes in two forms. One is the nominal use (or conversion) of a verb’s infinitive form, which applies not only to a simplex verb, as in yomi ‘reading, insight’ (< yomu ‘read’), but also to all classes of V-V compounds, as in Type 1 kasi-dasi [lend-take.out] ‘checking out (of a book)’, Type 2 yaki-agari [bake-] ‘the state of being fresh-baked’, and Type 3 kaki-naosi [write-] ‘rewriting’. Type 4 complexes, however, are averse to conversion or infinitive nominalization, as shown by the ungrammaticality of *tabe-te simai [eat- .] and *yon-de mi [read- ]. The other way of nominalization is the suffix -kata ‘the way, how to’ (Kageyama 1993; Kishimoto 2006). While applying freely to all of Type 1 (e.g. kasi-dasi-kata [lend-take.out-way] ‘how to

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

19

check out (a book)’), Type 2 (e.g. sikari-tuke-kata [scold--way] ‘the way (he) scolds’), and Type 3 (e.g. kaki-naosi-kata [write--way] ‘how to rewrite’), -kata nominalization has some difficulty with Type 4 complexes, as exemplified by the marginal acceptability of ?*tanon-de mi-kata [ask- -way] ‘the way (he) tries asking’ and *?kowasi-te simai-kata [destroy- .-way] ‘the way (he) destroyed (something)’. Since nominalization operates on a morphological object, the V-V complexes of Types 1, 2, and 3 are indeed identified as compound words rather than serial verbs (contrary to Nishiyama 1998).

2.3 Auxiliary verbs in Type 4 complex predicates This section outlines the essential properties of Type 4 complex predicates in the form of V1-te V2. Their V2 verbs, commonly called “auxiliary verbs” (hojodōshi), convey an array of functional meanings which are mostly rooted in pragmatic considerations in discourse context.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

2.3.1 Overview The gerundive inflection -te has two chief uses in syntactic structure, as a clause coordinator (e.g. Taiyoo ga sizun-de, tuki ga deta ‘The sun set and the moon came up’) and as a converb marker (an adverbial made from a verb as in nekoron-de hon o yomu ‘read a book lying down’). Only when the gerundive verb is followed by one of the eleven delexicalized verbs (and their stylistic variants) listed in Table 2.2 can the V1-te V2 sequence be identified as a complex predicate of Type 4 (see Nakatani 2016 for a literature review). These auxiliary verbs are by no means mutually exclusive but can be concatenated with each other sequentially, as in moo syokuzi o tukut-te kure-te iru [already dinner  cook- - ] ‘(She) has already cooked dinner for me’ and hoot-te oi-te simau [leave- - .] ‘unintentionally leave something/someone unattended’. Table 2.2 catalogs the auxiliary verbs by dividing them into three semantic classes: aspectual, attitudinal/aftereffect, and benefactive. The “aspectual” class designates perfective aspect (resultative with telic V1 + iru ‘be’) or imperfective aspect (progressive with atelic V1 + iru as well as expressing gradual change of state with different orientations of kuru ‘come’ and iku ‘go’). The “attitudinal/ aftereffect” class is concerned not so much with purely grammatical aspect as with pragmatically implied overtones stemming from interpersonal and inter-eventual relations in a discourse context. Such pragmatic implications appear comparable to the meanings characteristically found with certain vector verbs in Indian languages, such as ‘irreversible’ (Kashmiri gatshun ‘go’, Tamil poo ‘go’) and ‘regret, censurability, or undesirability’ (Hindi bait:h ‘sit’, Kashmiri gatshun ‘go’, Tamil

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

20

 

Table 2.2 Auxiliary verbs in Type 4 complex predicates auxiliary verbs 1. iru/ (formal) oru, (respect) irassyaru Aspectual 2. kuru 3. iku

4. aru

5. oku 6. simau

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Attitudinal/ Aftereffect 7. miru / (polite) goran (ni naru) 8. miseru / (polite) goran (ni ireru) 9. kureru / (polite) kudasaru 10. yaru / Benefactive (polite) ageru 11. morau / (polite) itadaku

original meanings grammaticalized meanings ‘(for an animate) to exist’

Progressive or resultative aspect, depending on V1’s lexical aspect

2. ‘to come’ 3. ‘to go’

Gradual change, with different orientations of coming to the speaker in ku(ru) or going away from the speaker in ik(u) ‘(for an inanimate) Perfect aspect, in preparation for to exist’, orig. ‘to sit future use. Often with passive down’ function. ‘to put, set aside’ Completion in preparation for a future action ‘to put away, store’ Irreversible completion with connotations of favorable or unfavorable consequences ‘to see’ Exploratory trial to see feasibility ‘to show, demonstrate’

Demonstration to show the subject’s capability

‘(for another) to Another’s action beneficial to the give to the speaker’ speaker ‘(for the speaker) to give to another’ ‘(for the speaker) to receive from another’

The speaker’s action beneficial to another The speaker’s receipt of benefit from another’s action

poo ‘go’, poot:u ‘put’) (cf. Abbi and Gopalakrishnan 1991). Finally, the “benefactive” class, comprising three subclasses distinguished according to the selection of subject and recipient, expresses favorable effects that the successful execution of V1’s action will bring about on the recipient. These beneficial meanings are akin to what Abbi and Gopalakrishnan (1991: 167) call “benefaction” meanings in Indian vectors (e.g. Hind le ‘take’ [own benefit] / de ‘give’ [other’s benefit], Marathi ghe ‘take’ /de ‘give’, Tamil ko:l ‘contain’ / kudu : ‘give’). The Japanese auxiliary verbs are “auxiliary” because, syntactically, they are restricted to the position subsidiary to a main verb (V1) and, semantically, they modify the main verb with diverse functional meanings. On the other hand, they are still “verbs” in that they fully conjugate and inflect on a par with independent lexical verbs and preserve some of the essential meaning inherent in their full-verb counterparts. Their lexical character figures prominently in suppletion. Thus,

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

21

identical suppletive forms for different registers, as in kureru ‘(for another) to give to the speaker’ (plain) vs. kudasaru ‘(for a superior person) to give to the speaker’ (polite), are deployed in both lexical and auxiliary usage (cf. Table 2.2). Furthermore, the lexical properties intrinsic to the deictic verbs of coming (kuru) and going (iku) and of giving (yaru, kureru) and receiving (morau) are inherited by the auxiliary verbs of the aspectual and benefactive classes. Additionally, the syntactic autonomy of the auxiliary verbs is attested to by the fact that an auxiliary verb alone can be the target of the syntactic rule of subject honorification (o-V ni naru), as in tabe-te o-simai ni naru [eat- -.  ] for tabe-te simau ‘has finished eating all’. The syntactic independence of the auxiliary verbs strongly suggests that Type 4 complexes have a biclausal structure, illustrated in (3).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(3)

Complementation structure of Type 4 complex predicates (Subject 1) [complement clause Subject 2 . . . V-te] auxiliary verb

In this structure, some auxiliary verbs may select an agentive subject of their own (Subject 1 in the matrix clause) that is coreferential with the subject of V-te (Subject 2). The auxiliary verbs that meet this characterization are oku () ‘get an action done in preparation for a future purpose’, miru () ‘try -ing’, and miseru () ‘show how to do’ in the attitudinal/aftereffect class as well as yaru () ‘(for the speaker) to give to another’ in the benefactive class. These auxiliary verbs exhibit the same behavior as “control” predicates like try to and be eager to. Originally a verb of existence for inanimate subjects, -aru ‘have done in anticipation of a future event’ has a unique property of passive-like function (e.g. Heya ga katazuke-te aru ‘The room has been tidied for a future purpose’: see Matsumoto 1990) in addition to functioning as a regular control predicate in active voice. Also exceptional is morau () ‘(for the speaker) to receive benefit from the action of another’, which requires disjoint reference between the matrix and embedded subjects. The other auxiliary verbs shown in Table 2.2 impose no semantic restriction on the subject of the whole clause, thus behaving like “raising” predicates such as begin to and be easy to. The lack of subject restrictions is observed not only with the three verbs of the aspectual class but also with simau ( , ) in the attitudinal class (e.g. Akanboo ga miruku o kobosi-te simatta ‘The baby spilled the milk’ vs. Miruku ga kobore-te simatta ‘The milk spilled’) and kureru ‘(for another) to give to the speaker’ (e.g. Isya ga byooki o naosi-te kureta ‘Luckily, the doctor cured my disease’ vs. Byooki ga naot-te kureta ‘Luckily, my disease was cured’). Apart from the three classes, there is a special class of V-te V complexes where the verbs of deictic motion (iku ‘go’, kuru ‘come’, kaeru ‘return’) in V2 select a gerundive verb in V1 that specifies manner or means of motion, as in mot-te kuru [have- come] ‘bring’, koi-de iku [row- go] ‘go by rowing’, and ture-te kaeru

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

22

 

[accompany- return] ‘bring/take home’. This motion class stands in marked contrast to the aspectual, attitudinal, and benefactive classes in that (i) the motion verbs in V2 preserve their original meanings of physical motion instead of being semantically weakened and (ii) the gerundive verbs in V1 function as adverbial converbs modifying the V2 verbs instead of holding the complementation relation depicted in (3). These two idiosyncrasies bring about a Janus-faced character of the deictic motion class: it resembles Type 4 in appearance but behaves like Type 1 compound verbs in semantic interpretation. In fact, some gerundive verbs in this construction alternate with infinitive verbs in compound structure, as in ture-te kaeru [accompany- return] and ture-kaeru [accompany-return], with the stylistic difference that the compound version sounds slightly old-fashioned (see Shibatani 2007a for the grammaticalization of the motion complexes).⁴ The motion verb constructions with gerundive converbs are discussed in detail in Chapter 6 (Matsumoto).

2.3.2 Functional unity as complex predicates Despite the lack of morphological integration, Type 4 complexes exhibit a functional unity as a predicate. This is shown by the impossibility of altering their internal composition by syntactic operations, as in (4), where V1 kowasi-te ‘destroy’ is moved or deleted from example (2).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(4)

a. *Kare wa kowasi-te zibun no gitaa o he  destroy- self  guitar  simat-ta. .- Cf. (2) ‘He shattered his own guitar (to an irreversible extent).’ b. #Kare wa zibun no gitaa o kowasi-te he  self  guitar  destroy- simat-ta. .- Acceptable only on the literal reading: ‘He put away his own guitar.’

The functional unity of V-te V complexes is enhanced by the contraction in colloquial speech of the gerundive marker -te (or -de) and certain auxiliary verbs that immediately follow it. As shown in (5), the contraction wipes out the phrasal ⁴ Besides the deictic motion class, a handful of lexicalized gerundive complexes are observed that are frozen as tight, uninterruptible units: mi-te toru [see- take] ‘grasp, understand’, kut-te kakaru [eat stick.to] ‘lash out at’, hut-te waku [fall- spring.out] ‘(bad thing) occur unexpectedly’, yot-te tatu [be.based- stand] ‘rely on’, kit-te suteru [cut- throw.away] ‘dismiss, discard ruthlessly’ (Kageyama 2016a: 275).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

23

boundary (#) after the inflectional ending, thus resulting in the univerbation of V1 and V2, and, after univerbation, particle insertion as in (2) is no longer possible. The full forms (formal, written) and reduced forms (casual, spoken) are stylistic variants with no semantic difference.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(5)

a. V-te iru [te # irɯ] ! V-teru [terɯ], V-de iru [de # irɯ] ! V-deru [derɯ] e.g. tabeteru [tabeterɯ] ‘is eating/has eaten’, sinderu [ʃiderɯ] ‘is dead’ b. V-te iku [te # ikɯ] ! V-teku [tekɯ], V-de iku [de # ikɯ] ! V-deku [dekɯ] e.g. hetteku [hettekɯ] ‘diminish more and more’, sindeku [ʃidekɯ] ‘die one after another’ c. V-te oku [te # okɯ] ! V-toku [tokɯ], V-de oku [de # okɯ] ! V-doku [dokɯ] e.g. oitoku [oitokɯ] ‘leave aside’, tundoku [tsɯdokɯ] ‘heap up’ d. V-te simau [te # ʃimaɯ] ! V-timau [tʃimaɯ] (older) / V-tyau [tʃaɯ], Vde simau [de # ʃimaɯ] ! V-dimau [dʒimaɯ] (older) / V-dyau [dʒaɯ] e.g. tabetyau [tabetʃaɯ] ‘eat up’, nondimau [nodʒimaɯ] / nondyau [nodʒaɯ] ‘drink up’ e. V-te ageru [te # agerɯ] ! V-tageru [tagerɯ], V-de ageru [de # agerɯ] ! V-dageru [dagerɯ] e.g. kasitageru [kaʃitagerɯ] ‘lend to you’, yondageru [yodagerɯ] ‘read for you’ f. (dialectal) V-te yaru ! V-taru [tarɯ], V-de yaru ! V-daru [darɯ] e.g. osietaru [oʃietarɯ] ‘tell (you)’, yondaru [jodarɯ] ‘read (for you)’

Heine (1993: 50–1) proposes an “overlap model” of grammaticalization, as shown in (6), where full verbal forms and reduced grammatical forms are in free variation at a midway stage (II). (6)

S I S II

S III

Morphosyntactic category Verbal

Verbal / Grammatical Grammatical

Phonological form

Full / Reduced

Full

Reduced

Since full and reduced forms are coexistent as stylistic variants, all instances of auxiliary reduction in (5) will be situated at Stage II, or more precisely at different points on the cline within the range of Stage II. A salient case is (5a), where the progressive/perfective auxiliary i- (‘be, exist’, etymologically ‘sit down’) is rendered null after contraction, with the consequence that seemingly aberrant morphology emerges in which the tense inflections (-ru ‘present’, -ta ‘past’) look as if they were adjoined directly to the gerundive inflection, as in tabeteru [eat-.] ‘is eating’ and sinderu [die..] ‘is dead’. Such ostensibly “ungrammatical” sequences are, in fact, encountered extremely commonly and do

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

24

 

not sound so casual or so vulgar as the reduced forms of the other auxiliaries. The grammaticalization of i- is thus regarded as most advanced of all the reducible auxiliary verbs. The auxiliary contraction, it should be stressed, takes place only when the designated auxiliary verbs in (5) occur in the V-te V construction and is extraneous to their full-verb counterparts, just as the gonna contraction in English applies to the future going to, but not to the lexical motion verb go. Note incidentally that the infinitive-based compound verbs of Types 1, 2, and 3 (Section 2.4) never permit their V2 members to undergo colloquial contraction or phonetic reduction. The variability of form in Type 4 auxiliary verbs and their full-verb counterparts as well as the disparity between Type 4 auxiliary verbs and V2s of infinitival compounds will militate against Butt’s (2010) claim that “light verbs” (if Japanese auxiliary verbs in V-te V complexes indeed fall in her light verbs’ category) are historically stable and always identical in form to their corresponding lexical verbs (see further Section 2.7). In closing this section, a remarkable innovation should be mentioned where one and the same auxiliary verb is repeated. While recursion may take place on different auxiliary verbs in semantically appropriate contexts, complete doubling of an auxiliary verb lacks no semantic motivation and yet is attested in casual speech, although not all speakers accept it. A few actual examples follow.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(7)

a. matigae-te simat -tyat-ta mitai make.mistake- .-(reduced).- look.like da.  ‘It looks like (he) made a mistake unknowingly (and there’s no hope of recovery)’ Y. Ashizawa: Ima dake no ano ko (Kindle book). Tokyo: Sōgensha b. kazoku ni kooden o todoke-te oi bereaved.family  funeral.offering  send-  -toi-te. -(reduced)- ‘Just send his bereaved family a funeral offering’ K. Suzuki: Kami no ryōiki ni idomu mono (Kindle book), p.100. Nihon Sōgi Renmei.

Among various combinations, the pattern of full form simau followed by reduced form tyau in (7a) sounds most natural. The doubling phenomenon can be taken as evidence that the Japanese auxiliary verbs are at a transitional stage from lexical to functional verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

25

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

2.3.3 Semantics While it is generally agreed that the prime cause of grammaticalization is desemanticization or semantic bleaching/generalization (Haspelmath 1999: 1062; Heine 2001: 583), no principled basis seems available for measuring the degrees of semantic bleaching that will trigger subsequent changes on the grammaticalization chain. The Japanese auxiliary verbs are suggestive in this regard. Although the delexicalized verbs listed in Table 2.2 have all undergone desemanticization to some extent or other, the degrees of semantic change on the whole are not so great as to undermine the principle of compositional semantics. For example, the verb iru ‘exist, stay’, which denotes the existence of animate subjects in its independent use, is predicated of the event or action denoted by V1 in its auxiliary usage, with the consequence that V-te iru expresses the existence of an event/action in the state of being in progress (i.e. progressive aspect) or being completed (i.e. resultative aspect). The diametrically opposite interpretations of progressive and resultative aspect fall out from the combined meanings of iru (existential, stative) and the gerundive -te (perfective) as it is applied to the event types of individual verbs in V1. The resultative meaning of V-te iru obtains if V1 is a telic (punctual) verb that has an explicit endpoint in event structure, where the gerundive signals the actual attainment of that endpoint (so-called “completive” aspect) and iru expresses continuation of the resultant state. The V-V complex sin-de iru ‘is dead’, for example, is interpreted as a combination of two subevents: ‘(He) has died’ and ‘(He) is in the state of being dead’. The progressive interpretation of V-te iru, on the other hand, obtains if V1 is an atelic (durative) verb that has a starting point but no explicit endpoint. Here, the gerundive -te denotes only the inception or execution of the initial phase of V1 (so-called “ingressive” aspect) and iru expresses the continuation of the event/action after the inception. The progressive meaning of nai-te iru ‘is crying’, for example, ensues from two combined events: ‘a crying event has started’ and ‘the crying event stays/ continues’. Compositional interpretations like this will plausibly apply to the auxiliary verbs in the attitudinal and benefactive classes as well. The lexical verb oku ‘to put, leave aside’ does not simply mean change of location but is pragmatically accompanied with an intended purpose of reserving the object for future use. The meaning of intended purpose, which is represented in the background of oku’s lexical meaning—technically as part of the “Telic role” of oku’s qualia structure in Pustejovsky’s (1995) theory of Generative Lexicon—is highlighted when the same verb is used in the V-te oku construction meaning ‘completion in preparation for a future action’. This construction strongly implies that the subject intends to carry out the action of V1 to reserve its effects for a future purpose, as in heya o katazuke-te oku ‘tidy a room for a future purpose’. Martin (1975: 529–30) distinguishes three meanings of V-te oku in pragmatic contexts: (i) ‘does it and puts it

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

26

 

aside; does it so that the result is on hand; gets it done’, as in Teeburu o yoyakusi-te oita ‘I have reserved a table’; (ii) ‘does it and leaves it that way’, as in Sono mama ni si-te oite kudasai ‘Leave things alone’; and (iii) ‘does it and lets it go at that (for the time being), does it for now (as a makeshift or temporary arrangement), lets it go at doing; does it once anyhow; does it once and for all; goes ahead/on and does it’, as in Tonikaku kii-te oite kudasai ‘Let him have a hearing at least’. In all these uses, V-te oku not only expresses the temporal completion of V1’s action but presents it as a preparatory step for an intended future goal. A striking resemblance is observed in a class of Indian vector/explicator verbs labeled “Anticipatory action done in advance” (Abbi and Gopalakrishnan 1991: Hindi rəkh ‘keep’, Marathi t:hev ‘keep’, Tamil and Malayalam vai ‘keep’). In contrast to oku, which takes a volitional subject in both full and auxiliary uses, the verb simau ‘put away, store away, lock up for safekeeping, finish’ is volitional as a lexical verb but may be either volitional or nonvolitional in the auxiliary use. Leaving aside this difference, the core meaning of the verb is virtually the same in both uses. If one stores a concrete entity away in a place, then that item will be out of reach and will not be available as long as it is put away. The ‘out-of-reach’ meaning (Telic role information in qualia structure) persists in the auxiliary use of simau, whereby V1’s completed action/event is conceived of as irretrievable, as in saihu o nakusi-te simatta ‘I’ve lost my wallet (and the event is irretrievable)’. When coupled with the speaker’s subjective stance on the desirability or undesirability of V1’s event, the irreversible completion of the event often brings about a variety of subtle emotive colorings. In normal contexts like Kodomo wa tukarete nemut-te simat-ta ‘The baby has completely fallen asleep as she was tired’ where a consequence (the baby’s falling asleep) follows as a natural result of a cause (her fatigue), simau is emotionally neutral, functioning as a marker of perfective aspect. In “counter-expectation” contexts where V1’s event takes place despite the speaker’s assumption that it will not occur easily as the natural course of events, however, the same auxiliary verb comes to carry, in addition to perfectivity, a host of emotional overtones such as regret, frustration, disappointment, relief, and even pride, as pointed out by Soga (1983), Ono (1992), Ono and Suzuki (1992), and Strauss (2003),⁵ among others. Heavily context-dependent as they are, the interpretations of pragmatic and emotional overtones can be captured by the interactions of such factors as the subject’s volitional controllability of

⁵ Strauss (2003) points out intriguing similarities in pragmatic overtones arising in “unexpected” or “out of the ordinary” contexts between the auxiliary verb constructions in Japanese (V-te simau) / Korean (V-a/e pelita) and the Spanish reflexive se. According to her observation, Spanish se and Korean auxiliary verb V-a/e pelita give rise to the same kind of emotional nuances (regret, relief, etc.) as the Japanese attitudinal V-te simau. These parallels are suggestive of the universal character of pragmatic and emotional meanings.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    

27

V1’s action and the anticipated favorable or unfavorable effects of V1’s event/ action on the speaker or other people involved. The emotions typically associated with simau are negative emotions like regret and frustration, which are evoked when V1’s action/event takes place contrary to the speaker’s expectation that it would not easily occur and if it should ever occur, it would bring about unfavorable effects on him or other people. An exemplary case will be Pikunikku no hi ga ame ni nat-te simat-ta ‘(To my disappointment) it rained on the day of picnic’. Such negative emotions are lexically fixed in the exclamation Simatta! ‘Shoot!/Crap!/Oops!’ (an instance of degrammaticalization), uttered spontaneously by a person who has carelessly done an action that he/she shouldn’t have done. ‘Positive’ emotions like relief and satisfaction, on the other hand, will emerge if V1’s action, which is supposed to be hard to carry out but beneficial to the speaker if properly carried out, is successfully implemented by the subject’s endeavor, as in Syukudai wa moo yat-te simat-ta ‘I’m already done with my homework (so I feel relieved)’ or (detective to suspect) Hayaku zihaku si-te simae! ‘Don’t hesitate to confess your guilt! (The speaker as well as the subject will be relieved if confession is made)’. A mixture of positive and negative emotions is also possible, as in Syoogakusei ga puro no kisi ni kat-te simat-ta ‘A schoolboy has beaten a professional shogi player. (Creditable for the small boy, but a shame and embarrassment for the professional player)’. Such pragmatic and emotive overtones, it should be underscored, vanish if simau is omitted. Interestingly, many South Asian languages also have vector/explicator verbs that are associated with negative emotions like regret, undesirability, and surprise at unexpectedness, while no cases accompanying positive emotions like relief and satisfaction are reported in Abbi and Gopalakrishnan (1991). The foregoing observations demonstrate that V-te oku and V-te simau are not purely aspectual markers of completion but are intimately linked with the speaker’s subjective judgment or expectation concerning future situations that is not explicitly stated but would most likely follow after the completion of the action. In other words, V-te oku and V-te simau serve as a bridge linking the completion of V1’s event/action to its anticipated future effects. This semantic function, which we might call “event bridging,” holds for V-te miru [V- see] as well. The meaning of this construction is not a mere attempt (‘try to V’) but an exploratory action ‘try V-ing’ to see what the effect or result will be (Martin 1975: 541). The transitive counterpart of miru ‘see’ is miseru ‘show, display’, which is also used in the V-te V construction, as in Kono ki ni nobot-te miseru ‘(in a context where people think I’m so cowardly I can’t climb the tree) Just you see how I can climb up this tree’ (Martin 1975: 545). Here the speaker expects to have a favorable effect after he succeeds in climbing. The event-bridging function, which extends to the benefactive class, can be attributed to the perfective meaning of the gerundive inflection -te/-de, because it is not found with Type 3 syntactic

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

28

 

compound verbs (Section 2.5), whose V1 is marked with the infinitive, which is aspectually neutral or imperfective. To recapitulate, the auxiliary verbs in Type 4 complex predicates have acquired functional meanings of a pragmatic nature when they are embedded in the V2 position of the V-te V construction. Diachronic studies show that this construction frame developed from reanalysis of serial verb constructions (cf. Chapter 3, Aoki and Frellesvig). These considerations suggest that the initial trigger of grammaticalization is not so much the semantic bleaching of individual source verbs but the establishment of a particular construction frame (see also Chapter 4, Kageyama, for the constructionalization of Type 2 compounds).

2.4 The lexical vs. syntactic distinction in compound verbs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In Japanese linguistics, it is customary to tease apart two major classes of infinitival compound verbs: lexical compound verbs (Types 1 and 2), as in (8), and syntactic compound verbs (Type 3), as in (9) (Kageyama 1989, 1993, 2016a). (8)

Lexical compound verbs (Types 1 and 2) yuri-okosu [shake-awake] ‘awaken by shaking, shake (someone) awake’, utikorosu [shoot-kill] ‘kill by shooting, shoot to death’, huri-yamu [rain-stop] ‘stop raining’, uki-agaru [float-go.up] ‘float up’, naki-sakebu [cry-shout] ‘cry and scream’

(9)

Syntactic compound verbs (Type 3) aruki-hazimeru [walk-] ‘begin to walk’, nomi-tuzukeru [drink] ‘keep on drinking’, kenasi-au [criticize-] ‘criticize each other’, tabe-sobireru [eat-] ‘miss eating’, nomi-sugiru [drink-] ‘drink too much’

The division is motivated by a systematic disparity in the range of elements that can appear in the V1 position. As shown by the robust contrast between (10) and (11), syntactic compound verbs (Type 3) can but lexical compound verbs (Types 1 and 2) cannot accommodate in V1 such syntactically motivated verbs as passive verbs, causative verbs, honorific verbs, Verbal Noun [VN] + suru (so-called ‘light verb constructions’), and idiom chunks. (10)

Lexical compounds exclude syntactically motivated verbs from V1. a. Passive -rare in V1: *tatak-are-okiru [beat--awake] Lit. ‘become awake by being beaten’ b. Causative -sase in V1: *aruk-ase-tukareru [walk--get.tired] Lit. ‘become tired from making (someone) walk’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

29

c. Honorific verb in V1: *o-tobi-ni nari-agaru [.jump..become-go. up] ‘jump up’ d. VN + suru in V1: *huzyoo-si-agaru [float-do-go.up] ‘float up’ e. Idiom chunk in V1: *saba o yomi-otosu [fudge-fail] ‘fail to fudge’ (cf. saba o yomu lit. ‘count the numbers of mackerel’ = ‘cheat in counting’)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(11)

Syntactic compounds may contain syntactically motivated verbs in V1. a. Passive in V1: [sikar-are-nareru] (scold--.) ‘be used to being told off ’ b. Causative in V1: [hatarak.ase-tuzukeru] (work--) ‘keep on making (someone) work’ c. Honorific verb in V1: o-tabe-ni [nari-sobireru] (-eat- [become]) ‘miss eating’ d. Light verb constructions in V1: [huzyoo-si-kakeru] (float-do-.. ) ‘be about to float up’ e. Idiom chunks in V1: saba o [yomi-sugiru] (fudge-) ‘fudge excessively’ (cf. saba o yomu lit. ‘count the numbers of mackerel’ = ‘cheat in counting’)

Granted that passives and other syntactically motivated verbs may participate in the formation of Type 3 compound verbs, a question immediately arises: Why can such syntactically motivated verbs occur legitimately in V1 without causing a violation of the morphological integrity principle? The common assumption to circumvent this issue is to hypothesize an underlying biclausal structure like (12), from which only the head verb (V1) in the complement clause is picked out and merged with the main verb in V2 by an operation such as incorporation (Kageyama 1993), restructuring (Fukuda 2012), or reanalysis (Kishimoto 2013). (12)

Complementation structure of Type 3 syntactic compound verbs

main clause

complement clause

…………….

V1

V2 integrated into a compound

In (11c), for example, only the infinitive verb nari ‘become’ in the honorific verb complex o-tabe ni nari participates in the compound formation with V2 sobireru ‘miss’, while the main verb o-tabe-ni [-eat-] ‘eat’ remains in situ in the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

30

 

complement clause. Likewise, the object saba o ‘mackerel’ in (11e) stays in the complement without being included in the compound verb. The lexical vs. syntactic distinction of compound verbs is thus justified primarily by the availability or unavailability of the V1 position to syntactically motivated elements. Because of the open-ended nature of syntactic structure, there is no upper limit to the number of syntactic compound verbs, although their heads are limited to thirty-odd designated verbs (Section 2.5). On the other hand, lexical compound verbs not only reject syntactic elements in the V1 position but are also limited in number (approximately 3,000) owing to idiosyncratic lexical restrictions on the combinations of two verbs. As a result, the output of lexical compound formation has to be registered in the lexicon. In this sense, they are not just morphological but “lexical” compounds. An additional distinction between the two classes of compound verbs is found in the ordering restriction when they show up sequentially. As shown by (13a), syntactic compound verbs may follow (i.e. show up outside of) lexical compound verbs, but the opposite ordering, as in (13b), is inadmissible. (13)

a. [[]-] [[huri-yami]-kakeru] [[rain-]-] ‘be about to stop raining’

b. *[[]-] *[[huri-kake]-yamu] [[rain-]-] lit. ‘stop being about to rain’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

2.5 Phasal verbs in Type 3 syntactic compound verbs In his renowned reference grammar of Japanese, Martin (1975) lists as many as fifty-six infinitive-attached auxiliary verbs, including nominal and adjectival auxiliaries. Unfortunately, his list mixes up syntactic and lexical aspectual verbs. By using the syntactic diagnoses discussed in Section 2.4, twenty-nine verbs are identified as the head verbs (V2) of syntactic V-V compounds and are classified according to the phasal meanings they designate in Table 2.3.⁶ All these phasal verbs are delexicalized within the compound structure inasmuch as their lexical meaning and argument structure information are lost partially or wholly. Even -hazimeru ‘begin (orig. tr.)’, -owaru ‘end (orig. intr.)’, and -tuzukeru ‘continue (orig. tr.)’, which are aspectual verbs par excellence, deviate from their full-verb counterparts in transitivity. Thus, V2 hazimeru, originally transitive, can combine with either a transitive or an intransitive verb ⁶ The list excludes suffixal verbs with grammatical meanings such as epistemic potentiality -e(ru) (originally meaning ‘get, obtain’; also realized as the archaic form u(ru) in the present tense), passive/ respect/spontaneous -(r)are, ability -(ra)re, causative -(s)ase, and outward expression -gar. In addition, the suffixal adjectives of ease and difficulty (-yasui ‘be easy’, -nikui ‘be hard’, -gatai ‘be hard’) and the negative adjective -nai can also take a complement clause, as in (12).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

31

Table 2.3 Phasal verbs in Type 3 syntactic compound verbs phasal meanings

examples

-dasu [ , vt.] ‘begin to V’, -hazimeru [, vt.] ‘begin to V’, -kakaru [ , vi.] ‘set about V-ing’ avertive -kakeru [, vt.] ‘be about to V but do not actually V’ continuative, repetitive -tuzukeru [, vt.] ‘continue to V’, -makuru [ , vt.] ‘V on and on intensely’ completive, terminative, -oeru [, vt.] ‘finish V-ing’, -owaru [, vi.] ‘stop V-ing’, depletive tukusu [, vt.] ‘V exhaustively’, -kiru [, vt.] ‘V completely’, -toosu [ , vt.] ‘V to the end’, -nuku [ , vt.] ‘V to the end’, -hateru [   , vi.] ‘V utterly’ incompletive, -sokonau [, vt.] ‘miss V-ing’, -sokoneru [, vt.] ‘fail to ineffective V’, -sonziru [, vt.] ‘fail to V’, -√sobireru [] ‘fail to V’, -kaneru [.] ‘be unable to V’, -wasureru [] ‘forget to V’, -nokosu [ ] ‘leave something without V-ing completely’, -ayamaru [, vt.] ‘make a mistake in V-ing’, -okureru [ , vi.] ‘be delayed in V-ing’, √aguneru [Classical Japanese ‘be satiated’] ‘hesitate to V’ excessive -sugiru [ , vi.] ‘V excessively’ retrial -naosu [, vt.] ‘V again (to obtain a desired result)’ habituative, -tukeru [, vt.] ‘be used to V-ing’, -nareru [ frequentative , vi.] ‘be accustomed to V-ing’, -akiru [ , vi.] ‘get weary of V-ing’ reciprocal -au [, vi.] ‘V reciprocally’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

inceptive

in V1. Furthermore, -aguneru ‘hesitate, be weary’ and -sobireru ‘fail to’, indicated by the symbol “√,” are restricted to the V2 position of compound structure as a consequence of the demise of their full-verb counterparts. Like English aspectual verbs (e.g. It began to rain vs. John tried to begin to sing), Japanese phasal verbs in Table 2.4 are generally categorized into raising-type, in which V2 imposes no semantic restriction on the subject, and control-type, where V2 selects an agent subject; see Kageyama (2016a) and Kishimoto (Chapter 5, this volume) for details. Apart from this subject selection, the phasal verbs in Table 2.4, like the auxiliary verbs in Type 4, are completely free of lexical combinatory restrictions and may be combined with any verb of any internal composition in V1 insofar as the two verbs are semantically harmonious. Semantically, these V2s express a wide spectrum of phasal meanings such as inception, continuation, completion, and habitual. The verb -kakeru [] merits special mention. The function of this verb can be identified with what Kuteva (2001: 77) calls “avertive” (meaning ‘was on the verge of V-ing but did not’). Although it is similar to “prospective” (be about to, be on the point of) in Comrie (1976: 64), it is different, crucially because avertive entails that V1’s event, which is about to happen, does

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

32

 

not actually occur, while prospective is noncommittal as to the actual realization of the verb’s event (Kuteva 2001: 92). -Kakeru indeed entails that V1’s event actually did not happen, as in Sini-kaketa ga, sinanakatta/*zissai sinda ‘(He) was just about to die, but he didn’t die/*and he died anyway’. It is fitting to note that most phasal verbs in Table 2.4 designate notions of imperfectivity in the sense that they elaborate on the internal phases of the event denoted by V1’s clause rather than presenting it as a complete “blob” (Comrie 1976: 18). This explains why verbs like -sugiru⁷ ‘do too much’ in the excessive category and -tukusu ‘exhaust’ in the excessive category, which appear to mark perfectivity, do not merely represent completion but concomitantly quantify over the amount of V1’s internal (object) argument resulting from the event’s termination, as in Yama-kazi ga mori o yaki-tukusita ‘The forest fire burned up all the trees’ (see Kageyama 1993; Yumoto 2005; Kishimoto (Chapter 5, this volume) for theoretical analyses).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

2.6 Type 1 (thematic) vs. Type 2 (aspectual) compound verbs Lexical V-V compound verbs are formed by directly adjoining an infinitive verb (V1) to the head verb (V2), as in V1 + V2 ! [V1-V2]V. The Compound Verb Lexicon (Kageyama and Kanzaki 2014), an online database, contains 2,756 lexical compound verbs commonly used in contemporary standard Japanese, and the number would increase only slightly if uncommon verbs used in literary works or irregular new coinages were included. Nonetheless, they are not freely productive and the range of possible combinations between two verbs hinges crucially on individual V2 verbs. For example, as many as 255 entries ending with the bound verb -komu ‘go in, put in’ as V2 are registered in the above-mentioned database, but only one entry ending with tagiru ‘do violently’ as V2 (i.e. nie-tagiru ‘boil over violently’). Among the long-standing issues in the theoretical treatments of lexical V-V compound verbs, two are of special importance. One concerns the compatibility of transitivity between two members. The basic observation is that a transitive verb is compounded with another transitive verb, as in (13a), whereas a mixed combination of a transitive and an intransitive (unaccusative) verb, as in (13b), is systematically ruled out. This restriction, dubbed the “Transitivity Harmony Principle” (Kageyama 1993), is discussed extensively by Matsumoto (1998), Nishiyama (2008), and Kageyama (2016a, 2016b), among others.

⁷ The verb -sugiru has an exceptional property of attaching not only to verb infinitives but also to adjective stems and adjectival nouns.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(14)

33

a. Kare wa doa o osi-aketa. [push-open] ‘He opened the door by pushing’ b. *Kare wa doa o osi-aita. [push-open] ‘He pushed the door open’ c. Ta tui-kai le men. [push-open] ‘He pushed the door open’ (Chinese)

Note that the ungrammatical composition of (14b) with a transitive V1 designating a causing action and an unaccusative V2 expressing the resultant state of an object entity corresponds to the canonical pattern of Chinese resultative compounds, as in (14c) (Kageyama and Shen 2018). While Masica (1976: 150) regards the Chinese-type resultative compounds as distinct from the Indian-type explicator compounds, Japanese, an explicator-compound language, has at least two exceptional verbs that display the Chinese-type “resultative” pattern when used as V2s of lexical compounds: -saru ‘leave, disappear’ and -komu ‘go inside’. In its literal meaning of ‘leave, go away’, the intransitive verb saru designates the disappearance of the subject entity, no matter whether it is used as an independent verb or as the V2 of compounds (e.g. hasiri-saru [run-leave] ‘run away’, hakobisaru [carry-leave] ‘go off carrying’). When it expresses a more abstract meaning of completion, however, -saru refers to the disappearance of the object entity (internal argument) of a transitive verb, as in kioku o kesi-saru [memory  wipe-] ‘wipe the memory out’ or ‘get rid of the memory completely’. Even more peculiar is the bound verb -komu designating movement into a closed space (translated simply as “into”). Lacking an independent use, this verb is compounded either with an intransitive verb (e.g. Yogoreta mizu ga kawa ni nagare-komu [dirty water  river  flow-] ‘Dirty water flows into the river’ [V1=unaccusative, non-volitional], Yakuza ga watasi no ie ni donari-komu[gangster  my house  yell-] ‘A gangster bursts into my house yelling’) [V1=unergative, volitional], or with a transitive verb (e.g. poketto ni satutaba o osi-komu [pocket  wad.of.bills  push-] ‘shove a wad of bills into the pocket’). In all these examples, -komu serves to delimit V1’s event by supplying it with a goal argument (and an internal argument if there isn’t any) . Compounds with these two verbs in V2 thus parallel the Chinese-type resultative compounds. Of special note is the fact that the completion meaning of -saru and the goal meaning of -komu are available only when they function as V2s of lexical compound verbs. This is tantamount to saying that these verbs are first compounded with V1 verbs and are then grammaticalized inside the compound structure, instead of being brought into compound structure after their full-verb counterparts are semantically bleached. The other outstanding issue of Japanese V-V compounds has to do with the semantic relations between the two members of a compound. Previous studies (Kageyama 1993; Matsumoto 1996, 1998; Yumoto 2005) have distinguished at least six semantic relationships that hold between the members of such compounds.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

34

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(15)

  Representative semantic relations between V1 and V2 a. Means of action: V2  V1-ing [] uti-otosu [shoot-make.fall] ‘shoot down’, yobi-atumeru [call-gather] ‘call together’ b. Manner of motion or change: V2  V1-ing [] hai-agaru [crawl-go.up] ‘crawl up’, tobi-noru [jump-go.onto] ‘jump onto’ c. Cause-result relation: V2  V1-ing [] syaberi-tukareru [speak-get.tired] ‘get tired from talking too much’, yoitubureru [get.drunk-fall.apart] ‘get dead drunk’ d. Apposition of similar actions (dvandva): V1 and V2 [] odoroki-akireru [be.surprised-be.stunned] ‘be completely stunned’ e. Complementation relation: [[V1] V2] [] mi-nogasu [see-] ‘overlook’, ami-ageru [knit-] ‘finish knitting’, f. Adverbial modification: V2 adds an adverbial meaning to V1. [] sikari-tukeru [scold-] ‘scold severely’, sizumari-kaeru [become. quiet-] ‘become completely quiet’

In (15a), for example, uti-otosu [shoot-make.fall] ‘shoot down’ is construed in the ‘means of action’ relation where the event in V2 otosu ‘make fall’ is implemented by means of the action in V1 utu ‘shoot’. While previous studies treated the six semantic relations as separate categories, Kageyama (2016a) makes a novel proposal to group them into two major classes, depending on which of the two member verbs determines the argument structure and core meaning of the whole compound. The determinant of argument structure and core meaning resides in V2 in the class of thematic compound verbs (Type 1) but in V1 in the class of aspectual compound verbs (Type 2). According to this criterion, the six semantic relations in (15) can now be reorganized into the two major types in (16). (16)

a. Type 1 thematic compound verbs—both V1 and V2 express literal meanings. —Means of action (15a) —Manner of motion or change (15b) —Cause-result relation (15c) —Apposition of similar actions (15d) b. Type 2 aspectual compound verbs—only V1 designates a literal meaning. —Complementation (15e) —Adverbial (15f)

The fundamental difference between the “thematic” and “aspectual” compounds resides in whether V2 has a full-fledged argument structure or not. The V2s of Type 1 thematic compounds, serving as the categorial and semantic heads, have

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    

35

both lexical meanings and full argument structures, whereas those of Type 2 aspectual compounds have virtually lost their lexical meanings and argument structures. It is predicted from this that the former should be more amenable to the Transitivity Harmony Principle than the latter. Statistical data (Kageyama 2016b) in fact show that Type 1 compounds exhibit a significantly higher degree of THP congruence (89.45%) than Type 2 compounds (71.39%). This means that a different mechanism of semantic interpretation is at work in the two types. In the Compound Verb Lexicon database, there are about 1,600 entries of Type 1 thematic compounds and about 800 entries of Type 2 aspectual compounds. For the determination of the class membership, a rule of thumb is to see whether a given compound verb can be plausibly paraphrased by the construction “V1-te V2” (V1 and (then) V2), where V1-te functions as a converb modifying the head verb V2 (Kageyama 2016a). If a given compound can be so paraphrased, its semantic core lies in V2. The paraphrase with left-to-right modification is expected of Type 1 thematic compounds because Japanese has robust head-final word order and both members of a compound maintain the same literal (thematic) meanings as when they are used as independent verbs. For example, doa o keri-akeru ‘kick the door open’ is paraphrasable as doa o ket-te akeru ‘kick the door and open it’/ ‘kick the door open’; hence keri-akeru is identified as a Type 1 thematic compound. Note that, in this example, the object ‘door’ is shared by the two member verbs ‘kick’ and ‘open (tr.)’. All the examples of (15a) through (15d), grouped together as Type 1 (thematic) compounds, converge on this manner of left-to-right modification and argument sharing. The sole exception is the bound verb -komu, discussed earlier. Because of its bound morphology, paraphrases like *ton-de komu ‘jump and (go) in’ and *osi-te komu ‘push and (put) in’ are precluded. Even so, -komu is considered the semantic center of the whole compound because it determines the telicity of the sentence. By contrast, compounds holding the complementation relation (15e) and adverbial relation (15f) do not easily accept a paraphrase by left-to-right modification. As an example of complementation relation, consider (matigai o) minogasu [(an error) see-miss] ‘fail to catch (an error)’, where the error goes unnoticed. This compound verb cannot be paraphrased as mi-te nogasu [see miss] ‘see and (then) miss’, where the error was once noticed and then ignored. A better paraphrase of the compound will be (matigai o) miru koto o nogasu [(an error) see   miss] ‘miss seeing/catching (an error)’, where V1 ‘see’ serves as a complement of V2 ‘miss’ and the object (‘error’) of the whole compound is licensed by V1 ‘see’. No argument sharing holds between the two members of this type of compound. Likewise, in the adverbial relation, (kodomo o) sikari-tukeru [(child ) scold-] meaning ‘scold (a child) severely’ is entirely different from the putative paraphrase sikat-te tukeru ‘scold (a child) and attach (something)’, which is nonsensical. In the compound sikari-tukeru, the object ‘child’ is licensed by V1 sikaru ‘sold’, whereas V2 -tukeru, literally meaning

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

36

 

‘attach, fix’ adds to V1 an adverbial meaning emphasizing the intensity and unidirectionality of the scolding action. Again, argument sharing is not observed between the two members. A rough count of the entries in the Compound Verb Lexicon database (Kageyama and Kanzaki 2014) suggests that there are sixty or more verbs (called “L-asp verbs” in Chapter 4) that appear in the V2 position of Type 2 compounds. Sample V2 verbs in this class together with their aktionsart meanings are shown in Table 2.4, where the first three rows (a, b, c) represent complementation relations and the rest (d–i) adverbial relations. While some of them are identical in both form and meaning to their full-verb counterparts (e.g. yamu ‘stop (said of rain, snow, sound, etc.), the great majority, while keeping the same morphology as their Table 2.4 Selected aktionsart meanings in Type 2 compound verbs aktionsart meanings

examples

huri-yamu (vi.) ‘stop falling (said of snow/rain)’, kaki-ageru (vt.) ‘write up’, hi-agaru (vi.) ‘(pond) dry up’, ni-tumeru (vt.) ‘boil down’, ki-√hurusu (vt.) ‘wear (a garment) until it is worn out’ b. incompletive, ii-√sasu (vt.) ‘stop speaking halfway’, uri-osimu (vt.) ‘abstain ineffective, suspension from selling’, kaki-otosu (vt.) ‘omit writing by mistake’, kasisiburu (vt.) ‘hesitate to lend’, nobi-nayamu (vi.) ‘do not make expected progress’, kiki-tigau (vt.) ‘hear wrongly’ c. inception saki-√someru (vi.) ‘begin to bloom’ d. intensive result, sizumari-kaeru (vi.) ‘become completely quiet’, koori-tuku complete change (vi.) ‘freeze up’, ne-√iru (vi.) ‘fall sound asleep’, kukarehateru (vt.) ‘get completely exhausted’, saki-midareru (vi.) ‘bloom in profusion’, otituki-harau (vi.) ‘be perfectly composed’, damari-√kokuru (vi.) ‘keep completely silent’, negiri-taosu (vi.) ‘knock the price down’ e. intensive continuation, huri-√sikiru (vi.) ‘(rain) fall incessantly’, naki-kurasu (vi.) repetition ‘cry all day’, hozikuri-kaesu (vt.) ‘dig again’, tukai-√komu (vt.) ‘use repeatedly’, tate-kaeru (vt.) ‘rebuild’, ii-√narawasu (vt.) ‘commonly say’ f. intensive action sawagi-tateru (vi.) ‘fuss about’, izikuri-mawasu (vt.) ‘fumble about’, waki-kaeru (vi.) ‘boil violently’, home-tigiru (vt.) ‘praise highly’ g. manner of action odori-kuruu (vi.) ‘dance crazily’, mati-kamaeru (vt.) ‘brace oneself and wait’, sikari-tukeru (vt.) ‘scold severely’, yomiageru (vt.) ‘read aloud’, uri-isogu (vt.) ‘try to sell in a hurry’, tukai-wakeru (vt.) ‘use different things according to the need’ h. spatial aspect (direction mi-orosu (vt.) ‘look down’, kosuri-ageru (vt.) ‘rub upwards’, of action) osi-kaesu (vt.) ‘push backwards’, sasayaki-kakeru (vt.) ‘whisper to’, nari-wataru (vi.) ‘resound throughout a place’ i. social deixis moosi-ageru (vt.) ‘say to a superior person’, harai-sageru (vt.) ‘(for public authorities) to sell to the private sector’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

a. completive, action to the end

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    

37

lexical counterparts, have developed a multitude of distinct functional meanings, so much so that some of them, marked with “√,” are now dedicated to the V2 position of compounds with highly specialized meanings (e.g. -someru in sakisomeru ‘begin to bloom’, -hurusu in ki-hurusu ‘wear a garment until it is worn out’, and -kokuru in damari-kokuru ‘keep completely quiet’). Interestingly, some examples are amenable to a paraphrase with a reversed modification relation where V2 semantically modifies V1, as in sikirini naku [continuously cry] for naki-sikiru⁸ [cry-.] ‘(small animal) cry continuously’ and naguru-yooni kaku [strike-manner write] for kaki-naguru [write-] ‘write roughly’ (cf. nominalized compound naguri-gaki ‘scribble’). The truth-value of sentences predicated by such compounds is unaffected even if the V2s are omitted. Thus, naki-sikiru is equivalent to naku ‘cry’ in its core meaning, as is kaki-naguru to kaku ‘write’. Such aktionsart meanings as intensity or manner of action appear to be shared by some vector verbs in South Asian languages (Abbi and Gopalakrishnan 1991: 169). Although Type 2 compounds look similar to Type 3 syntactic compounds in the functional meanings of V2 verbs, their semantic effects are different. The V2s of Type 3 compound verbs add phasal meanings of grammatical aspect (outer aspect) to the entire eventualities denoted by the embedded clauses of V1, whereas those of Type 2 lexical aspectual compound verbs directly manipulate the internal structure of V1’s aktionsart meanings. This point will be discussed fully in Chapter 4. Note finally that the lexical form alone of a V2 verb is not sufficient to determine its membership in the four types, for a single V2 verb may have dual or multiple membership, due to polysemy. The verb ageru ‘raise’, for example, belongs to three different types: Type 1 (hakobi-ageru [carry-raise] ‘carry to a higher place’), Type 2 (kaki-ageru [write-] ‘write up’), and Type 4 (kai-te ageru [write- /] ‘write on another’s behalf ’).

2.7 Conclusion: Semilexical categories and grammaticalization The findings in the preceding sections will be summarized in Table 2.5, where it is shown that the forms of V-V complexes are intimately bound up with their semantic effects. Type 1 and Type 2 compounds, which are morphologically integrated, are located at the left end of the scale, whereas Type 4 complexes, with no morphological integration, are located at the right end. Type 3 syntactic compounds come in between, because they have biclausal syntax and yet are morphologically ⁸ Aoki and Frellesvig (Chapter 3, this volume) report that such a reversed order is attested with some aktionsart verbs in texts of Old Japanese, as in siki puru [extend fall] ‘keep on raining’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

38

 

Table 2.5 Form-meaning correspondence in V-V complexes

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Types 1 and 2 lexical Type 3 syntactic compounds compounds Form

stronger

Semantic effects

Unification of the internal semantic structures of V1 and V2

M  Imposition of grammatical aspect on the whole of V1’s event

Type 4 complex predicates weaker Bridging V1’s event to a future purpose/event by speaker’s subjective judgment

integrated. Corresponding to this formal difference, the three groups involve disparate semantic effects of V2 verbs. The group with the strongest morphological bond (lexical compounds of Type 1 and Type 2) is naturally designed to manipulate directly the internal semantic structures of the member verbs to create a coherent semantic structure (see Chapter 4 for formalization). Type 3 compound verbs, on the other hand, impose meanings of grammatical (viewpoint) aspect on the propositional content of V1 because their V2s take a clausal complement headed by V1 in syntactic structure. Finally, the attitudinal and benefactive classes in Type 4 complexes imply future effects of V1’s event because the gerundive -te is a kind of perfectivity marker and therefore serves to link the completion of V1’s event to an unexpressed future purpose or future outcome. In a nutshell, the stronger the morphological bond, the more closely integrated the meanings of two-member verbs are. Conversely, the weaker the morphological bond, the more distanced the semantic connections between two verbs become. In this way, the difference in morphosyntactic structure among the four types of V-V complexes and the distribution of different semantic functions among them are neatly correlated with each other by the form-meaning iconicity. Leaving aside Type 1 thematic compounds, the V2s in the other three types carrying delexicalized meanings and more or less grammatical functions are regarded as “semilexical categories” located at intermediate steps between genuine lexical verbs and genuine auxiliaries on the verb-to-auxiliary grammaticalization cline. Such intermediate categories have been largely neglected in the previous literature on auxiliaries and auxiliation. In their otherwise highly useful lexicon of grammaticalized morphemes in the world’s languages, for example, Heine and Kuteva (2002) present an abundant catalog of grammaticalized target meanings paired with their source meanings in individual lexical items. The mere list of two poles, source and target, however, is at variance with their theory of continuous and chainlike development of grammaticalization, as the authors themselves admit (p. 6). The previous studies on the grammaticalization of aspectual notions as epitomized by the Verb-to-TAM auxiliation (Heine 2001) emphasize only the continuity of semantic change without much discussion of intermediate categories between sources (full lexical categories) and targets (grammatical affixes). This

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

39

gap is filled by the various types of V-V complexes in Japanese. Particularly relevant are the phasal verbs in Type 3 and the auxiliary verbs in Type 4, which display appreciable differences not only in the kinds of functional meanings but also in susceptibility to colloquial contraction and compatibility with nominalization (Section 2.2). Type 3 phasal verbs, which are amenable to nominalization on a par with lexical verbs but reject phonetic reduction, are closer to lexical verbs than Type 4 auxiliary verbs, which are difficult to nominalize and are subject to morphological erosion. Type 3 and Type 4 thus differ markedly in degree of “lexicalness.” Now, to these two categories is added a new category, Type 5, that has been left out of consideration so far. It comprises the suffixal auxiliary verbs expressing the speaker’s respectful attitude in formal style or pejorative attitude in vulgar (and mostly dialectal) speech.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(17)

Type 5 auxiliary verbs a. politeness: V-masu (etym. mairasu ‘(make some) serve a superior person’; e.g. huri-masu ‘(it) rains’, tabe-masu ‘eat’) b. disparagement, contempt, or condemnation: V-yagaru (etym. agaru ‘go up’; e.g. sabori-yagaru ‘playing truant’), V-kusaru (etym. ‘go rotten’; e.g. si-kusaru ‘do’), V-sarasu (etym. pejorative for ‘do’; ii-sarasu ‘say’), V-ketukaru9 (etym. pejorative for ‘be’, ‘go’, ‘say’; ii-ketukaru ‘say’)

Although they superficially pattern like Type 3 phasal verbs in attaching to infinitive verbs and inflecting for tense, the Type 5 auxiliary verbs are idiosyncratic in many respects. Lexically they lack full-verb counterparts; morphologically they have a defective conjugational paradigm lacking some conjugational forms like the infinitive, negative, and conditional; and semantically they are motivated only for the sociolinguistic purpose of presenting a proposition in a style that is apposite to the interpersonal relationship of the speaker and addressee(s) in particular discourse contexts. These idiosyncrasies strongly suggest that the Type 5 verbs have only a very low degree of “verbhood.” Because they are relevant only to speech style, they are outside the scope of word formation processes like nominalization, as evidenced by the sheer ungrammaticality of *sabori-yagari ‘playing truant’ (infinitive nominalization) and *(ame no) huri-masi-kata ‘the way it’s raining’ (-kata ⁹ Dictionaries stipulate that -ketukaru allows either the infinitive or the gerundive (-te) in V1. Semantic considerations, however, suggest that the infinitive is the rule and the gerundive -te appearing before this auxiliary verb corresponds to the aspectual -te iru ‘be’ (progressive, resultative), from which i- ‘be’ is elided (i.e. contraction in (5a)). Thus, ne-te ketukaru ‘is sleeping’ with an atelic verb ‘sleep’ is interpreted only as progressive, and sin-de ketukaru ‘is dead’ with a telic verb ‘die’ only as resultative. This aspectual distinction does not appear when V1 takes the infinitive form, which is aspectually neutral, as in uso o ii-ketukaru ‘tell a lie’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 

40

nominalization). Nominalization operates on major lexical categories representing a propositional content and does not extend to the functional categories of politeness/disparagement, tense, and modality that are layered outside of the proposition. In fact, the Type 5 auxiliary verbs of politeness and disparagement can occur only after (and never before) lexical verbs of simplex or compound structure, Type 3 phasal verbs, and Type 4 auxiliary verbs in predicate agglutination, as exemplified in (18).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(18)

[tabe - hazime - te simai]proposition eat   . ‘(He) has already started eating’

masi/yagat /

- ta 

More generally, the word order of sentence (18) is assimilated to the regular pattern of predicate layering in Japanese as schematically represented in Figure 2.1. Between fully lexical verbs and full auxiliaries of modality lie a variety of semilexical categories tentatively labeled as follows: Type 2 aktionsart verbs and Type 3 phasal verbs are “semilexical verbs” (the difference being morphological or syntactic), Type 4 auxiliary verbs of aspectual, attitudinal/aftereffect, and benefactive classes are now labeled “semi-auxiliary verbs,” and Type 5 suffixal verbs of politeness and disparagement are bona fide “auxiliary verbs.” Outside the sphere of semilexical verbals, tense affixes are attached as clitics, and noninflecting modals like daroo (epistemic possibility) and mai (negation of epistemic or deontic necessity) show up after them as simple “auxiliaries,” which are bound morphemes but are not cliticized to their host. In syntactic structure, these categories are hierarchically layered, as in Figure 2.1, where a lower category must precede a higher category in linear order, with the proviso that the order of Type 3 preceding Type 4 could be reversed in pragmatically marked contexts.

syntactic morphological

Type 1 (thematic) LEXICAL V

Type 2 (aktionsart)

Type 3 (phasal)

SEMI-LEXICAL V

Type 4 (attitudinal, etc.)

Type 5 (polite, etc.)

TENSE

MODAL

SEMI-AUXILIARY V

AUXILIARY V less lexical

CLITIC

AUX

Figure 2.1 Semilexical categories in predicate agglutination

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    

41

The relative ordering among these categories is taken as mirroring the transitional stages of the verb-to-auxiliary grammaticalization rather faithfully. The consecutive decrease in the tokens of semilexical categories from Type 2 (60+) to Type 3 (29) to Type 4 (11+) and Type 5 (4) adds credibility to our claim that these categories reflect different stages of verb-to-auxiliary grammaticalization. The findings in this chapter will be instrumental in broadening the scope of typological and theoretical research on auxiliaries and grammaticalization. From the viewpoint of typology, it is found that many functional meanings that have been considered characteristic of vector/explicator verbs in South Asian languages (see Chapters 9–13) are spread over three distinct classes of semilexical verbals in Japanese: semilexical verbs of Type 2 (aktionsart meanings), semi-auxiliary verbs of Type 4 (periphrastic aspect, aftereffect/attitudinal, benefactive), and Type 5 auxiliary verbs (politeness, disparagement) which have no corresponding lexical verbs. This finding apparently supports Masica’s (1976) idea of vector/explicator compound verbs as an areal feature of Asian languages viewed from the synchronic point of view. Diachronically, however, the rise of the five types of V-V complexes in Japanese is an endogenic process occurring over the past 800 years or so, as demonstrated in Chapters 3 and 4 of this volume. It is thus tempting to hypothesize that vectors and similar functional verbs constitute a natural class of “semilexical verbals,” prominently (though perhaps not exclusively) found in Asian languages. From the standpoint of theories concerning linguistic categories, the fact that the Japanese semilexical verbals diverge not only in meaning but also in form (i.e. the availability of phonetic reduction and the availability of full-verb counterparts) questions the validity of Butt’s (2010) theory of “light verbs,” which are defined as delexicalized verbs that are paired with formally identical lexical verbs. Based on a variety of nonlexical verbs in Urdu and other Asian languages, Butt (Butt and Geuder 2001; Butt and Lahiri 2013) advances the claim that light verbs are always identical in form to their corresponding lexical verbs and thus constitute a discrete category on their own which is outside the standard verb-to-auxiliary grammaticalization chain in which a full verb undergoes desemanticization and phonetic erosion to become bound affixes. This claim is partly intended to challenge Hook’s (1974, 1991) view that Hindi vectors are intermediate between full verbs and auxiliaries (cf. also Hopper and Traugott 2003: 112–14). Provided that the Japanese semilexical verbs in question are indeed subsumed under Butt’s “light verb” category, the robust division of labor and form observed between Types 2, 3, and 4, as summarized in Table 2.5, strongly suggests that the semilexical verbal categories in Japanese do not form a unitary class but are divided into a few subclasses that are graded according to their properties as verbs. This point is further clarified in Table 2.6, where Type 5 auxiliary verbs are added for comparison.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

42

 

Table 2.6 Different stages of grammaticalization in delexicalized verbs Type 2

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Degree of delexicalization Formal correspondence to lexical verbs Combinatory restrictions on V1

Type 3

Type 4

Type 5 higher

Yes, but with many Yes, but with a Tends to few exceptions collapse by lexical erosion idiosyncrasies heavy light almost none

No correspondence none

In Table 2.6, four different stages of delexicalization are differentiated on the basis of two parameters: (i) formal identity with full-verb counterparts and (ii) combinatory restrictions on main verbs in V1. The polite/pejorative auxiliary verbs of Type 5, located at the most advanced stage of delexicalization, lack lexical counterparts and are free from combinatory restrictions on V1 verbs. In the second most advanced class are Type 4 semi-auxiliary verbs representing periphrastic aspect, attitudinals, and benefactives. Due to phonetic erosion in colloquial speech, some of them have lost the formal identity with their lexical counterparts. Most phasal verbs of Type 3 are form-identical to their full-verb counterparts, and yet a few verbs are obsolete today and are frozen as V2s in the compound structure. Lack of formal identity is more conspicuous in Type 2 compounds, where many descendants of Old Japanese aktionsart verbs are fossilized in the lexical compound structure (cf. Chapter 4). These facts are at variance with Butt’s claim that light verbs are always form-identical to their corresponding full verbs (Butt and Geuder 2001; Butt and Lahiri 2013), and that if one member of a lexical-light verb pair falls out of use in historical change, then the other member will cease to exist at the same time (Butt and Tantos 2004). It is thus shown that the delexicalized verbs of Type 2 to Type 5 are located on intermediate points between full verbs and (noninflecting) auxiliaries along the verb-to-auxiliary grammaticalization chain, as in (19). The order of these intermediate categories corresponds to the hierarchical syntactic structure of Japanese predicate agglutination shown in Figure 2.1. (19) full verbs > Type 2 > Type 3 > Type 4 > Type 5 > tense > modal Of special interest is that all the four intermediate categories converge on a broad notion of aktionsart and aspect (including not only temporal aspect but also spatial and social aspect encompassing a host of semantic and pragmatic elements directly involved in the unfolding of V1’s eventuality), but to the exclusion of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    

43

modality. There is indeed an auxiliary verb oru of literary style, which was originally a lexical verb of existence and is grammaticalized to an aspectual marker of imperfectivity (continuation of activity or state), often accompanied by a derogatory or insulting nuance when applied to subjects other than the speaker. The connotation of humiliation and condemnation is stronger in yoru, a phonetic/ dialectal variant of oru in Osaka and other western dialects, used as in Taihuu ga ki-yoru ‘A typhoon is coming (damn it!)’. This added overtone is distinct from deontic or epistemic modality and from evidentiality, for that matter. Both oru and yoru retain their original function as markers of imperfective aspect. Although it is not known exactly how the humiliation meaning arose, it is likely that the speaker’s subjective feeling of irritation about unexpected continuation or prolongation of V1’s action was a contributory factor. In conclusion, the systematic correlations between the functional meanings and syntactic positions of Japanese semilexical verbal categories demonstrate the intermediate stages of grammaticalization. This does not mean, however, that each and every individual verb starts out as an independent lexical verb and is grammaticalized gradually along the cline from Type 2 to Type 5. This scenario is highly unlikely. According to Nihon Kokugo Daijiten (The Grand Dictionary of the Japanese Language), the auxiliary uses of simau ‘put away’, oku ‘put’, and miru ‘see’ in Type 4 complexes, for example, are first attested at about the same periods as their lexical counterparts of equivalent meanings: simau (. 1687, . 1686), oku (both late eighth c.), miru (. late ninth c., . 905–14). At least for these verbs, a “chain-like” succession of semantic bleaching and formal reduction cannot be corroborated (see further Chapter 4 for the rise of Type 2 aktionsart verbs). All these considerations indicate that it is not individual lexical items that proceed along the grammaticalization path, as assumed in the conventional itembased studies on grammaticalization and auxiliation (Heine 1993; Kuteva 2001); rather, it is the semilexical categories set in particular V-V constructions that are aligned along a lexical-to-grammatical cline of decategorization.

Acknowledgments This work was supported in part by the grant for the research project “Syntactic, Semantic, and Morphological Characteristics of the Japanese Lexicon” at the National Institute for Japanese Language and Linguistics (2010–15) and in part by research funding from Doshisha University (2017–18). I benefited greatly from discussion with Peter Hook on the nature of vectors. I am also grateful to Yo Matsumoto and John Haig for feedback on earlier drafts.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

3 Verb verb complex predicates in Old and Middle Japanese Hirofumi Aoki and Bjarke Frellesvig

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.1 Introduction In this chapter we outline and discuss salient differences between verb verb complexes in Old and Middle Japanese and the verb verb complexes of Modern Japanese.¹ We do so contrastively with reference to the four main types of complexes found in present-day Modern Japanese, as set out in Chapter 2 (Kageyama, this volume): Type 1 lexical thematic compounds; Type 2 lexical aspectual compounds; Type 3 syntactic aspectual compounds; Type 4 syntactic complex verbs. This section gives some background and conventions. Section 3.2 addresses complex verbal predicates in Old Japanese (OJ) and Early Middle Japanese (EMJ). We describe an overall functional similarity with Modern Japanese (NJ) in the system of complex predicates which comprises both thematic complexes (like NJ Type 1) and aktionsart verb complexes (like NJ Types 2–4), but also point to the existence, primarily in OJ, of an aktionsart verb construction which is not found in NJ, in which V1 is an aktionsart verb and V2 is the main verb. We also present hypotheses about the origin of complex predicates in Japanese and in particular

¹ The history of the Japanese language may be divided into the following four main periods: Old Japanese (OJ; 700–800), Early Middle Japanese (EMJ; 800–1200), Late Middle Japanese (LMJ; 1200–1600), and Modern Japanese (NJ; 1600–). Modern Japanese is abbreviated ‘NJ’ (‘New Japanese’) in order to avoid confusion with ‘MJ’, which we sometimes use for all of ‘Middle Japanese’ (EMJ and LMJ). We sometimes also make reference to ‘present-day’ Japanese. OJ (and EMJ) is, like NJ, considered a typical SOV or verb-final language. OJ is, like NJ, head-final, has postposed particles, verbal suffixes in derivational and inflectional morphology, and pervasive subject and object pro-drop. OJ/EMJ has an extensive inventory of inflecting verbal suffixes, which are not found in NJ, to express aspect, tense, and mood. Verb inflection proper in OJ/EMJ is for syntactic and modal categories (conclusive, adnominal, provisional, conditional, imperative, etc.). OJ/EMJ do not have a nominative case particle; subjects are sometimes bare and sometimes marked by one of the two genitive case particles no and ga. Of these two genitive particles, ga has become a nominative case particle in NJ, whereas no remains a genitive in modern Japanese. See further Frellesvig 2010 on premodern Japanese. In the Japanese examples we cite, we use a phonemic transcription appropriate for the period of the texts (see Frellesvig 2010: 176 for a simple transcription guide). We keep glossing in examples to a necessary minimum, for ease of reading. For reasons briefly explained at the end of this section, we use the Infinitive as the citation form of Japanese verbs.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

45

about the development of aktionsart verb constructions. Section 3.3 deals with the issue of wordhood of complex predicates in OJ/EMJ. We discuss some of the points which have been made in order to argue that V V complexes in OJ/EMJ do not form morphological compounds. While we show that some of these arguments have little force, we do agree that it is likely that at least some V V complexes in OJ/EMJ did not form compounds. We point out that a relationship similar to that found between the two verbs in the small number of NJ Type 3 Syntactic Compounds and Type 4 Syntactic Complex Verbs was far more widespread in OJ/EMJ. We propose that the change toward robust compound formation took place as a consequence of the Infinitive verb form losing its subordination function in the course of the Late Middle Japanese (LMJ) period, and at the same time we propose that there is a significant difference in terms of full grammaticalization potential between the NJ Type 4 Syntactic Complexes (Vte V), on the one hand, and, on the other hand, the NJ Types 2–3 (Lexical and Syntactic) Aspectual Compounds (V-V) and the OJ/EMJ aktionsart verb constructions (V V). We identify LMJ as the turning point in the changes that took place between OJ/EMJ and NJ. Like NJ, OJ/EMJ has a number of different structures which all in some sense involve two adjacent verbs. They may be summarized as follows:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

Different types of verb verb sequences in Old Japanese a. Biclausal structures Coordinate [ V1 ] [ V2 ] Subordinate [ [ V1 ] V2 ] b. Monoclausal structures Deverbal prefixes attached to verbs2 PFX-V Auxiliary verb constructions3 V + V{aux} Lexicalized (fixed or univerbated) compounds VV Complex verbal predicates Coordinate complexes V1 + V2 Thematic complexes V1{manner} + V2 Aktionsart complexes V1 + V2{aktionsart} V1{aktionsart} + V2

² Deverbal prefixes are semantically empty and simply serve to express some kind of emphasis. In OJ they seem limited to kaki- (< kaki ‘scratch’) and uti- (< uti ‘strike’), but their number increases significantly in the course of EMJ and LMJ, accompanied by phonological reduction (e.g. uti- > uC-). These prefixes probably derive from full grammaticalization of V1 in thematic complex verbal predicates (see Sections 3.2.1 and 3.2.6). ³ The two main auxiliary verb constructions in OJ are Vte (Gerund) + existential verb (first of all ari ‘be, exist’, but also other existential verbs), which expresses the periphrastic stative, and Vi (Infinitive) + si ‘do’ expressing generic activity or used for verb focusing.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

46

    

In this chapter we are concerned with complex predicates and only address other structures to the extent to which they are relevant to our discussion of complex predicates. We use the term “complex verbal predicate” in a broad sense, to refer to a predicate of a single surface-level clause which consists of two verbs which both contribute semantically to the predication.⁴ This broad definition will include both the compound verbs of NJ and serial verb constructions,⁵ but it excludes deverbal prefixes, auxiliary verb constructions (in which the auxiliary verb does not contribute semantically to the predication), univerbated compounds, and, obviously, as biclausal structures. In complex V1 V2 predicates in OJ/EMJ, V1 invariably has the form of the Infinitive (or the word-forming stem which is homophonous with the Infinitive; cf. Section 3.3). Thus, V1 in the V1 V2 structures we discuss is always in the Infinitive form (unless explicitly otherwise noted). For ease of recognition, we therefore use the Infinitive as the citation form of Japanese verbs.⁶

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.2 Clausality: complex predicates The OJ texts hold many examples of adjacent verbs where V1 is in the Infinitive. In this section we will discuss the problems encountered when trying to determine whether two (or more) adjacent verbs make up a complex predicate or whether they belong to different clauses. First, we have a number of examples where verbs in sequence with little or no doubt belong to different clauses. In (2),⁷ which describes the actions of the narrator on the death of his young son, we have eight consecutive and adjacent infinitives (two, asi-suri and mune-uti, with incorporated objects). Apart from tati wodori, which seem to form a complex predicate, ‘jump up’, rather than ‘stand up and (then) jump’, these are simply verbs in sequence, describing a sequence of events, or perhaps representative events (not necessarily occurring in the sequence of the verbs), jointly subordinate to te ni moteru a ga kwo tobasitu ‘(I) lost my child whom I was holding in my hand’. It is possible that there is a closer connection within each pair of verbs occuring together in a verse line (asi-suri ⁴ When characterizing complex predicate structures as ‘monoclausal’, we do so with regard to surface structure, not deeper levels of derivation which in some cases should be analyzed as involving multiple clauses which merge in the course of derivation (see Chapter 2). ⁵ Verb Verb complexes in OJ/EMJ and Verb-Verb compounds in NJ conform to most definitions of serial verb constructions (Nishiyama 1998; Aikhenvald 2006), whereas, for example, Butt (1995: ch. 7) differentiates between serial verbs, complex predicates, and compounds. ⁶ e.g. tati ‘stand up’, mi ‘see, look’, kake ‘hang (tr.), si ‘do’, and ki ‘come’. Most dictionaries of premodern Japanese use the conclusive verb form as the dictionary form, e.g. tatu, miru, kaku (NJ kakeru), su (suru), and ku (kuru). ⁷ All OJ texts were written in Chinese characters, as shown in (2), used phonographically and logographically. When citing examples from OJ, we render phonographically written text portions in italics (e.g. wodori ‘jump’ in (2)) and logographically written text in plain type (e.g. tati ‘stand up’) when citing full examples (but use italics for in-text citation).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

47

sakyebi; pusi apugi; and mune-uti nakeki) than between the verbs in different verse lines, but we have no good reasons for claiming that such a relation is reflected in syntactic structure (e.g. [ [ mune-uti ] nakeki ]), rather than simply being semantic/conventional ([ mune-uti ] [ nakeki ]). (2)

立 乎杼利 足須里 佐家婢 伏 仰 [tati wodori] [asi-suri] [sakyebi] [pusi] [apugi] stand.up jump foot-rub scream lie.down look.up 武祢宇知 奈気吉 手 爾 持流 安我 古 登婆之都 [mune-uti] [nakeki] te ni mot-eru a ga kwo tobasi-tu chest-beat lament hand  hold- I  child let.fly- ‘Jumping up, rubbing my feet (as a sign of sadness), screaming, lying down, looking up at the sky, beating my chest, and lamenting, (I) lost my child whom I was holding in my hand.’ (MYS8 5.904)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The kind of coordinate verb sequencing we see in (2) is common in OJ/EMJ. In many such cases it can be difficult to distinguish between a coordinate complex predicate or dvandva compound and two coordinate clauses, as, for example, in (3), where mimoro ga upe ni could equally well be the argument of nobori, of tati, or of nobori tati, with a number of different structural parsings, but without a great deal of difference in meaning. We will not discuss coordinate complex predicates to any significant extent in the following, but note the remark on (8b) in Section 3.2.1. (3)

mimoro ga upe ni nobori tati Mimoro  top  climb stand ‘(If I was to) ascend and stand on the top of Mimoro’ (NSK 97)

In (2) and (3) (except for tati wodori), the adjacent verbs retain their full lexical meaning and exhibit (a high degree of) coordination. Therefore, the structural ambiguity is in a sense not very significant. However, there are many examples of V1 V2 sequences where one verb is subordinate to the other. Such cases can present more significant problems of structural interpretation: [ [ V1 ] V2 ] (biclausal) or [ V1 V2 ] (monoclausal V V complex). We will give a few examples which all have V1 V2 sequences and therefore potentially could be either biclausal or monoclausal. In many cases, a determination can be made only on the basis of the semantic interpretation, as many clauses because of extensive pro-drop have unexpressed arguments. However, in (4), the two verbs have separate arguments: ⁸ We use the following standard abbreviations for OJ sources: MYS: Man’yōūshu, cited from the NKBT edition (vv. 4–7); NSK: Nihon shoki kayo, cited from NKBT v.3; SM: Shoku Nihongi senmyo, cited from Kitagawa (1982).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

48

    

[idumi no kapa no payaki se wo . . . watari] and [sawo sasi], and we are clearly dealing with two clauses, one being subordinate within the other. (4)

sasi watari (MYS 13.3240) [ idumi no kapa no payaki se wo [sawo sasi ] Izumi  river  fast rapids  pole plunge.in watari ] cross ‘Cross the fast rapids of the Izumi River, plunging in a pole’

In (5), the sole noun phrase, sapogapa no kwoisi, is the argument of V1 pumi and we again have two clauses, one subordinate within the other. (5)

pumi watari (MYS 4.525) [ [ sapogapa no kwoisi pumi ] watari ] Saogawa  pebble step.on cross ‘Cross, by stepping on the pebbles in the Saogawa River’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.2.1 Thematic verbal complexes In (6), the sole noun phrase, asipikwi no yatu wo, is the argument of V2 kwoye, and V1 pumi is subordinate, expressing the manner of crossing. Semantically, this example is like the (Type 1) Lexical Thematic Compounds of NJ, and it seems that we have a complex predicate in which V1 pumi modifies V2 kwoye in the same way that V1 modifies V2 in a Lexical Thematic Compound in NJ. (6)

pumi kwoye (MYS 19.4164) [ asipikwi no yatu wo pumi  eight hill step.on ‘Walk across many mountains’

kwoye] cross

OJ/EMJ have large numbers of such examples, e.g. from OJ arapi susugi ‘wash’ (‘wash’ + ‘rinse’), omopi sini ‘die from longing’ (‘long for’ + ‘die’), yaki porobosi ‘burn down’ (‘burn (tr.)’ + ‘destroy’), in which V1 modifies V2 in what we very broadly can describe as ‘manner’ (encompassing a variety of more particular semantic relations). Functionally, such examples are no different from NJ Type 1 compounds such as koi-sitau ‘long for’ (‘love’ + ‘yearn for’), yake-sinu ‘burn to death’ (‘burn (intr.) + ‘die’), koroge-otiru ‘roll down, fall out of ’ (‘roll’ + ‘fall’), or kiri-taosu ‘cut down’ (‘cut’ + ‘fell’). We interpret these examples as having complex predicates, but it must be noted that strictly speaking it is not possible to determine that V1 is always part of a complex predicate, rather than forming its

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

49

own one-word subordinate clause, i.e. [ asipikwi no yatu wo [pumi] kwoye ] ‘cross many mountains, by stepping (on foot)’, or [ [ yaki ] porobosi ] ‘destroy, by burning’. It should also be noted, however, that examples such as (8b, c) in the following section have an aktionsart verb having scope over what seems to be a monoclausal complex predicate: pazimete taking scope over [ megumi moye ] (which seems to be a coordinate complex), and kanetu taking scope over [ tobi tati ], a thematic complex.

3.2.2 Aktionsart verb constructions Finally, in (7) we have a clear example of a complex predicate, with watari (whose lexical meaning is ‘cross’, as in several of the examples) having no independent lexical meaning, but here expressing ‘continuative’ or ‘indefinite extent’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(7)

kwopwi watari (MYS 15.3603) [ yuyusiki kimi ni kwopwi wataru kamo ] awesome lord  yearn   ‘I will continue to yearn for you, my awesome lord!’

Such examples represent aktionsart verb constructions in which one of the verbs, here kwopwi, is the main verb (the primary determinant of argument structure) and the other, watari, is grammatical(ized) and augments the lexical meaning of the first in terms of aktionsart meanings.⁹ Functionally, such examples are largely like the Type 2 and 3 (aspectual) compounds of NJ, but we use the term “aktionsart” here to allow for inclusion of a wider range of meanings than those traditionally associated with “aspect.” The aktionsart verbs in Japanese are characterized by retaining their use as full lexical verbs alongside their more grammatical function as part of complex predicates.¹⁰ At the surface level, these aktionsart complex predicate constructions are monoclausal, although some of them derive from deeper-level multiclausal complementation structures which merge in the course of derivation.

⁹ It should be noted that the term ‘aktionsart’ within scholarship on V V complexes in Japanese (as in this volume) is not used in the narrow sense of ‘lexical aspect’ in which it has been adopted in much English-language literature, but in the more traditional wider sense of ‘(grammatical) manner of action’, which the word itself (aktions-art) suggests. ¹⁰ In the Japanese grammatical tradition, the aktionsart verbs are included in, and make up the larger part of, a category labeled “hojodōshi” (lit. ‘helping verb’, usually translated as ‘auxiliary verb’). However, this categorization is unfortunate as it includes both the aktionsart verbs that take part in forming complex predicates and auxiliary verbs proper; and terminologically it is, of course, misleading because most the verbs included under this label (i.e., the aktionsart verbs) are not auxiliary verbs in the conventional use of that word.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    

50

We give a few more examples of aktionsart verb constructions in OJ: (8)

a. tamapi ‘respect’ (MYS 2.159) wa go opo-kimi no yupu sare-ba myesi I  great-lord  evening become. see. tamapu rasi ake kure-ba twopi tamapu rasi []  dawn come- visit []  ‘When evening falls, my great lord will surely (deign to) see. When dawn comes, he will surely (deign to) visit’ b. pazime ‘inceptive’ (SM 46) yorodu no mono mo megumi moye pazime-te 10,000  thing too grow.rich sprout - ‘Myriad things are beginning to grow rich and bud’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. kane ‘negative potential’ (MYS 5.893) yononaka wo usi to yasasi to omope-domo this.world  sad  shameful  think- tobi tati kane-tu tori ni si ara-ne-ba fly stand.up - bird  -- ‘Although I find this world sad and shameful, I cannot fly away—as I am not a bird‘ d. ki Adventive (‘come’) (MYS 9.1757) tukupane no yokyeku wo mire-ba nagaki ke ni Tsukuba.Peak  goodness  look- long days  omopi tumi ko-si urepe pa yami-nu think pile - sadness  stop- ‘When I look at how fine Tsukuba Peak is, the despondency that I came to feel for long days has stopped.’ The OJ inventory of aktionsart verbs includes the following (see Frellesvig et al. 2010: 35–6 for a fuller inventory), here grouped in semantic categories, giving in brackets the lexical meaning of each aktionsart verb. (9)

Some OJ V2 aktionsart verbs11 Explorative: mi (‘see’) Reciprocal: api (‘meet’) Adventive: ki (‘come’)

¹¹ As aktionsart verbs by definition are verbs which can both function as full lexical verbs and have some more grammatical function as part of a complex predicate, a few of the verbs traditionally included as aktionsart verbs in OJ might be excluded on the grounds that they are grammaticalized to the extent that they only ever occur in their grammatical function together with a main verb and never as full lexical verbs on their own, e.g. kate, kane, and some in the current list.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

51

Inceptive: some (‘begin’); ide (‘go out’); kake (‘hang up’); pazime (‘begin, start (tr.)’) Potential: e (‘get’); kane (‘fail’); kate (‘prevail’) Continuative: yuki (‘go’); tugi (‘convey’); topori (‘pass through’); watari (‘go across’) Progressive: wori (‘be sitting’) Social deixis: imasi (‘be’); itadaki (‘receive’); maturi (‘offer’); myesi (‘see’); tamapi (‘give’); tamape (‘receive’) Degree: kwosi ‘put over’; kwoye ‘go over’; masari ‘exceed’; sugusi ‘put beyond’; sugwi ‘go beyond’; tari ‘suffice’. As has been pointed out by, for example, Aoki (2013) and Kageyama (2013), the categories expressed by aktionsart verbs have remained largely the same over time from OJ to present-day Japanese, although the actual inventories of verbs used as aktionsart verbs have changed.

3.2.3 V1 aktionsart verbs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

A significant difference between OJ and present-day Japanese is the existence in OJ of a complex predicate construction not found in NJ in which V1 is an aktionsart verb and V2 is the main verb. We give some examples in (10)–(13). (10)

ari Iterative/ Stative (‘be, exist’) (MYS 20.4408) sima-dutapi i-kogi watari-te ari meguri island-pass.along -row cross-  go.around ‘I keep going around and around the islands, crossing by boat between them’

(11)

api Reciprocal (‘meet’) (MYS 19.4264) api noma-mu ki  drink- saké ‘The saké we will drink together’

(12)

ki Adventive (‘come’) (MYS 17.3981) asipikwi no yama ki pyenari-te  mountain  be.between-1 ‘The mountains having come between us’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

52 (13)

     siki Exhaustive/Continuative (‘stretch, extend’) (MYS 4.786) paru no ame pa iya siki puru spring  rain  more.and.more  fall ‘Although the spring rain keeps falling’

ni although

(14) is a list of some V1 aktionsart verbs. All of these are also used as V2 aktionsart verbs. A good amount of detailed descriptive work remains to be done with regard to the V1 aktionsart verb construction in OJ, but the verbs given in (14) are good candidates, out of what is probably a somewhat larger set.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(14)

Some OJ V1 aktionsart verbs ari Iterative/Stative (‘be, exist’), e Potential (‘get’), api Reciprocal/ Proximative (‘meet’) ape Additive (‘join, vt.’), ide Elative (‘emerge, go out’), ki Adventive (‘come’), siki Exhaustive/Continuative (‘stretch, extend’), tati Inceptive (‘stand up, set out’), topori Permeative (‘pass through’), tugi Contiguative (‘pass on’)

The V1 aktionsart verb construction is not recognized as such in descriptive grammars of OJ, although some V1 aktionsart verbs are mentioned as deverbal “prefixes,” in particular api, ari, e, and dictionaries will sometimes remark on “helping” (= aktionsart) verblike usage in V1 position of some verbs. However, this is a significant construction which merits explicit recognition and further study. It should also be noted that the V1 aktionsart verb construction is not a good typological fit with the overall structure of Japanese. Japanese is a strong SOV language, with suffixation and postverbal morphology, so the preverbal position of an aktionsart verb is highly unusual (and not, of course, found in NJ). We discuss this point and the place of the V1 aktionsart verb construction within the diachrony of complex predicates in Japanese we list (see Section 3.2.7).

3.2.4 Summary of types of complex predicates Thus, we observe the following types of complex predicates in OJ. Functionally, (15)-I is like the Type 1 Lexical Thematic V-V compounds of NJ, while (15)-II functionally is like the Type 2–4 complexes, which differ morphologically and syntactically, but all involve aktionsart verbs.¹²

¹² Note, however, that some of the structures conventionally included in Type 4 (syntactic complex verbs) in NJ in fact seem not to be complex predicates, but simple auxiliary verb constructions (e.g. Vte i/ari/simai/oki/iki etc.); see further Section 3.3.7. This difference is obscured in the Japanese grammatical tradition, which includes under the label hojodōshi (lit. ‘helping verb’) both aktionsart verbs and auxiliary verbs proper.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

        (15)

I. Thematic verbal complexes V1 manner V1 modifies V2, mainly in terms of manner; V2 is the primary determinant of argument structure and case assignment. II. Aktionsart verb constructions A. V1 aktionsart verb V1 is grammatical(ized) and augments the lexical meaning of V2 in terms of aktionsart meanings; V2 is the primary determinant of argument structure. B. V2 aktionsart verb V1 is the primary determinant of argument structure; V2 is grammatical(ized) and augments the lexical meaning of V1 in terms of aktionsart meanings.

53

V1(manner) V2

V1(aktionsart) V2

V1 V2(aktionsart)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.2.5 The development of complex verbal predicates in Japanese In very general terms, it may be assumed that complex predicate constructions (as well as the other monoclausal structures mentioned in (1)) arose, directly or indirectly, through diachronic reinterpretation of biclausal structures in which two adjacent verbs had the same subject. The fact that (Old) Japanese is verb-final and has a high frequency of both subject and object pro-drop will facilitate this, as the verbs of two adjacent clauses, especially with the same subject, often will appear next to each other, resulting in surface strings which are open to reinterpretation as single clauses with a complex predicate. This may informally be noted as [ [ V ] V ] > [ V V ] (or, for coordinate structures [ V ] [ V ] > [ V V ] ). Although the available materials are not sufficient to chart the details of the course of developments, OJ does provide material which allows us to suggest a more detailed and concrete set of innovations which may be thought to have been involved in the development of the OJ/EMJ and LMJ/NJ systems of complex predicates. The developments we have in mind may be summarized as follows. (16)

Biclausal subor- > Thematic > V1 aktionsart > V2 aktionsart dinate structures complexes verb verb [ [ Vi ]{manner} Vj ] > [ Vi{manner} Vj ] > [Vi{aktionsart}Vj ] > [ Vj Vi{aktionsart} ]

On our proposal, the V1 aktionsart verb type complex predicate is a transitional, but pivotal construction in the genesis of aktionsart verb constructions in Japanese. All four stages remain populated throughout the OJ/EMJ period, although numbers and occurrences of V1 aktionsart verb constructions decreased

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

54

    

significantly through EMJ and disappeared as productive constructions. Apart from the V1 aktionsart verb construction, the chain of developments in (16) is thus not a chain of replacements, but shows the sources of each construction.

3.2.6 Thematic complexes: Adverbialization

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

As we have noted, it can be difficult to determine the clausal structure of the constructions which are functionally equivalent to the Type 1 lexical thematic compounds in NJ. The transition from biclausal to monoclausal was a fairly small step, which can be noted as in (16): [ [ Vi ]{manner} Vj ] > [ Vi{manner} Vj ]. This type of development is in line with a similar type of development in which subordinate verb forms become adverbs, through what may be described as “deverbalization”: loss of argument structure and clause projection. An example is the verb pazime ‘start, begin (tr.), initiate’, which in addition to its use as a full lexical verb and V2 aktionsart verb ((9)) is found in adverbial expressions which range from depleted to no clausal status: a subjectless, depleted subordinate (adverbial) verb phrase kyepu wo pazimete (MYS 8.1530) ‘beginning with today’; an Infinitive-derived adverb, pazime ‘(at) first; from/at the beginning’; and a Gerund-derived adverb: pazimete ‘(at) first; from/at the beginning’. All of these adverbial uses must have originated in reanalyses of short or single-word subordinate clauses, accompanied by (some) loss of independent argument structure. This may also be thought to be the origin of the use of V1 in thematic complex predicates: subordinate modifying uses of verbs incorporated into the clause of the following verb.

3.2.7 Aktionsart verbs OJ has both V1 and V2 aktionsart verbs, whereas NJ only has V2 aktionsart verbs, as in (17). This distribution suggests that OJ in this respect represents a transitional stage of some kind. (17)

Old Japanese V1 aktionsart (fewer; less frequent) V2 aktionsart (more; more frequent)

Modern Japanese V2 aktionsart

We posit two innovations in the course of the development of aktionsart verbs in Japanese: see (18).¹³ The first, (18a), is the development of V1 aktionsart verbs ¹³ An entirely different alternative explanation of the V1 aktionsart verb construction is that it reflects an older, different basic word order. This has been indirectly proposed, for example, by Vovin (2009) in the section in his reference grammar of OJ on verb ‘prefixes’ (which include api, ari, e; like other grammars of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

55

from thematic complexes and the second, (18b), is the development of V2 aktionsart verbs from V1 aktionsart verbs. The innovation proposed here under (18a) is presented as a reinterpretation of thematic complex predicates of the structure [ Vi{manner} Vj ], but it would be equally applicable to biclausal [ [ Vi ]{manner} Vj ], in which case it would be noted [ [ Vi ]{manner} Vj ] > [Vi{aktionsart}Vj ] and involve an innovation regarding clausal structure, as well as the driving categorical reinterpretation. On this proposal, the V1 aktionsart verb construction is a transitional type which disappears in later stages of the Japanese language, but which played a pivotal role in the development of the system of aktionsart verb constructions in Japanese. (18)

a. Categorial reinterpretation: from lexical ‘manner’ modification to grammatical aktionsart augmentation [ Vi{manner} Vj ] > [Vi{aktionsart}Vj ]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Structurally motivated shift in surface position: from preverbal to postverbal aktionsart verb. [Vi{aktionsart}Vj ] > [ Vj Vi{aktionsart} ] The first innovation, which resulted in V1 aktionsart verbs, is a semantically motivated categorial reinterpretation from lexical “manner” modification to grammatical “aktionsart” augmentation,¹⁴ with no overt effect on surface structure, but with invisible effects on the categorial properties of individual items and as a result also on their position in underlying syntactic structure; this is an abductive innovation in the technical terminology of language change theory (see Andersen 1989 for an overview of types of innovations). It is a straightforward case of “grammaticalization” in the literal meaning of the word: the reinterpretation of something lexical as being more grammatical. The second major innovation involved is a structurally motivated shift in surface position from preverbal to postverbal aktionsart verb. This is in a sense simply the realization in linear order of the structural implications of the first innovation, namely the realization in surface structure of a more grammatical element in its underlying syntactic position (given the overall typological properties of Japanese); this is a deductive innovation (Andersen 1989).

OJ, Vovin’s reference grammar does not recognize V1 aktionsart verbs as such), where he in passing and with no further elaboration or comment remarks that the preverbal position of these morphemes may reflect “the last stages of transition from a SVO to a SOV language” (2009: 589). This would conform to the idea that “today’s morphology is yesterday’s syntax,” but, first, nothing else within OJ suggests a transition from SVO to SOV typology, and, second, while the notion that “today’s morphology is yesterday’s syntax” surely is relevant in some cases, it should not be assumed to be the default diachronic explanation for observed morpheme order (cf., e.g., Comrie 1980; see also Andersen 1987:1). ¹⁴ Recall that ‘aktionsart’ here is used in the sense of ‘(grammatical) manner of action’ (cf. 3.2.2). There is clearly a close semantic connection between (adverbial) manner and (grammatical) aktionsart.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

56

    

This, thus, is the overall scenario we propose: a semantically motivated categorial reinterpretation, with structural implications, which are realized gradually, through the presence of both preverbal and postverbal aktionsart verbs, but with the older, structurally unmotivated order disappearing over time to give us NJ. What we have reflected in OJ is the transitional stage in the course of the gradual realization of the change, as shown by the presence in OJ of both preverbal (V1) and postverbal (V2) aktionsart verbs. Compare the following three examples, which all include forms of topori ‘pass through’ in combination with nure ‘be soaked’. In (19a), the Gerund toporite is used adverbially, but is semantically bleached, to mean ‘through and through; completely. In (19b), we see V1 aktionsart verb use of topori for Permeative, and in (19c) we have V2 aktionsart verb use in the same sense. (19)

a. koromo no swode pa toporite nure-nu garment  sleeve  passing.through get.soaked- ‘The sleeves of (my) clothes are completely soaked’ (MYS 2.135)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. amawotomyera ga [ swode topori nure-ni-si ] diver.girl  sleeve . get.soaked-- koromo garment ‘The clothes of the diver girls, where the sleeves are soaked through’ (MYS 7.1186) c. sigure puri nure toporu tomo rain.showers fall get.soaked . even.if ‘even if (I am) soaked through, with rain showers falling’ (MYS 9.1760) We noted in Section 3.3 that the V1 aktionsart verb construction is a bad typological fit in Japanese. However, in our understanding of the development of the Japanese system of aktionsart verbs, the V1 aktionsart verb construction is a transitional, but necessary type in the development of the V2 aktionsart verb construction, which itself, once established, has remained part and parcel of Japanese throughout its attested history.

3.2.8 Similar developments It is possible to find similar structurally or typologically motivated shifts in position within the history of Japanese which may be understood as involving the same set of innovations as the aktionsart verbs. A case in point is the Prohibitive (negative command) na, which seems to have changed from a preverbal adverb to a postverbal grammatical particle. In OJ, Prohibitive na can occur in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

57

both preverbal and postverbal position,¹⁵ but in NJ it is only found in postverbal position. See (20) for examples of both. (20)

Prohibitive na (with yuki ‘go’) Old Japanese Preverbal prohibitive na yuki (so) Postverbal prohibitive yuku na

Modern Japanese Postverbal prohibitive yuku na

This may be explained by the same kind of innovations which we posit in the development of aktionsart verb constructions: Categorial reinterpretation (lexical to grammatical: from adverb to grammatical particle), followed by a structurally motivated shift in surface position (from preverbal to postverbal). In fact, several grammatical suffixes in OJ seem to have originated in adverbs which were reinterpreted in exactly the same way, from lexical (preverbal) adverb to grammatical element, which then subsequently shifted to postverbal position, accompanied by acquisition of morphology, e.g. ube ‘indeed’ > be- (a modal postverbal clitic with adjectival morphology), or ani ‘no!, nay!, (not) at all’ (probably cognate with the Korean negative adverb ani) > -(a)n- the negative verb suffix. From much further afield, another example of a structurally motivated shift in position following grammatical reinterpretation are forms of the Polish verb ‘to be’, which over several centuries shifted from sentence clitics, regularly placed in clause-second position, to verbal desinences marking person and number (see Andersen 1987).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.3 Constituency (wordhood) We have noted complex predicates as two separate words (or, rather, with a space between the two verbs). In this section we will discuss the wordhood of the OJ/EMJ complex predicates. Traditionally, the two verbs in a complex predicate have been interpreted as forming a compound, in parallel with the Type 1–3 compounds in NJ.¹⁶ However, the view has also proposed for some time that the V V complex predicates of OJ/EMJ differ from the compound verbs of NJ in respect of wordhood and that they should not be considered compounds. In particular, it has been pointed out that, as opposed to NJ (e.g. Yoshizawa 1952; Kindaichi 1953; Hyakutome 2001, 2002): (21)

a. There are examples in OJ/EMJ where the order of V1 and V2 seems to be reversible without a change in meaning. b. V1 and V2 can be separated by (focus or restrictive) particles in OJ/EMJ. c. V1 V2 sequences do not exhibit compound accentuation in EMJ.

¹⁵ When preverbal, OJ na is often reinforced by postverbal so, which is a fossilized pre-OJ imperative of si ‘to do’. ¹⁶ This convention was, for example, adopted in Frellesvig et al. 2010.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

58

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The reason that this is an issue is that the verb stem used for compound formation is homophonous (segmentally identical) with the Infinitive. Thus, two adjacent verbs where the first has a shape which can be either the Infinitive or the wordforming stem are structurally ambiguous.¹⁷ A surface string such as /torimoti/ (cf. tori ‘take’ and moti ‘hold’) can be parsed as [ [ tori ] moti ] (a biclausal structure), [ tori moti ] (a complex V V predicate), [ tori-moti ] (a complex compound), or [ torimoti ] (a univerbated compound). Before briefly addressing each of the points in (21), it should be made clear that it is not—or at least should not be—in dispute that OJ had VV compound verbs, some of which were clearly lexicalized, e.g. torimoti (< tori ‘take’ + moti ‘hold’), which can mean ‘perform administrative business’ in which the structure and meaning of the two verbs tori and moti are lost (Abe 2011). The existence of compound verbs in OJ and EMJ is also clearly shown by the existence of nouns derived from VV compounds, e.g. OJ kapyerimi ‘looking back, hesitation, concern’ < kapyeri-mi ‘look back’ < kapyeri ‘return (intr.)’ + mi ‘see, look’, or omopide (attested early EMJ) ‘keepsake, memory’ < OJ omopide ‘remember, come to think of ’ < omopi ‘think’ + ide ‘emerge’. Furthermore, in some verb classes (consonant base verbs ending in /p (~w), t, k, b, g, m, n, s/), the word-forming stem, but not the Infinitive, acquired a variant (onbinkei ‘euphonic stem’) in the course of early EMJ. It is significant that we in some cases find the onbinkei variant preceding an aktionsart verb and clearly making up a compound, as in (22), where we have omou-tamawe ( Vtoki) and may be on the way to full grammaticalization as suffixes. It thus seems that the Vte V construction lends itself more easily to further grammaticalization than the OJ/EMJ Vi V complex predicate construction did. One example is the development of the Vte oki construction, which today is used both for ‘advance preparation’ (zyugyoo no mae ni tango no imi o sirabete oku ‘check out the meaning of vocabulary before the class’) and to express ‘reproach’ (makete oite hansee siyoo to sinai ‘you don’t try to reflect, even if you lost’). The former sense (advance preparation), which metaphorically retains some of the lexical meaning of oki ‘put’, was expressed by Vi oki already in EMJ, but the latter sense (reproach) only developed from the end of LMJ once the expression had been reformed as Vte oki. Thus, with the new shape Vte oki, new meanings were added and that can be taken to show advanced grammaticalization (Isshiki 2016). This suggests an important difference between, on the one hand, the NJ Type 4 Vte V complex predicates and, on the other hand, the OJ/EMJ Vi V aktionsart verb construction and the NJ Types 2–3 aktionsart V-V verb compounds. It has been observed (Butt and Lahiri 2013) that “light verbs” (the term Butt and Lahiri use for a class of verb in South Asian languages which shows a great deal of similarity to the aktionsart verbs in this chapter) are robustly stable over time: They tend to retain both their lexical and grammatical uses over time and do not sit as intermediary points on a grammaticalization cline. The OJ/EMJ Vi V aktionsart verb construction and the NJ Types 2–3 aktionsart V-V verb compounds conform well to this observation and generalization, whereas the Vte V construction does not. Vte V has been found throughout the history of attested Japanese, but in OJ/ EMJ either in clearly biclausal structures ([ [ Vte ] V ]) or in auxiliary verb constructions [ Vte V{aux} ],²² but not in complex predicates of the form [ Vte V ].

²² As we have mentioned, the main OJ auxiliary verb construction was the periphrastic stative Vte ari (ari ‘exist’). This construction was contracted and univerbated and gave rise to an inflecting suffix, EMJ -tari stative aspect, which in turn gave rise to the NJ past tense flective -ta, thus exhibiting the kind of full grammaticalization proposed here as a tendency of [ Vte V ] constructions in Japanese.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

68

    

That construction is a LMJ innovation.²³ It may be hypothesized that it is a grammatical construction very different from the OJ/EMJ Vi V aktionsart verb construction and the NJ Types 2–3 aspectual V-V verb compounds, and that it attracts verbs which are particularly susceptible to full grammaticalization; and conversely, that once a verb is a V2 in a Type 4 [ Vte V ] complex predicate, it is likely to take part in advanced and eventually full grammaticalization, ending up as a real auxiliary verb. It is worth pointing out that while Japanese has had Vte V{aux} constructions throughout its attested history, these were few before LMJ, whereas the development of complex predicates into compound verbs was accompanied by an increase in auxiliary verb constructions.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3.4 Summary In Section 3.2 we outlined the system of complex predicates in OJ/EMJ. While OJ/ EMJ had a set of complex verbal predicates which is functionally much like that of NJ in some respects, comprising Thematic complexes and aktionsart verb constructions, we pointed out the existence of V1 aktionsart verb constructions in OJ/ EMJ, which is not found in LMJ or NJ. We proposed a hypothesis about the origin of complex predicates in Japanese in which the V1 aktionsart verbs played an important role, positing a number of innovations, starting with reinterpretation of biclausal [ V1 [ V2 ] ] structures as monoclausal [ V1 V2 ], followed by other innovations. As already mentioned, aktionsart complex predicates (which overlap with what Butt and Lahiri (2013) call “light verb” constructions) are remarkably stable over time, once they are present in a language. On our hypothesis, Old Japanese offers a view of the origin and provenance of such constructions, which may be of wider application in future research on the origin of such constructions in other languages. In Section 3.3 we discussed the wordhood of complex verbal predicates in OJ/ EMJ. We showed that some of the arguments traditionally invoked to suggest that OJ/EMJ V V complexes were not compounds had little force, while agreeing that two adjacent verbs in a complex predicate in OJ/EMJ did not merge phonologically into a compound word. We discussed the development of compound-word formation, attributing this to the loss of subordinating function of the Infinitive. We finally observed that while the NJ Type 1–3 compounds reflected the functional system of OJ/EMJ, the NJ Type 4 is an innovation which is not a prototypical aktionsart verb construction, but rather inclines toward an auxiliary verb construction. We also noted that the change of complex predicates from adjacent

²³ Again marking out LMJ as the pivotal period in this change from the OJ/EMJ type complex predicates to the NJ system of V-V compounds (Types 1–3) and syntactic complexes (Type 4), as well as in many other significant morphosyntactic changes in the history of the Japanese language.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

       

69

verbs into compound words was accompanied by an increase in the number of Vte V{aux} auxiliary verb constructions.

Acknowledgments This chapter forms part of the research project “Construction of Diachronic Corpora and New Developments in Research on the History of Japanese” at the National Institute for Japanese Language and Linguistics in Tokyo. Frellesvig is grateful to Stephen W. Horn for discussion and comments on earlier versions of the chapter. We would also like to thank Taro Kageyama for his insightful and constructive comments and suggestions.

Sources

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

NKBT = Nihon Koten Bungaku Taikei. Tokyo: Iwanami Shoten. SNKBT = Nihon Koten Bungaku Taikei. Tokyo: Iwanami Shoten. SNKBZ = Shin Nihon Koten Bungaku Zenshū. Tokyo: Iwanami Shoten.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

4 Grammaticalization and constructionalization in Japanese lexical compound verbs Taro Kageyama

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4.1 Introduction Based on morphological, syntactic, and semantic criteria, Chapter 2 (Kageyama, this volume) classifies Japanese V-V complexes into four major types (plus a fifth type representing the speaker’s polite or disparaging attitude). Among them, Type 2 “lexical aspectual” compounds, characterized by a cluster of conspicuous features ranging from morphology and lexicon to semantics and syntax, appear to occupy a unique position in the architecture of V-V complexes not only in Japanese but also in other Asian languages. In terms of morphology, many of the second members (V2s) are no longer capable of independent usage, at least in the same meaning as they manifest in compounds, but they nonetheless preserve the verb category, exhibiting the same inflection as regular verbs. Syntactically, unlike Type 3 “syntactic” compound verbs that originate from biclausal complementation structures, the compounds of this class are incapable of accommodating in the V1 position such syntactic elements as passive -rare, causative -sase, or honorific o-V ni nar-. Hence the Type 2 compound verbs are monoclausal and “lexical” (i.e. formed by combining two verbs directly and registered in the lexicon). Their lexical nature is reflected in their limited productivity: some V2s are restricted to one or two particular verbs in V1, though others have slightly greater freedom of combination. Semantically, V2s are desemanticized to such a great extent that their interpretations often appear non-compositional, thus causing problems for learners in both first and second language acquisition. In particular, the argument structure of the whole compound verb is determined by V1, whereas V2s, with the loss of their argument structure, serve as semantic modifiers enriching V1’s event structure with a wide array of aktionsart (lexical aspectual) concepts depicting how V1’s event does or does not unfold. As shown in Table 2.4 (Chapter 2), these added concepts include not just temporal aspect (e.g. inception, incompletion, continuation,

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

  

71

completion, suspension, retardation), but intensity of action and result, spatial aspect (e.g. orientation and extent in space), and social deixis (e.g. orientation in interpersonal relations such as higher or lower status) as well. Additionally, many of the Type 2 compound verbs characteristically carry a literary or rhetorical flavor. In light of the robust head-final structure of Japanese, Type 2 compound verbs have a highly marked status, contrasting with the unmarked status of Type 1 “thematic” compound verbs. In Type 1 compounds, the argument structure of the whole compound is principally regulated by the head verb in V2, as in kabe ni kugi o uti-tukeru [pound-attach] ‘pound a nail into a wall’, where the ‘wall’ and ‘nail’ arguments are selected by V2 tukeru ‘attach and fix’. In Type 2 lexical aspectual compounds, on the other hand, it is V1 that controls the argument structure of the whole compound, as in kodomo o sikari-tukeru ‘scold a child severely and unilaterally’, where the accusative object ‘child’ is selected by V1 sikari ‘scold’ and V2 tukeru intensifies the harshness and unilateral direction of V1’s scolding action. Intriguingly, this meaning of intensification, observed in the compound structure, is not manifested by tukeru when it is used as an independent verb or as the V2 of a Type 1 compound. A proper understanding of the atypical properties of Type 2 compound verbs and of their raison d’être in Japanese grammar requires a comprehensive approach encompassing both synchronic and diachronic perspectives. The present chapter attempts to elucidate the nature of these peculiar compounds, bringing to light some diachronic data that are challenging to grammaticalization studies. In the traditional “item-based” approach to grammaticalization, morphological and functional changes are supposed to take place gradually and unidirectionally along a universal pathway like Hopper and Traugott’s (2003) cline of grammaticalization in (1). (1)

content item > grammatical word > clitic > inflectional affix

It is true that Japanese offers a great deal of material that conforms to the grammaticalization cline, as in the allative case particle (postposition) e derived from the noun pye¹ ‘vicinity’ in Old Japanese, and Narrog and Ohori (2011) argue that the grammaticalization processes in this language are straightforward and transparent by virtue of its agglutinative character. The notion of agglutination alone, however, will not account for the asymmetry observed inside compound verbs. In V-V compound verbs, some of the first verbs which lost their original meaning and argument structure have ended up as prefixes, as in uti- (original meaning of ‘strike’ ¹ In this chapter, pre modern Japanese is given in a phonemic transcription appropriate to the date of citation. Following Frellesvig (2010), kōrui “e” and “o” in OJ are respectively represented as ye and wo, and otsurui “i” as wi. Words from Early Middle Japanese without a specific date are transcribed in their eleventh-century shape; see Frellesvig (2010: 176) for an easy guide to converting between historical kana spelling and phonemic transcription.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

72

 

changed to intensify V2’s action) in uti-kyooziru [PRE-enjoy] ‘enjoy greatly’, and but- (original meaning of ‘beat’ (buti-) changed to emphasize the violent manner of V2’s action) as in but-taoreru [PRE-fall] ‘collapse with a thump’. Notably, decategorization to affixes like these is restricted to the V1 position of Type 1 (thematic) compounds. The second verbs in Type 2 (aspectual) compounds, though lacking argument structure, are immune to phonetic reduction and decategorization because the verb category is required to carry verbal tense and other inflections. Heine (2001: 579) proposes four interrelated mechanisms that work in a chain to bring about grammaticalization effects; he writes:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(2)

desemanticization (or “bleaching,” semantic reduction): loss in meaning content; ii. extension (or context generalization): use in new contexts; iii. decategorization: loss in morphosyntactic properties characteristic of the source forms. Including the loss of independent word status (cliticization, affixation). iv. erosion (or “phonetic reduction”): loss in phonetic substance.

i.

Where on this chain can the change observed in Type 2 compound verbs be located? Since the third (decategorization) and fourth (erosion) mechanisms are irrelevant, only the first and second pertain. The semantic dilution to adverbial meanings can be identified as “desemanticization” (2i), and the manifestation of the new meanings in the context of compound structure as “extension” (2ii). However, it remains to be seen whether these two mechanisms came into play in that order in the historical development of this particular type of compound. This chapter aims to disentangle the complex interactions of form and meaning in Type 2 aspectual compounds by paying attention to their historical origin. In the absence of morphological clues for uniquely identifying this class of compounds, our arguments inevitably rely on semantic interpretation. Section 4.2 will explicate the typical patterns of semantic interpretation in which the semantic structures of two verbs are integrated into a coherent semantic structure. Based on this semantic foundation, the next two sections enter into diachronic exploration, presenting a panoramic view of the development of Type 2 compound verbs. Specifically, Section 4.3 clarifies how the “auxiliary” verbs in Old Japanese (OJ) and Early Middle Japanese (EMJ) are related to the V2s of Type 2 compounds in Present-day Japanese (PDJ), and Section 4.4 traces the earliest appearances of representative V2s of current Type 2 compounds. In Section 4.5, the findings in these two sections are put together, arguing that Type 2 compound verbs emerged in Late Middle and Early Modern Japanese as a result of the interplay of reanalysis, lexicalization, and desemanticization, which jointly led to the “constructionalization” of a novel compound structure. In Section 4.6, it is suggested that the rise of this

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

73

marked class of compound verbs was motivated by language users’ rhetorical creativity and expressivity.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4.2 Semantic integration in lexical aspectual compound verbs This section will explicate the lexical meanings of Type 2 (aspectual) compound verbs with a view to providing the semantic underpinnings necessary for tracing their historical development in subsequent sections. The vital difference between Type 1 and Type 2 compound verbs lies in the way one member of a compound modifies the other (Kageyama 2016a). In Type 1 (thematic) compounds, V2s are full-fledged verbs and determine the argument structure, case, and core meaning of the whole compound, with V1s modifying them in semantic relations such as means of action, manner of motion, or cause-result. This left-to-right modification is a natural consequence of the head-final structure of Japanese. Type 2 (aspectual) compound verbs, by contrast, do not follow the left-to-right modification pattern; instead, the first verb (V1) determines the argument relations and core meaning of the whole compound, while the second verb (V2) semantically modifies it from right to left—an aberrant manner of modification but one that is observed outside the realm of V-V compound verbs as well (cf. Kageyama 2010). The modification relations in Type 2 compounds can be roughly divided into two patterns. One is what Kageyama (1993) called the “complementation relation.” Consider ure-nokoru [be.sold-] ‘remain unsold’, whose meaning cannot be captured by the converb paraphrase ure-te nokoru, lit. ‘be sold and remain’. More plausible is the paraphrase ure-zu ni nokoru ‘remain without being sold, remain unsold’, where the first verb ure-zu [be.sold-] is in the negative. The negative meaning in this paraphrase is ascribed to V2’s meaning of imperfectivity designating incomplete attainment of V1’s event, as in ‘The being-sold event remains (to be finished)’, which in turn implies that a particular entity remains unsold. The other pattern of modification is what Kageyama (1993) and Matsumoto (1998) called the “adverbial relation.” A notable example is (Hana ga) sakihokoru [bloom-.] ‘Flowers blossom in full glory’, where V2 hokoru lit. ‘be proud’ emphasizes the exuberance of flowers in full bloom. For this example, a paraphrase like *(Hana ga) sai-te hokoru, lit. ‘(Flowers) bloom and are proud’ is inappropriate because flowers cannot be proud. A more cogent paraphrase is obtained by reversing the order of the two verbs, as in (Hana ga) hokorasige ni saku, lit. ‘(Flowers) blossom proudly’, where the meaning of V2 is conveyed by a morphologically related adverbial form meaning ‘proudly’ or ‘in full glory’. Employing simplified lexical conceptual structures (LCS) as a formal representation of a verb’s event structure, the two patterns of semantic modification can be schematically formulated as in (3) and (4), where the semantic predicates in LCS and V2’s semantic effects are shown in small capitals.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

74

 

(3)

a. External specification in complementation relation V1 ure- ‘be sold’: [ [x  ]] V2 nokoru ‘remain’: [e  ] (e stands for V1’s event.) ! Integration: ure-nokoru [[ [x  ]]  ] (i.e. ‘The being-sold event remains unfinished’)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Internal specification in adverbial relation V1 saki- ‘bloom’ : [ [x   ]] V2 hokoru ‘be proud’:    ! Integration: saki-hokoru [ [x      ]] (i.e. ‘(Flowers) bloom in full glory’) The semantic effect of the complementation interpretation will be generalized as “external specification,” whereby V2 adds a semantic specification to the outside of V1’s LCS. Thus, in (3a), the semantic structure of V1 ure- ‘be sold’ as a whole is embedded in the ‘e(vent)’ slot of V2 nokoru ‘remain’. The adverbial interpretation, on the other hand, has the semantic effect of “internal specification.” In (3b), for example, V2’s adverbial meaning ‘proudly, in full glory’ is added inside V1’s LCS as a semantic adjunct. In both cases, we will refer to the V2 verbs (called “aktionsart verbs” in Chapter 2) as L-asp verbs (lexical aspectual verbs) because they directly manipulate V1’s lexical meaning. In Kageyama and Kanzaki’s (2014) database registering over 2,700 lexical compound verbs, 833 verbs belong to Type 2 (aspectual), as against 1,651 verbs in Type 1 (thematic). A scrutiny of the L-asp verbs in the database suggests that the instances of external specification as in (4) are not so common as those of internal specification. (4)

a. Inception V1 saki- ‘bloom’: [ [x   ]] V2 -someru ‘begin’: [e ] ! Integration: saki-someru [[ [x   ]] ] (i.e. ‘x starts blooming’) b. Completion V1 naki- ‘cry’: [x ] V2 -yamu ‘(for rain/snow/cry) to stop’: [e   ] ! Integration: naki-yamu [[x ]   ] (i.e. ‘x stops crying’) c. Retardation V1: ii- ‘say’: [x ] V2 -siburu: ‘hesitate, be unwilling’ [x   e ] ! Integration: ii-siburu [x   [x ]] (i.e. ‘x hesitates to speak’)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

75

The L-asp verbs in (4) all denote purely phasal meanings of grammatical aspect. In (4a), -someru ‘(blooming) begin’ designates the beginning of V1’s event; in (4b), -yamu ‘(for rain/snow/cry) to stop’ adds an endpoint to the durative activity of V1; and in (4c), -siburu, lit. ‘be not willing to do’ expresses the subject’s restraint in initiating V1’s event. Compared with the conceptual simplicity of external specification, internal specification in the adverbial relation is far more complicated, assigning a multiplicity of elaborate and decorative shades of meaning to the internal semantics of V1. The rather long lists of representative examples in (5) through (8) are intended as illustrations of the semantic diversity of L-asp verbs. In each example, observe how V2’s semantic information is manifested as an adverbial meaning and assimilated into V1’s LCS to yield a coherent semantic structure denoting a single event. (5)

Specification of endpoint and completion a. -tumeru, lit. ‘stuff something in (so that no empty space is left)’, specifies the culmination of V1’s event to the extreme end. ni (‘boil’)-tumeru: [[x   y]  [ [y      ]]] (i.e. ‘x boils down y’)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. -saru, lit. ‘leave, go off (so that nothing is left behind)’, specifies the complete disappearance of a theme leaving no remnants behind. kesi (‘erase’)-saru: [[x ]  [ [y     ]]] (i.e. ‘x erases y completely’) c. -todokeru, lit. ‘deliver, send’, specifies that V1’s act is carried out to the end. mi (‘see’)-todokeru: [x   y   ] (i.e. ‘x sees y to the end’) d. -hateru, lit. ‘come to an end’, specifies the extreme endpoint of V1’s change of state . yature (‘become haggard’)-hateru: [ [x  -    ]] (i.e. ‘x becomes completely haggard’) (6)

Specification of suspension and incompletion a. -sasu (etymology unknown) specifies that the subject stops his action halfway. kaki (‘write’)-sasu: [[x ]   [ [y  ]]] (i.e. ‘x stops writing y in the middle’) b. -nayamu, lit. ‘worry’, specifies the lack of progress in V1’s event. nobi (‘grow’)-nayamu [x      ] (i.e. ‘x fails to grow as expected’) c. -kurasu, lit. ‘live, spend a day’, specifies prolonged continuation of V1’s event. mati (‘wait’)-kurasu: [x    ] (i.e. ‘x waits all day long’)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

76

 

(7)

Specification of intensity of action and state a. -midareru, lit. ‘be dispersed, be in disorder’, specifies the profuseness of V1’s state. saki (‘bloom’)-midareru: [ [x     ]] (i.e. ‘x (flowers) bloom all over in profusion’) b. -mawasu lit. ‘turn around, tweak’ specifies the persistent repetition of V1’s action. izikuri (‘fumble’)-mawasu: [x   y ] (i.e. ‘x fiddles with y persistently’) c. -tigiru, lit. ‘tear off ’, specifies the maximum intensity of V1’s action. home (‘praise’)-tigiru: [x   y   ] (i.e. ‘x praises y to the skies’) d. -tukeru, lit. ‘attach, affix’, specifies the fierce intensity and direction of V1’s action. donari (‘shout’)-tukeru: [x    y] (i.e. ‘x shouts fiercely at y’) e. -tateru, lit. ‘give off (steam/wind/smoke) vigorously’, specifies the vigor and excessiveness of V1’s action. sawagi (‘fuss’)-tateru: [x  ] (i.e. ‘x makes a terrible fuss’)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

f. -kaeru, lit. ‘go and come back, turn over’, specifies the completeness of V1’s state. sizumari (‘become quiet’)-kaeru: [ [x   ]] (i.e. ‘x (a place) becomes completely quiet’) (8)

Specification of spatial orientation a. -orosu, lit. ‘take down’, specifies downward direction of V1’s action. mi (‘look’)-orosu: [x    y] (i.e. ‘x looks down’) b. -kaesu, lit. ‘reverse, send back’, specifies repetition of V1’s action in a reverse direction. nirami (‘glare’)-kaesu: [x    y] (i.e. ‘x glares back at y, who has already glared at x’) c. -wataru, lit. ‘cross, go over’, specifies spatial extension of V1’s action/state. hare (‘sky become clear’)-wataru: [ [x    ] (i.e. ‘The whole sky becomes clear (cloudless)’)

Besides the verbs listed above, there are many more verbs that belong to the same category of aspectual and intensity specification. This section has postulated two modes of semantic modification in Type 2 compound verbs: external specification in the complementation interpretation and internal specification in the adverbial modification. This semantic distinction will play a key role in identifying L-asp verbs in the diachronic exploration in subsequent sections.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

77

4.3 Auxiliary verbs in Old and Early Middle Japanese This section examines the semantic and syntactic properties of the V-V complexes in Old Japanese and Early Middle Japanese,² with a view to clarifying which V2 verbs from these earliest periods persist today in which types of compound verbs in Present-day Japanese. The V2 verbs in question correspond to what are called “auxiliary verbs”³ in traditional Japanese grammar and “aktionsart verbs” in Chapter 3 (Aoki and Frellesvig).

4.3.1 V-V complexes as layered serial verbs Since Kindaichi (1953), it has been known that the V-V complexes in OJ and EMJ were not equipped with morphological integrity as compound words. Kindaichi’s (1953) arguments for the lack of morphological wordhood in OJ/EMJ V-V sequences were based on phonological and syntactic grounds. His phonological argument based on accentual patterns will not be discussed here because quite a few researchers, including Komatsu (1971), Seki (1977), and Aoki and Frellesvig (Chapter 3, this volume) have voiced criticism of it. More cogent is Kindaichi’s syntactic argument that emphatic particles were allowed to intervene between the members of V-V complexes, as shown in the Man’yōshū poem in (9a).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(9)

a. 慰もる心はなしにかくのみし恋ひや渡らむ月に日に異に nagusamoru kokoro pa nasi ni comforting heart  not.exist  kaku nomwisi kwopwi ya watara-mu this.much only yearn  continue-will tukwi ni pi ni kyeni month  day  even.more ‘I am too disturbed . . . I yearn to see my love again, but days and months pass by in vain’.4 (Man’yōshū 11.2596a)

² Following Frellesvig (2010), the history of Japanese is divided into the following stages: Old Japanese (OJ) 700–800 Nara period Early Middle Japanese (EMJ) 800–1200 Heian period Late Middle Japanese (LMJ) 1200–1600 Kamakura and Muromachi periods Modern Japanese (ModJ; NJ in Frellesvig 2010) 1600– Edo and subsequent periods ³ The term “auxiliary verb” is employed here following the tradition of domestic Japanese grammar to refer to aspectual, honorific, and other functional verbs that take a clausal complement. Despite the modifier “auxiliary,” they are verbs because of their capacity to inflect in the same way as regular lexical verbs. ⁴ English translation from H. H. Honda (1967). The MANYOSHU, a New and Complete Translation. Tokyo: Hokuseido Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

78

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. あのもの射もころしきりもとどめなんや ano mono i mo korosi kiri mo todome-nanya. that creature shoot  kill cut  finish.off-why.don’t.you ‘Why don’t you kill the creature by shooting or by cutting?’ (Heike Monogatari, 1219) In (9a), the exclamatory particle ya, which is limited to occurring only at a clausal level, splits the verbal sequence into kwopwi ‘yearn’ and watara-mu ‘will continue’, resulting in a violation of the universal principle of “morphological integrity” (the syntax cannot break up the internal structure of words; cf. Los et al. 2012: 86). This demonstrates that V-V sequences in OJ lack morphological integrity as a word. What is interesting is that particle interpolation in OJ and EMJ was observed not only with phasal and honorific auxiliary verbs but also in V-V combinations that apparently correspond to modern Type 1 (thematic) compound verbs. According to Aoki (2013), this phenomenon was continually documented in both verse and prose up until Late Middle Japanese of the Kamakura and Muromachi periods. (9b) is one such example, where the complex i korosi [shoot.arrow kill] is interpreted as a sequence of two actions, ‘shoot an arrow and kill’, rather than as the single integrated event of ‘kill by shooting an arrow’ denoted by the modern counterpart i-korosu. The fact that the V-V sequences in OJ and EMJ lacked morphological integrity, however, does not preclude the possibility that they had a functional unity as a complex predicate. While Aoki (2013) and Aoki and Frellesvig (Chapter 3, this volume) only assume that two verbs were loosely connected with each other, I argue that these sequences are best characterized as “complex predicates” in the form of verb serialization, because they share the same characteristics as serial verbs observed in other languages (Aikhenvald 2006). First, a V-V complex can carry on its second member at most one tense marker, whose scope extends to the first member; second, a negative morpheme added to V2 takes scope over the whole combination of V1 + V2; and third, V1 and V2 share one subject (and one object if transitive). Language-internal evidence for the functional unity of a complex predicate can be found in the negative polarity construction “e (‘possibly’) + X + negation,” where X can be either a single lexical verb or a sequence of two or more verbs that are not separated by a clause boundary. (10)

a. e nenzi sugusu mazi possibly tolerate continue be.able. ‘would not possibly be able to continue to tolerate’ (c. 1001–14, Genji Monogatari) b. e oki agari tamapa-de possibly get.up rise - ‘couldn’t possibly stand up (from the floor)’ (c. 900, Taketori Monogatari)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

79

The example in (10a), nenzi sugusu ‘continue to tolerate’, apparently expresses a phasal meaning, and that in (10b), oki agari ‘stand up’, a thematic meaning. In both examples, the verb sequences are sandwiched by the negative polarity adverb e ‘possibly’ and a negative morpheme, thus functioning as a predicate. Exactly what kind of syntactic structure serial verbs assume is debatable. My suggestion is that the OJ/EMJ V-V complexes of both phasal and thematic meanings are organized in the same kind of clause embedding as are modern Type 3 syntactic compound verbs. Parallel behavior between OJ/EMJ V-V complexes and modern Type 3 syntactic compounds is observable in the embedding of passives and causatives in the V1 position, as illustrated in (11). (11)

a. kokoro yori poka ni yo ni ari to mind than other  world  exist  sir-are pazimuru o know- begin  ‘ . . . the fact that it became known that she is still alive in this world against her will’. (c. 1001–14, Genji Monogatari)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. siwite owas-imas-ase some-te-keri. ‘forcibly come-- begin-- (Koremitsu) began to force (him) to visit the place’. (c. 1001–14, Genji Monogatari) The verb pazime ‘begin’ in (11a) was not very productive in the earliest days but has acquired full productivity in modern syntactic V-hazimeru compounds, whereas the verb some ‘start’ in (11b) was fairly productive in OJ and EMJ but now survives only in a small number of lexicalized Type 2 compounds such as mi-someru ‘fall in love with (a girl) at first sight’. Other aspectual verbs in EMJ that have proved to allow passives or causatives in V1 include sokonawi (PDJ sokonau) ‘fail’, awe ‘endure’, meaning ‘be unable to’ when used in the negative (obsolete in PDJ except in tori-aezu ‘for the moment’), and ariki (PDJ aruku) ‘walk (to various places)’. It follows that those V2s that allow passives or causatives in V1 take a vP complement in syntactic structure. Our suggested structure gains support from honorification on V1s. In (12), the honorific auxiliary verb tamawi ‘an honorable subject does an action’ and the humble auxiliary verb tamawe ‘an inferior subject does an action’ occur right after the V1 verbs. (12)

a. sinobite ne tamawi some-te-keri secretly sleep  begin-- ‘(they) secretly began to sleep together’ (951, Yamato Monogatari) b. kono pousi mi tamawe tukisi pazime yori this priest see  come.to beginning from ‘from the first time she caught a glimpse of this priest’ (c. 970, Utsuho Monogatari)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

80

 

The occurrence of syntactic honorific/humble verbs after V1 verbs, in parallel with the appearance of honorific verbs in the V1 position of modern Type 3 syntactic compounds, counts as robust evidence that the OJ/EMJ complex predicates are formed of a biclausal syntactic structure of the form ‘[clause V1 /] V2’. It is worth pointing out that the occurrence of an honorific verb on V1 is not a privilege of aspectual auxiliary verbs alone but is also found with what would strike modern speakers as thematic compound verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(13)

a. mi look

atukawi take.care.of

tatematuru 

(c. 1069–77, Sagoromo Monogatari)

b. mi look

tatematuri 

atukau take.care.of

(c. 1001–14, Genji Monogatari)

In (13a), the humble verb tatematuri, lit. ‘offer up’, shows up after the whole complex mi atukawi [look take.care.of], which is arguably analyzed as a thematic complex meaning ‘look after, nurse’, whereas in (13b), the same humble verb intrudes into the same complex. The intrusion of honorific and humble verbs in both aspectual and thematic complexes is the same kind of phenomenon as the particle insertion pointed out by Kindaichi (1953). Since the intrusion of particles and auxiliary verbs is indicative of lack of morphological integrity, it can safely be concluded that V-V complexes of both thematic and phasal meanings in OJ and EMJ are identified as complex predicates in the form of layered serial verbs. This conclusion, coupled with the scope relations in negative polarity constructions in (10), leads to the hierarchical syntactic structure shown in (14), where V⁰ stands for the minimum verb category that is syntactically visible and the second members of both phasal and thematic complexes project “vP,” so that they may occur in either order. (14)

vP

V03

vP

V02

VP

…

V01

tatematuri ‘ofer up (humble)’/ atukawi ‘take care of ’

atukawi ‘take care of ’/ tatematuri ‘ofer up (humble)’

mi ‘look’ Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

81

This syntactic structure implies that the meaning of a whole complex predicate obtains not by the amalgamation of the lexical conceptual structures of member verbs into one coherent semantic representation along the lines suggested in Section 4.2 for modern lexical compounds, but by the unification of the individual verbs’ meanings in the manner of external specification. Building on this mode of semantic interpretation, we now turn to individual V2 verbs in OJ and EMJ to see which auxiliary verbs have changed into which type of compound verbs in the modern language.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4.3.2 Auxiliary verbs in Old Japanese In their corpus-based study of V-V complexes in Old Japanese, Frellesvig et al. (2010) give forty-four verbs as a provisional inventory of OJ auxiliary verbs that are semantically light or productive or freely combining V2s.⁵ Of these, some verbs, such as -tamapi ‘give (respect)’, -tamape ‘receive (humble)’, -maturi ‘serve (humble)’, -myesi ‘see (respect)’, -yame ‘stop (tr.)’, -kate ‘prevail, endure, be able’,⁶ kose/koso ‘(please) do for me’, -ape-/-au ‘be unable’ (cf. the fossilized adverbial tori-aezu ‘for the time being’ in PDJ), -tari ‘suffice’, and -yame ‘make stop’ (cf. PDJ tori-yame ‘stop’, where tori- is considered a prefix), fell out of use as productive V2s in PDJ, and -e ‘be able’ has been grammaticalized to a syntactic suffix of potentiality. Excluding them, I classified the rest of the OJ auxiliary verbs according to the status of their corresponding V2 usage in current Japanese. The results are summarized in Table 4.1. Table 4.1 reveals that the OJ auxiliary verbs can be divided into five major groups (A–E) based on the complex developments they subsequently underwent. Notice first that not many OJ auxiliaries converge uniquely on one of the four modern types (Groups A, B, C, D). The verbs that straightforwardly correspond to modern “lexical” types are seven verbs in Group A (Type 1) and four verbs in Group B (Type 2). While the former retain some degree of productivity, the latter are fossilized as such. On the other hand, only a very few auxiliary verbs (three verbs each in Groups C and D) grew into productive Type 3 or Type 4 complexes in PDJ. In sharp contrast to Groups A–D, which show a one-to-one correspondence with a particular modern type, Group E involves divergent expansions, where polysemy caused as many as 40% (eleven verbs) of the OJ auxiliary verbs to split into two or three different types in PDJ. Combining the verbs in Groups A, B, and ⁵ For the reasons spelled out in Chapter 3 (Aoki and Frellesvig), verbs are cited in their infinitive form when the attestation is from before 1900, unless otherwise noted. ⁶ According to Nihon Kokugo Daijiten (A Grand Dictionary of Japanese), the verb kate was attested only as V2 in the writings of the eighth and ninth centuries, but was later extended to an independent verb meaning ‘endure’ in the tenth century, thus presenting a case of degrammaticalization (cf. Norde 2009).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

82

 

Table 4.1 The vicissitudes of auxiliary verbs in Old Japanese Old Japanese ide ‘exit’ ire ‘put in’ kwoye ‘go over’ sugwi ‘go beyond’ A

Present-day Japanese V-deru ‘go/come out’ V-ireru ‘put in’ V-koeru ‘go over’ V-sugiru ‘pass’

topori ‘go through’ V-tooru ‘go through’ toposi ‘put through’ V-toosu ‘put through’ tugi ‘continue, convey’

V-tugu ‘convey’

pate ‘end’

V-hateru ‘do completely/to the end’ V-someru ‘begin to’

dasi ‘remove, take out’ kake ‘hang’

VP-dasu ‘begin to’

Current status Type 1 (nagare-deru ‘flow out’) Type 1 (sosogi-ireru ‘pour in’) Type 1 (tobi-koeru ‘jump over’) Type 1 (toori-sugiru ‘pass and keep going’) Type 1 (simi-tooru ‘seep into’) Type 1 (sasi-toosu ‘pierce to the other side’) Type 1 (katari-tugu ‘tell and pass from one generation to the next’)

Type 2 (akire-hateru ‘be completely exasperated’) some ‘begin’ Type 2 (saki-someru ‘(flowers) start blooming’) B sugusi ‘put beyond’ V-sugosu ‘do beyond a Type 2 (ne-sugosu ‘oversleep’) certain limit’ yami ‘stop’ V-yamu ‘stop’ Type 2 (naki-yamu ‘stop crying’)

C

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

kane ‘fail’ itadaki ‘receive’ ki ‘come’ D yuki ‘go’ iri ‘enter’

Type 3 (huri-dasu ‘begin to rain’)

VP-kakeru ‘be about to’ Type 3 (taore-kakeru ‘be about to collapse’) VP-kaneru ‘be unable Type 3 (hikiuke-kaneru ‘be unable to to’ accept’) V-te itadaku ‘receive a Type 4 benefactive (kasite-itadaku favor (from a superior)’ ‘receive the favor of lending me’) V-te kuru Type 4 aspectual (hue-te kuru ‘be gradually increasing’), Type 4 motional (mot-te kuru ‘bring’) V-te yuku/iku Type 4 aspectual (hue-te yuku/iku ‘go on increasing’), Type 4 motional (mot-te yuku/iku ‘take’)

V-iru ‘go in’ / ‘do completely’ kapyeri ‘come back’ V-kaeru ‘return’ (intr.) / ‘do in a different way’ kapyesi ‘put back’ E tati ‘stand’ tate ‘make stand’

Type 1 (seme-iru ‘attack and enter’), Type 2 (osore-iru ‘be much obliged’) Type 1 (ture-kaeru ‘take home’), Type 2 (kaki-kaeru ‘rewrite into a different format’) V-kaesu ‘return’ (tr.) / Type 1 (uti-kaesu ‘hit a ball and return ‘do in the opposite way’ it’), Type 2 (toi-kaesu ‘ask back’) V-tatu ‘stand’ / ‘come Type 1 (sobie-tatu ‘stand as if towering to do’ high’), Type 2 (ni-tatu ‘(water/broth) come to the boil’) V-tateru ‘make stand’ / Type 1 (tuki-tateru ‘stick upright’), ‘do violently’ Type 2 (abaki-tateru’debunk thoroughly’)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

   tuki ‘stick’

V-tuku ‘stick’ / ‘do thoroughly’

watari ‘go across’

V-wataru ‘go across’ / ‘do completely/all over’ V-watasu ‘put across’ / ‘do all over an area’

watasi ‘put across’ tuke ‘affix’

V-tukeru ‘affix’ / ‘do thoroughly/harshly’ / ‘be accustomed to’

mi ‘see’

V-miru ‘see’, V-te miru ‘try and see’ V-oku ‘put’, V-te oku ‘do for a future purpose’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

oki ‘put’

83

Type 1 (simi-tuku ‘seep in and stick’), Type 2 (kogoe-tuku ‘become thoroughly chilled’) Type 1 (oyogi-wataru ‘swim across’), Type 2 (hare-wataru ‘(the sky) clear up’) Type 1 (uri-watasu ‘transfer by selling’), Type 2 (mi-watasu ‘look out over (an area)’) Type 1 (ue-tukeru ‘plant in a specified place’), Type 2 (itame-tukeru ‘hurt harshly’), Type 3 (tukai-tukeru ‘be accustomed to using’) Type 1 (nozoki-miru ‘peek inside’), Type 4 (nozoi-te miru ‘try peeking’) Type 1 (kaki-oku ‘write a memo and leave it’), Type 2 (sute-oku ‘leave as it is’), Type 4 (yon-de oku ‘read for a future purpose’)

E, we find that the great majority (nearly 80%) of the OJ complex predicates are now lexicalized as morphological compounds of Type 1 (thematic) and/or Type 2 (aspectual). Between Type 1 and Type 2 in Group E, the aktionsart meanings of modern Type 2 verbs appear to have emerged as a result of metaphorical extension from the physical meanings of Type 1 verbs. This is verified by comparing the dates of first attestations of concrete and abstract meanings. To give only a few examples, 1179 seme-iri ‘attack and enter’ vs. 1254 osore-iri ‘be much obliged’; 715 oyogi-watari ‘swim and cross’ vs. 13c. hare-watari ‘(sky) become clear all over’; 1001–14 simi-tuki ‘seep in and stick’ vs. c. 1711 kogoe-tuki ‘become completely frozen’; 1563 ue-tuke ‘plant seedlings in a place’ vs. 1608 teri-tuke ‘(sunlight) shine fiercely’ (cf. Nihon Kokugo Daijiten). The historical survey given in this subsection should not be taken as implying that the lexical compound verbs in PDJ are all descendants of OJ serial verbs. Although there are indeed relics of OJ complexes registered as fixed combinations in the lexicon of modern native speakers, lexicalization alone cannot account for the vast diversity of the modern compound verbs or, more importantly, for the change from syntactic serial verbs to morphological compounds. Data from EMJ will now be brought to bear on the issue.

4.3.3 Auxiliary verbs in Early Middle Japanese In traditional studies of Japanese grammar, there is a fairly large body of literature discussing the V-V complexes of the Heian period (800–1200) under the rubric of “compound verbs.” Higashitsuji (1975), for example, investigated V2 verbs that

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

84

 

are attested with high frequencies in Genji Monogatari (The Tale of Genji, c. 1001), a representative literary work written in EMJ. From the list of verbs discussed by Higashitsuji (1975), I culled the auxiliary verbs with aspectual and other related meanings and classified them into two groups depending on whether or not they are retained in PDJ. First, the V2 verbs in (15) were used extensively in EMJ but became extinct in current Japanese. Each entry is accompanied by a simplified event structure to show that the V2 verb has the semantic effect of external rather than internal specification in parallel with the OJ auxiliary verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(15)

Aspectual V2s in EMJ that are extinct in PDJ a. -atume ‘collect’ [accumulation of action], e.g. kaki-atume ‘write a lot’ [[e1 (writing event) + e2 (writing event) + . . . ] ] b. -sosi ‘go beyond a limit’ [excessive action (no independent use)], e.g. iwisosi ‘speak too much’ [e (speaking event)   ] c. -masi ‘increase’ [incremental increase of action], e.g. hagemi-masi ‘work increasingly harder’ [e (working event)  ] d. -sirowi ‘do each other’ [reciprocal (origin unknown)], e.g. ihi-sirowi ‘say to each other’ [e1 (speaking event by agent1)      e2 (speaking event by agent2)] e. -kutugaweri ‘turn upside down’ [intensive action], e.g. mede-kutugaweri ‘praise lavishly’ [e (praising event)   ] f. -suguri ‘excel, surpass’ [excellence of action], e.g. piki-suguri ‘excel in playing the koto (musical instrument)’ [e (playing event)   ] g. -awe ‘endure’ [inability when followed by negative], e.g. kazowe-awezu ‘be too many to count’ [e (counting event)   ] h. -miti ‘become full’ [emphatic state], e.g. sumi-miti ‘many people live together (in a place)’ [e (living event)    ] i. -todomari ‘stay’ [remain without undergoing V1], e.g. oti-todomari ‘survive without going to ruin’ [e (ruining event)  ] j. -mayowi ‘stray’ [emphatic action], e.g. puki-mayowi ‘(wind) blow heavily in various directions’ [e (blowing event)     ]

The extinct verbs in (15) stand in contrast to the verbs in (16), which have been handed down to PDJ. (16)

Aspectual V2s in EMJ that have persisted to this day (here shown in their present-day conjugations) a. -kaneru [OJ kane] ‘be unable’ [failure/inability, no independent usage], e.g. sadame-kanu ‘be unable to decide’ [x     e (deciding event)]; PDJ Type 3 kessin-si-kaneru ‘be reluctant/unable to make a decision’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

  

85

b. -ayamaru ‘make an error’ [x  e    ]; PDJ Type 3 mi-ayamaru ‘make a mistake in’ c. -toosu ‘put through’ [x  e    ]; PDJ Type 3 kakusi-toosu ‘get away with not telling the truth’ d. -tuzuku ‘continue’ [e ]; PDJ Type 2 huri-tuzuku ‘continue raining’ e. -someru [OJ some] ‘begin’ [no independent use attested], e.g. EMJ arisome ‘begin to exist’ [e (existing event) ]; PDJ Type 2 saki-someru ‘begin to bloom’ f. -sasu ‘do and stop halfway’ [etymology unknown], e.g. EMJ sousi-sasi ‘begin telling (humble) and stop halfway’ [x    e]; PDJ Type 2 kaki-sasu ‘stop writing in the middle’ g. -magau ‘make an error’, e.g. EMJ tobi-magawi [x       e]; PDJ Type 2 mi-magau ‘mistake’ h. -tagau ‘cross’ [e1 and e2   ], e.g. yuki-tagawi ‘cross each other on the way (and do not meet)’; PDJ Type 2 yuki-tagau/yuki-tigau ‘cross each other on the way’ i. -oosu ‘do to the end’ (origin obscure) [x  e   ]; PDJ Type 2 nige-oosu ‘succeed in escaping’ j. -tirasu ‘scatter in a mess’ [x   e    y]; PDJ Type 2 wameki-tirasu ‘yell incomprehensibly without regard to other people’ k. -nokoru ‘remain’ [e  ]; PDJ Type 2 ure-nokoru ‘remain unsold’ l. -midareru ‘become disorderly’ [e     ]; PDJ Type 2 saki-midareru ‘(flowers) bloom in great profusion’ m.-naoru ‘do again’ [x  e ]; PDJ Type 2 tati-naoru ‘recover from a bad situation’ n. -kau ‘cross’ [e1 and e2   ]; PDJ Type 2 yuki-kau ‘(for a multitude of people/vehicles) come and go in various directions’ Of the fourteen verbs in (16), only the first three remain productive as V2s of Type 3 syntactic compounds, while all the rest are lexicalized to limited numbers of Type 2 compounds with fixed verbs in V1. A glance at Higashitsuji et al.’s (2003) comprehensive list of thousands of “compound verbs” in the Heian period suggests that the lexical selection between V1 and V2 was much freer in EMJ than in PDJ. This observation, coupled with the possibility of particle insertion, indicates that the V-V combinations in this period were serial verbs rather than morphological compounds. Their syntactic structure corresponds to their semantics. In (16) as well as in (15), all the V2s add an array of meanings related to grammatical aspect to the whole of V1’s events by way of external specification. The meaning of -some ‘begin’ in ari-some ‘begin to exist’,

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

86

 

for example, is represented in simplified form as [e begins], where e stands for the event of V1, and -some was indeed capable of taking any kind of verb—transitive, unergative, unaccusative—as its complement, as attested by the 132 verbs registered under the entry of -some in Higashitsuji et al.’s (2003) list. If this line of reasoning is plausible, it can be concluded that the V2s in EMJ were not so tightly glued together with the verbs in V1 as to directly manipulate the internal composition of V1’s event structure. In other words, they had not yet attained the status of Type 2 compound verbs. As suggested by Aoki and Frellesvig (Chapter 3, this volume), the advent of morphological compound structure cannot be seen until Late Middle Japanese (fourteenth to sixteenth centuries). In the next section, the origins of representative V2s in PDJ will be explored by dating them back to their earliest appearances.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4.4 The sources of L-asp verbs This section aims to clarify how the L-asp verbs in modern Type 2 compound verbs acquired their current meanings by tracing them back to their first attestations. For this purpose, I tried collecting objective data from the online versions of Nihon Kokugo Daijiten (The Grand Dictionary of Japanese; hereafter abbreviated to NKD), the largest dictionary of Japanese and comparable to the Oxford English Dictionary in that it presents major attested examples in chronological order. For confirmation, the examples and descriptions in NKD were double-checked, whenever possible, with the online versions of Nihon Kotenbungaku Zenshū (The Complete Collection of Japanese Classical Literary Works). The historical data are classified according to the degree of similarity in the dates of the earliest attestations of two types of usage with the same aspectual meaning—one as an independent lexical verb and the other as the V2 member of a compound verb. Theoretically, four cases of temporal relationships are conceivable: (I) the independent usage antedated the V2 usage; (II) the V2 usage antedated the independent usage; (III) the independent usage and the V2 usage emerged almost concurrently; and (IV) only the V2 usage was recorded, with no firm documentation of the independent usage. Case I encompasses the largest number of examples among the four cases, apparently in line with the tendency toward unidirectional grammaticalization. For example, pate (PDJ hateru) ‘come to an end, finish’ as an intransitive lexical verb was first attested in the late eighth century (Man’yōshū 10.1843), and its V2 usage in the late tenth century (Utsuho Monogatari). It is immediately apparent that these two usages have a non-negligible time gap of about 200 years between them. The temporal discrepancy between the two uses can be even greater with other verbs, as shown in Table 4.2.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

87

Table 4.2 Examples of the independent usage antedating the V2 usage original verb

independent usage with an aspectual meaning

1. siki ‘occur repeatedly’ 2. pate ‘come to an end’

720 ‘occur repeatedly’ late eighth c. puri-siki ‘rain frequently’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

3. yami ‘stop’

late eighth c. ‘for an event to come to an end’ late eighth c. ‘(an event) stops’ late ninth c. ‘accustom oneself ’ 974 ‘make an error’

4. narapasi ‘make customary’ 5. tigawi/tigawe/ tagawe ‘differ, do not match’ 6. midare ‘get late eighth c. ‘get disorderly’ disordered’ 7. tataki ‘strike’ 1774 ‘beat down (the price)’ 8. sari ‘leave, depart’ 712 ‘leave, be gone’ 9. nadumi ‘do not 712 ‘do not advance proceed as expected’ because of obstruction’ 10. pokori ‘be proud 720 ‘pride oneself ’ of ’ 11. wosimi ‘begrudge’ eighth c. ‘grudge, spare’ 12. kapyesi ‘send back’ late eighth c. ‘send back’ 13. siburi ‘hesitate’ 974 ‘hesitate, be unwilling’ 14. apase ‘put late ninth c. ‘do together’ concurrently’ 15. toge ‘achieve’ early tenth c. ‘achieve a goal’ 16. wake ‘divide’ 1001 ‘divide’ 17. mapasi ‘move around’ (tr.)

1001 ‘move around’ (tr.)

18. agumi ‘get distressed’ 19. susume ‘set forward’

late fourteenth c. ‘get irritated’ late fourteenth c. ‘make move ahead’

V2 usage with the same aspectual meaning

970–90 siri-pate ‘realize completely’ 951 puri-yami ‘rain stops’, naki-yami ‘stop weeping’ 1198 ii-narawasi ‘say customarily’ 1085 kiki-tagawe ‘mishear’ 974 saki-midare ‘bloom in confused profusion’ 1917 uri-tataku ‘undersell’ 1878 nugui-sari ‘wipe out completely’ 1639 yuki-nazumi ‘have difficulty going on’ mid-fourteenth c. saki-fokori ‘bloom beautifully, as if proud’ 1780 ake-osimi ‘day does not break (as if dark night were lingering)’ 1603 yomi-kaesi ‘read again’ late nineteenth c. ii-siburi ‘hesitate to say explicitly’ 1477 umare-awase ‘happen to be born in a certain era’ 1534 ‘do until the end’ 1674 mori-wake ‘dish out food on several plates’ late fourteenth c. oi-mawasi ‘chase around’, 1683 kone-mawasi ‘knead repeatedly’, 1711 sikari-mawasu ‘scold roundly’ early seventeenth c. si-agumi ‘be unable to do as expected’ 1939 kaki-susumeru ‘move ahead with writing’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

88

 

The nineteen verbs in Table 4.2 can be roughly divided into two groups. The first group, comprising the examples in 1 through 7, is distinguished from the second group, consisting of the examples from 8 to 19, in terms of the enormity of the temporal discrepancy between the independent use and the V2 use. The first five examples (1. siki ‘occur repeatedly (< extend)’, 2. pate ‘come to an end’, 3. yami ‘stop’, 4. narapasi ‘make customary’, 5. tigawi/tigawe/tagawe ‘differ, do not match’) established their V2 usage within a relatively short period of time after the appearance of the corresponding independent verbs. Their V2 usage can, therefore, be deemed a natural extension of the lexical meanings. This extension is unsurprising because the independent verbs denote aspectual or aspectinducing meanings from the start. The second group in Table 4.2, on the other hand, displays wider discrepancies between the first attestations of the independent use and of the V2 use, presumably because their original lexical meanings are not aspectual. This is tantamount to saying that the examples from 8 through 19 acquired their aspectual and other functional meanings only when they established their V2 usage in V-V compounds, and not through a simple metaphorical extension from their corresponding lexical verbs. A notable fact about these verbs is that their first appearances as V2s can be traced back only to the fourteenth century (Late Middle Japanese), and not earlier. This fact can be taken as strongly suggesting that the Type 2 lexical compound verbs did not emerge until LMJ or later. So far, only the first group in Table 4.2 is supportive of the unidirectionality hypothesis of grammaticalization. The enormous temporal gaps in the second group suggest that the V2 usage arose independently of the corresponding lexical verbs. The hypothesis of unidirectional grammaticalization is confronted with an even more serious problem with Case II, where the V2 use antedates the corresponding independent use, as shown in Table 4.3. Table 4.3 Examples of the V2 usage antedating the independent usage original meaning

independent usage with a bleached meaning

iri ‘enter’

early twelfth c. ‘enter an abstract 970–90 ne-iri ‘fall asleep’ state’ early twelfth c. ‘finish’ 976–87 kari-age ‘finish harvesting’ late tenth c. ‘stuff to the capacity’ early tenth c. omowi-tume ‘think persistently’ (intr.) early thirteenth c. ‘come near’ 1001 kure-kakari ‘(the sun) be about to set’ 1308 ‘go too far, overdo’ 1001 ne-sugusi ‘oversleep’

age ‘raise’ tume ‘stuff ’ kakari ‘hang’ sugosi/sugusi ‘pass time’ agari ‘rise’

1604 ‘be finished’

V2 usage with much the same bleached meaning

late fourteenth c. pi-agari ‘(for the tide) to ebb away’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

89

Table 4.4 Examples of the V2 usage and the independent usage appearing concurrently original meaning

independent usage with a bleached meaning

V2 usage with much the same bleached meaning

kurasi ‘spend time’

late eighth c. ‘spend/pass a day’ late eighth c. yuki-kurasi ‘go/ walk all day long’ watari ‘cross, go over’ 1001 ‘cover a large area’ 1001 are-watari ‘be ruined all over’

Table 4.5 The V2 usage evolving only in the compound structure independent usage

bleached meaning in compound structure

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

some ‘begin’ (attested only as V2) oposi/opose ‘finish’ (attested only as V2) komi ‘go in’ (attested only as V2) 712 tati ‘stand’ 712 sasi ‘stick, point’

late eighth c. mi-some ‘meet for the first time’ late tenth c. nuwi-oose ‘keep on sewing to the end’ 1785 oi-komi ‘grow old and decrepit’ early tenth c. omowi-tati ‘begin to think’ early tenth c. iwi-sasi ‘begin to speak but stop halfway’ 720 tuke ‘attach’ 1608 teri-tuke ‘shine fiercely’ 720 kiri ‘cut, separate’ late tenth c. sadame-kiri ‘decide conclusively’, early thirteenth c. tanomi-kiri ‘rely entirely’ 720 nayami ‘worry’ 893 mote-nayami ‘have difficulty handling, cannot handle’ late eighth c. tate ‘give off steam’ late eighth c. ywobi-tate ‘shout loudly’ late eighth c. parapi ‘remove’ 1660 de-harai ‘(people involved) be all out’ late eighth c. kapyeri ‘turn back, rotate’ late eighth c. sini-kapyeri ‘die over and over again’, 974 kie-kaweri ‘disappear completely’ late eighth c. wazurawi ‘worry’ 974 ide-wazurawi ‘hesitate to leave’ 942 tigiri ‘tear off ’ c. 1221 (zyuzu o) momi-tigiri ‘rub (a Buddhist rosary) over and over again’ 1275 kokuri ‘rub hard, tear’ 1439 tuki-kokuri ‘thrust violently’, 1822 damari-kokuri ‘remain completely silent’ sixteenth c. koke ‘roll over’ 1767 neburi-koke ‘sleep hard’

In these examples, it is highly likely that an independent lexical verb grew out of the V2 usage. If this view is correct, the verbs in Table 4.3 present a case of degrammaticalization where a bound element (auxiliary verb) changes to an independent verb. In Case III, where the independent use and the V2 use emerged almost simultaneously, very few examples are found, as shown in Table 4.4. These examples are noncommittal as to the directionality of grammaticalization. Finally, the L-asp verbs in Case IV, exemplified in Table 4.5, apparently lack independent lexical sources to which their origin can be attributed. In other

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

90

 

words, they are not attested as independent verbs carrying the same bleached meanings as they convey in compound structure. In view of the fact that some of them are labeled “suffix” or “auxiliary verb” in the entries of NKD, it is likely that the L-asp verbs in this group acquired the bleached meanings only in the compound structure, independently of their corresponding lexical verbs, or even started out as dedicated V2s in V-V complexes. In summary, the historical documentation in Tables 4.2–4.5 strongly suggests that it was not the case that each and every L-asp verb attained an aktionsart meaning by having recourse to its corresponding lexical verb. Rather, many L-asp verbs developed their delexicalized meanings only in combination with particular V1 verbs.

The findings in the preceding sections point to the existence of two different paths along which Type 2 compounds developed. One is a direct path from OJ and EMJ serial verbs, and the other is a detour path through Type 1 (thematic) compounds. Figure 4.1 sketches the historical development of Japanese V-V complexes, where the two paths leading to Type 2 compounds are shown by thick arrows (2) and (3). The path indicated by arrow (2) shows the downward shift (“downgrading” in Newmeyer’s (2001) terminology) of OJ/EMJ serial verbs to morphological compounds. This change, involving the OJ/EMJ auxiliary verbs that denoted grammatical aspect, is achieved by reanalysis, as in (17), and the resulting compound verbs are as a whole registered in the lexicon (lexicalization). OJ/EMJ serial verbs

weaker

MORPHOLOGICAL BOND

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4.5 The rise of Type 2 compound verbs as constructionalization

(2) LMJ ‘downgrading’: Reanalysis to compounds and lexicalization (1)MMJ Reanalysis to compounds

stronger

Type 2 L-asp compounds

Type 3 syntactic compounds auxiliary verbs

Type 4 V-te V complex verbs, Type 4 motional V-te V

(4) V2s of unknown etymology

(3) LMJ to EModJ ‘Upgrading’: Loss of argument structure and desemanticization to aktionsart meanings Type 1 thematic compounds

Figure 4.1 Paths to Type 2 aspectual compounds

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(17)

91

[vP [VP . . . [V0 V1]] [V0 V2]] ! [V0 [V1] [V2]]

The reanalysis, presumably enhanced by the increased agglutinative character of the infinitive (ren’yōkei) inflection in Middle Japanese (cf. Frellesvig 2010), offers an answer to the long-standing question of why rendaku or sequential voicing, which applies inside compound nouns fairly regularly, does not apply to the second members of V-V compound verbs. Reanalysis cum lexicalization also explains why the descendants that are available today of OJ/EME auxiliary verbs are confined to a relatively small number of lexically fixed combinations. An exemplar case is -someru [OJ -some] ‘begin’, which served as a syntactic phasal verb in OJ and EMJ in combinations with a variety of V1s, as in naki-some ‘(little birds) begin to chirp’, saki-some ‘(flowers) begin to bloom’, kobore-some ‘(tears) begin to drop’, ohasi-some ‘begin to show up’, midare-some ‘begin to be chaotic’, and kiki-some ‘hear for the first time’, but currently remains only in a few lexicalized compounds of Type 2 (saki-someru ‘(flowers) begin to bloom’, misomeru ‘fall in love at first sight’). Also included in this category are frequentative -sikiru ‘occur repeatedly’ (huri-sikiru ‘(rain) fall incessantly’), terminative -yamu ‘stop’ (naki-yamu ‘stop crying’), multiple occurrence -kau lit. ‘cross’ (iki-kau ‘(people and vehicles) come and go’), incompletive -nokoru ‘remain’ (yake-nokoru ‘remain unburned’), and wrong result -magau ‘err’ (mi-magau ‘mistake’). The semantic effects of all these erstwhile auxiliary verbs are subsumed under the external specification in complementation relation (3a). It can be assumed that the reanalysis of these OJ/EMJ aspectual verbs, shown in the downward arrow (2) in Figure 4.1, contributed to forming the prototype of L-asp compound structure, which was later fortified by a growing number of new V2 verbs that emerged from Type 1 (thematic) compounds through the upward route of arrow (3). From the viewpoint of grammatical change, the upward path is more intricate and more intriguing than the downward path. The “upgrading” has two facets. One is the semantic change in which V2’s thematic meaning shifted to aspectual or other functional meanings with the concomitant loss of its argument structure. The other is the change in morphological composition. To understand this second point, it is necessary to explain the nature of the “morphological bond” shown along the vertical axis of Figure 4.1. The OJ/EMJ serial verbs are shown at the top because they have no morphological unity, whereas Type 1 compounds are shown at the bottom because their two members are assumed to be more tightly glued together than those of Type 2 compounds. Evidence for the disparity in morphological strength derives from three sets of observations. First, according to Aoki (2013), particle insertion inside V-V complexes, which was characteristically observed in Old and Early Middle Japanese, first faded with thematic-type complexes in the Kamakura period (thirteenth c.) but persisted with aspectual-type complexes until the Muromachi period (fourteenth to fifteenth c.). Relics of such aspectual V-V complexes with particles inside remain

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

92

 

today in fossilized forms like tori-mo-ae-zu [take--be.able-neg] ‘for the moment’ and omoi-mo-yora-zu/nai [think--come.near-neg] ‘cannot think, unexpectedly’. This historical fact suggests that the morphological bond of aspectual complexes (corresponding to Type 2 compounds) was somehow weaker than that of thematic complexes (corresponding to Type 1). Second, in current Japanese the two types of compounds have an ordering restriction in which a Type 1 compound may be embedded inside a Type 2 compound, but the reverse order is prohibited. The contrast between (18a) and (18b) demonstrates this point. (18)

a. Mati wa [[ne-sizumari (Type 1)]-kaet (Type 2)]-ta town  [[sleep-get.silent]-.]- ‘The whole town became completely silent after people fell asleep’. b. *[[ne-kaeri (Type 2)]-sizumat (Type 1)]-ta [[sleep-.]-get.silent]-

From this it follows that Type 2 (aspectual) compounds appears at a higher level than Type 1 compounds in hierarchical morphological structure, as illustrated in (19). (19)

Type 2 compound

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Type 1 compound

thematic V1

L-asp verb

thematic V2

Third, consider verb reduplication of the form “V-ni [] V,” as in utai ni utau [sing  sing] ‘sing on and on’. In Kageyama (1993), this phenomenon was looked upon as a hallmark of syntactic compound formation on a par with passive, causative, honorifics, and other syntactically motivated elements, and was employed as a diagnostic to distinguish syntactic from lexical compound verbs. More recently, however, Kageyama (2013) found that this verb reduplication may also take place in the V1 position of Type 2 lexical compounds. (20) shows some actually occurring examples taken from Japanese websites (all retrieved in 2013). (20)

a. gitarisuto tosite nobori ni nobori-tumete guitarist as climb  climb-stuff ‘He climbed and climbed to the top as a guitarist’. http://cjcj.ti-da.net/e3529604.html [original url; now deleted]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

93

b. Tookyoo-orinpikku ga kaisai-sare, iti-oku Tokyo-Olympics  hold- one-hundred.billion zen-kokumin ga waki ni waki-tat-ta. whole-nation  boil  boil-- ‘When the (1964) Tokyo Olympics was held, the blood of all Japanese people boiled violently in excitement’. www.kikou.co.jp/095.htm

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. kinoo wa sinderu ka gotoku nemuri ni yesterday  dead  as.if sleep  nemuri-koke-te i-masi-ta. sleep-.- be-- ‘Yesterday I continued sleeping soundly as if I were dead’. http://ameblo.jp/kanari-kawashiro/day-20090729.html [original url; now deleted] Other examples include sae ni sae-wataru [be.sharp  be.sharp-.] ‘(one’s mind) to be especially sharp’, saki ni saki-midareru [bloom  bloom-. ] ‘bloom all over in confused profusion’, and neri ni neri-ageru [elaborate  elaborate-] ‘complete with an extensive elaboration’. The V2s in these examples are all L-asp verbs of Type 2 compounds. In fact, Type 1 (thematic) compounds never permit verb reduplication in their V1 position. The discrepancy in the degree of morphological integration between Type 1 and Type 2 is reinforced by the ambiguity of -komu, a bound verb that adds a goal argument to V1’s argument structure if it belongs to Type 1, as in Kawa ga mizuumi ni nagare-komu [flow-into] ‘The river flows into the lake’, while it adds the meaning of ‘being deeply engaged in V1’s action/state’ if it belongs to Type 2, as in kangae-komu [think-into] ‘become immersed in thinking’. The two meanings are reflected in the ambiguity of the compound hasiri-komu [run-into], which denotes ‘run into (a place)’ under the Type 1 interpretation and ‘be fully engaged in running practice’ under the Type 2 interpretation. As expected, verb reduplication is outright ungrammatical on the Type 1 reading (21a) but is fully acceptable on the Type 2 reading (21b). (21)

a. *Sensyu wa gooru ni hasiri ni hasiri-kon-da. [Type 1] athlete  goal  run  run-into- Lit. ‘The athlete(s) ran and ran into the goal’. b. Sensyu wa siai ni sonaete hasiri ni athlete  game  in.preparation run  hasiri-kon-da. [Type 2] run-- ‘The athlete(s) got ample running practice in preparation for the game’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

94

 

The acceptability of verb reduplication in Type 2 compounds should not be taken as indicative of their syntactic status. As a matter of fact, the reduplicated verb sequence “V-ni V” is qualified as a word because it cannot be broken up by any syntactic rule. (22)

*Sensyu wa athlete  Cf. (21b)

hasiri run

ni 

siai game

ni 

sonaete in.preparation

hasiri-kon-da. run--

The word status of a reduplicated verb, coupled with the hierarchical relation of the two types of compounds, as in (19), predicts then that a reduplicated Type 1 compound may occur in the V1 position of a Type 2 compound. The prediction is borne out by the attested example in (23) involving a Type 1 compound oti-tuki‘calm down’ and an L-asp verb -harau ‘do completely’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(23)

oti-tuki ni oti-tuki harat-te calm-  calm- .- ‘(He) is completely calm and serene’. http://ncode.syosetu.com/n0919s/121/ (Retrieved 2013)

i-ru. be-

Having established that Type 2 compounds with L-asp heads occupy a higher position in morphological structure than Type 1 compounds, we are in a position to consider the significance of the upward change (arrow 3) in Figure 4.1. On this route, the V2s of Type 1 (thematic) compounds, whether descendants of OJ/EMJ serial verbs or new creations, change to L-asp verbs through desemanticization. The crux of this change is that, as argued in Section 4.4, many of the L-asp verbs in current use appear to have acquired their functional meanings only when they were embedded in the particular morphological structure of V-V compounds. Consider the intensive meaning of -kaeru lit. ‘turn back’, as in sizumari-kaeru ‘be completely quiet’ and akire-kaeru ‘be thoroughly flabbergasted’. One of the earliest appearances of this form is waki-kapyeri ‘(of hot water) boil violently’ (1275), which, at the time of its initial use, must have been taken as complementation structure with a literal meaning like ‘The boiling event returns (=repeats itself)’. Repetition, then, is naturally extended to denote the violent manner of an action or completeness/permanence of a state. The complementation structure is in accord with the relatively early date of the first attestation (13th c. sizumari-kaeru ‘be completely quiet’). On the other hand, purely adverbial meanings like ‘violently’ and ‘completely’ gained impetus much later. Some examples in modern conjugation: 1798 akire-kaeru ‘be thoroughly flabbergasted’, 1802 syoge-kaeru ‘be completely dejected’, 1870 kutabire-kaeru ‘be tired out’ (obsolete). Crucially, the lexical verb kaeru never means ‘do violently/completely’ in its independent use; the intensive meaning manifests itself only when kaeru is compounded with suitable V1 verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  

95

If it is combined with unsuitable verbs, as in hane-kaeru [bounce-go.back] ‘bounce back’ or moti-kaeru [hold-go.back] ‘take back’, the resulting compounds are identified as Type 1 (thematic) compounds, with -kaeru retaining its literal meaning. The adverbial meaning of V2 kaeru is thus construction-specific, and the whole compound verb is qualified as a “constructional idiom” in the sense of Jackendoff (2010: 272–4). Using the formalism of Booij (2010), the construction schema of Type 2 -kaeru compounds will be represented as in (24).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24)

[V V1 [L asp kaeru]] , [V1’s event takes place completely/violently]

Of special importance is that the semantic specialization of kaeru to the particular adverbial meaning, which itself is viewed as a familiar case of metonymy, took place only in the V2 position of morphological compound structure. Not all L-asp verbs are so straightforwardly linked to their original meanings by metaphor or metonymy. Consider the transitive verb -morasu ‘leak, let out’, which may be used as an L-asp verb in Type 2 compounds in PDJ, as in ii-morasu [speak-] ‘fail to say what one wanted to say’ and kaki-morasu [write-] ‘fail to write what one should write’. Here, -morasu expresses the subject’s failure to accomplish the action of V1 due to his absentmindedness. According to Hyakutome (2015), however, the same verb in EMJ (The Tale of Genji) did not have this functional meaning but conveyed the literal meaning of letting something out not only when used as an independent verb but also as the V2 of a V-V complex. Thus, ii-morasu in EMJ would have meant ‘to let (a secret, etc.) out to the addressee by mentioning it’, and kaki-morasu ‘to let (a message, etc.) out by writing it’. These interpretations correspond to Type 1 (thematic) compounds in modern language, and it was not until Late Middle Japanese (Heike Monogatari, Taiheiki, Tsurezuregusa, etc.) that the modern usage of ‘fail to accomplish’ materialized, as in uti-morasi [slayleak] ‘fail to kill (an enemy)’ and mi-morasi [see-leak] ‘fail to see/recognize’. How can the thematic meaning of V-morasu be linked to the aspectual interpretation of the same compound? Although the thematic interpretation of ‘to let out’ looks diametrically opposite to the modern aspectual meaning of ‘to fail to accomplish’, the former expressing perfectivity and the latter imperfectivity the two meanings are, however, actually associated with each other in a natural way as a category extension of the object argument of the verb morasu ‘to leak, let out’. In the thematic interpretation of ii-morasu ‘to let (something) out by saying’, the object of morasu is identified as the object of V1 ‘say’, while in the aspectual interpretation, it is V1’s event itself, whereby ii-morasu is construed as lit. ‘let the saying action out’ or ‘fail to say, miss saying’. Again, the shift to the aspectual meaning was the privilege of the particular compound structure. Hyakutome’s (2015) suggested date of the first appearance of the aspectual meaning in V-morasu thus agrees with Aoki’s (2013) suggested period of the rise of morphological compound structure (Muromachi period; LMJ, fifteenth c.).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

96

 

While the semantic change in -morasu involves external specification, there are also examples involving the semantic change from thematic meaning to adverbial modification by internal specification. A case in point is tigiru, lit. ‘tear off ’, which is treated as a suffix in Table 4.5. Only when used as an L-asp in V-V complexes can this verb designate an intensive repetition or continuation of an action, an adverbial meaning which was first attested in 1221 (nenzu wo momi-tigiru ‘rub the prayer beads hard and repeatedly’; Ujishui Monogatari), followed by examples in the 1500s (NKD, under tigiru). There is, however, some reason to assume that it took another century or so for the Type 2 compounds to gather momentum. Pertinent to this assumption is the information gained from Nippo Jisho (Portuguese title: Vocabulario da Lingoa de Iapam com Adeclaração em Portugues (Vocabulary of the language of Japan with definitions in Portuguese)), a dictionary compiled by Jesuit missionaries and published in Japan in 1603–4. This dictionary supposedly reflects the language of the late 1500s fairly accurately. From Imaizumi’s (1971) thorough classification of about 32,000 words registered in Nippo Jisho, I collected a total of 557 V-V complexes (compound verbs) and classified them into thematic class (410 words) and aspectual class (147 words). The thematic class encompasses 100 compound verbs that were to develop later into Type 4 complexes of the form ‘V-te V’, such as sagasi-mi [search-see] for PDJ sagasi-te miru ‘try searching’ and ue-oki [plantput] for PDJ ue-te oku ‘plant and leave as such’, while the aspectual class contains sixty-one compound verbs that correspond to current syntactic compounds of Type 3 such as V-au ‘V each other’ and V-kakeru ‘be about to V’. Leaving out these verbs, we end up with a total of 396 ‘lexical’ compound verbs, with 310 verbs (78%) in the thematic class and 86 (22%) in the aspectual class. Notably, the proportion of lexical aspectual compounds (22%) is considerably smaller than that of modern L-asp compounds (34%) as against thematic compounds (66%), according to Kageyama and Kanzaki’s (2014) database. The relatively small proportion of lexical aspectual compounds in Nippo Jisho can be attributed to the fact that they mostly have complementation-type semantic interpretations, while L-asp verbs of adverbial meanings were not yet fully developed. In fact, many current L-asp verbs with a rhetorical flavor are first attested in Modern Japanese after the publication of Nippo Jisho. Drawing on Nihon Kokugo Daijiten, the first attestations of selected Type 2 compounds (in PDJ forms) are shown in (25). (25)

1683 1691 17th c. 1750 1759 1798

nirami-kaesu [glare.at-.] ‘glare back at someone’ kaki-naguru [write-] ‘write roughly’ si-agumu [do-] ‘be unable to achieve as one hopes’ home-tigiru [praise-.] ‘praise to the skies’ tukai-wakeru [use-] ‘select and use depending on the purpose’ akire-kaeru [be.dumbfounded-.] ‘be utterly dumbfounded’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

   1810 1822 1836 1886 1887 1889 1896 1902 1905 1911

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

1926 1948 1948 1965

97

yoi-harau / yop-parau [be.drunk-.] ‘get dead drunk’ damari-kokuru [be.quiet-.(?)] ‘be perfectly still and silent’ ure-nokoru [be.sold-] ‘remain unsold’ nemuri-kokeru [sleep-.(?)] ‘sleep fast and soundly’ kangae-komu [think-.] ‘become deeply involved in thinking’ ii-siburu [say-.] ‘be reluctant to say’ ii-yodomu [say-.] ‘do not speak clearly’ uri-osimu [sell-] ‘be reluctant to sell’ yomi-tobasu [read-.] ‘skip over in reading’ yomi-susumu [read-.(intr.)] ‘move ahead in reading (a book)’ kure-nazumu [(sun)set-] ‘(The day) lingers on’ wasure-saru [forget-] ‘forget completely’ nobi-nayamu [grow-] ‘do not grow as expected’ yomi-susumeru [read-.(tr.)] ‘read (a book) further ahead’

Note that such L-asp verbs of recent innovation are by no means frozen but may be creatively expanded to new V1s, as exemplified by V-siburu [V-. ]: 1950 kotae-siburu [answer-.] ‘be reluctant to answer’, 1963 moe-siburu [burn-.] ‘do not burn as one hopes’, and many other ad hoc examples found on the Internet such as ki-siburu [come-. ] ‘(school absentee) be unwilling to go to school’. These considerations lead us to conclude that the morphological structure of L-asp compound verbs was established from Late Middle Japanese to Early Modern Japanese. Additional data in support of this conclusion may be found in nominalization of infinitive verbs. In (26), some examples of compound nouns headed by the nominalized verbs -some ‘begin, do for the first time’ and -tate ‘do harshly’ are randomly selected from NKD. They all date back only to LMJ or EMJ. (26)

a. 1586 1603 1780 1841 1888

kai-some [buy-begin] ‘first-time purchase of a new year’ ai-some [meet-begin] ‘first-time meeting with someone’ asobi-some [play-begin] ‘first-time playing of a new year’ oki-some [get.up-begin] ‘first-time awakening of a new year’ saki-some [bloom-begin] ‘beginning of blooming’

b. 1477 1714 1780

ii-tate [speak-give.off.steam] ‘strong assertion’ oi-tate [chase-give.off.steam] ‘hurrying along as if by chasing’ ue-tate [plant-give.off.steam] ‘planting of trees/flowers in abundance’ kime-tate [decide-give.off.steam] ‘one-sided decision’

1800

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

98

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

How could the upward shift in Figure 4.1 be accounted for in existing theories of grammaticalization? If we focus on the semantic facet of the change, the upward shift is motivated by desemanticization of V2s and hence might well be explained away as an instance of grammaticalization. Note, however, that the desemanticization here is not merely the weakening of a concrete meaning to an abstract one. More important is the loss of V2’s argument structure. Although semantic change is likely to take place gradually, it would be hard to conceive of the loss of a verb’s argument structure as a gradual change. There is another reason why the upward shift in Figure 4.1 cannot be accounted for straightforwardly as grammaticalization. This point becomes clear if we pay attention to the formal facet of the change in the layered morphological structures of compound verbs, in particular the shift from a lower position with strong morphological bond (Type 1 in Figure 4.1) to a higher position with weaker morphological bond (Type 2 in Figure 4.1). In terms of the strength of morphological bond, this upgrading is regarded as a case of degrammaticalization, particularly what Norde (2009) calls “debonding,” where a bound morpheme becomes less bound (see also Trousdale and Norde 2013). Besides, the possibility of a third route should not be ignored. It will be recalled from Table 4.5 that NKD leaves the origins of quite a few verbs as unknown or as suffixes or auxiliary verbs from the start. These L-asp verbs whose origins as independent verbs are not documented threaten the fundamental idea of grammaticalization as well as Butt’s (2010) theory of light verbs.

4.6 Concluding remarks on creativity and grammatical change The lexical aspectual compounds of Type 2 are endowed with multifarious properties in their morphology and semantics which resist a unitary account along the lines of metonymic extension and unidirectional grammaticalization. The complexities of their synchronic properties are reflections of their complex diachronic development through multiple paths. As shown in Figure 4.1, there are at least two routes that led to the emergence of this peculiar class of compounds: (i) the downgrading of syntactic aspectual verbs to morphological compounds by lexicalization and (ii) the upgrading of thematic V2s to L-asp verbs. While the former is considered a garden-variety type of reanalysis and lexicalization, the latter, regarded as debonding (Norde 2009), runs counter to Hopper and Traugott’s (1993) cline of grammaticalization (Section 4.1) and also to Haspelmath’s (2004) view that grammaticalization is a change by which the parts of a construction acquire stronger internal dependencies. At the same time, however, the upward shift is reminiscent of Roberts and Roussou’s (2003: 205) minimalist treatment of grammaticalization as an instance of “upward reanalysis, which gives rise to new functional material.” Although Roberts and Roussou (2003) targeted the development of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

  

99

functional categories in syntactic structure (such as modal auxiliaries in English) from lexical heads (such as modal verbs in Old English), the parallelism with the rise of L-asp verbs within morphological structure is evident. While studies on grammaticalization following the tradition of Antoine Meillet and Jerzy Kuryłowicz tend to treat bare items (lexemes) as the primary target of research and morphosyntactic constructions as secondary, researchers are increasingly becoming aware of the necessity of taking morphosyntactic environments or “constructions” into account (Bybee 2001; Traugott 2001; Traugott and Trousdale 2013; Trousdale and Norde 2013). Bybee et al. (1994: 11) remark: “It is the entire construction, and not simply the lexical meaning of the stem, which is the precursor, and hence the source, of grammatical meaning.” Even these construction-oriented researchers, however, seem to presume that change of bare items is a prerequisite for constructionalization (recall Heine’s (2001) chain of grammaticalization, where desemanticization precedes contextual extension; cf. Section 4.1). The diachronic data surveyed in this chapter point to the opposite direction of change, where an abstract morphological schema of L-asp compounds was established first and then new verbs took advantage of it to give rise to novel compounds. There is good reason to consider that this must have been the case. First, the bare items (individual verbs) that participated in the development of Type 2 compounds were so variegated in meaning and form that it is hard to imagine that each of them individually underwent gradual semantic change and gradual loss of argument structure over many centuries. Second, the semantic effects produced by L-asp verbs are not the usual kind of change to purely grammatical concepts such as tense and modality but are an innovative mode of semantic modification whereby V1’s lexical semantic structure is augmented by internal specification of a particular adverbial meaning—presumably more or less comparable to vectors in South Asian languages (see Chapter 1, this volume). Third, this novel mode of semantic specification is executed not only by the L-asp verbs whose lexical sources are well documented but also by those whose origin as independent verbs is left obscure in reliable dictionaries (cf. Table 4.5). All these considerations lead us to analyze the rise of Type 2 compound verbs as constructionalization, i.e. the creation of a particular type of morphological construction associated with a particular semantic schema. After this compound schema was established in the 1600s, a wide assortment of new L-asp verbs as exemplified in (25) cropped up in the 1800s and 1900s. At this juncture, it is instructive to note that similar functional meanings are expressed by Dutch verb-particle constructions, where the combination of a verb and a separable particle behaves as a complex predicate that is functionally analogous to compound verbs. According to Los et al. (2012), Dutch separable particles convey a broad range of meanings, apart from resultatives.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

100

 

(27)

a. Modifying particles (temporal): de groenten voor-koken [the vegetables for-cook] ‘to precook the vegetables’, formally represented as [koken (x), (y), {before E}]) b. Relator particles: het publiek toe-spreken [the public to-speak] ‘to address the audience’, formally represented as [spreken (x) {to (het publiek)}]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. Aktionsart particles: door-zingen [through-sing] ‘continue singing’, formally represented as [continue Vactivity without intended goal (x) ([for y time])] (Los et al. 2012: 69–80) Since separable particles and L-asp verbs are distinct in morphological categories, their syntactic effects naturally diverge. As can be seen from the formal representations in (27), Los et al. (2012) conceive of the Dutch particles as syntactic modifiers which add aspectual and other kinds of meanings to the whole event of the main verb, often introducing an extra argument and affecting the transitivity of the base verb. For example, the particle toe ‘to’ in (27b) is combined with the intransitive verb spreken, yielding a transitive verb toe-spreken with a new direct object argument ‘audience’. Japanese can express the same meaning with the Type 2 compound verb hanasi-kakeru ‘speak to’, but the L-asp verb -kakeru lit. ‘hang’, devoid of argument structure, only emphasizes the direction of speaking without changing the transitivity of V1. Such syntactic differences notwithstanding, the semantic and cognitive affinity between Dutch separable particles and Japanese L-asp verbs is remarkable. In both languages, the meanings of subsidiary predicates, i.e. Japanese L-asp verbs and Dutch particles, cluster around analogous sets of aspectual and other functional concepts. It is tempting to hypothesize that there is a universal repertoire of aspectual and functional concepts available in morphological as well as syntactic structure—something more or less comparable to Heine and Kuteva’s (2002) universal lexicon of grammatical concepts—and that individual languages make use of their syntactic and morphological means to materialize some or all of these semantic concepts to an optimal extent. The last question to be addressed is: What motivated the creation of the marked construction of Type 2 compound verbs? What was its driving force? A fact that might be linked to this question is the Japanese propensity for putting focus on the step-by-step process of an event rather than on its result. A notable feature of the aspectual V2s in EMJ, such as -kane ‘be unable’, -ayamati ‘make an error’, -tuduki ‘continue’, -some ‘begin’, -sasi ‘do and stop halfway’, -magawi [wrong action], -tirasi ‘scatter in a mess’, -nokori ‘remain’, and -midare ‘become disorderly’, is that they are mostly associated with imperfective aspect, while only a few verbs (such as -toosi ‘put through’, -oosi/-oose ‘do to the end’, -naori ‘do again’ [retrial], -kawi ‘cross’) express perfective aspect. Even the perfective awe, lit. ‘endure’, is generally

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

  

101

put in the negative to give rise to an imperfective meaning like ‘be unable to’. Compound verbs with these V2s of imperfectivity enable Japanese speakers to elaborate on the details of the intermediate processes by which an event unfolds. The overwhelming tendency toward imperfectivity or process-focusing in Japanese L-asp verbs should now be contrasted with the overwhelming tendency of English verb-particle constructions toward perfective or resultative aspect, as exemplified by write up, put down, carry through, etc. The contrast in imperfectivity/perfectivity between Japanese and English has indeed been observed pervasively outside of the realm of V-V compounds as well. For example, telic or telicity-inducing constructions such as resultative constructions, locative alternations, and dative alternations are abundant in English, but are rare or unavailable in Japanese. Closely related to these facts is the availability of “conative” readings in many, though not all accomplishment verbs, as in moyashita keredo moenakatta, lit. ‘(I) burned it but it didn’t burn’—a phenomenon which is foreign to English but is shared by some Asian languages (Kato 2018). Although this phenomenon is sometimes referred to as “cancellation” of results, a better way of characterizing it will be that the speaker is more interested in narrating details of the initial and intermediate stages leading to the culmination of an accomplishment event rather than in announcing its end or result. Such a propensity in Japanese has been noted by a number of domestic researchers and is characterized as “-orientation” by Ikegami (1991) and as “process focus” by Kageyama (1996). The predominance of imperfective L-asp verbs could plausibly be regarded as a reflection of this general propensity of Japanese toward process-focused expressions. If so, it would not be fitting to attribute the rise of this special class of L-asp compounds to purely linguistic principles and constraints. More pertinent is the distinction in word formation between “productivity” and “creativity,” which Lyons (1977: 549) defines as in (28). (28)

a. Productivity: a design-feature of the language system b. Creativity: the language user’s ability to extend the system by means of motivated, but unpredictable principles of abstraction and comparison

According to this distinction, the formation of Type 1 thematic compounds is attributed to productivity, and that of Type 2 L-asp compounds to creativity. Speakers’ creativity will account for why Type 2 compounds are often associated with a rhetorical flavor. Lastly, the grammatical extension of the L-asp structure merits special mention. The L-asp category, once established, contributed to introducing lexical voice change in morphological structure. In conformity with the close relationship between aspect (stativity) and voice (passive) observed crosslinguistically, the Japanese L-asp appears to have developed into an abstract functional category of lexical passive. Jacobsen (2016) and Kageyama (2016b) discuss the semantics of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

102

 

Japanese intransitive verbs that can be translated into English only by using the passive form, as exemplified in (29). Unlike anticausative verbs like ore-ru ‘breakintr’ from or-u ‘breaktr’, the intransitive verbs like those in (29), called “decausative” verbs (Kageyama 1996, 2016b), are strongly associated with implicit agents that correspond to the subjects of their transitive variants.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(29)

a. b. c. d. e.

mituk-ar-u [find-intr-prs] ‘be found’ (first attestation 1767) kim-ar-u [decide-intr-prs] ‘be decided’ (1790) mook-ar-u [earn-intr-prs] ‘be earned’ (1800) tukam-ar-u [catch-intr-prs] ‘be caught/arrested’ (1809) some-agar-u [dye-rise-prs] ‘finish being dyed’ (1891)

A salient characteristic of these verbs is that their valency change is signaled by the decausative suffix -ar, which is presumably cognate with the existential verb ar-u ‘be’. In fact, when the existential verb ar-u is employed as a resultative V2 of Type 4 complexes in syntactic structure, it displays the same behavior of intransitivization by agent defocusing and theme focusing as the lexical decausative verbs in (29), as in Syokki ga arat-te aru ‘The dishes have been washed (and they are there for future use)’. Particularly noteworthy is (29e), where the intransitive verb agaru (rise) ‘be finished’ is used as the V2 of a Type 2 compound verb. This L-asp verb, when compounded with accomplishment verbs of creation in the V1 position, exhibits a productive pattern of intransitivization (decausativization) from transitive Type 2 compounds of the form V-ageru (V-raise) ‘finish V-ing’ (Kageyama 2016b). Decausative intransitivization is also in keeping with the Japanese propensity for dwelling on the culminative process of an event by defocusing a human agent.

Acknowledgments This work was supported in part by the grant for the research project “Syntactic, Semantic, and Morphological Characteristics of the Japanese Lexicon” (National Institute for Japanese Language and Linguistics, 2010–15) and in part by research funding from Doshisha University (2017). My special thanks are due to two reviewers: Bjarke Frellesvig, who kindly made numerous suggestions concerning the examples from premodern Japanese and encouraged me to examine the data in Nippo Jisho (1603), and John Haig, who made perceptive comments on some obscure points in the draft.

Sources of pre-modern Japanese Nihon Kokugo Daijiten (A Grand Dictionary of Japanese) Nihon Kotenbungaku Zenshū (The complete collection of Japanese classical literary works) —both available at JapanKnowledge (https://japanknowledge.com)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

5 Syntactic verb-verb compounds in Japanese Hideki Kishimoto

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

5.1 Introduction Compound verbs, which are derived by combining two verbs into one morphological unit, are found in many Asian languages. Japanese, for one, abounds with such compound verbs (Kuno 1978b). In the literature on Japanese, it is often claimed (Kageyama 1989, 1993, 2016a; Yumoto 2005; Kishimoto 2009, and others) that V-V compounds can be divided into syntactic and lexical types. In Kageyama’s classification (Chapter 2, this volume), V-V complexes are divided into four types. Lexical V-V compounds are classified into Type 1 (“thematic” compounds) and Type 2 (“lexical aspectual” compounds). Both types of lexical VV compounds lack analyzable structures and behave as single words in syntactic terms, despite their complex morphology (Chapter 2, this volume). On the other hand, syntactic V-V compounds, which fall into Type 3 in Kageyama’s classification, have syntactically transparent structures despite their morphological coherence.¹ It is generally agreed that syntactic V-V compounds involve complementation, i.e. the first verb is embedded under the second verb syntactically (see, e.g., Shibatani 1973; Kageyama 1993; Nishigauchi 1993; Koizumi 1999). The major aim of this chapter is to elucidate some notable properties of syntactic V-V compounds in Japanese. In particular, the present chapter shows

¹ Both syntactic and lexical V-V compounds constitute morphologically coherent constituents, so particles cannot be inserted between the component verbs (Kageyama 1993). *Sensei ga hon o {kaki wa hazime-ta/kaki wa age-ta}. teacher  book  {write  -/write  -} ‘The teacher {began to write/wrote up} a book.’ When suru-support is implemented, a difference in acceptability arises between syntactic and lexical V-V compounds. wa si-age-ta}. (ii) Sensei wa ittan hon o {?kaki wa si-hazime-ta/*kaki teacher  once book  {write  do--/write  do--} ‘The teacher once {began to write/wrote up} a book.’ (i)

When the verb suru is inserted to the left of V2, particle insertion is possible with syntactic V-V compounds, which shows that their component verbs are syntactically separate. By contrast, particle insertion is never allowed with lexical V-V compounds, which count as single lexical items syntactically.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

104

 

that the syntactic structures obtained in syntactic V-V compounds differ according to whether they belong to the raising or the control class. In Japanese, some, but not all control verbs appearing as V2s in syntactic V-V compounds allow so-called “long-distance” passivization. Kageyama (1993) attempts to account for the facts by positing distinct syntactic configurations for the passivizable and unpassivizable V-V compounds. On the contrary, the passivizability of V2 in syntactic V-V compounds does not directly depend on syntactic structures. Instead, I suggest that the difference in applicability of long-distance passivization between raising and control V-V compounds is attributed to whether V2 has an accusative-case feature to case-license an object; raising V2s are not passivizable because they do not carry accusative-case features, but control V2s carrying accusative-case features are passivizable. It is also shown that long-distance passivization is not applicable to control V-V compounds when they denote an uncontrollable event. I suggest that V2s in the raising class are construed as delexicalized verbs, whereas V2s in the control class behave as full lexical verbs. The discussion proceeds as follows. Section 5.2 discusses how syntactic V-V compounds are distinguished from lexical ones. In Section 5.3, it is shown that syntactic V-V compounds take distinct complement structures, which vary according to whether V2 is classified as a raising or a control verb. Section 5.4 takes up the question of how and why certain types of compound verbs allow longdistance passivization. This section suggests that long-distance passivization is allowed on control V-V compounds, but not raising V-V compounds, since V2 needs to bear a case feature to license an object, and that the passivizability of V2 in control V-V compounds is also contingent upon whether or not they describe a controllable event. Section 5.5 discusses the peculiar behavior of sugiru-compounds. A conclusion is presented in Section 5.6.

5.2 Some classifications V-V compounds are divided into two major classes, lexical and syntactic V-V compounds.² Syntactic and lexical V-V compounds look the same in morphological terms, but they do indeed exhibit a number of distinct properties (cf. Chapter 2, this volume). To mention just a few differences, lexical V-V compounds (such as toritukeru [take-attach] ‘attach’ and hasiri-komu [run-.] ‘run into’) constitute single words syntactically and possess many of the idiosyncratic properties of complex lexical items stored in the lexicon; for example, lexical V-V compounds have limited productivity and show lexical gaps; they sometimes undergo a semantic shift, with the result that they do not have compositional meanings (e.g. kaziri-tuku ² In this chapter, two kinds of lexical V-V compounds (i.e. Types 1 and 2) are coalesced into one class, as both possess basic properties as lexical words.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

105

[bite-.] has both the literal meaning (‘bite into’) and the figurative meaning (‘stick to’)). In contrast, syntactic V-V compounds have complex structures whose internal constituents are visible to the syntax, and the meaning of the whole expression is constructed compositionally from the meanings of their parts. As discussed by Kageyama (1993), syntactic V-V compounds are formed on a certain set of second verbs, many of which specify aspectual meanings in one way or another but some of which carry nonaspectual meanings (see also Chapter 2, this volume).³ (1)

a. Aspectual: inception (hazimeru ‘begin’, dasu ‘start’, kakeru ‘be about to’), continuation (tuzukeru ‘continue’), termination (owaru ‘end’, oeru ‘finish’), incompletion (sokoneru ‘fail’, wasureru ‘forget’)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Others: habitual (nareru ‘get used to’, tukeru ‘get used to’), excessiveness (sugiru ‘exceed’), reciprocity (au ‘meet’), repetition (naosu ‘fix, repeat’), potentiality (uru ‘can do’, eru ‘can do’) In both syntactic and lexical V-V compounds, V1 appears in the adverbal form (ren’yōkei), and V2 is combined with tense. Nishiyama (1998) regards V-V compounds as involving verb serialization, but the morphological form of V1 suggests that V-V compounding involves some kind of subordination, i.e. V1 and V2 have different syntactic statuses, while such indication is generally lacking in serial verbs (cf. Aikhenvald 2006). Furthermore, V1 and V2 form a morphologically tight unit, and, therefore, the verbal complexes can be best described as involving compounding, at least, in morphological terms. The sequence of syntactic V-V compounds most typically consists of two verbs, but can include more than two verbs, as long as they are aligned observing selectional restrictions (e.g. yomi-naosi-hazimeru [read--] ‘begin to read again’ versus *yomihazime-naosu [read--]). Syntactic V-V compounds include syntactically visible component verbs, but lexical V-V compounds do not. Thus, the two types of V-V compounds can be discerned by considering whether syntactic operations (such as passivization, subject honorification, and soo suru ‘do so’ replacement) are applicable to the component verbs. Example (2) represents a case involving soo suru replacement.⁴

³ Some of the listed verbs can also appear as the V2s of lexical V-V compounds. For instance, the verb dasu appearing as V2 in syntactic compounds carries the meaning of ‘start’. But when the same verb appears as V2 in lexical compounds, it expresses a motional sense, as in osi-dasu [push-.] ‘push out’. ⁴ Some construction-specific constraints may apply to the operational tests, but if any of these operations are allowed, compound verbs are identified as the syntactic type.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

106

 

(2)

Ken ga sen o hiki-{dasi/owat/naosi}-ta. Ken  line  draw-{//}- Eri mo [soo si]-{dasi/owat/naosi}-ta. Eri also [so do]-{//}- Lit. ‘Ken {started/finished/repeated the act of} drawing a line. Eri {started/ finished/repeated the act of} doing so.’

The well-formed second sentence in (2) includes verbal compounds derived by applying soo suru replacement to the first verb hiku ‘draw’ in the preceding syntactic V-V compounds. In contrast, soo suru replacement cannot apply to the first verb haru ‘paste’ appearing in the lexical V-V compound hari-tukeru ‘attach’, as shown in (3). (3)

Ken ga kabe ni memo o hari-tuke-ta. Ken  wall on memo  paste-attach- *Eri mo [soo si]-tuke-ta. Eri also [so do]-attach- Lit. ‘Ken paste-attached a memo on the wall. Eri do-attached so’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In a similar vein, the syntactic operations of turning verbs into honorific or passive forms are generally applicable to the first verbs in syntactic V-V compounds (e.g. kaki-hazimeru ‘begin to write’), but not lexical V-V compounds (e.g. kaki-ageru ‘write up’), as in (4). (4)

a. Sensei ga hon o {o-kaki-ni-nari-hazime-ta/ teacher  book  {-write---begin-/ *o-kaki-ni-nari-age-ta}. -write----}. ‘The teacher {began to write/wrote up} a book.’ b. Hon ga {kak-are-hazime-ta/*kak-are-age-ta}. book  {write---/write---} ‘The book began to be written (up).’

If the complex of V1 and V2 constitutes a single lexical item, no syntactic operation may apply to only one of the two components owing to V1-V2’s lexical integrity (see Anderson 1992). Given that the components of syntactic compound verbs are not constrained by the lexical-integrity principle, despite their morphological coherence, it can be concluded that syntactic V-V compounds have a constituent structure in which the component verbs are visible syntactically. Additionally, as noted by Kageyama (1993), only (simple) native Japanese verbs can be compounded to form lexical V-V compounds. Thus, compound verbs are

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

107

categorized as syntactic if they allow VN-suru (i.e. suru combined with a verbal noun) to appear as their first verb, as in (5). (5)

Ken wa posutaa o insatu-si-{dasi/wasure/oe}-ta. Ken  poster  print-do-{//}- ‘Ken {reprinted/forgot to print/finished printing} the poster.’

As discussed by Kageyama (1993) and Kishimoto (2006), VN-suru forms are held to be created syntactically by virtue of incorporating a VN into suru.⁵ If so, the complex verbs containing VN-suru cannot be lexical V-V compounds, i.e. they are not morphologically complex lexical items registered as single words in the lexicon. The acceptability of the VN-suru forms in (5) suggests that compound verbs with the V2s dasu/wasureru/oeru ‘start/forget/finish’ fall into the class of syntactic V-V compounds. Syntactic V-V compounds have syntactically transparent structures. Their structures vary depending on whether they fall into the “raising” or “control” class (Kageyama 1993; Kishimoto 2009).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(6)

a. Ken ga hon o yomi-{dasi/kake}-ta. Ken  book  read-{/..}- ‘Ken {started/was about} to read the book.’

(Raising)

b. Ken ga hon o yomi-{owat/naosi}-ta. Ken  book  read-{/}- ‘Ken {stopped/repeated the action of} reading the book.’

(Control)

Several tests can be used to distinguish between raising and control compounds (see Kageyama 1993). The difference can be confirmed, for instance, by considering whether a clausal idiom (or a subject idiom) can appear in the compound verb constructions, as in (7). (7)

a. Kono mise de kankodori ga naki-{dasi/kake}-ta. this store at cuckoo  sing-{/..}- ‘This store {started/was about} to get fewer customers.’ b. *Kono mise de kankodori ga naki-{owat/naosi}-ta. this store at cuckoo  sing-{/}- Lit. ‘The cuckoo {stopped/repeated the act of} singing at this store’.

Given that raising, but not control verbs allow clausal idioms to appear as their complements (Carnie 2006), the contrast in acceptability between (7a) and (7b) ⁵ Many verbal nouns are Sino-Japanese words, but they can be native Japanese, Foreign, or Mimetic words.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

108

 

suggests that the compound verb construction in (7a)—comprising dasu ‘start’/ kakeru ‘be about to’—involves raising, whereas the compound verb construction in (7b)—comprising owaru ‘end’/naosu ‘fix, repeat’—involves control. Another way of distinguishing between the two classes of syntactic compound verbs is to see whether inanimate subjects are allowed. (8)

a. Ame ga huri-{dasi/kake}-ta. rain  fall-{/..}- ‘It {started/was about} to rain.’ b. *Ame ga huri-{sokone/wasure/nokosi}-ta. rain  fall-{//}- Lit. ‘It {failed/forgot/remained} to rain.’

As seen in (8), a meteorological expression taking an inanimate subject can appear in raising V-V compound constructions, but generally not in control V-V compound constructions. This diagnostic is slightly complicated, however, by the fact that inanimate subjects are allowed to appear in certain control V-V compound constructions, as in (9).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(9)

Kane ga nari-owat-ta. bell  ring-- ‘The bell stopped ringing.’

Example (9) is construed as involving pseudo-control in the sense of Pustejovsky (1995). Pseudo-control is possible when an inanimate entity is identified as an autonomous entity (i.e. a pseudo-agent) instigating the event described by the verb. The distinction of raising versus control compounds is largely determined according to what type of V2 occurs in syntactic V-V compounds. A nonexhaustive list of V2s classified into raising and control classes is provided in (10). (10)

a. Raising verbs: kakeru ‘be about to’, dasu ‘start’, sugiru ‘exceed’ b. Control verbs: naosu ‘repeat’, sokoneru ‘fail’, tukusu ‘exhaust’, tukeru ‘have a habit of ’ c. Ambiguous: hazimeru ‘begin’, tuzukeru ‘continue’

Raising verbs primarily specify aspectual meanings. Control verbs include both verbs specifying aspectual meanings and thematic verbs denoting (sub)events. Some V2s, including hazimeru ‘begin’ and tuzukeru ‘continue’, are ambiguous in that they can be used as either raising or control verbs (cf. Perlmutter 1970).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

109

In Section 5.3, I suggest that some notable behavioral differences between raising and control predicates come from the fact that control verbs select agent or experiencer arguments as their subjects, while raising verbs do not select subjects.

5.3 The layered structures For syntactic V-V compounds, it is generally agreed that the first verb is embedded under the second verb syntactically (see, e.g., Shibatani 1973, 1978; Kageyama 1993; Koizumi 1999; Kishimoto 2009; Fukuda 2012). Nevertheless, there has not been much discussion about the structural organization of the embedded clause (headed by V1). In the following, I suggest that syntactic V-V compounds take distinct complement structures, depending on whether they belong to the control or the raising type. Specifically, I propose that control V-V compounds have complements that include projections only up to vP, but that raising V-V compounds have a structure where aspectual projections like AspfreqP and AspdurP intervene between the verbal projections projected from the two component verbs. (AspfreqP, which specifies frequencies of events, can be assumed to project above AspdurP, which specifies the temporal length of a single event (see Cinque 1999). (11) a.

TP

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

SBJ

b. SBJ

vP2 SBJ

vP1 VP1

VP2 v2

v2

AspfreqP V2 AspdurP Aspfreq

V2 v1

V1

vP2

v'2 VP2

OBJ

TP

vP1 SBJ

Aspdur v'1

VP1 OBJ

v1 V1

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

110

 

I suggest that the subjects of raising V-V compounds are generated in vP₁ projected from the lower main verb (V1), whereas control V-V compounds generate their subjects in vP₂ projected from the upper verb (V2). If subjects are amenable to raising, they end up in Spec-TP, as indicated by the arrows in (11) (see Kishimoto 2010). I propose that syntactic V-V compounds require only one subject syntactically. As argued next, control predicates do not project PRO even though vP is present in the embedded clause, which implies that the agent theta role which V1 carries is suppressed when V1 is embedded under a control verb V2, and hence the subject of V1 is not projected syntactically.⁶ The present proposal on the embedded structure for syntactic V-V compounds is verified by appeal to various adjuncts, such as agent-oriented adverbs (e.g. kenmeini ‘hard’, wazato ‘deliberately’), instrumental PPs (e.g. zyoogi de ‘with a ruler’), manner adverbs (e.g. nanameni ‘diagonally’), durative adverbs (e.g. zutto ‘constantly’), and frequency adverbs (e.g. nan-kai mo ‘many times’, syuu ik-kai ‘once a week’).⁷ These adjuncts can be assumed to modify distinct projections, as in (12).⁸

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

a. b. c. d. e.

Agent-oriented adverbs modify agentive subject/PRO. Instrumental PPs modify vP. Manner adverbs specifying how a change of state takes place modify VP. Durative adverbs modify AspdurP. Frequency adverbs modify AspfreqP.

It is a commonly accepted assumption that adjuncts are adjoined to projections that they modify (Zubizarretta 1987; Radford 2009). It follows from this that adjuncts cannot appear in a clause if their modifying projections are not present. With the above remarks in mind, let us consider whether various adjuncts can be associated with the embedded clauses in control V-V compounds. As seen in (13), the instrument PP zyoogi de ‘with a ruler’ and the manner adverb nanameni ‘diagonally’ can modify the embedded clause, but the durative adverb zutto ‘constantly’ and the frequency adverb nan-kai mo ‘many times’ cannot.

⁶ This view is reasonable given that no projection case-licensing  is included in the lower clause. ⁷ Temporal adverbs cannot modify the lower clauses in either type of syntactic V-V compounds owing to the absence of a tense projection, as in (i). (i) *Ken ga [koko de kinoo hanasi]-{owat-ta/dasi-ta}. Ken  here in yesterday speak-{-/-} ‘Ken {ended/started} [speaking yesterday here].’ The adverb kinoo ‘yesterday’ cannot be associated with the embedded clause containing V1, and (i) is not acceptable in the relevant sense. ⁸ Some adjuncts not discussed here include depictive predicates and resultative predicates. Needless to say, subject-oriented depictives modify subjects, and object-oriented depictives and resultative predicates modify objects.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -    (13)

111

a. Ken ga [{zyoogi de/nanameni} sen o hiki]-{owat/oe}-ta. Ken  [{ruler with/diagonally} line  draw]-{/}- ‘Ken {stopped/finished} drawing lines {with a ruler/diagonally}.’ b. Ken ga [{*zutto/*nan-kai mo} sen o Ken  [{constantly/many.times also} line  hiki]-{owat/oe}-ta. draw]-{/}- ‘Ken {stopped/finished} drawing lines {constantly/many times}.’

There is good reason to believe that zyoogi de and nanameni are adjoined to distinct verbal projections. This can be seen from the fact that in the pseudo-cleft sentences (involving vP-clefting) in (14), zyoogi de can occur in focus position as well as in the presupposition clause, but nanameni can occur only in focus position.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(14)

a. [Ken ga si-ta] no wa [{zyoogi de/nanameni} [Ken  do-] that  [{ruler with/diagonally} sen o hiku] koto da. line  draw] fact  ‘What Ken did was draw a line {with a ruler/diagonally}.’ b. [Ken ga {zyoogi de/*nanameni} si-ta] no [Ken   {ruler with/diagonally} do-] that [sen o hiku] koto da. [line  draw] fact  ‘What Ken did {with a ruler/*diagonally} was draw a line.’

wa 

The pseudo-cleft constructions in (14) involve vP focusing, suggesting that vP and VP projections appear in the focus position, while the antecedent clause includes vP and other higher projections, but not VP, as represented in (15) (see Kishimoto 2016 for the details of the derivation of vP-focus pseudo-clefting). (15)

[Presupposition [TP SBJ . . . [vP ]] ] no wa [Focus [vP [VP ]] ] koto da.

Given the structure in which vP is included in both antecedent and focus components, it is easy to see that the instrumental zyoogi de can appear in either the antecedent clause or focus position.⁹ The manner adverb nanameni can occur in focus position, because VP, to which it is adjoined, is projected there; the same ⁹ In (15), the subject in TP is originated from vP in the antecedent clause, while  is filled in vP in the focus position. The agentive adverb kenmeini ‘hard’ shows the same behavior as the manner instrument PP zyoogi de in a vP-cleft construction like (14) in that they can appear either in the antecedent clause or in the focus position.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

112

 

manner adverb, however, cannot be placed in the antecedent clause, which does not include VP (located below vP). Interestingly, control constructions do not allow agent-oriented adverbs (like kenmeini ‘hard’) to modify the embedded clause, as in (16), despite the fact that vP is projected in the embedded clause. (16)

Ken ga [(*kenmeini) yama-miti o aruki]-{owat/oe}-ta. Ken  [(hard) mountain-path  walk]-{/}- ‘Ken {stopped/finished} walking the mountain path (hard).’

On the other hand, the agent-oriented adverb kenmeini can modify the embedded clauses in (17), which can be assumed to contain . (17)

a. [ kenmeini aruku] no [ hard walk] that ‘It is a good thing to walk hard.’

wa 

ii good

koto thing

da. 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Kare wa [ kenmeini aruki]-taku-nakat-ta. he  [ hard walk]-want-- ‘He did not want to walk hard.’ The availability of the intended interpretations in (17) illustrates that  can be a target for modification by kenmeini ‘hard’, but, in (16), kenmeini cannot modify the embedded clause. Note that vP is projected from the embedded verb, as evidenced by the modifiability of the instrumental zyoogi de in (14a). Then, the data in (16) and (17) suggest that the embedded vP of control compounds does not include , as represented in (11a). If the lower clauses of control V-V compounds do not comprise , it is necessary to answer the question why they exhibit properties characteristic of control predicates projecting . I suggest that syntactic V-V compounds behave as control predicates when the upper V2s select thematic subjects, as in (18)—the structure equivalent to (11a). (18)

[

[vP2

SBJ [VP2

[vP1

[VP1

V1] v1] V2] v2] T ]

When clausal idioms are embedded in control V-V compounds, their subjects are placed in the vP₂ associated with V2 and assigned either an experiencer or agent theta role accordingly. Note, however, that idioms have noncompositional meanings that are interpreted without theta-role assignment (see Chomsky 1981; Jackendoff 1997). Thus, clausal idioms are not allowed to appear in control V-V compound constructions, even though they do not involve control in a strictly

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

113

syntactic sense.¹⁰ In a similar vein, inanimate subjects are not allowed in control V-V compound constructions (except when they involve pseudo-control), because the subjects, which are assigned either an experiencer or agent theta role by the upper verbs, need to be animate. Raising predicates, by contrast, take complement clauses that include aspectual projections positioned above vP. This can be seen from the fact that durative and frequency adverbs, as well as instrumental PPs and manner adverbs, can be interpreted as modifying the lower clause. (19)

a. Ken ga [{zyoogi de/nanameni} sen o Ken  [{ruler with/diagonally} line  hiki]-{dasi/kake}-ta. draw]-{/..}- ‘Ken {started/was about} to draw lines {with a ruler/diagonally}.’ b. Ken ga [{zutto/nan-kai mo} sen o Ken  [{constantly/many.times also} line  hiki]-{dasi/kake}-ta. draw]-{/..}- ‘Ken {started/was about} to draw lines {constantly/many times}.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The facts of the aspectual adverbs zutto ‘constantly’ and nan-kai mo ‘many times’ indicate that the raising verbs select a complement clause that includes the aspectual projections AspfreqP and AspdurP.¹¹

¹⁰ The fact that subject idioms are not consonant with control constructions is sometimes accounted for in terms of the adjacency condition (e.g. Carnie 2006), but this account cannot be extended to syntactic V-V compounds. ¹¹ The raising verbs, but not the control verbs, can take aspectual auxiliary verbs in the te-form as their complements, as in (i). (i) Ken wa [[kusuri o herasi-te] iki]-{dasi/sugi/?*oe/?*naosi}-ta. Ken  medicine  reduce- go-{///}- ‘Ken {started/exceeded/finished/repeated the act of} reducing medicine.’ I assume that -te heads infinitival TP since it originated as a perfective marker tu historically (cf. Nakatani 2016). The contrast in acceptability in (i) is due to the fact that the aspectual verb iku ‘go’ is a functional predicate, which occupies a position higher than vP, perhaps in AspdurP. Since the raising predicates can have aspectual projections included in the complement clauses, they can take -te clauses as their complements. By contrast, the control verbs do not allow projections positioned above vP to be included in their complement clauses, and hence cannot take te-complement clauses. The contrast in acceptability found in syntactic V-V compounds is not observed for V-te V complexes. (ii) Ken ga [[kusuri o herasi-te] it-te] {simat/mi}-ta. Ken  medicine  reduce- - {./}- ‘Ken {ended up/tried} reducing medicine.’ The auxiliary verbs simau and miru select an infinitival TP complement with -te, which includes aspectual projections. Accordingly, both sentences in (ii) are acceptable regardless of whether the uppermost verb is a control or raising predicate.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

114

 

Moreover, raising compound verbs do allow modification by subject/ agent-oriented adverbs in the embedded clause. (20)

Ken ga [kenmeini hanasi]-{dasi/kake}-ta. Ken  [hard speak]-{/..}- ‘Ken {started/was about} to speak hard.’

This is expected, since the surface subject of raising compound verbs originates in the lower clause. In essence, the data regarding adjunct modifications illustrate that raising V-V compounds have an embedded constituent structure comprised of aspectual projections alongside verbal projections (AspfreqP-AspdurP-vP-VP), while the control compounds involve only vP complementation (vP-VP). Notably, control V-V compounds do not project  even though vP is present in the lower clause.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

5.4 Long-distance passivization In this section, I will turn to the question of why syntactic V-V compound constructions are divided into two types according to whether passivization may apply to V2. By looking at a number of case-marking phenomena that are found when passivization applies to V2 in syntactic V-V compounds, I suggest that the raising type of V-V compound does not allow long-distance passivization due to the absence of an accusative-case feature on the V2, and that long-distance passivization is prevented from applying to the control type of V-V compound if it describes an uncontrollable event.

5.4.1 Problems with a movement account To begin, observe that some but not all V2s in control V-V compound verbs allow passivization when V1s are transitive (i.e. V1 selects a direct object), as noted by Kageyama (1993, 2016a). (21)

a. *Kono kanzi ga yomi-sokone-rare-ta. this Chinese.character  read--- Lit. ‘This Chinese character was failed reading.’ b. Kono hon ga yomi-naos-are-ta. this book  read--- ‘This book was read again.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

115

The clauses in (21) have the same transitive V1, but show a difference in applicability of passivization on V2 (i.e. long-distance passivization). Passivizable control verbs include naosu ‘fix’, tukusu ‘exhaust’, oeru ‘finish’, hazimeru ‘begin’, and tuzukeru ‘continue’. Unpassivizable control verbs include owaru ‘end’, sokoneru ‘fail’, and ayamaru ‘fail’. Kageyama (1993) attempts to account for the (un)availability of long-distance passivization in terms of a syntactic constraint on movement, i.e. the minimality condition (Rizzi 1990), by positing the structures in (22a) and (22b) for (21a) and (21b), respectively (see also Kageyama 2016a; Nishigauchi 1993). (22) a. *[tp

[vp2

[vp1

PRO SBJpass

V1] V2-pass]T]

X

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. [tp

[vp2

[vp1

SBJpass

V1] V2-pass]T]

The two types of control constructions differ as to whether the lower VP (=vP in the present perspective) includes  (cf. Kageyama 2016a). According to Kageyama, the raising of the passive subject to the surface subject position (Spec-) is blocked in (22a) owing to the presence of an intervening , and hence the sentence is unacceptable. In contrast, (22b) is acceptable, because the passive subject can be moved to Spec- in the absence of  in the lower clause. In Kageyama’s analysis, the failure of long-distance passivization in (22a) is attributed to the presence of  in the complement clause. As discussed previously, however, there is evidence that control predicates do not include . If so, the (un)passivizability of control verbs cannot be treated in terms of the minimality condition. His analysis is also called into question by the observation that two types of passive clauses can be formed on ditransitive verbs. (23)

a. Kodomo ga omotya o child  toy  ‘The child was given the toy.’

atae-rare-ta. give--

b. Omotya ga kodomo ni toy  child  ‘The toy was given to the child.’

atae-rare-ta. give--

As argued by Hoji (1985), a direct object is located in a position structurally lower than that of an indirect object in Japanese. If one argument cannot move across another argument via A-movement owing to the minimality condition, it is expected that the passive clause in (23b), where the direct object is promoted to the passive subject, should be unacceptable. Nevertheless, (23b) is acceptable,

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 

116

contrary to expectation.¹² Another problem with Kageyama’s proposal is that raising verbs are predicted to be passivizable, but they cannot, in fact, be passivized.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24)

*Kono ronbun ga kaki-das-are-ta. this paper  write--- Lit. ‘This paper was started to write.’

(Raising)

Kageyama’s account can only take care of control verbs. Kageyama in fact notes that a completely different account is required to deal with the nonapplicability of long-distance passivization to raising V-V compounds. Given the previously noted difficulties the prior analysis faces, it is apparent that the minimality condition imposed on movement cannot provide an account of the contrast in passivizability observed in the syntactic V-V compounds. In Section 5.4.2, I suggest that the difference in applicability of long-distance passivization between the control V-V compounds in (21b) and the raising V-V compounds in (24) can be attributed to whether V2 has an accusative-case feature to case-license an object, and that some control verbs resist passivization (e.g. (21a)) on the grounds that they denote uncontrollable events (even if they carry an accusative-case feature). If V2 case-licenses an object of V1, there is a sense in which passivization may apply to V2, because passivization is made available via absorption of a case feature (see Baker, Johnson, and Roberts 1989). Under the present perspective, the applicability of long-distance passivization is determined without reference to A-movement. This view gains empirical support from the fact that syntactic V-V compounds give rise to the same contrast in acceptability even when no A-movement is involved. In what follows, I will discuss three types of empirical evidence in support of the present view. As a first piece of evidence in support of the present analysis, let us consider kata-nominals in (25), which are derived by suffixing the nominalizer kata ‘way’ to syntactic V-V compounds.¹³ (25)

ano hito no utai-{sokone/naosi/dasi}-kata that man  sing-{//}-way ‘the way of that man’s {failing to sing/re-singing/starting to sing}’

¹² Takano (2010) suggests that pro/ does not block A-scrambling in Japanese, which may provide another argument against Kageyama’s analysis. ¹³ Kata-nominals are often not easy to construct from syntactic V-V compounds that have kakeru ‘be about to’ as V2, but it is still possible to derive kata-nominals from some syntactic V-V compounds with kakeru, as in (i). (i)

ano ie no katamuki-kake-kata that house  lean-..-way ‘the way of the house’s about to lean’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

117

The hallmark of kata-nominalization is that the hierarchical position of genitive arguments is reflected in their surface order, owing to the absence of scrambling applying to them. Furthermore, kata-nominals do not induce A-movement even if their verbal complements have undergone passivization (Kishimoto 2006). The absence of A-movement is verified by considering how arguments are ordered in active and passive kata-nominals. As seen in (26), when the two arguments of oku ‘put’ are marked with genitive case, they are realized in the same order in katanominals regardless of whether passivization applies.¹⁴ (26)

a. teeburu no kabin no {oki-kata/ok-are-kata} table  vase  {place-way/place--way} ‘the way of {placing the vase /the vase’s being placed} on the table’ b. *kabin no teeburu no {oki-kata/ok-are-kata} vase  table  {place-way/place--way} ‘the way of {placing the vase /the vase’s being placed} on the table’

Note that the verb oku ‘put’ allows the direct object kabin ‘vase’ to be promoted to a passive subject, but not the PP teeburu ‘table’, as in (27).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(27)

a. Kabin ga teeburu ni ok-are-ta. vase  table  place-- ‘The vase was placed on the table.’ b. *Teeburu ga kabin o ok-are-ta. table  vase  place-- ‘The table was placed the vase on.’

If A-movement is invoked in the passive nominal, the argument kabin should be realized as a passive subject appearing to the left of teeburu. This expectation is not fulfilled, however, because kabin must follow teeburu, showing that no A-movement is involved in the passive nominal. Notably, when long-distance

¹⁴ The judgments come from a group of speakers who allow the two internal arguments to be marked with genitive case under kata-nominalization. While some speakers do not seem to allow indirect internal arguments to be marked with genitive case, examples like (i) are acceptable for many speakers. (i) a. kiiboodo no yubi no oki-kata keyboard  finger  put-way ‘the way of putting fingers on the keyboard’ b. kan’yoo-syokubutu no mizu no yari-kata ornamental-plant  water  give-way ‘the way of giving water to ornamental plants’ In particular, when appropriate contexts are given, nominal constructions in which more than one argument is marked with genitive case can be readily accepted by many speakers.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

118

 

passivization applies to the kata-nominals in (25) (derived from syntactic V-V compounds), they display the same syntactic behavior as their clausal counterparts, as shown in (28). (28)

a. *kono kanzi no kaki-sokone-rare-kata this Chinese.character  write---way Lit. ‘the way of this Chinese character’s being failed to write’ b. kono ronbun no kaki-naos-are-kata this paper  write---way Lit. ‘the way of the paper’s being written again’ c. *kono ronbun no kaki-das-are-kata this paper  write---way Lit. ‘the way of this paper’s being started to write’

The fact that no A-movement is involved in passive kata-nominals shows that the difference in the applicability of long-distance passivization seen in (22) cannot be attributed to a syntactic constraint imposed on A-movement. Tough-constructions provide another piece of evidence for the present claim. When a tough-construction formed on syntactic V-V compounds takes a dativenominative case-marking pattern, it displays a syntactic behavior akin to what is observed for long-distance passivization.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(29)

a. ?*Ken ni hiru-gohan ga tabe-sobire-yasukat-ta. Ken  lunch  eat--- ‘It was easy for Ken to fail to have a lunch.’ b. Ken ni kono hon ga yomi-naosi-yasukat-ta. Ken  this book  read--- ‘It was easy for Ken to read this book again.’ c. ?*Ken ni osake ga nomi-dasi-nikukat-ta. Ken  sake  drink--- ‘It was difficult for Ken to start drink sake.’

No such difference in acceptability is observed for the tough-construction taking a ‘nominative-accusative’ case-marking pattern, however. (30)

a. Ken ga hiru-gohan o tabe-sobire-yasukat-ta. Ken  lunch  eat--- ‘It was easy for Ken to miss his lunch.’ b. Ken ga kono hon o yomi-naosi-yasukat-ta. Ken  this book  read--- ‘It was easy for Ken to read this book again.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

119

c. Ken ga osake o nomi-dasi-nikukat-ta. Ken  sake  drink--- ‘It was difficult for Ken to start to drink sake.’ When the tough-construction has the dative-nominative case-marking pattern, it describes the subject’s judgment on the degree of difficulty. On the other hand, when the tough-construction takes the nominative-accusative case-marking pattern, it describes the likelihood (or tendency) of the described event taking place (see, e.g., Saito 1982). Despite the difference in their meaning and case-marking patterns, the grammatical status of arguments in the tough-construction remains invariant. This is confirmed by the fact that, in both variants, the subject-oriented reflexive zibun ‘self ’ can take the initial argument as its antecedent.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(31)

a. Keni ni zibuni no kodomo ga Keni  selfi  child  ‘It is easy for Ken to scold his children.’

sikari-yasu-i. scold--

b. Keni ga zibuni no kodomo o Keni  selfi  child  ‘It is easy for Ken to scold his children.’

sikari-yasu-i. scold--

The object status of the second argument can be confirmed by looking at whether a formal noun koto ‘fact’ devoid of a substantive lexical meaning can be inserted (see Sasaguri 1999; Kishimoto 2005; Takubo 2007). Observe first that the formal noun koto can be inserted into the object, but not the subject position, regardless of the case-marking pattern, as in (32). (32)

a. Ken (*no koto) Ken ( fact) ‘Ken scolded Eri.’

ga 

Eri Eri

(no (

koto) fact)

o 

sikat-ta. scold-

b. Ken (*no koto) ni Eri (no koto) ga sikar-e-ru Ken ( fact)  Eri ( fact)  scold-- ‘Ken can scold Eri.’ The same distribution is observed for the tough-clauses in (33); the formal noun koto can be added only to the theme argument irrespective of its case marking. (33)

a. Ken (*no koto) ni Eri (no koto) ga sikari-yasu-i. Ken ( fact)  Eri ( fact)  scold-- ‘It is easy for Ken to scold Eri.’ b. Ken (*no koto) ga Eri (no koto) o sikari-yasu-i. Ken ( fact)  Eri ( fact)  scold-- ‘It is easy for Ken to scold Eri.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

120

 

The data illustrate that the nominative argument in (33a) serves as an object in just the same way as the accusative argument in (33b). The two variants of the tough-construction possess different syntactic structures even though their arguments assume the same grammatical functions. This can be seen from (34), where a transitive clause takes an inanimate subject. (34)

a. Konpyuuta ga netu o moti-yasu-i. computer  heat  hold-- ‘It is easy for computers to get hot.’ b. *Konpyuuta ni netu ga moti-yasu-i. computer  heat  hold-- ‘It is easy for computers to get hot.’

Inanimate subjects can appear in the tough-construction when it takes the nominative-accusative case-marking pattern, but not when it takes the dativenominative case-marking pattern. This is indicative of the fact that the toughpredicates carrying the meaning of subjective judgment select an experiencer, whereas the same tough-predicates do not select any argument if they express the meaning of tendency or disposition (Saito 1982; Sugioka 1986). The data suggest that the constituent structures of tough-clauses vary depending on which case-marking pattern they take; the nominative-accusative variant involves raising, and the dative-nominative variant, control, as represented in (35).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(35)

a. [TP SBJ- [adjP [vP SBJ- - V]-yasu/niku]T] b. [TP - [adjP - [vP  - V]-yasu/niku ]T]

In both variants of the tough-construction, the arguments have the same grammatical relations, which suggests that, in the tough-constructions, no A-movement promoting an argument takes place even when their case-marking pattern is changed.¹⁵ ¹⁵ One reviewer remarked that ii-dasu ‘bring up, propose’ can license a dative-nominative case-marking pattern, despite the generalization that the raising verbs are not capable of assigning accusative case. (i)

Ken ni wa sono koto ga ii-dasi-nikukat-ta. Ken   that matter  say--- ‘It was difficult for Ken to start talking about the matter.’ The dative-nominative case-marking pattern is licensed by ii-dasu in (i), since this compound verb serves not as a syntactic but as a lexical V-V compound, as confirmed by the failure to license soo suru replacement. (ii) *Ken ga sono koto o ii-dasa-nakat-ta node, Ken  that matter  say--- because si]-dasa-nakat-ta. do--- ‘Because Ken did not bring up this matter, Mari did not start doing so.’

Mari Mari

mo also

[soo so

There are lexical V-V compounds where V2 carries an aspectual meaning (Chapter 2, this volume). Accordingly, it is not always possible to distinguish syntactic compounds from lexical compounds solely in terms of the meanings of V2s. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

121

Syntactic V-V compounds appearing in the dative-nominative variant of tough-clauses display the same syntactic behavior as those appearing in ordinary clauses with regard to long-distance passivization. Importantly, the dativenominative case pattern is made available by the tough-adjective changing the case frame of its complement verb with no A-movement. This fact suggests that the possibility of long-distance passivization in syntactic V-V compounds is determined independently of the question of whether an argument is moved across another argument via A-movement. Yet another form of confirmation may be derived from potential verb formation. In Japanese, verbs can be turned into potential forms by adding the potential affixes -(r)are/-(r)e. When the potential forms of syntactic V-V compounds have a dative-nominative case-marking pattern, the contrast in acceptability is observed. (36)

wa nantoka zikannai ni kotae ga a. ?*Ken ni Ken   somehow time in answer  kaki-owar-e-ta. write--- ‘Somehow, Ken was able to finish writing answers in time.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Sensei ni wa umaku zikanwari ga teacher   well class.schedule  kumi-{naos-e/oe-rare}-ta. make-{-/-}- ‘The teacher was able to {rearrange/finish making} the class schedule well.’ c. ?*Ken ni wa taimingu yoku uta ga utai-das-e-ta. Ken   timing good song  sing--- ‘Ken was able to start singing with good timing.’ No difference in acceptability emerges when they take a nominative-accusative case-marking pattern, however. (37)

a. Ken ga nantoka zikannai ni kotae o Ken  somehow time in answer  kaki-owar-e-ta. write--- ‘Somehow, Ken was able to finish writing answers in time.’ b. Sensei ga umaku zikanwari o teacher  well class.schedule  kumi-{naos-e/oe-rare}-ta. make-{-/-}- ‘The teacher was able to {rearrange/finish making} the class schedule well.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

122

  c. Ken ga taimingu yoku uta o utai-das-e-ta. Ken  timing good song  sing--- ‘Ken was able to start to sing a song with good timing.’

The stative case-marking pattern is not available if V2s appear in isolation, i.e. as main verbs, because they are nonstative verbs. This shows that the dativenominative case-marking pattern in (36b) is sanctioned by the stative potential verb created via potential verb formation. In the potential verb constructions, the grammatical function of the arguments is not altered even if their case-marking pattern is changed. This is evidenced by the fact that the initial experiencer argument, whether it is marked with nominative or dative case, can be the antecedent of subject-oriented reflexive zibun, as in (38). (38)

a. Keni ni zibuni no kodomo Keni  selfi  child ‘Ken can scold his children.’

ga 

sikar-e-ru. scold--

b. Keni ga zibuni no kodomo Keni  selfi  child ‘Ken can scold his children.’

o 

sikar-e-ru. scold--

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Furthermore, the formal noun koto (with a genitive case marker) can be added to the theme argument, but not the experiencer argument, regardless of the casemarking patterns, as in (39). (39)

a. Ken (*no koto) ni Eri (no koto) ga sikar-e-ru. Ken ( fact)  Eri ( fact)  scold-- ‘Ken can scold Eri.’ b. Ken (*no koto) ga Eri (no koto) o sikar-e-ru. Ken ( fact)  Eri ( fact)  scold-- ‘Ken can scold Eri.’

The fact shows that the theme argument counts as an object regardless of whether it is marked with accusative or nominative case. The data show then that no promotion of an object into a subject takes place in potential verb formation, that is, that potential verbs invoke no A-movement moving one argument across another when they effect a change in case marking. The fact that the potential morpheme changes case frames without invoking A-movement shows that the applicability of long-distance passivization to V2 is not determined by the constraint on A-movement.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

123

5.4.2 A case account Syntactic V-V compounds show the same contrast in acceptability observed for long-distance passivization even when case-marking change is effected with no A-movement. It is a commonly accepted view that passivization is induced by the passive morpheme absorbing the case feature to which it attaches (Baker, Johnson, and Roberts 1989). If this is the case, it is reasonable to postulate that the applicability of long-distance passivization is determined by the case properties of V2, to which the passive morpheme is attached, i.e. when V2 rather than V1 case-licenses an object of V1, long-distance passivization may be invoked. In light of this consideration, I propose that long-distance passivization may apply to V2 bearing an accusative-case feature, and further, that the case frames of the two verbs are merged into one, as in (40).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(40) [[ V1] -V2 ]

!

[[ V1]-V2 ]

When V1 is a transitive verb, the passivizable compound verb has two verbal heads that can potentially case-license an accusative object, i.e. V1 and V2. In (40), the accusative-case feature of V1 is eliminated under matching with the accusative-case feature of V2. This means that the accusative feature of V2 is used to case-license a direct object. On the other hand, the dative-case feature of ditransitive V1 is transferred to V2. Under the present account, V2s are responsible for accusative case marking on the object in passivizable control V-V compound constructions. The proposed analysis gains support from the facts regarding the syntactic V-V compound noritukusu. This compound verb can take an accusative argument even when the lower main verb cannot mark its object with accusative case, as observed by Yumoto (2005). (41)

a. Mari ga kooen no norimono {ni/o} nori-tukusi-ta. Mari  park  ride {/} take-- ‘Mari got on all the rides in the park.’ b. Mari ga kooen no norimono Mari  park  ride ‘Mari got on the rides in the park.’

{ni/*o} {/}

not-ta. take-

c. Kono kooen no norimono ga nori-tukus-are-ta. this park  ride  take--- ‘All the rides in this park were gotten on.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

124

 

The syntactic V-V compound nori-tukusu can undergo passivization, as in (41c).¹⁶ Note that the ni-marked argument of the verb noru ‘take’ cannot be promoted to a passive subject, while passivization can apply to the object of tukusu (when the verb is used as a main predicate), as shown in (42). (42)

a. *Kono kooen no norimono ga this park  ride  ‘Rides in this park were gotten on.’

nor-are-ta. take--

b. Iinkai ga giron {o/*ni} tukusi-ta. committee  discussion {/} exhaust-  ‘The committee held a thorough discussion.’ c. Giron ga tukus-are-ta. discussion  exhaust-- ‘A thorough discussion was held.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Since tukusu, but not noru, can mark its object with accusative case, it is plausible to say that the accusative case marking of the theme argument is sanctioned by V2 in the passivizable control V-V compound nori-tukusu. There is further evidence pointing to the conclusion that in passivizable control compounds, the upper control verb bears an accusative-case feature used for licensing a direct object even if the lower verb bears an accusative-case feature. To be concrete, observe that long-distance passivization may apply to the compound verb aruki-naosu. (43)

a. Ken ga kono rooka o Ken  this corridor  ‘Ken walked this corridor again.’

aruki-naosi-ta. walk--

b. Kono rooka ga Ken niyotte aruki-naos-are-ta. this corridor  Ken by walk--- ‘This corridor was walked again by Ken.’ Both upper verb naosu and lower verb aruku can mark their objects with accusative case when used as independent verbs, as in (44).

¹⁶ Kageyama (1993) claims that V2 tukusu assigns a theta role to an object of V1, since V2 characterizes the exhaustivity of the object, but Yumoto (2009) suggests that the fact should not be taken care of by theta-role assignment. The latter view is favored over the first, because the theta-role assignment analysis incorrectly predicts an expression like *basu o tukusu [bus  exhaust] ‘exhaust the bus’ to be well-formed.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -    (44)

a. Ken ga kono rooka Ken  this corridor ‘Ken walked this corridor.’ b. Ken ga kono kikai Ken  this device ‘Ken fixed this device.’

o  o 

125

arui-ta. walk- naosi-ta. fix-

Importantly, even though the verb aruku can mark its path argument with accusative case, this argument cannot be promoted to a subject by direct passivization, as shown in (45a). By contrast, the accusative argument taken by naosu can readily be rendered as a passive subject, as shown in (45b). (45)

a. *Kono rooka ga Ken niyotte this corridor  Ken by Lit. ‘This corridor was walked by Ken.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Kono kikai ga Ken niyotte this device  Ken by ‘This device was fixed by Ken.’

aruk-are-ta. walk--

naos-are-ta. fix--

The contrast in acceptability between (45a) and (45b) shows that naosu, but not aruku, bears an accusative-case feature that can be absorbed by the passive morpheme under passivization. Given that long-distance passivization is possible with aruki-naosu, it is apparent that in the control V-V compound aruki-naosu, the accusative-case feature possessed by the upper control verb naosu (V2) is used to case-license a direct object even when the lower main verb aruku (V1) has the potential to mark an argument with accusative case. Turning now to the question of unpassivizable V-V compounds, the V2s listed in (46) resist long-distance passivization, as noted by Kageyama (1993) and Kishimoto (2009); the V2s in (46a) are raising predicates and the V2s in (46b) are control predicates. (46)

a. kakeru ‘be about to’, dasu ‘start’, sugiru ‘exceed’ (Raising) b. tukeru ‘get used to’, sobireru ‘fail’, askiru ‘get bored’, okureru ‘delay’, kaneru ‘cannot’, sokoneru ‘fail’, sokonau ‘fail’ (Control)

The raising verbs in (46a) do not have the ability to case-license accusative arguments, as shown by the fact that accusative case marking is not available for an object when noru ‘take’ is combined with a raising verb. (47)

a. Ken ga basu {ni/*o} Ken  bus {/} ‘Ken was about to take a bus.’

nori-kake-ta. take-..-

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

126

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. *Basu ga Ken ni nori-kake-rare-ta. bus  Ken by take-..-- ‘A bus was about to be taken by Ken.’ The impossibility of accusative marking on the object in (47a) is naturally expected if V2 in raising V-V compounds does not have an accusative-case feature. As argued by Kishimoto (2009), this fact follows from Burzio’s generalization (Burzio 1982). Burzio’s generalization states that if a verb does not assign a theta role to its subject, it does not assign accusative case to its object. Raising V2s do not take subjects to which they assign theta roles, and accordingly, they do not bear accusative-case features. It also follows from this that long-distance passivization does not apply to raising V-V compounds owing to the absence of an accusative feature on V2, as shown in (47b). Some unpassivizable control V2s (such as akiru ‘get bored’, okureru ‘delay’, and owaru ‘end’) receive a similar account, since they are intransitive (i.e. they are intransitive when used as independent verbs). These intransitive control V2s are naturally expected not to undergo passivization owing to the absence of an accusative-case feature on the V2s (Kishimoto 2009). (In control V-V compound constructions with intransitive V2s, as well as raising V-V compound constructions, it can be assumed that transitive V1 rather than V2 licenses accusative case on an object.) Note, however, that control V2s like sokoneru ‘fail’ and sokonau ‘fail’ resist passivization even though they are transitive (i.e. these verbs can be used as transitive main verbs). There is evidence suggesting that some unpassivizable control V2s can license accusative case on an object, as seen in (48a). (48)

a. Ken ga ip-pon basu {ni/o} nori-sokonat-ta. Ken  one- bus {/} take-- ‘Ken failed to take a bus (=Ken missed one bus).’ b. *Basu ga nori-sokone-rare-ta. bus  ride--- Lit. ‘The bus was failed to take’.

Recall that noru ‘take’ does not license an accusative object. Thus, when the object is marked with accusative case, its case marking must be licensed by the V2 sokoneru ‘fail’. Nevertheless, passivization fails to apply to sokoneru, as shown in (48b). At this point, it is important to see that V-V compounds comprising the V2s listed in (46b) describe uncontrollable events (i.e. events over which the subjects

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

127

have no control).¹⁷ This property of unpassivizable control V-V compounds differs from that of passivizable control V-V compounds. The difference is confirmed by a contrast in acceptability between passivizable and unpassivizable control verbs with regard to imperative formation.¹⁸ (49)

a. Hon o yomi-{naosi/tukusi} nasai! book  read-{/}  ‘Read the books {again/thoroughly}!’ b. *Hon o yomi-{tuke/sokone/aki/kane} nasai! book  read-{..//./}  Lit. ‘{Get used to/Fail to/Get bored/Do not be able to} read the books!’

Given the semantic deviance of (49b), which stands in contrast to the acceptability of (49a), it can be stated that the control verbs in (49b) describe uncontrollable events. In the V-V compound construction in (48a), the subject cannot be a participant having control over the described event. As noted by Kageyama (2009b) and others, transitive verbs can be passivized only if the subjects are conceived of as participants having control over the described events. If so, (48b) is expected to be unacceptable. Since the transitive V2 sokoneru ‘fail’ cannot be passivized just like intransitive V2s ¹⁷ The V-owaru ‘end’ in (i) can be taken to express a controllable event while taking an accusative object.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(i)

Ken ga kono hon o Ken  this book  ‘Ken stopped reading this book.’

yomi-owat-ta. read--

Although Kageyama (1993) does not list owaru ‘end’ as an unpassivizable V2, this verb resists longdistance passivization, unlike oeru ‘finish’. (ii) Kono hon ga {yomi-oe-rare-ta/*yomi-owa-rare-ta}. this book  {read---/read---} Lit. ‘This book was {finished/stopped} reading’. Note that the verb owaru resists passivization even when it is used as a main predicate, which stands in contrast to oeru ‘finish’. (iii) Zyugyoo ga sukosi hayame-ni {oe-rare-ta/*owa-rare-ta}. class  a.little earlier {finish--/end--} Lit. ‘The class was {finished/stopped} a little earlier.’ This peculiar behavior of owaru presumably comes from the fact that the verb is originally an intransitive verb, although it can now take an accusative object. Given that the accusative object does not act like an ordinary object, it can be postulated that the V2 owaru fails to undergo long-distance passivization on the grounds that the verb does not have an ordinary accusative feature that can be absorbed by the passive morpheme. ¹⁸ The same contrast is observed when the V-V compounds are embedded under an intentional predicate like tumori-da ‘intend’. (i)

a. Ken ga hon o yomi-{naosu/tukusu} Ken  book  read-{/} ‘Ken intends to read the books {again/thoroughly}.’

tumori intention

da. 

b. *Ken ga hon o yomi-{tukeru/sobireru/akiru/kaneru} tumori da. Ken  book  read-{..//./} intention  ‘Ken intends to {get used to/fail to/get bored/cannot} read the books.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 

128

like okureru ‘delay’, it should be apparent that there are cases where long-distance passivization is not applicable to control V-V compounds even if an object is assigned accusative case by V2.¹⁹ It can be stated here that control V-V compounds describing uncontrollable events do not undergo long-distance passivization, independently of whether their V2s have the ability to assign accusative case.²⁰ Let us now proceed to discuss where a dative-case feature comes from in passivizable control V-V compounds. Under the present view, if V1 is a ditransitive verb, V2 inherits the dative-case feature of V1. A compound verb like atae-naosu ‘give again’ provides corroboration for this claim. Note that with atae-naosu, the dative as well as the accusative argument can be promoted to a passive subject, as seen in (50b–c). (50)

a. Oya ga kodomo ni omotya o parent  child  toy  ‘The parents gave the child the toy again.’

atae-naosi-ta. give--

b. Kodomo ga omotya o atae-naos-are-ta. child  toy  give--- ‘The child was given the toy again.’ c. Omotya ga kodomo ni atae-naos-are-ta. toy  child  give--- ‘The toy was given to the child again.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

When a ditransitive verb appears as V1, V2 bears the dative-case feature to license an indirect argument. Thus, (50c) is legitimate. This dative-case feature originates ¹⁹ One reviewer has drawn attention to this fact. There might be speaker variation as to what type of control verb allows accusative case marking on the object when V1 is a verb like noru ‘take’. The same restriction on the case-marking change applies to the tough-constructions involving  and potential verbs. This shows that the (un)passivizability of control verbs has a tight correlation with the distribution of case-marking changes as well. ²⁰ Although syntactic V-V compounds where V2 is wasureru ‘forget’ are not consonant with imperative formation, they allow long-distance passivization. (i)

a. *Kono nezi o sime-wasure nasai! this screw  tighten-  ‘Forget to tighten the screw.’ b. Kono nezi ga sime-wasure-rare-ta. this screw  tighten--- Lit. ‘This screw was forgotten to tighten.’

The unacceptability of (ia) shows that forgetting to do something is not an intentional act. Nevertheless, (ib) carries the implication that the tightening of the screws (i.e. the event described by V1) is possible if the agent is attentive enough. Perhaps, this implication makes it possible to conceive the event described by the compound verb as controllable; accordingly, long-distance passivization is applicable to V2 wasureru. Apparently, unpassivizable V-V compounds formed from V2s in (46b) do not carry this kind of implication. It should also be mentioned that some uncontrollable verbs allow negative imperatives, and thus a contrast in acceptability is observed between *Tori-sokone nasai! ‘Fail to take (it)!’ and Tori-sokoneru na! ‘Don’t fail to take (it)!’ This fact seems to indicate that there is an intermediate level of controllability in which one cannot choose to do something but one can still prevent it from happening, as suggested by John Haig (pers. comm.).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

129

from V1. This is confirmed by the syntactic behavior of naosu ‘fix, repeat’, which can take a ni-marked argument even when it serves as a control verb, as in (51) (see Yumoto 2005). (51)

Kare ga nihongo no ronbun o eigo ni kaki-naosi-ta. he  Japanese  paper  English to write-- ‘He translated his paper from Japanese to English.’

The ni-marked result phrase eigo ni ‘into English’ in (51) must be selected by the upper V2 because kaku ‘write’ cannot select it, as shown in (52). (52)

Kare ga nihongo no ronbun o eigo ni {naosi-ta/*kai-ta}. he  Japanese  paper  English to {fix-/write-} Lit. ‘He {changed/wrote} his paper from Japanese to English’.

The verbs naosu and kaku show distinct behavior with regard to passivization (when used as main verbs). The examples in (53b–c) show that the accusative object of the verb naosu can be turned into a passive subject, but the dative argument cannot.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(53)

a. Ken ga ronbun o eigo ni Ken  paper  English to ‘Ken translated his paper into English.’

naosi-ta. fix-

b. *Eigo ga (Ken niyotte) ronbun English  (Ken by) paper ‘English was translated his paper by Ken.’

o 

c. Ronbun ga (Ken niyotte) eigo paper  (Ken by) English ‘The paper was translated into English.’

ni to

naos-are-ta. fix-- naos-are-ta. fix--

The impossibility of promoting the ni-marked argument to a passive subject in (53b) is presumably due to the ni marking on the argument Eigo ‘English’ serving as a postposition—the analog of to/into. Since passivization is not allowed in (53b), it is easy to see that in the passivizable V-V compound in (50a), which takes a ditransitive verb as V1, V2 inherits the dative-case feature from V1, so that the dative argument can be promoted to subject by long-distance passivization applying to V2, as in (50b–c).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

130

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

5.5 Delexicalization Many works on syntactic V-V compounds (e.g. Kageyama 1993; Koizumi 1999; Yumoto 2005), explicitly or implicitly, take both V1 and V2 to be lexical verbs regardless of whether they belong to the control or the raising class. There are also works suggesting that some V2s have the status of functional categories (e.g. Fukuda 2012; Nishiyama and Ogawa 2014; Oprina 2014). In this section, I suggest that whereas control V2s behave as lexical verbs, raising V2s are delexicalized, having the status of functional verbs. Delexicalization is a process of grammaticalization invoking a shift from a lexical to a functional category. In the case of syntactic V-V compounds, I suggest that some V2s are rendered as functional verbs, i.e. they are delexicalized while retaining their categorial status as verbs, involving a change from [+Verb, +Lexical] to [+Verb, -Lexical].²¹ This process is not detectable by way of morphology, for V2s retain their inflectional forms. As noted by Shibatani (1973), dasu has the sense of ‘take out’, but carries an aspectual meaning if used as V2 in syntactic V-V compounds. Some authors (e.g. Nishiyama and Ogawa 2014; Oprina 2014) take this semantic fact to indicate that V2 has the status of a functional category (i.e. a ‘delexicalized’ verb under the present perspective). Nevertheless, verbs could specify aspectual meanings while retaining their lexical/syntactic status, and thus this fact does not necessarily show that the aspectual V2s count as functional verbs rather than lexical verbs. There is a way of measuring whether or not delexicalization has taken place on V2. Evidence that the raising V2 has delexicalized verb status may be adduced from the fact that the raising class of compound verbs, but not the control class, allows potential verbs (suffixed with the potential affix -(r)e or -(r)are) to be embedded under them. (54)

a. Kare ga sono uta o umaku he  that song  well uta-e-{dasi/kake}-ta. sing--{/..}- ‘He {started/was about} to be able to sing that song well.’

²¹ In grammaticalization studies, the term ‘delexicalization’ is often used as (nearly) synonymous with ‘grammaticalization’, which may induce a number of different kinds of grammatical and semantic changes on lexical words (cf. Brinton and Traugott 2005). In this chapter, ‘delexicalization’ refers to a more specific process that invokes a loss of the lexical properties of words. The central claim is that lexical words may be rendered as functional words while retaining their categorial status (e.g. a lexical verb may turn into a functional verb) independently of whether or not they undergo a shift in meaning. Note that many, if not all V2s carry bleached meanings, which are different from the meanings expressed by their main verb counterparts, even when they are not delexicalized (in the intended sense).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

131

b. *Kare ga sono uta o umaku uta-e-{oe/owat}-ta. he  that song  well sing--{/}- ‘He {finished/ended up} being able to sing that song well.’ Potential verbs are generally prohibited from standing as complements to lexical predicates that select verbs as their complements, i.e. the direct embedding of potential verbs under lexical predicates is not allowed. Note that the potential affix has the ability to change the case-marking pattern of a predicate to which it attaches. Since a number of different case-marking patterns are derived in combination with tense, I presume that the potential affix has the property that it needs to be connected to finite T with no intervening lexical predicate for its occurrence to be sanctioned. Consequently, it is not possible to embed potential verbs under the causative sase and the desiderative tai, as exemplified in (55). (55)

a. *Sensei ga kodomo o hasir-e-sase-ta. teacher  child  run--- ‘The teacher made the child be able to run.’ b. *Kodomo ga hasir-e-ta-i. child  run--want- ‘The child wants to be able to run.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Then, the facts of potential verb embedding in (54) fall into place, given that raising verbs, but not control verbs, serve as functional verbs. (56)

a. [TP [vP2 [vP1 . . . V1- ] V2 [+Lexical] ] T ] b. [TP [vP2 [vP1 . . . V1- ] V2 [-Lexical] ] T ]

(Control) (Raising)

The contrast in acceptability between (54a) and (54b) depends on whether the potential affix is linked to T with no intervening lexical predicate. In the control construction (54a), the intervening V2 is a lexical predicate (specified as [+Lexical]), so that the potential predicate V1 is not directly associated with the tense, as illustrated in (56a). In the raising construction (54b), the potential predicate V1 is directly linked to T with no intervening lexical verb, as in (56b). This analysis taking raising V2s to bear the feature [-Lexical] is reasonable, since the same kind of behavior is observed for functional elements such as the modal soo-da ‘likely’, the negative nai ‘not’, and the polite marker masu, which are allowed to intervene between a potential verb and tense, as in kak-e-soo-da ‘likely to be able to write, kak-e-nai ‘not be able to write’, and kak-e-masi-ta ‘was able to write’. Note that raising predicates can be turned into potential forms, as in hasiridase-e-ru ‘be able to start to run’. This fact suggests that, despite their functional properties, raising V2s possess categorial status as verbs, and thus have the feature

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

132

 

[+Verb, -Lexical]. Then, the generalization is that when V2 is a delexicalized verb (i.e. a functional verb) specified as [+Verb, -Lexical], it can take a potential verb as its complement. In regard to the question of why raising verbs are delexicalized, unlike other predicates taking verbs as their complements, I suggest that delexicalization is motivated by the fact that raising verbs do not select any argument, while specifying aspectual meanings. Control verbs are not delexicalized and retain the status of lexical verbs, because they select subjects. Evidence in favor of this view may be found in the fact that V2s in syntactic V-V compounds display the same syntactic behavior as auxiliary verbs taking verbs in the gerundive te-form as complements (like -te iku/kuru/miru [-te //] ‘go/come/see -ing’), which are V-V complexes of Type 4 in Kageyama’s taxonomy (Chapter 2, this volume). To be concrete, the examples in (57) show that when potential predicates appear in the te-complements selected by auxiliary verbs, a difference in acceptability arises, depending on the choice of auxiliary verb.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(57)

Ken ni tanin no sinri ga Ken  others  mental.state  {ki-ta/*mi-ta}. {-/-} ‘Ken {came/tried} to read others’ mental states.’

yom-e-te read--

The difference in acceptability observed in (57) accrues from the fact that kuru ‘come’, which specifies an aspectual meaning, is a raising predicate, and miru ‘try’, a control predicate. This difference can easily be checked by embedding a clausal idiom or an inanimate subject. (58)

a. Kono mise de mo kankodori ga nai-te {ki-ta/*mi-ta}. this store at also cuckoo  sing- {-/-} ‘There {came/*tried} to be fewer customers at this store as well.’ b. Ame ga hut-te {ki-ta/*mi-ta}. rain  fall- {-/-} ‘It {came/*tried} to rain.’

The data in (57) and (58) show that when the auxiliary verb counts as a control predicate, it does not allow a potential predicate to appear as its complement. In this connection, note that the motion verbs kuru ‘come’ and iku ‘go’ are ambiguous between the literal meaning of physical motion and the bleached meaning of gradual change when they are combined with a gerundive verb (see Chapters 2 and 6 of this volume for details). In (59a), the verb kuru takes a goal argument on its own, as it is used in its literal meaning.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -    (59)

a. Ken ga gakkoo ni nimotu o Ken  school  luggage  ‘Ken brought his luggage to the school.’ b. *Ken ga gakkoo ni nimotu Ken  school  luggage ‘Ken held the luggage to the school.’

mot-te hold-

o 

133

ki-ta. come-

mot-ta. hold-

The locative argument in (59a) must be selected by kuru because the main verb motu ‘hold’ does not take a locative argument, as seen in (59b). This shows that the verb kuru in its use of (59a) is a lexical verb and does not have the status of a delexicalized predicate. This motional auxiliary verb does not allow a potential predicate to occur as its complement, as shown in (60). (60)

*Ken ga gakko ni nimotu o mot-e-te ki-ta. Ken  school  luggage  hold-- come- Lit. ‘Ken came to school while being able to hold his luggage.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(60) differs minimally from (59a) in the choice of embedded predicate. Given that only a delexicalized predicate can take a potential predicate as its complement, the unacceptability of (60) shows that the motional auxiliary verb kuru is not delexicalized. By contrast, when an aspectual rather than a motional meaning is expressed by an auxiliary verb taking a te-complement, a potential predicate can appear as its complement, as shown in (61). (61)

Rensyuu no okage de watasi mo sukosizutu umaku practice  thank with I also gradually well oyog-e-te ki-ta. swim-- - ‘Thanks to practice, I am gradually beginning to swim well.’

This fact suggests that the auxiliary verb kuru carrying an aspectual meaning functions as a delexicalized (raising) predicate.²² Under the present view, the failure of the motion auxiliary verb kuru to undergo delexicalization is naturally expected, because it is conceived of as conveying a substantive lexical meaning, while selecting a locative argument. In the present perspective, raising predicates are potentially qualified to undergo delexicalization, since they specify aspectual meanings and do not select ²² V2 iku ‘go’ can take a potential predicate as V1 when it expresses an aspectual meaning, as in (i). (i)

Ronbun ga tugitugi-to kak-e-te paper  one.after.another write-- Lit. ‘(I) became able to write papers one after another’.

it-ta. -

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 

134

arguments. Nevertheless, it is also true that some raising predicates retain their lexical status. As discussed in Section 5.4.1, tough-adjective constructions have raising structures when they take a nominative-accusative case-marking pattern. The impossibility of embedding a potential predicate under nikui in (62) shows that this raising adjective is not delexicalized, perhaps because it has a substantive lexical content, i.e. the meaning of ‘difficulty’. (62)

Ken ga hon o {yomi/*yom-e}-niku-i. Ken  book  {read/read-}-- ‘It is hard for Ken to {read/be able to read} books.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The fact suggests that whether or not a given predicate is identified as being delexicalized cannot be determined solely by the distinction between ‘raising’ and ‘control’, although the possibility of delexicalization coincides with the ‘control/ raising’ distinction in the case of syntactic V-V compounds. In short, just as lexical predicates do not allow potential verbs to appear as their complements, so too V2s in the control class, which retain the status of lexical verbs, cannot take potential verbs as their complements. By contrast, raising predicates, whether they are second verbs in syntactic V-V compounds or auxiliary verbs taking te-complements, do allow potential verbs to appear as their complements, since they serve as delexicalized functional verbs.²³ Incidentally, although Fukuda (2012) stipulates that the four aspectual verbs hazimeru ‘begin’, owaru ‘end’, oeru ‘finish’, and tuzukeru ‘continue’ are classified as functional categories, the facts discussed here indicate that they are not identified as functional categories in a unitary manner.

5.6 The peculiar raising verb sugiru Syntactic V-V compounds formed on the verb sugiru ‘exceed’ fall into the raising class according to the diagnostic tests for raising and control (Kageyama 1993; Yumoto 2005; Kishimoto 2009). Nevertheless, syntactic sugiru-compounds

²³ Causativized verbs, which take lexical verbs (referring to controllable events) as their complements, allow us to distinguish delexicalized raising predicates from fully lexical control predicates, as in (i). (i)

a. *Sensei ga kodomo ni guraundo o teacher  child  ground  hasiri-{kake-sase/das-ase}-ta. run-{..-/-}- ‘The teacher made the child {be about/start} to run on the ground.’ b. Sensei ga kodomo ni guraundo o hasiri-{oe-sase/owar-ase}-ta. teacher  child  ground  run-{-/-}- ‘The teacher made the child {finish/stop} running on the ground.’

The difference in acceptability between (ia) and (ib) suggests that a causative sentence cannot be constructed from a syntactic V-V compound where V2 is a delexicalized raising predicate.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -   

135

possess properties that are not shared with other raising V-V compounds. It is shown in the following that the verb sugiru ‘exceed’ appearing as V2 in syntactic V-V compounds behaves differently from other V2s. Compound verbs having sugiru as V2 can be either syntactic or lexical, depending on whether the verb sugiru carries the meaning of ‘exceed’ or ‘pass’. The compound verb tabe-sugiru ‘eat too much’ in (63a), where V2 has the meaning of ‘exceed’, is construed as a syntactic V-V compound, and toori-sugiru ‘pass by’ in (63b), where V2 has the meaning of ‘pass’, is a lexical V-V compound. (63)

a. Ken ga gohan o Ken  rice  ‘Ken eats too much rice.’

tabe-sugi-ru. eat--

b. Kuruma ga toori-sugi-ta. car  go.through-pass- ‘Cars passed by.’ The difference in their syntactic statuses can be checked by applying soo suru ‘do so’ replacement to the first verbs. Soo suru replacement confirms that tabe-sugiru in (63a) is a syntactic V-V compound, while toori-sugiru in (63b) is a lexical V-V compound.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(64)

a. Ken ga gohan o tabe-sugi-ta. Ken  rice  eat-- Mari mo [soo si]-sugi-ta. Mari also so do-- ‘Ken ate too much rice. Mari did so, too (=ate too much rice).’ b. Ken ga toori-sugi-ta. Ken  go.through-pass- #Mari mo [soo si]-sugi-ta. Mari also so do-pass- ‘Ken passed by. Mari did so, too (≠ passed by).’

The bipartite distinction between the syntactic and the lexical V-V compounds is further corroborated by the fact that the V2 sugiru expressing the sense of ‘exceed’ can be combined with any type of verb, while the V2 sugiru with the meaning of ‘pass’ can only be combined with a limited set of verbs. The verb sugiru used in syntactic V-V compounds in the sense of ‘exceed’ displays the syntactic behavior of a raising predicate, as a clausal idiom can be embedded under sugiru. (65)

Kono mise de kankodori ga naki-sugi-ta. this store at cuckoo  sing-- ‘The number of customers at the store has fallen exceedingly.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

136

 

The V2 sugiru ‘exceed’ has the status of a delexicalized predicate, and thus embedding of a potential verb under sugiru is tolerated. (66)

Konkai wa Ken ga umaku oyog-e-sugi-ta. this.time  Ken  well swim--- ‘Ken was able to swim exceedingly well this time.’

Syntactic compounds formed on sugiru ‘exceed’ take unique complement structures. While syntactic sugiru-compounds allow a durative adverb tuzukete ‘continuously’, which specifies the duration of an event, to modify the embedded clause, a frequency adverb syuu san-kai ‘three times a week’, which specifies the number of events taking place, is not permitted to modify the embedded clause, as in (67). (67)

Ken ga [{tuzukete/*syuu san-kai} koohii o Ken  [{continuously/week three-times} coffee  nomi]-sugi-ta. drink]-- ‘Ken drank too much coffee {continuously/three times a week}.’

On the other hand, verbal modifiers are allowed to modify the embedded main verbs, just like other raising compound verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(68)

Ken ga [{enpitu de/nanameni} sen o Ken  [{pencil with/diagonally} line  ‘Ken drew too many lines {with a pencil/diagonally}.’

hiki]-sugi-ta. draw]--

The data suggest that the complement clause of sugiru-compound verbs includes AspdurP, alongside verbal projections up to vP, but not AspfreqP.²⁴ The fact leads to the conclusion that syntactic sugiru-compounds have the vP complement structure given in (69).

²⁴ The second verb sugiru ‘exceed’ allows even a negator to precede it. The complement clause taken by sugiru can include NegP, projected below the AspdurP. Thus, a contrast in acceptability is observed between (ia) and (ib) with regard to the licensing of NPIs. ni mo ika-na(sa)]-sugi-ru. (i) a. Ken wa [doko Ken  anywhere to also go--- ‘Ken does not go anywhere too much.’ ni ika-na(sa)]-sugi-ru. b. *Ken wa mettani [daigaku Ken  rarely university to go--- ‘He rarely goes to the university very much.’ The NPI doko ni mo ‘to anywhere’ is selected by the verb iku ‘go’, and thus is licensed under the scope of the negator appearing in the complement clause. By contrast, since AspfreqP is not allowed in the embedded clause, the frequency NPI mettani ‘rarely’ must be located in the upper clause, and hence is not licensed, falling outside the scope of negation.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

 -    (69)

vP2 VP2 AspdurP vP1 SBJ

OBJ

v2 V2

Aspdur v'1

VP1

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

137

v1 V1

As discussed in Section 5.3, the complement clauses of raising V-V compounds include the aspectual projections AspfreqP and AspdurP above the verbal projection vP-VP for V1. Syntactic sugiru-compounds possess complement structures different from those constructed from other raising compounds in that AspfreqP licensing the occurrence of frequency adverbs is not projected. Nevertheless, syntactic sugiru-compounds fall into the raising class, and the V2 sugiru has the status of a delexicalized predicate. Note further that the excessive sugiru ‘exceed’ can combine with a noun or an adjective to form a complex compound as long as the preceding predicative element expresses a certain kind of degree (Martin 1975). This property is not shared with other V2s (e.g. kakeru ‘start’, tukusu ‘exhaust’). (70)

a. Ken wa {kodomo/karu/sizuka/tabe}-sugi-ru. Ken  {child/light/quiet/eat}-- ‘Ken is {too childish/too light/too quiet/too big an eater}. b. *Ken wa {kodomo/omo/sizuka}-tukus-u. Ken  {child/heavy/quiet}-- Lit. ‘Ken is exhaustively {a child/heavy/quiet}.

The examples in (70) illustrate that ordinary V2s in compound verbs impose a categorial restriction such that their preceding predicate is confined to a verb, but the V2 sugiru does not. In fact, sugiru appearing as V2 in syntactic compounds is the only predicate that allows elements other than verbs to be combined with it (Sugioka 1986; Himeno 1999; Yumoto 2009).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

138

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

5.7 Concluding remarks This chapter has discussed some notable properties of syntactic V-V compounds in Japanese (falling into Type 3 in the classification of Chapter 2, this volume), providing some fresh empirical data. On the basis of the syntactic behavior of various adjuncts, it has been shown that syntactic V-V compounds take distinct complement structures, depending on whether they are classified as the raising or the control type. It has been argued that long-distance passivization on V2 is available when V2 bears an accusative-case feature that case-licenses a direct object, and that long-distance passivization does not apply to raising V2s, which do not bear accusative-case features. Furthermore, it has been suggested that whether or not control V-V compounds can undergo long-distance passivization is also determined according to whether they express a controllable event. Finally, it has been shown that syntactic compound verbs formed on sugiru ‘exceed’ belong to the raising class but take complement structures distinguished from those possessed by other raising V-V compounds. The discussion has brought several facts to light. First, while syntactic V-V compounds form morphologically tight units, they involve complementation syntactically, i.e. they have analytic syntactic structures, which stand in contrast to lexical V-V compounds, which behave as single predicates syntactically. Secondly, the categorial status of V2 differs depending on the type of V-V compound. Control V2s are categorized as lexical verbs, but the components of compound verbs are subject to a reanalysis, in the sense that control V2s inherit certain case properties of V1s. Thirdly, V2s in raising compounds are delexicalized despite their verbal inflection, which indicates that raising V2s have status as a functional verb. The interesting fact is that delexicalization may apply to V2s even when they retain their morphological characteristics (cf. Hopper and Traugott 2003; Brinton and Traugott 2005). The discussion reveals that in Japanese (and possibly in other languages that have compound verbs as well), the V2 position, but not the V1 position, provides a syntactic context to which a number of grammatical processes such as delexicalization may apply.

Acknowledgments Part of the material in this chapter was presented in Linguistics and Language Conference (LILA 2014) (June 2014), Formal Approaches to Japanese Linguistics 7 (FAJL7) (June 2014), GLOW in Asia X (May 2014), and a NINJAL colloquium (April 2014). I am grateful to Mamoru Saito, Serkan Şener, Balkız Öztürk, K. A. Jayaseelan, Keiko Murasugi, R. Amritavalli, Hisatsugu Kitahara, Dylan Tsai, Yoshiki Ogawa, Noriko Yoshimura, Haruo Kubozono, Hisashi Noda, John Whitman, Edith Aldridge, and the audiences of the conferences for helpful discussions and suggestions. I am thankful to Taro Kageyama and the two anonymous reviewers for their comments on an earlier version of the chapter. Needless to say, the author is solely responsible for any remaining inadequacies.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

6 The semantic differentiation of verb-te verb complexes and verb-verb compounds in Japanese Yo Matsumoto

6.1 Introduction Japanese has two different formal types of complex predicates involving two verbs: V-te V complex predicates (Type 4 in Chapter 2, this volume) and V-V compound verbs (Types 1, 2, 3). In this chapter, I discuss the formal and functional nature of the former in comparison to the latter. V-te V complexes are exemplified in (1), while V-V compounds are exemplified in (2).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

V-te V complexes a. Kare wa baggu o dokoka he  bag  somewhere ‘He took a bag somewhere.’

ni 

mot-te hold-

it-ta. go-

b. Boku wa kodomo ni hon o yon-de age-ta. I  child  book  read- - ‘I read a book for the child (I gave the child the benefit of reading a book).’ (2)

V-V compounds: Kare wa baggu o dokoka he  bag  somewhere ‘He took a bag away to somewhere.’

ni 

moti-sat-ta. hold-leave-

(1a) and (1b) exemplify different uses of the V-te V complexes. (1a) is a case of “complex motion predicates” (Matsumoto 1996, 1997, 2017a, 2018), and (1b) is a case of benefactive use of -te age(-ru) (Shibatani 1978, 2003; Yamada 2004; Sawada 2014, etc.). V-V compounds also exhibit different types (see Kageyama 1993, 2013, Chapter 2, this volume; Kishimoto Chapter 5, this volume; Matsumoto 1996). The particular example given in (2) is semantically similar to that in (1a) and is cited here for comparison. The nature of V-V compound verbs

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

140

 

is examined extensively in other chapters and so I will discuss them only in contrast to the V-te V complexes (e.g. I will not discuss the distinction between Kageyama’s Types 1 and 2). For my own views on the nature of V-V compound verbs, see Matsumoto (1996, 1998) and Chen and Matsumoto (2018). The coexistence of two such different forms of multiverbal complexes in one language calls for the identification of semantic and functional differences with which formal difference is associated, which is the aim of this chapter. One crucial property of the V-te V complexes is that the first verb is in the nonfinite -te form. This form is often called gerundive (e.g. Martin 1975), and treated as a converbal verb form in typological literature (e.g. Alpatov and Podlesskaya 1995; Shibatani 2007b). The suffix -te has been historically related to perfective –tu, whose meaning is claimed to be reflected in the current uses of – te. -te triggers certain morphophonological changes to the verb form, depending on the final consonant of the verb stem to which it is suffixed (e.g. stem-final /r/ and /w/ are changed to /t/, leading to a gemminate, as in sit-te ( speaker’s “IN-group” person > hearer > others

Not all verbs of giving are deictically conditioned. Other verbs, such as atae(-ru) ‘give’, and watas(-u) ‘hand over’, are not deictic. Interestingly, nondeictic verbs of giving are found as V2s in lexical compounds, but deictic ones are not, as shown in (36). (36)

a. wake-atae(-ru) [divide-give-] ‘give (a divided portion) to’, kasi-atae(-ru) [lend-give-] ‘lend’, yuzuri-watas(-u) [yield-hand.over-] ‘yield over’, b. *wake-age(-ru) [divide-give-], *kasi-age(-ru) [lend-give-], *yuzuri-kure(-ru) [yield-give-]

Apparent counterexamples can be found with yar(-u) ‘give to a person other than the speaker’, a somewhat vulgar variant of age(-ru). This verb can occur in V2 in lexical compounds in (37).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(37)

mi-yar(-u) [look-give-] ‘give a glance at’, omoi-yar(-u) [think-give-]] ‘give thought to, consider’

However, these can be used for actions directed to the speaker, and so deictic meaning is absent in such cases. On the other hand, all of those deictic verbs of giving can occur as V2 in -te complexes to indicate the giving of benefit (Sakuma 1936; Inoue 1976; Shibatani 1978; Kuno 1986; Yamada 2004; Sawada 2014, etc.), but none of the nondeictic verbs of giving can, as shown in (38). Note that (38c) has only a sequential reading of ‘read and then gave’. (38)

a. Ken wa boku ni hon Ken  I  book ‘Ken read a book for me.’

o 

yon-de read-

kure-ta. -

b. Boku wa kare ni I  he  ‘I read a book for him.’

o 

yon-de read-

age-ta -

hon book

c. Boku wa kare ni hon I  he  book ‘I read a book and gave it to him.’

o 

yon-de read-

atae-ta give-

Other verbs are also deictic in that they code social deixis: the social differences relative to the speaker (see Levinson 1983). Honorific verbs, exemplified in (39), are such examples. These represent the speaker’s sense of respect toward either the subject or other argument of the verb (normally representing the hearer).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

160

 

(39)

ossyar(-u) ‘say’, nasar(-u) ‘do’, sasi-age(-ru) ‘give’, kudasar(-u) ‘give’, itadak(-u) ‘receive, eat’, gozar(-u) ‘be’, irassyar(-u) ‘be, come, go’, mair(-u) ‘come, go’, goran ni nar(-u) ‘look’

These honorific, socially deictic verbs do not participate in V-V compounds. In older Japanese, a presently obsolete honorific verb tamaw(-u) ‘give’ was used in syntactic V-V compounds (e.g. yomi-tamaw(-u), [read-(Hon)-] ‘read (Hon)’, but they are now replaced by -te complex forms like yon-de kudasar(-u) [read- (Hon)-].¹² In contrast, many of those honorific verbs occur in the V-te V complexes in order to form honorific variants of V-te V complexes, as shown in (40).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(40)

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

V-te ik(-u) V-te ku(-ru) V-te i(-ru) V-te ar(-u) V-te age(-ru) V-te moraw(-u) V-te kure(-ru) V-te mi(-ru) V-te mise(-ru)

V-te irassyar(-u) V-te irassyar(-u) V-te irassyar(-u) V-te gozar(-u) V-te sasi-age(-ru) V-te itadak(-u) V-te kudasar(-u) V-te goran ni nar(-u) V-te goran ni ire(-ru)

Note that the nonhonorific complexes in (40c, d, h, and i) do not have deictic verbs as V2. By taking the V-te V form, these nonhonorific complexes can have the same form as their socially deictic variants, which are required to be in this form. Thus, it is generally true that deictic verbs are excluded from lexical compounds but are preferably used as V2s of V-te V complexes. Those deictic verbs are different from nondeictic verbs in that they are not purely propositional in meaning. They do code propositional contents, given that they describe motion, shift in ownership, etc. However, these contents are perspectivized (described from the viewpoint of the speaker (Kuno 1986) or set in the interactional contexts of utterance (used to convey the speaker’s sense of respect or care to the subject or to the hearer).¹³ The schematic structures of V-te V complexes and lexical V-V compound verbs in terms of purely propositional vs. perspectivized/interactional contents are given in (41). ¹² There are sporadic examples of compound verbs used as honorific verbs. Examples are mesiagar(-u) [take-ascend-] ‘eat (Hon)’ and sasi-age(-ru) [hold.up-lift-] ‘give (to a higher person)’. In both cases the meanings of the whole verb are somewhat untransparent. It may seem that expressions such as o-yomi kudasar(-u) [Hon-read (Hon)-] ‘give (me) an honor of reading’ are instances of V-V compounds with honorific V2. But these are o-N V complexes, with a nominalized element as the first member, as evidenced by phenomena described in footnote 9. ¹³ Some of the V-te V complexes are interactional in another sense. The imperative forms of benefactive constructions are very often used to express the speech act of request (Mori 2016). Note that V-te kudasar(-u) [V- give.to.the.speaker-] has a special imperative form V-te kudasai, which is used exclusively to express a request.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    (41)

V0

161

V0 ‘perspectival/ interactional’

V0

V0

kakete

kuru

purely V0 propositional kake-

V0 agaru

The V1 of a V-te V complex is unspecified with respect to perspectivity, and a deictic verb can appear in that slot. (The same is true of the V1 of a syntactic compound verb.) There are, however, some V-te V complexes that have nothing to do with perspectival or interactional V2. The V-te V complexes which do not have deictic verbs (or have socially deictic variants) are V-te simaw(-u), V-te ok(-u), and V-te hosi(-i). Interestingly, these are predicates that have perfective/resultative -te and/or create a different-subject structure. Thus, the three factors in combination appear to account for why -te is selected in the attested V-te V complexes.

6.5.3 Summary and discussion

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Table 2 summarizes the findings in 6.4 and 6.5 and shows which of the three factors are relevant to each V-te V complex. It shows that none of the three is crucial by itself. However, each of the V-te V complexes has at least one of the factors that favor a V-te V complex.

Table 6.2 V-te V complexes and three factors favorable to the choice of their form different subject

perfective/resultative -te

deictic V2

-te ik(-u) -te ku(-ru) -te i(-ru) -te ar(-u)

* * * */✓

*/✓ */✓ */✓ ✓

✓ ✓ has a variant has a variant

-te ok(-u) -te simaw(-u) -te moraw(-u) -te kure(-ru) -te age(-ru)

* * ✓ * *

✓ ✓ ? ? ?

* * has a variant ✓ ✓

-te hosi(-i) -te mi(-ru) -te mise(-ru) syntactic compounds

✓ * * *

* ✓ ✓ */✓

* has a variant has a variant *

lexical compounds

*

*/✓

*

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

162

 

Different-subject structure is not allowed in V-V compounds but is allowed in V-te V complexes. Deictic verbs are not allowed in V-V compounds, but are typically used in the V2 of V-te V complexes. The perfective/resultative reading is not common in the V1s of V-V compounds but is typical of those of V-te V complexes. The separation of deictic information in the structure of a verbal complex can be made clearer in the embedding of one multiverbal complex in another. Lexical compounds can be embedded in the V-te V structure in the V1 position to form a V-V-te V structure, such as kake-agat-te ku(-ru) [run-ascend- come-]. However, it is not possible to embed a V-te V complex within a V-V lexical compound (since V1 cannot be a unit larger than a morphological word), and so it is not possible to say *kake-te ki-agar(-u) [run- come-ascend-]. This means that deixis is always at the end as far as lexical compound verbs and monoclausal V-te V complexes are concerned. The structure of kake-agat-te ku(-ru) is given in (42), together with the kinds of information coded. (42)

V0 V0

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

[V - V]-te purely propositional kake-agatte

V0 perspectival/ interactional

kuru

Deictic verbs code meanings that are not purely propositional but convey perspectival and interactional meanings. Therefore, it is natural that a position outside of the positions for purely propositional contents is assigned to such meanings. This observation is to some extent consistent with what has been claimed about the general tendencies in the order of elements in Japanese verbal structure. Semantically motivated layering of verbal structure can be seen generally in Japanese. Kageyama (2013) argues that lexical compound verbs in Japanese have a layered structure more complex than is assumed in this chapter, and those verbs representing some sort of aspectuality occur outside the combination of verbs with lexical contents (e.g. ne-sizumari-kaer(-u) [sleep-become.silentreturn-] ‘become utterly silent due to sleeping’). It has also been suggested that verb suffixes and other postverbal elements are ordered in semantically and pragmatically motivated ways (Nitta 1991; Minami 1993; Narrog 2010, etc.). Narrog argues that those elements that take a wider semantic scope tend to occur in an outside position (i.e. toward the end). One has to note that deictic verbs are not always at the end when biclausal V-te V complexes and syntactic compound verbs are involved, as shown in (43). In these examples, a V-te V complex is embedded as V1 in another V-te V complex or a syntactic compound.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

    (43)

163

a. hasit-te ki-hazime-ta. run- come-begin- ‘began to come running’ b. kare ni hasit-te ki-te he  run- come- ‘want him to come running’

hosi-i want-

Note that the V1 of a biclausal V-te V complex and a syntactic compound verb is the head of an embedded complement clause. This suggests that the tendency of deixis occurring at the end of a clause is observed strictly within a single clause, and the sentences in (43) are not counterexamples to this generalization.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

6.6 Concluding remarks The two kinds of multiverbal complexes are different morphologically, syntactically, and semantically. Morphologically tighter V-V compounds require samesubject relation between the two verbs and exclude perspectival or interactional meanings (except V1 in syntactic compounds). Loosely concatenated V-te V complexes allow different subjects, typically have perfective/resultative V1, and have V2 as a preferred slot for perspectival/interactional meanings. These observations suggest that language does not have more than one option meaninglessly; different multiverbal complexes serve different purposes.¹⁴ What kinds of implications do our findings have in the syntax and semantics of verbal complexes? The present findings from Japanese can be taken to show how different types of multiverbal complexes exhibit division of labor. One way in which multiverbal complexes can be different involves different degrees of formal and semantic integration. It is often argued that the degree of syntactic/morphological integration is correlated with semantic relatedness (e.g. Foley and Van Valin 1984). The Japanese facts presented are consistent with the generalization of the iconicity of form–meaning relation. Different-subject structures involve a less semantic integration than do same-subject structures. In addition, the combination of propositional and perspectivized elements in V-te V complexes constitutes a less integrated combination than the combination of purely propositional contents. Thus, morphologically tighter complexes can be said to represent semantically more integrated structures. The second way multiverbal complexes can be different can come from the semantics of different verb forms involved, as suggested by the perfective/resultative ¹⁴ The historical development of the division of labor between V-V compounds and V-te V complexes in view of the present findings is an interesting topic for study. The latter is a more recent option and, for the appearance of some of the complexes, see Miyaji (1981), Kojima (1999), Fukushima (2011), Mori (2016), and Aoki and Frellesvig (Chapter 3, this volume).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

164

 

tendency in V-te V complexes. This can be more clearly seen in other languages. Altaic languages have a rich set of forms available for nonfinite verbs, differing in such parameters as the temporal relationship of the nonfinite verb to the main verb, and the identity of subjects of the two verbs (see Nedjalkov 1997 for Evenki, and Janhunen 2012 for Mongolian). In Mongolic languages, such different forms of converbs can participate in forming different kinds of complex predicates (Janhunen 2012; Badema 2012), and the choice of forms reflects the temporal nature of the converb forms to some extent (see Badema 2012 for Khorchin and Baranova 2013 for Kalmyk; cf. Ibrahim 1995 for Uyghur and Uzbek). Finally, we have identified the sharing of subjects as an important difference that multiverbal complexes can exhibit. This parameter of the presence/absence of the same-subject requirement is an interesting, crosslinguistically valid parameter of multiverbal structures in general. A similar parameter has already been discussed in serial-verb constructions. Crowley (2002) discusses the distinction between same-subject serialization (‘hit kill’) vs. switch-subject serialization (‘hit die’). Aikhenvald (2006) argues that all serializing languages have same-subject serialization, and that they differ in whether they have “switch-function serialization” in addition. Languages that seem to lack switch-subject serialization include Kambera (Malayo-Polynesian; Klamer 1998: 279), Lavukaleve (Papuan; Terrill 2003: 373), and ǂHoan (Khoisan; Collins 2002), while those which allow it include Thai (Tai-Kadai; Thepkanjana 1986) and Sranan (Creole; Sebba 1987). Bril (2004) also examines what she calls “nuclear juncture” complex predicates (which correspond to compounds) in different Oceanic languages in terms of same-subject/ different-subject subtypes and identifies several languages that require Subject Sharing and those which do not. The former include Saliba (Malayo-Polynesian, Papuan Tip: see Margetts 1999: 102), Paamese (Malayo-Polynesian, Vanuatu: Crowley 1987), and Tahitian (Malayo-Polynesian, French Polynesia: Paia and Vernaudon 2004). Compounds in Mapudungun (isolate, Chile: Baker and Fasola 2009: 601–2) and Tepehua (Totonacan, Mexico: Watters 2017) also appear to be Subject Sharing, while those in Mandarin Chinese do not have such constraint (e.g. Li and Thompson 1981). The relation of the two Japanese multiverbal complexes to such parameters in typological literature is worth further investigation.

Acknowledgments This chapter is based on my paper presented at the NINJAL Conference on the Mysteries of Verb-Verb Complexes in Asian Languages. I would like to thank participants in the conference for their comments on this work. My special thanks are also due to Taro Kageyama, Kentaro Nakatani, and John Haig for their comments on an earlier draft of this chapter. This chapter also contains materials related to my work supported by JSPS KAKENHI (Grant Number 19H01264) and by the NINJAL project “Contrastive Studies of Japanese Prosody and Grammar: Verb Semantics.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

7 Verb-verb complexes in Irabu Ryukyuan Michinori Shimoji

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.1 Introduction This chapter examines the V-V complex structure of the Irabu dialect of Ryukyuan (henceforth Irabu), which is spoken on Irabu island, Okinawa, Japan. Ryukyuan is the only known sister language of Japanese, and the two languages constitute the Japonic Family. Irabu is a northwest variety of the Miyako Ryukyuan language, which belongs to the Southern Ryukyuan group of the Ryukyuan branch of the Japonic language family. Miyako Ryukyuan is characterized by a number of lexical, phonological, and grammatical innovations that distinguish it from other Southern Ryukyuan languages, and Miyako Ryukyuan is typologically quite distinct from Japanese and other Ryukyuan languages (see Shimoji 2010 for a summary of typological characteristics of Ryukyuan, including Miyako), but it is generally conservative in terms of the organization of V-V complex structures, retaining many major features that are still observed in Modern Japanese, as will be discussed in the remainder of the chapter. There is one detailed reference grammar of Irabu (Shimoji 2017). Like most other Ryukyuan languages, Irabu has no written tradition. The data used in this study are thus based exclusively on the spoken language. Irabu has five subvarieties, i.e. Irabu, Nakachi, Kuninaka, Nagahama, and Sawada. Our focus is on Nagahama. To the best of my knowledge, the generalizations presented in this chapter also hold true for all the Irabu varieties, and for most Miyako Ryukyuan dialects as well. While there are a number of lexical, structural, and semantic similarities between the Irabu V-V complexes and those of Japanese, there are also a number of divergences, which will be taken up and discussed later in this chapter. Two major types of Irabu V-V complexes are identified: phrasal V-V complexes, which further divide into Auxiliary Verb Constructions (AVCs) and Serial Verb Constructions (SVCs), and one-word V-V complexes, which further divide into syntactic compounds and lexical compounds. This chapter is organized as follows. Sections 7.2 and 7.3 are basic phonological and morphological descriptions. Section 7.4 gives an overview of the Irabu V-V complex structures, introducing the relevant terms and concepts to be discussed in later sections. In Sections 7.5 and 7.6, the description of one-word V-V complexes (compounds) will be given, comparing lexical and syntactic compounds. In

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

166

 

sections 7.7 to 7.9, the description of phrasal V-V complexes (SVCs and AVCs) will be given, comparing these constructions with each other and these two constructions on the one hand with one-word constructions on the other. In Section 7.10, the Irabu V-V complexes will be compared with Japanese V-V complexes (Types 1 to 4; see Chapter 2 of this volume for the classification of Japanese V-V complexes in Japanese).

7.2 Overview of Irabu phonology

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Irabu has five underlying vowel phonemes /a, i, u, e, o/, and eighteen consonant phonemes /p, t, k, b, d, g, f, v, s, (h), c [ts], z [dz], m, n, ž [z], r, w, j/. There is a predictable epenthetic phoneme /ï/, which is inserted after a consonant to fix impermissible phonotactic patterns. For example, the underlying //zmiz// ‘worm’ has the phonotactically impermissible word-initial heterosyllabic CC and the word-final obstruent coda. Hence /ï/ insertion operates to yield the surface /zïmizï/ (CV.CV.CV). Irabu has no lexical accent. The word-level prosody is characterized by a footbased alternating rhythm of tone features (/H/ and toneless; Shimoji 2009). Even though the prosodic organization is relevant in discussion of the distinction between a word and a phrase, I do not go into detail on this topic mainly owing to the limitation of the length of the chapter (see Shimoji 2009 and 2017 for details); the distinction between a word and a phrase can be made by referring to other criteria, which will be discussed in this chapter.

7.3 Verb morphology A verb inflects word-finally. I call the portion of the verb word other than the inflection a stem. A stem is minimally a root. A stem may be a compound and/or may be derived by affixation. Verbs are classified into two inflectional classes, finite and nonfinite verbs.

7.3.1 Finite verbs Table 7.1 shows the inflectional paradigm of finite verbs. Finite verbs inflect for one or more of the following three inflectional categories: negative polarity, tense, and mood. Finite verbs fall into (a) the optative forms and (b) the declarative forms. The optative forms indicate the speaker’s future-oriented modalities such as desiderative (e.g. ibi-baa ‘want to plant’), imperative (affirmative ibi-ru ‘(you) plant’ and negative ibi-rna ‘(you) don’t plant’), and intention (affirmative ibi-di ‘will plant’ and negative

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

167

Table 7.1 Finite inflection in Irabu Ryukyuan Form

Mood

Tense Polarity Class 1 ibi- ‘plant’

Declarative Unmarked nonpast past Realis

Optative

intentional imperative

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

desiderative

nonpast past

affm neg affm neg affm neg affm neg affm neg affm neg

ibi-r ibi-n ibi-tar ibi-t-tar ibi-r-m ibi-n-Ø ibi-ta-m ibi-t-ta-m ibi-di ibi-djaan ibi-ru ibi-rna ibi-baa

Class 2 tur- ‘take’

kak- ‘write’

tur-Ø tur-a-n tur-tar tur-a-t-tar tur-Ø-m tur-a-n-Ø tur-ta-m tur-a-t-ta-m tur-a-di tur-a-djaan tur-i-Ø tur-na tur-a-baa

kafï-Ø kak-a-n kafï-tar kak-a-t-tar kafï-Ø-m kak-a-n-Ø kafï-ta-m kak-a-t-ta-m kak-a-di kak-a-djaan kak-i-Ø kafï-na kak-a-baa

ibi-djaan ‘will not plant’). The declarative forms are used in a general statement (i.e. in a declarative sentence). The declarative forms comprise the unmarked form, which inflects for polarity and tense (e.g. affirmative non-past ibi-r ‘plant’ and negative non-past ibi-n ‘not plant’), and the realis form (e.g. ibi-r-m ‘(surely) plant’), which additionally indicates the speaker’s perceived certainty after tense marking. A finite verb occurs as the predicate of a main clause, but the declarative unmarked form may also occur as the predicate of an adnominal clause. The inflectional morphology differs according to the inflectional class of the verb stem, which is determined by the stem-final segment: Class 1 stems end in /i/ (e.g. ibi- ‘plant), while Class 2 stems end otherwise (basically in a consonant, e.g. tur- ‘take’, kak- ‘write’, etc.). Table 7.1 lists the finite inflection of the Class 1 stem ibi- ‘plant’ and the Class 2 stems tur- ‘take’ and kak- ‘write’. A Class 2 stem may be extended by the insertion of the ‘thematic vowel’ -a in order to further carry certain inflectional affixes such as the optative intentional (tur-a-di and kak-a-di as opposed to ibi-di).

7.3.2 Nonfinite verbs Converbs and gerunds are nonfinite verb forms which do not inflect as either declarative or optative. They indicate various kinds of clausal dependency, functioning like a conjunction marker. Table 7.2 is a partial list of nonfinite inflections. Converbs are found in adverbial subordinate clauses such as conditional (e.g. ibi(r)tigaa ‘if (x) plants’), purposive (ibi-ga ‘in order to plant’), and causal (ibi-(ri)ba ‘since

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 

168

Table 7.2 A partial paradigm of nonfinite inflections Form

Class 1 ibi- ‘plant’

Class 2 tur- ‘take’ Converbs causal ibi-ba/ibir-iba tur-i-ba conditional ibi-tigaa/ibir-tigaa tur-tigaa negative conditional ibi-dakaa tur-a-dakaa simultaneous ibi-ccjaaki/ibir-ccjaaki tur-ccjaaki purposive ibi-ga tur-ga Gerunds affirmative ibi-i tur-i-i negative ibi-da tur-a-da

kak- ‘write’ kak-i-ba kafï-tigaa kak-a-dakaa kafï-ccjaaki kafï-ga kak-i-i kak-a-da

(x) plants’), etc. The gerund heads a coordinate dependent clause, as illustrated in (1a), or functions as the lexical-verb component of a phrasal V-V-complex structure, as in (1b).¹

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

a. ba=a [buuz=zu ibi-i]=du 1= sugar.cane= plant-= ‘I planted sugarcanes, and went home.’

jaa=nkai house=

b. ba=a buuz=zu=baa 1= sugar.cane== ‘I have planted sugarcanes.’2

njaan-Ø]. .-

[ibi-i plant-

par-tar. leave-

As in the case of finite inflections, a Class 2 stem is extended by a thematic vowel when carrying certain inflectional affixes. The thematic vowel for nonfinite verbs may be -a or –i, depending on the individual inflectional affix that is attached to the stem. A stem with a thematic vowel will be called a ‘thematic stem’, whereas one without will be called an ‘athematic stem’.

7.3.3 Stem and inflection Table 7.3 lists the stem forms that are relevant to the description of V-V complexes in Irabu.

¹ The present chapter uses the following signs for different kinds of morpheme boundaries: ‘-’ (hyphen) for affix boundaries, ‘=’ for clitic boundaries, and ‘+’ for stem boundaries within a compound word. ² Note that the perfect auxiliary njaan is affirmative in meaning although negative in form (see Section 7.8.1 for details).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

169

Table 7.3 Stem forms for V-V complexes

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Phrasal V-V complex V₁ of a phrasal V-V complex (gerund) Class 1 ibi-i ibi- ‘plant’ plant- ‘plant:’ Athematic stem Class 2 tur-i-i tur- ‘take’ take-- ‘take:’ Thematic form Class 2 kak-i-i kak- ‘write’ write-- ‘write:’ Thematic form

One-word (compound) V-V complex V₁ of a syntactic V₁ of a lexical compound compound ibi+pazïmi-r ibi+kair-Ø plant+begin- plant+change- ‘begin planting’ ‘transplant’ tur+pazïmi-r take+begin- ‘begin taking’ Athematic form kafï+pazïmi-r write.+begin- ‘wrote’ Basic form Athematic stem

tur+kair-Ø take+change- ‘replace’ kacï+kair-Ø write.+change- ‘rewrite’ lnfinitive form

Class 1 stems have a single athematic form throughout any inflection and other morphological contexts such as V₁ of a (lexical/syntactic) compound verb, verbal noun stem, etc. Class 2 stems have both thematic and athematic forms depending on the morphological context in which they occur. While the /i/ thematic form is used for the V₁ stem (which further carries the gerund inflection -i), the athematic form is used for V₁ of a compound (lexical or syntactic). For Class 2 stems which end in /k/ or /g/, such as kak- ‘write’ and kug- ‘paddle’, a further distinction is made within the athematic form: the basic form, which is formed by replacing the root-final /k/ and /g/ by /f/ and /v/ respectively (/kak/ ! /kaf/), with the additional insertion of /ï/ after the stem-final obstruent (/kaf/ ! /kaf ï/), and the infinitive form, which is formed by replacing the root-final /k/ and /g/ by /c/ and /z/ respectively (/kak/ ! /kac/), with the additional insertion of /ï/ after the root-final obstruent (/kac/ ! /kacï/). For these Class 2 stems which have a formal distinction between the basic and infinitive stem forms, the basic form is used for V₁ of a syntactic compound and also a lexical compound, whereas the infinitive form tends to be limited to a lexical compound.

7.4 Irabu V-V complexes 7.4.1 An overview of Irabu V-V complexes This section sets out to describe V-V complexes in Irabu. As summarized in Table 7.4, the major division is drawn between phrasal V-V complexes and one-word

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 

170

Table 7.4 Irabu V–V complexes Type

productivity of V₂

focus marking on V₁

+

causative embeddability marking on V₁ of phrases in V₁ + +

auxiliary verb construction (AVC) serial verb construction (SVC) syntactic compound lexical compound

–

–

+

Æ

+ –

+ –

+ –

– –

+

V-V complexes. Phrasal V-V complexes comprise the AVC and SVC. One-word V-V complexes comprise the syntactic compound and the lexical compound. The four subclasses of V-V complexes are characterized and compared according to four features which indicate the presence or absence of the postlexical nature of a V-V complex, and each feature will be discussed later in this section. The four types should be understood as prototypes, and we encounter constructions which are somewhere between AVCs and SVCs, and which are somewhere between syntactic and lexical compounds.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.4.2 Phrasal vs. one-word V-V complexes Before examining the two major types of V-V complexes, let us briefly examine the most prominent difference between the two constructions, which concerns whether focus marking on V₁ is possible. This criterion, a manifestation of the lexical integrity principle, is crucial when examining the wordhood of a construction, since a word is never broken up by the insertion of a focus marker whereas a phrase can be. Irabu has focus-marking morphology, with three dedicated focus markers: declarative focus marker du, yes-no interrogative focus marker ru, and content interrogative focus marker ga. They are all clitics. The following example is in response to a question like ‘Who was crying?’, with the focus on the subject NP, which is focus-marked by the declarative focus du. (2)

Argument focus unu jarabi=nu=du nak-i+u-tar. this child== cry-+- ‘THIS CHILD was crying.’ (in response to ‘Who was crying?’)

As a general principle, focus marking is on the leftmost constituent of focus domain (Shimoji 2017). When the predicate of a clause is focused, as in (3), where the predicate is a phrasal V-V complex, the leftmost constituent of the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

171

predicate is focus-marked. In (3a) it is the gerund form (V₁), and in (3b) it is the adverb zïïtii ‘continuously’. Note that in (3c), where the leftmost constituent is a compound verb, the whole compound, not the V₁ stem alone of the compound, is focus-marked. (3)

Predicate focus a. unu jarabi=a nak-i-i=du this child= cry--= ‘This child was CRYING.’

u-tar. -

b. unu jarabi=a zïïtii=du nak-i+u-tar. this child= continuously= cry-+- ‘This child was CONTINUOUSLY CRYING.’ c. unu jarabi=a nacï+bamik-i-i=du this child= cry.+scream--= ‘This child was CRYING AND SCREAMING.’

u-tar. -

Since a phrasal V-V complex consists of two words, with V₁ being a syntactic constituent, it may be focus-marked (3a). By contrast, V₁ of a compound V-V complex is never focus-marked, as in (3c), since V₁ is part of a word and is thus not a syntactic constituent.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.5 One-word V-V complexes 7.5.1 Lexical compound verbs A lexical compound is largely compositional in meaning but is collocational, and it is necessary for each V-V complex to be listed as a lexical item. Some lexical compounds are idiomatic, as in niv+bždar (sleep+become.low), which means ‘fall asleep’. Many lexical compounds are cognates of Japanese lexical compounds (e.g. usï+tausï ‘get someone down’, which shows a regular sound correspondence with osi+taosu in Japanese), but many have different lexical sources, as in niv+bždar above, which semantically corresponds to Japanese ne+kom-u. In a lexical compound, V₁ must take a specific stem form (Table 7.3) according to the class of the stem which occurs in the V₁ slot. Since the Class 1 stem is uniformly athematic, it takes the athematic form when functioning as V₁. (4)

Lexical compounds where V1 is a Class 1 stem a. karagi+ukusї-Ø (turn.over+get.up-) ‘raise (window, etc.)’ b. kurugi+uti-r (turn.round+fall-) ‘tumble down’ [J. koroge-otiru] c. mii+macïgav-Ø (see+make.mistake-) ‘misread’ [J. mi-matigau]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

172

 

A Class 2 stem also takes the athematic form. The following set of examples illustrates lexical compounds where V₁ is filled by a Class 2 stem which does not have a formal distinction within the athematic forms (the distinction between the basic and infinitive stem form). (5)

Lexical compounds where V1 is filled by a Class 2 stem a. tur+kai-r (take+change-) ‘replace’ (V1: tur- ‘take’) [J. tori-kaeru] b. usї+cїvsï-Ø (push+crush-) ‘crush’ (V1: us- ‘push’) [J. osi-tubusu] c. tuv+uri-r (jump+go.down-) ‘drop down’ (V1: tub- ‘jump’) [J. tobioriru]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

A Class 2 verb which has a formal distinction between the basic and infinitive stem form (i.e. a Class 2 verb that ends in /k/ or /g/) usually takes the infinitive rather than the basic form. Some lexical compounds permit both forms, as in (7a) and (7b). (6)

Lexical compounds where V1 is filled by a Class 2 stem (which ends in /k/ or /g/) a. kacї+kai-r (write.+change-) ‘rewrite’ (V1: kak- ‘write’) [J. kakikaeru] b. nacї+bamifï-Ø (cry.+scream-) ‘cry and scream’ (V1: nak- ‘cry’)

(7)

V1: pžk- ‘pull’ a. pžcï+mudusï (pull.+return-) ‘pull back’ [J. hiki-modosu] b. pžfї+mudusї-Ø (pull.+return-) ‘pull back’

It is also common for V₁ to be in the archaic infinitive form, which is made up of a Class 2 verb root (of any kind) plus /i/. In (8), for example, the Class 2 stem tur‘take’ shows up in the archaic infinitive form tui (/r/ is deleted by rule; turi ! tui). (8)

tu-i+fau-Ø take-+eat- ‘pick up and eat’

Likewise in (9), the Class 2 stem pžk- is in the archaic infinitive form pžki, rather than the expected athematic (current) infinitive form pžcï (pžk ! pžc ! /ï/ insertion ! pžcï). (9)

pžk-i+rri-r pull-+let.into- ‘pull in’

The archaic infinitive /i/ corresponds to the infinitive suffix -i of Japanese (or ren’yō suffix) and goes back to the infinitive suffix *-i of proto-Japonic. For

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

173

example, (9) is a cognate of hiki+ireru ‘pull in’ in Japanese, where V₁ ends in the infinitive form hik-i. The /i/ is productively used in most Japanese dialects and in Ryukyuan, but has mostly disappeared in Irabu (and other Miyako Ryukyuan dialects) owing to the sound change that caused */i/ to become /ï/ (e.g. kaki ‘writing’ > kacï) and further to be fused with the final C (e.g. *tori ‘taking’ > turï > tur, *asobi ‘playing’ > asubï > asïv, etc.), leading to the merger of the former infinitive form (e.g. tori ‘taking’) and the former stem form which was used for the adnominal clause predicate (i.e. rentai form; toru ‘take’). The resulting stem form tur thus covers both the one which used to be covered by the infinitive form (e.g. a stem for a compound) and the one which used to be covered by the adnominal form (e.g. a stem for the declarative unmarked inflection). (10) (former infinitive)

*tori >

turï > tur athematic form

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(former adnominal) *toru > turu > tur For the Class 2 stems that end in /k/ or /g/, the distinction between the former infinitive and adnominal is retained, with the current infinitive form (e.g. *kaki > kacï) and the basic form (e.g. kaku > kaf ï). There are also lexical compounds where V₁ can be considered to be in the archaic infinitive: mucjagar ‘pop up’ (< muc-i ‘carry’ + agar ‘go up’), tacjagar ‘stand up’ (tac-i ‘stand’ + agar ‘go up’), pžkjagir ‘pull up’ (< pžk-i ‘pull’ + agir ‘lift’), etc. Here, */Ci+a/ (2 moras) is now reanalyzed as /Cja/ (1 mora).

7.5.2 Syntactic compound verbs In syntactic compounds, V₁ is relatively less restricted than in lexical compounds, whereas V₂ is chosen from a restricted set of verb stems that encode phasal and procedural meanings. Table 7.5 lists the verbs that are identified as V₂ of syntactic compounds. Generally speaking, a single verb stem may fill a V₁ slot to combine with a range of V₂, as exemplified by num ‘drink’ in Table 7.4, as long as the semantics allows such a combination. Conversely, a given V₂ stem may combine with different V₁ stems. For example, a syntactic compound with the V₂ stem uwar ‘finish’ may have the forms (a) num+uwar ‘finish drinking’, (b) jum+uwar ‘finish reading’, kafï +uwar ‘finish writing’, (c) jurav+uwar ‘finish calling’, (d) cïfï+uwar ‘finish listening’, (e) mii+uwar ‘finish watching’, etc. That is, generally speaking, V₂ has morphological productivity, although the productivity of specific V₂ stems of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

174

  Table 7.5 V₂ of syntactic compounds Form pazїmir uwar cїzїkir maar paccjakir nausï bassir nnasï sïgir

Gloss begin finish keep wander fail to try again forget complete do excessively

Example num+pazїmi-r num+uwar-Ø num+cїzїki-r num+maar-Ø num+paccjaki-r num+nausï num+bassi-r num+nnasï num+sïgi-r

Meaning ‘begin drinking’ ‘finish drinking’ ‘keep drinking’ ‘drink here and there’ ‘fail to drink’ ‘drink again’ ‘forget drinking’ ‘complete drinking’ ‘drink too much’

syntactic compounds varies, with some V₂ stems showing very limited productivity. Thus, uwar is much more productive than patir ‘finish’, which never combines with (c) to (e) above, since patir designates the using up of some amount of some entity.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.6 Comparisons of syntactic and lexical compounds The difference between syntactic and lexical compounds is not clear-cut. What we have discussed are prototypical examples of syntactic and lexical compounds. Four features are relevant in characterizing the prototypes of lexical and syntactic compounds: degree of lexicalization, stem form, (in)ability to accommodate syntactic operations such as causative marking, and phrasal embedding into V₁.

7.6.1 Degree of lexicalization Prototypically, lexical compounds have a severe restriction with regard to the combination of V₁ and V₂, i.e. they are highly lexicalized. By contrast, prototypical syntactic compounds have a productive V₂ stem, which may combine with a wide range of V₁ stems.

7.6.2 Stem form As noted in section 7.3.3, V₁ of a lexical compound may be the infinitive form if there is a formal distinction between the infinitive and basic stem forms. It is also common for a lexical compound to contain the remnant of the earlier infinitive ending. By contrast, in syntactic compounds, V₁ of a Class 2 stem which ends in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

175

/k/ or /g/ (as in (11) and (12)) mostly opts for the basic rather than the infinitive stem form. And the archaic infinitive is very rare in V₁ of syntactic compounds. (11)

V1: kak- ‘write’ a. kafï+pazïmi-r write.+begin- ‘begin writing’ (V1 as basic stem form: common) b. kacï+pazïmi-r write.+begin- ‘begin writing’ (V1 as infinitive stem form: rare)

(12)

V1: kug- ‘paddle’ a. kuv+pazïmi-r paddle.+begin- ‘begin paddling’ (V1 as basic stem form: common) b. kuzï+pazïmi-r paddle.+begin- ‘begin writing’ (V1 as infinitive stem form: rare)

7.6.3 Structure of the stem

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The schematic structure of a verb is shown in (13), where the stem is indicated by square brackets. (13)

[Nucleus (-)(-)]STEM-

Thus, a stem is minimally a stem nucleus, as in (14a), and may be extended by the affixation of the causative suffix () and the passive suffix (), as illustrated in (14b) and (14c). (14)

usï ‘push’ as a stem nucleus a. taroo=ja kuruma=u=du Taro= cart== ‘Taro pushed the cart.’

[usï]-tar. push-

b. taroo=ja ziroo=n=du kuruma=u Taro= Jiro== cart= ‘Taro made Jiro push the cart.’

[us-asï]-tar. push--

c. taroo=ja ziroo=n=du kuruma=u [us-as-ai]-tar. Taro= Jiro== cart= push--- ‘Taro was made to push the cart by Jiro.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

176

 

A stem nucleus and a root are not the same notion, since a stem nucleus may consist of two roots, as illustrated in (15). That is, a stem nucleus may be a compound. The V-V complex of a lexical compound is always a single stem nucleus. (15)

taroo=ja nisïdu=u=du [usï+tausï]-tar. Taro= thief== push+get.down- ‘Taro pushed the thief down.’

A prototypical syntactic compound consists of a stem nucleus which is internally complex, i.e. a stem nucleus borne by V₁ being embedded within a larger stem nucleus (V₁+V₂), exhibiting a phraselike hierarchical structure. For example, in (16) the verb stem usï ‘push’ constitutes a single stem nucleus, which is embedded within a whole compound stem usï+pazïmi ‘begin pushing’, which is also a stem nucleus, serving as a stem that takes tense inflection to stand as a verb word. (16)

kuruma=u=du [[usï]+pazïmi]-tar cart== push+begin- ‘(x) began pushing the cart.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Since V₁ of usï+pazïmi ‘begin pushing’ is a stem nucleus, and since a stem nucleus may consist of two roots, as in usï+tausï ‘push down’, it is possible for V₁ of a syntactic compound to be a lexical compound of two roots. (17)

taroo=ja nisïdu=u=du [[usï+tausï]+pazïmi]-tar. Taro= thief== push+get.down+begin- [[V1+V2]V1+V2] ‘Taro began pushing the thief down.’

7.6.4 Causative suffixation on V₁ Since the voice suffixes attach to a stem nucleus as a whole, not a root within a stem nucleus, structures like (18) are permissible, but structures like (19) are impermissible in lexical compounds. (18)

a. taroo=ja ziroo=n=du nisïdu=u Taro= Jiro== thief= ‘Taro made Jiro push the thief down.’

[usï+taus-asï]-tar. push+knock.down--

b. nisïdu=u ziroo=n=du [usï+taus-ai]-tar. thief= Jiro== push+knock.down-- ‘The thief was pushed down by Jiro.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     (19)

177

*[us-asï+tausï]-tar push-+knock.down-

By contrast, the V₁ stem of a syntactic compound may take the causative suffix, since it constitutes an independent stem nucleus (Section 7.6.3). (20)

taroo=ja ziroo=n=du kuruma=u Taro= Jiro== cart= ‘Taro began to make Jiro push the cart.’

[[us-asï]+pazïmi]-tar push-+begin-

As a stem nucleus, the whole compound stem V₁+V₂ of a prototypical syntactic compound may also be suffixed by the causative suffix.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(21)

taroo=ja ziroo=n=du kuruma=u Taro= Jiro== cart= ‘Taro made Jiro begin pushing the cart.’

[[usï]+pazïmi-sïmi]-tar push+begin--

Note that the form of the causative suffix in (21) is different from that in (20). There are two variant forms of the causative suffix: -as, which is attached to a Class 2 stem nucleus, and -sïmi, which is attached to a Class 1 stem nucleus. Since the verb root usï- serves as the stem nucleus of (20), the causative suffix -as is selected. Given that the class of a stem nucleus is determined by the structural head if it is a compound, as in [[usï]+pazïmi] in (21), the compound stem is a Class 1 stem, and the causative suffix -sïmi is attached to it. It is possible for each stem nucleus of the internally complex stem nucleus to take the causative suffix in the same word. (22)

taroo=ja ziroo=n=du tuu=ju [[us-asï]+pazïmi-sïmi]-tar. Taro= Jiro== door= push-+begin-- ‘Taro ordered Jiro to make (someone) begin pushing the door.’

7.6.5 Phrasal embedding V₁ of a syntactic compound may embed a phrasal expression, specifically the Light Verb Construction (LVC) which consists of the light verb as- ‘do’ and its complement adverbial and/or direct object. For example, V₁ of the syntactic compound uuv+pazïmi-r ‘start swimming’ can be reworded by using the LVC which functions to refer to the action denoted by the V₁ anaphorically. (23)

[ai(=du) asï]+pazïmi-tar. that.way(=) do+begin-= ‘(He) started doing that.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

178

 

It is also possible for the V₁ of a syntactic compound to be interrogated by using an LVC which consists of the WH word (as a direct object) and the light verb. (24)

[nau=ju=ga asï]+pazïmi-tar=ga? what== do+begin-= ‘What did (you) start doing?’

By contrast, V₁ of a lexical compound cannot be filled by an LVC. For example, V₁ of the lexical compound usï+tausï (push+knock.down) ‘push (someone) down’ designates the manner or the preliminary phase of the event denoted by V₂, but it cannot be reworded anaphorically (‘knock down by doing that’), nor can it be questioned by the manner interrogative word nausi ‘how’. (25)

*ai(=du) asï+tausï-tar. that.way(=) do+get.down- (intended meaning) ‘(I) knocked (it) down by doing that way.’

(26)

*nausi=ga asï+tausï-tar=ga? how= do+knock.down-= (intended meaning) ‘How (did you) knock(ed) (it) down?’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.7 General remarks on phrasal V-V complexes 7.7.1 Structure In a phrasal V-V complex, each of V₁ and V₂ constitutes a word. V₁ is the gerund form (one of the nonfinite verbs), while V₂ is either finite or nonfinite, depending on the type of clause in which the complex occurs. In the following examples, V₂ is a finite verb form (declarative unmarked past), as it occurs in a main clause. (27)

tuu=ju=du door==

us-i-i push--= V1 ‘(I) tried pushing the door.’

mii-tar. - V2

(28)

tuu=ju=du door==

aki-tar. open(tr.)- V2

us-i-i push-- V1 ‘(I) pushed the door to open.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

179

7.7.2 Biclausal constructions and phrasal V-V complexes Phrasal V-V complexes have developed from a biclausal structure where a series of gerundive clauses is joined to a final main clause, or a clause-chaining structure, a structure which is very common in both Ryukyuan (Shimoji 2017) and Japanese (Myhill and Hibiya 1988). (29) is a directional auxiliary verb construction, whereas (30) is a clause-chaining structure where two clauses are chained. (29)

agu=u friend=

jurab-i-i t-tar. call-- come- V1 V2 ‘(I) brought my friends’ [lit. (I) called (my) friends and came back.]

(30)

agu=u friend=

jurab-i-i, t-tar. call-- come- gerundive clause V main clause V ‘(I) called (my) friends, and came back.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In the following examples, (31) is a serial verb construction in which two lexical verbs constitute a single phrasal predicate, whereas (32) is a clause-chaining structure where two clauses are chained. (31)

taka=nu=du hawk==

tub-i-i fly-- V1 ‘The hawk flew away.’

(32)

taka=nu=du hawk==

par-tar. leave- V2

tub-i-i, fly-- gerundive clause V ‘The hawk flew, and left.’

par-tar. leave- main clause V

There are three criteria for distinguishing between the two structures. First, a phrasal V-V complex is a tight syntactic knit, while two verbs in a clause chain are not. This is easily tested by seeing whether it is possible to insert a word between the two verbs in each case. Each of the following examples can only be interpreted as consisting of two predicates in a clause chain, since the adverb sugu ‘right away’ intervenes between the first verb and the second verb in each construction. (33)

agu=u jurab-i-i, sugu=du t-tar. friend= call-- right.away= come- ‘(I) called (my) friends, and (I) came back right away.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

180

 

(34)

taka=nu tub-i-i, sugu=du hawk= fly-- right.away= ‘The hawk flew, and left right away.’

par-tar. leave-

By contrast, if a phrasal V-V complex is to be modified by an adverb, it must be placed before the whole phrase. (35)

agu=u sugu=du jurab-i-i friend= right.away= call-- ‘I brought my friends right away.’ (AVC)

(36)

taka=a sugu=du tub-i-i hawk= right.away= fly-- ‘The hawk flew away right away.’ (SVC)

t-tar. come-

par-tar. leave-

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Second, a biclausal structure and a phrasal V-V complex are distinct in terms of prosody. In Irabu, V₁ and V₂ in a phrasal V-V complex constitute a single prosodic phrase (indicated by curly brackets in (37) and (38)). So, the SVC in (37) below is a single prosodic phrase, with V₂ having lower pitch than V₁. If the two verbs constitute a clause chain, as in (38), the two verbs never form a single prosodic phrase, and V₂ is slightly higher in pitch than V₁. (37)

{taka=nu=du} hawk==

{tub-i-i fly-- V1 ‘The hawk flew away.’

(38)

{taka=nu=du hawk==

par-tar}. leave- V2

tub-i-i} fly-- gerundive clause V ‘The hawk flew, and left.’

{par-tar}. leave- main clause V

The insertion of an adverb is only possible in the prosodic phrasing pattern exemplified in (38), and not in (37). A third criterion in making the distinction between a phrasal V-V complex and a biclausal one concerns semantics. In the biclausal (33), for example, a sequence of two events, the event of the subject calling friends and another event of the subject coming back are described as two separate events. So, it is possible that the subject went to call friends but came back without bringing his friends with him. By contrast, such an interpretation is impossible in (35), where the complex predicate jurabii ttar designates a single event that consists of two phases, i.e. the subevent of calling a friend and another subevent of coming with them. In

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

181

other words, the direct object agu ‘friends’ is shared by the entire V-V complex, an important characteristic of a phrasal V-V complex. The closest translation in English for such an event in (35) is ‘bring friends’.

7.8 Auxiliary verb construction (AVC) An AVC is defined as a ‘mono-clausal structure minimally consisting of a lexical verb element that contributes lexical content to the construction and an auxiliary verb element that contributes some grammatical or functional content to the construction’ (Anderson 2006: 7). Irabu has a class of auxiliary verbs, which are divided into three groups: (a) aspectual, (b) attitudinal, and (c) benefactive. AVCs correspond to Type 4 in the classification of Japanese V-V complexes (Chapter 2, this volume).

7.8.1 Aspectual AVC

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

There are five aspectual auxiliary verbs, progressive ur, perfect (‘disappearance’ perfect njaan and ‘emergence’ perfect ar), and directional (fïï ‘come’ and ifï ‘go’). The progressive auxiliary designates the ongoing phase of an event described by an activity or accomplishment verb, while it designates the resulting state if the verb is achievement. (39)

jarabi-mmi=a uug-i-i=du child-= swim--= ‘The children are swimming.’

ur-Ø=doo. -=

(40)

kii=nu nar=ra mmna uti-i=du tree= nuts= all drop-= ‘The nuts on the tree have all dropped.’

ur-Ø. -

The ‘disappearance’ perfect auxiliary njaan is a grammaticalized form of the lexical verb njaan ‘not exist’ and designates the completion of an event with a focus on the disappearance, demise, extinction, loss, etc. of the entity involved in the event. Thus, it is not a simple denotation of the perfect, as designated by the progressive auxiliary in (40), but the perfect which views the event as completed by the disappearance of the entity being acted on (either the direct object or the intransitive subject). In (41), the event of nuts dropping from a tree is viewed as completed, with the speaker’s focus being on the disappearance of the nuts on the tree.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

182

 

(41)

kii=nu nar=ra mmna uti-i tree= nuts= all drop- ‘The nuts on the tree have all dropped.’

njaa-n. .-

The auxiliary njaan may be used for an event in which no physical disappearance or loss is involved. Here, the notion of disappearance or loss is extended to the disappearance or loss of a normal or expected situation, as in (42), where the event of the watch stopping is viewed as completed by the construal that the watch has lost its expected function. (42)

tokee=nu=du tumar-i-i njaa-n. watch== stop-- .- ‘The watch has stopped (against my expectation).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Thus, the auxiliary njaan can often entail the speaker’s regret or negative attitude toward the completion of the event, since the auxiliary creates the construal that the expected situation is lost. In this sense, the auxiliary is intermediate between aspectual and attitudinal auxiliaries. The other perfect form, the ‘emergence’ perfect ar, is a grammaticalized form of the existential verb ar ‘exist’, the antonym of the nonexistential verb njaan. It focuses on the phase of emergence of an entity which has come into being due to the action being described. In (43), the speaker’s focus is on the nuts which are dropped on the ground. The action of the nuts being dropped from the tree is viewed as completed by the fact that all the nuts are now on the ground. (43)

kii=nu nar=ra mmna uti-i ar-Ø. tree= nuts= all drop- - ‘The nuts on the tree have all dropped.’ (looking at the ground)

The directional auxiliary verbs fïï ‘come’ and ifï ‘go’ are similar in use to those of Japanese, and they are actually cognate forms of Japanese kuru and iku, respectively. In (44), the speaker is on the ground and looks at the cat go up the tree, while in (45) the speaker is on the tree, waiting for the cat to come up. (44)

maju=nu=du kii=nkai nuur-i-i cat== tree= climb-- ‘The cat climbed up on the tree.’

ifï-tar. go-

(45)

maju=nu=du kii=nkai nuur-i-i cat== tree= climb-- ‘The cat climbed up on the tree.’

t-tar. come-

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

183

7.8.2 Attitudinal AVC The attitudinal auxiliaries are the conative miir and the prospective ufï. The conative miir ‘try doing something’ is a cognate of the Japanese miru ‘see’ and has the same conative ‘try’ meaning. In addition, the Irabu miir has an experiential meaning, as in (46). (46)

ba=a ai=nu munu=u=ba fa-i-i=ja mii-n. 1= that.way= thing== eat--= - ‘I have never eaten such a thing.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The prospective ufï is a cognate of the Japanese oku. It may express the prospective of the speaker, as in (47), but such a usage is not common and is probably a usage influenced by Japanese. It more commonly designates counterfactual meanings, as in (48). (47)

fïsur=ru=baa num-i-i=du medicine== drink--= ‘(I) took medicine in advance.’

ufï-tar. -

(48)

kama=n u-tigaa sïn-i-i=du that.place= exist- die--= ‘If (I) had been there, (I) would have died.’

ufï-Ø=pazï. -=maybe

7.8.3 Benefactive AVC The benefactive auxiliary is fiir, which express actions that are directed to a person for his/her benefit. (49)

a. ba=a maccja=nkai=du 1= shop== ‘I went to a shop.’

ifї-tar. go-

b. ba=a ui=ga kaari=n maccja=nkai=du ik-i-i fii-tar. 1= 3= stead= shop== go-- - ‘I went to a shop instead of him.’ c. kari=a ba=ga kaari=n maccja=nkai=du 3= 1= stead= shop== fii-tar. - ‘He went to a shop instead of me.’

ik-i-i go--

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

184

 

(50)

ba=a kai=n=du 1= 3== ‘I have him go to a shop.’

maccja=nkai shop=

ik-asï-tar. go--

The benefactive AVC is also observed in Japanese (see Chapters 2 and 6, this volume). One interesting difference between Irabu and Japanese is that in Irabu the benefactive auxiliary is invariably fiir (which is cognate with kureru in Japanese), whereas in Japanese the choice of auxiliary depends on the deictic center: (i) ageru ‘(speaker) do for someone’s benefit (from speaker’s point of view)’, (ii) jaru ‘(speaker) do for the benefit of someone who is lower in social rank than the speaker’, (iii) kureru ‘(nonspeaker subject) do for speaker’s benefit’, and (iv) morau ‘(speaker) have something done for speaker’s benefit.’ Note that (i) and (ii) correspond to fiir in (49b). (iii) corresponds to fiir in (49c). (iv) has no structural equivalent in Irabu. The same meaning of (iv) is expressed by causative, as in (50), and this lack of an auxiliary that functions like (iv) and the use of causative for (iv) are common in Ryukyuan.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.8.4 Serial verb construction (SVC) Phrasal V-V complexes in which both V₁ and V₂ are lexical verbs are called SVCs.³ The combination of V₁ and V₂ is conventionalized and collocational to varying degrees. V₁ and V₂ may constitute a series of sequential events, or V₁ serves to modify V₂ as a manner adverbial, although the distinction is not clear-cut. For example, (51) is more like a sequential event, while (52) is more like an event (denoted by V₂) modified by a manner (V₁). (51)

aa+gara=u=ru foxtail.millet+hull==

tur-i-i sїti-r? take-- do.away.with- V1 V2 ‘Do (I have to) take the hulls of foxtail millet and do away with them?’

(52)

nuuma=n horse=

nuur-i-i par-tar=ca. ride-- leave-= V1 V2 ‘(He) left riding on a horse.’

³ The term SVC is often restricted to those V-V complexes where V₁ lacks any overt marker of syntactic dependency (Aikhenvald 2006: 1), which should exclude Irabu phrasal V-V complexes, as they mark syntactic dependency through the gerund inflection. However, nonfinite marking on a verb within a series does not necessarily exclude the possibility of calling the construction an SVC, as such exclusion would entail a loss of generalization (similar claims are found in typological studies like Payne 1997: 311). In fact, typical features of SVCs such as monoclausality, argument sharing, single prosodic unity, and encoding sequential events/manner (Foley and Olson 1985) exactly hold for Irabu SVCs (see Shimoji 2017 for a detailed description of Irabu SVCs from a typological perspective).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

185

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The following examples are not meant to be an exhaustive list of SVCs, but to illustrate SVCs which are frequently used in Irabu discourse. (53)

suu=ju vegetable=

makas-i-i pull-- V1 ‘Pull the vegetable out.’

tur-i. take- V2

(54)

kai like.that uri=a 3=

(55)

ami fїї-Ø atu=n=na ssudur=nkai ik-i-i, rain fall- after== ssudur= go-- nuur+zuu=ju=mai tur-i-i fau-Ø. nuur+plant==too pick-- eat- V1 V2 ‘After it rains (I) go to Ssudur (place name) and pick and eat nuurzuu plant.’

(56)

kata+bata=u=baa half+body==

jak-i-i fau-Ø. burn-- eat- V1 V2 ‘(They) (would) burn and eat the half of the body.’

(57)

icu=u=baa, mmja, fїzї=sii kir-Ø. thread==  mouth=INST cut- mata, umacї=sii jak-i-i kir-Ø. and fire= burn-- cut- V1 V2 ‘(One) cuts threads with his mouth; also (one can) cut (threads) by burning (them) with fire.’

bazakar-i-i, ff-a-Ø=tti as-i-i, show.claw-- bite--= do-- tur-i-i rri-i . . . take-- put.in- V1 V2 ‘(The crab) showed its claw, moving menacingly, (and I) took it and put it into (the basket).’

7.9 Comparison between AVC and SVC As in the case of the distinction between syntactic and lexical compounds, there are intermediate cases where a phrasal V-V complex exhibits certain AVC-like

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

186

 

features and SVC-like features. Crosslinguistically, an SVC is a typical diachronic source for AVCs (Anderson 2006: 11), and if V₂ in an SVC has undergone a significant semantic bleaching or abstraction, and hence the productivity that enables V₂ to combine more freely with V₂, the construction begins to be like an AVC.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.9.1 Degree of lexicalization In AVCs, all V₂ stems are delexicalized auxiliary verbs, since they do not determine the transitivity of the clause. By contrast, in SVCs, both V₁ and V₂ are lexical verbs in the sense that they retain transitivity. The transitivity must be consistent between V₁ and V₂, a situation called ‘transitivity harmony’ in Kageyama (1993) and Frellesvig et al. (2010), and they must share the argument(s). Some SVCs have V₂s that productively combine relatively freely with a variety of different V₁ stems. For example, fau ‘eat’ as V₂ may be combined with a number of sequential or manner-like elements: turii fau ‘take and eat’, nii fau ‘eat by boiling’, jakii fau ‘eat by frying’, etc. However, unlike V₂ of an AVC, fau ‘eat’ clearly retains its lexical meaning. It is transitive and takes the arguments which are shared by V₁. By contrast, in an AVC, it is V₁ that determines the transitivity of the whole construction, with V₂ being a delexicalized auxiliary element. The V₂ stem of an SVC par ‘leave’ is one step further on the grammaticalization pathway from an SVC to an AVC. It retains its status as a lexical verb, since it is an intransitive verb and cannot co-occur with transitive V₁. This is in sharp contrast to other directional verbs like ifï ‘go’ or fïï ‘come’, which serve as auxiliary verbs of AVCs and can co-occur with either transitive or intransitive V₁. The verb par is, however, combinable with a range of intransitive verbs designating an abstract, aspectual meaning of perfect, as in sïnii par ‘die out’, kjaarii par ‘(fire) go out’, etc.

7.9.2 Focus marking When information-structurally focused, focus marking on V₁ of a typical AVC is obligatory, whereas V₁ of a typical SVC is only optional. In (58), the interrogation focus scopes over V₁ of the progressive AVC, and in response to it, the answer sentence (59) has the focus on V₁, which must be overtly focus-marked. (58)

kari=a nau=ju=ga 3= what== ‘What is he doing?’

as-i+ur-Ø=ga? do-+-=

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     (59)

nivv-i-i=du sleep--= ‘(He) is SLEEPing.’

187

ur-Ø. -

By contrast, focus marking on V₁ of an SVC is optional, and typically it is not focus-marked. (60)

nausi=ga fau-tar=ga? how= eat-= ‘How (did you) get (it) down?’

(61)

jak-i-i(=du) fau-tar. burn--= eat- ‘(I) ate (it) BY BURNING.’

The optionality of focus marking on V₁ in an SVC may be related to the fact that the entire complex is treated like a single lexical item: a lexical item is commonly immune to syntactic operations, and focus marking is clearly a syntactic operation.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.9.3 Causative suffixation Phrasal V-V complexes are considered to consist of two distinct stem nuclei, given the fact that they constitute distinct words, and that a stem nucleus is, by definition, smaller than a word. The prediction can be made that V₁ of an AVC and an SVC may be marked by the causative suffix. This prediction is borne out in the case of AVCs. The V₁ of an AVC of any kind can be marked by the causative. (62)

kanu buuciri=u=ba tat-as-i-i=du ur-Ø. that bad.boy== stand---= - ‘(The teacher) has ordered the bad boy to stand (in the corridor).’

(63)

munu=u fii-da sïn-as-i-i njaa-n=ni. thing= give-. die--- .-=eh ‘By failing to give food, (you) caused (the animal) to die, right?’

By contrast, the prediction fails in the case of SVCs, where it is the entire phrase, not just V₁, that can take the causative suffix.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

188

 

(64)

a. taka=nu=du hawk==

par-tar. leave- V2

b. taka=u=du hawk==

par]-asï-tar. leave-- V2

tub-i-i fly-- V1 ‘The hawk flew away.’ [tub-i-i fly-- V1 ‘(I) made the hawk fly away.’

c. *taka=u=du hawk==

[tub]-as-i-i fly--- V1

par-tar. leave- V1

(64c) is possible if V₁ and V₂ are two separate predicates, i.e. a clause chain. In such a case, the sentence means ‘(I) let the hawk fly, and left’, as illustrated in (65). (65)

ba=a 1=

taka=u=du hawk==

tub-as-i-i, fly--- gerundive clause V

par-tar. leave- main clause V

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

‘I let the hawk fly, and left.’ It is important to note that an SVC is similar to a lexical compound in two respects, which should be interrelated. First, in both V-V complexes the entire V-V complex, whether it is a phrasal or a one-word construction, is treated as a single morphological locus of causative marking. Second, in an SVC the whole complex is a conventional collocational unit, hence a lexical item, just as in the case of lexical compounds. This fact suggests that causative suffixation does not apply inside a lexical item.

7.9.4 Phrasal embedding V₁ of an AVC or SVC may embed a phrasal expression. V₁ of an AVC may embed an SVC. For example, the two verbs in the SVC (66) can be embedded as V₁ of the progressive AVC (67). (66)

jak-i-i(=du) fau-tar. burn--= eat- ‘(I) ate (it) BY GRILLING.’ (SVC)

(67)

kari=a [jak-i-i 3= grill-- ‘He is eating by grilling’

fa-i-i]=du eat--=

ur-Ø. -

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

189

Unlike lexical compounds (Section 7.6.5), an SVC may also embed a phrasal expression within V₁, allowing an LVC to occur within it. Thus, it is possible for V₁ to be replaced by the phrasal anaphoric expression ai=ja sii=du ‘doing like that’. Note that in (68), the focus marking is optional, since it is an SVC, where the focused V₁ is only optionally focus-marked. (68)

[ai=ja s-i-i](=du) fau-tar. that.way= do--= eat- ‘(I) ate (it) that way (lit. by doing like that)’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.10 A comparison with Japanese Since we have a basic understanding of V-V complexes in Irabu, it is now possible to compare Irabu V-V complexes with those in Japanese, as summarized in Table 7.6. AVCs are unambiguously analyzed as structural as well as lexical equivalents to Type 4 in Japanese in that both are phrasal, with V₂ being delexicalized auxiliary elements which are mostly cognate forms (e.g. Japanese aru and Irabu ar). In Japanese, the gerund form is formed with -te, which is also common across Japanese dialects. In Irabu (and in many Southern Ryukyuan languages) the gerund form is formed with -i, which is not cognate with -te (Karimata 2017). The gerund forms thus developed independently in the two languages with different lexical sources for the gerundive marker. SVCs are argued to be phrasal (rather than biclausal) constructions and must be dealt with as instances of a V-V complex in Irabu, but there is no structural equivalent in the classification of Japanese V-V complexes suggested by Kageyama (Chapter 2, this volume). In Japanese, the following construction may be an equivalent to an SVC in Irabu. (69)

sakana=o fish=

ja-i-te tabe-ta. burn-- eat- V1 V2 ‘(I) ate fish by grilling (it).’

Table 7.6 V-V complexes in Irabu and in Japanese Types suggested for Irabu AVC SVC Lexical compound Syntactic compound

Types suggested for Japanese (Chapter 2) Type 4 — Types 1, 2 Type 3

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

190

 

One possible analysis of this construction would be to exclude it from the classification of V-V complexes, regarding it as a biclausal construction where V₁ belongs to a gerundive clause and V₂ belongs to the main clause, constituting a clause chain. However, such an analysis is problematic, since V₁ and V₂ may form a single prosodic phrase, as in (70), where prosodic phrase boundaries are indicated by curly brackets. (70)

{sakana=o} fish=

{ja-i-te burn-- V1 ‘(I) ate fish by grilling (it).’

tabe-ta}. eat- V2

If the V₁ and V₂ (excluding the direct object) constitute a single prosodic phrase, the initial syllable of V₁, /ja/, has initial lowering. If the entire VP (O + V1 + V2) constitutes a single prosodic phrase, the initial lowering of /ja/ disappears; in such a case the initial lowering shifts to the prosodic-phrase-initial /sa/. On the other hand, V₁ and V₂ may belong to separate prosodic phrases, as in (71), where the initial lowering feature of /ja/ obligatorily disappears, since V₁ is not prosodic-phrase-initial.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(71)

{sakana=o fish=

ja-i-te} {tabe-ta}. burn-- eat- V1 V2 ‘(I) grilled fish, and ate (it).’

The prosodic phrasing of (71) marks it as biclausal, both in structure and meaning. In terms of structure, it is possible for a clause-linkage marker to be inserted between V₁ and V₂, such as sono ato ‘after that’ and sore=kara ‘then’. The insertion would be impossible in (70) with a single prosodic phrasing, as the phrasing makes it a phrasal V-V complex. In terms of meaning, it is possible to interpret the sentence in (71) as two separate events performed by different agents, such as ‘(A) grilled fish, and (B) ate (it)’. But such an interpretation is impossible for (70). All these indicate that (70) is monoclausal, while (71) is biclausal. The upshot is that the identical form (69) may have two structures, a phrasal V-V complex (70) and a biclausal V-V sequence (71), and the difference surfaces with distinct prosodic phrasing. We cannot distinguish between the two by referring to distinct morphosyntactic properties like inflection or the unique delexicalized V₂ (as in AVCs). Rather, the same lexical string may be a phrasal construction or a biclausal one, and the difference surfaces with distinct prosodic phrasing patterns. Such a situation is not surprising, given that a biclausal construction with the gerundive clause (i.e. clause chain) is the source structure from which a phrasal V-V complex developed in Japanese (and in Ryukyuan). In Japanese and in Irabu,

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

191

the source structure and the target structure of the grammaticalization pathway, i.e. a biclausal V-V sequence (clause-chaining structure) and a phrasal V-V complex (SVC), are simply coexistent. In Irabu, the latter construction is called SVC, and the structural equivalent for it in Japanese is an example like (70), with the single prosodic phrase pattern. One-word V-V complexes exist both in Japanese and in Irabu. Syntactic compounds in both languages have many structural and lexical similarities. However, the Irabu syntactic compounds constitute a much smaller class than the Japanese Type 3. We often encounter cases where a given Japanese Type 3 compound is not translatable as a syntactic compound in Irabu. For example, the reciprocal compound with the V₂ stem au ‘do each other’ (as in nagur-i+au ‘strike each other’) cannot be translated into Irabu with any syntactic compound. Lexical compounds abound in both languages. We can find lexical compounds of Irabu which are considered to be cognates with Japanese lexical compounds, and we can also find lexical compounds which must be unique to Irabu, as in niv+bždar (sleep+become.low) ‘fall asleep’, fau+kamar (eat+become.bored) ‘get sick of eating’, fau+pinarasï (eat+reduce) ‘reduce by eating’, etc.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

7.11 Concluding remarks This chapter has examined V-V complexes in Irabu and identified two major classes of V-V complexes: phrasal V-V complexes (AVCs and SVCs) and oneword V-V complexes (lexical and syntactic compounds). The chapter has aimed to be descriptive, and the focus was mostly on the synchronic Irabu data. However, the Irabu data also contribute to diachronic studies, especially the ongoing debate in Japanese historical linguistics with regard to whether one-word V-V complexes existed in Old Japanese. Some scholars claim that V-V complexes in OJ were all phrasal, with V₁ being infinitive (Yoshizawa 1952; Kindaichi 1953, etc.). Here, it is particularly noteworthy that many Japanese Type 1 lexical compounds correspond to Irabu SVCs with cognate forms for V₁ and V₂, which leads us to believe that the same V-V complexes developed into lexical compounds in Japanese (as Type 1) but into SVCs in Irabu. For example, the V-V complex tobi+mawaru ‘jump around’ is a lexical compound in Japanese. In Irabu, the cognate stem forms tub- and maar participate in an SVC and not in a lexical compound, i.e. tub-i-i#maar (‘#’ indicates the word boundary). It is easy to find similar examples, and Table 7.7 shows only a partial list of such examples. In (vi), the Irabu V-V complex shows fluidity between an SVC and a lexical compound. Based on the crosslinguistically common process of grammaticalization whereby a phrase becomes a one-word construction and on the assumption that a change in the opposite direction is not likely, it is possible to claim that, before Japanese and Ryukyuan branched off, i.e. in proto-Japonic, there must have been a

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

192

 

Table 7.7 Irabu SVCs which are cognates with Japanese Types 1 and 2 Japanese V₁+V₂ (compound) os-i+ake-ru push-+open+ (ii) mow down nag-i+taos-u mow-+get.down- (iii) pluck up musir-i+tor-u pluck++take- (iv) suck up su-i+tor-u suck-+take- (v) jump onto tob-i+nor-u jump-+ride- (vi) cry and scream nak-i+wamek-u cry-+scream-

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(i)

Meaning push and open

Irabu V₁#V₂ (phrasal) us-i-i#aki-r push--#open- nag-i-i#tausï-Ø mow--#get.down muss-i-i#tur-Ø pluck--#take- sïpžž-i-i#tur-Ø suck--#take- tub-i-i#nuur-Ø jump--#ride- nak-i-i#bamifï-Ø (SVC) cry--#scream- nacï+bamifï-Ø (lexical CV) cry.+scream-

situation where V₁ and V₂ formed a phrasal V-V complex, with V₁ being the infinitive form. In Japanese, some of the phrasal V-V complexes developed into one-word V-V complexes (lexical compounds), while others remained phrasal, and the latter were rearranged as new phrasal V-V complexes with the development of the –te gerund suffix attached to the former infinitive word form (e.g. ide ku > de-te kuru ‘come out’, Aoki 2013). A similar diachronic process must have occurred in Ryukyuan, probably as a parallel development. Some of the source V-V complexes, which were phrasal with the infinitive V₁, became lexical compounds, and others remained phrasal; for the phrasal V-V complexes, V₁ came to be inflected as the -i gerund form, with the infinitive of the source structure being retained in the gerundive form as the /i/ thematic vowel (e.g. tur-i-i ‘take-- ‘taking’). The infinitive stem form in the source structure is also fossilized as the ‘archaic infinitive’ in a limited number of lexical compounds in contemporary Irabu (Section 7.5.1), but most underwent the sound change whereby the infinitive became the athematic stem form in Irabu.

Acknowledgments I am grateful to my language consultants for Irabu, especially Mr. Yukio Okuhama. My deep thanks also go to Dr. Christopher Davis and the editors of this volume for detailed and helpful comments on earlier drafts of the chapter. The present chapter was supported by Grant-in-aid for Young Scientists B (16K16843) and Grant-in-aid for Scientific Research S (17H06115).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

8 Korean verb-verb sequences Hyun Kyung Hwang and John Whitman

8.1 Introduction This chapter examines sequences of verbs V₁V₂ in Korean like (1), where V₁ ends with the infinitive suffix -e/a.¹ (1)

Taymin=i kemok=ul mil-e Taemin= big.tree= push- ‘Taemin pushed over the big tree.’

nemettuli-ess-ta. (NINJAL 2014)2 topple--

Our focus is “thematic” V₁-e/a V₂ sequences like mil-e nemettuli ‘push- topple’ in (1) where both verbs retain their lexical meaning, but we also provide a brief classification of nonthematic, “light” or “vector” V₁-e/a V₂ sequences like (2), where the lexical meaning of V₂ is not retained:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(2)

Taymin=i makkelli=lul mantul-e Taemin= makkeolli= make- ‘Taemin tried making makkeolli.’

po-ass-ta. --

We also compare thematic V₁-e/a V₂ sequences to bare-root V-V compounds like (3), where V₁ is an unsuffixed verb root (Martin 1997; Lee 2006):

¹ The shape of infinitive ending -e/a is determined by vowel harmony: inflecting stems whose stemfinal vowel is /a/ or /o/ take -a; all others, -e. In addition to the thematic and non thematic V-V sequences that are the topic of this chapter, the infinitive is used for various types of clausal and subclausal (e.g. VP) coordination. It is also used to form the base for a small number of lexical derivational patterns; for example, nemettuli- in (1) is derived from nem-e ‘go.over-’ plus the bound verbal intensifier ttuli-. As its label suggests, the infinitive is not tensed, but it can be used as a matrix imperative or indicative verb in the panmal pattern introduced in (39) of Section 8.5. For fuller descriptions of the function of the infinitive, see Martin (1993) and Cho and Whitman (2020). ² Enclitic particle boundaries are represented by an equals sign =, suffix boundaries by a hyphen -, and compound boundaries by a plus sign +. See Cho and Whitman (2020) for justification of the analysis of particles as enclitics.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

194

     

(3)

a. Sey salam sai=ey kyelyenhan sisen=i o+ka-ss-ta. 3 person between= determined look= come+go-- ‘Determined looks went back and forth between the three people.’ (Chang Yunjin, Chŏldaeja ŭi kwihan 6) b. Aliatuney=uy mom=i nal+ttwi-ess-ta. Ariadne= body= fly+jump-- ‘Ariadne’s body jumped violently.’ (Mina Iori, Ariadŭne wa sarang ŭi migong) c. Yeyppun noyey sonye=ka cipanil=ul tol+po-ass-ta. pretty slave girl= housework= turn+see-- ‘A pretty slave girl took care of the housework.’ (Yi Chae-gyu, Italia ŭi kkot, T’osuk’ana esŏ yesul ŭl mannada)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Bareroot compounding is not productive in modern Korean, but the pattern provides a point of contrast with thematic V₁-e/a V₂ sequences, which we show in Section 8.2 are not compounds with single-word status. In Section 8.3, we review the diverse functions of nonthematic V₁-e/a V₂ sequences. Section 8.4 contrasts other forms of V₁ in V-V sequences, including sequential -e/ase, transferative -ta, gerundive –ko, and conjunctive -(u)mye (see Section 8.4 for a brief description of the function of these forms). All of these are used in contexts where V₁-e/a V₂ sequences are not possible, as in (4)–(7): (4)

Swumi=nun mwukewun mwun=ul mil-e *(se) yel-ess-ta. (NINJAL 2014) Sumi= heavy door= push- opentr-- ‘Sumi pushed and opened the heavy door.’

(5)

Na=nun hanul=ul olli-e*(-ta) I= sky= raise-- ‘I looked up at the sky.’

(6)

Na=nun tayhak ywuk nyenkan nol-*a/ko I= university six year play-/ ‘I spent six years of college playing around.’

(7)

Na=nun tayhak ywuk nyenkan=ul nol-*a/mye I= college six years= play-- ‘I spent six years of college playing around.’

po-ass-ta. look--

cinay-ess-ta. pass--

cinay-ess-ta. pass--

In Section 8.5, we look at the constraints on the thematic subclass of V₁-e/a V₂ sequences. Section 8.6 places the status of V₁-e/a V₂ sequences in the context of the four-way typology of Japanese V-V sequences outlined in by Kageyama (Chapter 2 this volume), and concludes the chapter.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

195

8.2 Thematic V₂ sequences are not compounds Early work on thematic V₁-e/a V₂ sequences (Ch’oe 1929; Sohn 1976) described them as compounds, but more recent researchers have pointed out that they are not.³ We present five arguments here. First, thematic V₂ sequences can be split up by delimiter particles (Chung 1993; Yi 1997; Zubizarreta and Oh 2007): (8)

a. Taymin=i kemok=ul mil-e=to/man Taemin= big tree= push-=even/only ‘Taemin even/only pushed over the big tree.’

nemettuli-ess-ta. topple--

b. Swumi=nun mikkuleci-e=to/?nun tteleci-ess-ta. Sumi= slip-=even/ fall-- ‘Sumi even slipped and fell/Sumi SLIPPED and fell.’ This contrasts with bare-root V-V compounds, which cannot be split up by delimiter particles:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(9)

a. *Sey salam sai=ey kyelyenhan sisen=i o=to 3 person between= embarrassed look= come=even ka-ss-ta. go-- ‘Embarassed looks went back and forth between the three people.’ b. *Yeyppun noyey sonye=ka cipanil=ul pretty slave girl= housework= po-ass-ta. see-- ‘A pretty slave girl took care of the housework.’

tol=to turn=even

This property of Korean bare-root V-V compounds follows from the fact that verb roots in Korean are bound morphemes. They must surface followed by one of the closed class of verb suffixes. As a consequence, verb roots cannot stand alone as words, nor can they host delimiter particles, which have the status of enclitics (Cho and Whitman 2020). A second argument comes from the fact pointed out by Ko and Sohn (2015) that the subclass of thematic V₁-e/a V₂ sequences that they call “HSVCs” (high serial verb constructions) allows V₂ and its complement to be scrambled over the matrix subject:

³ Ch‘oe (1929/1961: 283) describes thematic V₁-e/a V₂ sequences as han nath uy kyepssi (한 낱의 겹씨 ‘compounds comprising a single word’) and pokhapsa (複合詞 ‘compound words’).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

196

     

(10)

a. Kemok=ul mil-e Taymin=i big.tree= push- Taemin= ‘Taemin pushed over the big tree.’ b. Yene=lul cap-a John=i salmon= catch- John= ‘John caught the salmon and ate it.’

nemettuli-ess-ta. topple-- mek-ess-ta. eat-- (Ko and Sohn 2015: 88)

Again, this is completely impossible with bare-root compounds, as predicted by the bound status of V₁: (11)

*Cipanil=ul tol yeyppun noyey sonye=ka housework= turn pretty slave girl= ‘A pretty slave girl took care of the housework.’

po-ass-ta. see--

A third argument comes from VP-pronominalization. V₁ can be replaced by the VP proform kulihay ‘do so’:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

Taymin=i khu-n namwu=lul mil-e nemettuli-ess-ta. Taemin= big- tree= push- topple-- Nayeng=i cak-un namwu=lul kulihay nemettuli-ess-ta. Nayoung= little- tree= doing.so topple-- ‘Taemin pushed over the big tree. Nayoung thus toppled the little tree.’

A fourth argument comes from their accentuation. In Korean varieties that have lexical pitch accent, such as South Kyŏngsang Korean (SKK), V₁ and V₂ form two separate accentual phrases, each with the same accent as the independent verb: (13)

V₁-e/a V₂ accentuation in South Kyŏngsang Korean⁴ a. L H L HL LHL HL mantul-ta mek-ta mantul-e mek-ta make- eat- make- eat- ‘make and eat’ b. L H et-ta get-

HL mek-ta eat-

LH HL et-e mek-ta get- eat- ‘beg, be treated for one’s food’

⁴ A male speaker in his thirties was recorded in a quiet office using a portable recorder (MARANTZ PMD661) with a microphone (SHURE SM10A). The speaker was born and raised in the city of Changwon, South Kyŏngsang.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

197

These patterns are shown in the pitch tracks in (14): (14) (a)

200

200

Pitch (Hz)

H

H

L

L

man 0

tul

0

L

ta

mek

ta

0 0.8326 0

200

Pitch (Hz)

H

0.8388

H

L

L L

man 0

(b)

tu

mek

le

0

ta 1.059

Times (s)

200

200

Pitch (Hz)

H

H

L

L

et 0

ta

0

mek 0.7819

200

Pitch (Hz)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

H

0

0.8388

H L

L

e 0

ta

0

te

mek

0

ta 1.115

Times (s)

For certain verbs, if V₁ is focused, the accent of V₂ may be deleted. This is possible for mantul-e mek- ‘make and eat’ in (13a), but not for et-e mek- ‘beg for one’s food’ in (13b). That is, in addition to the LHL#HL pattern in (13a), mantul-e mek- ‘make and eat’ may have the pattern LHL#LL, if mantul-e ‘make-’ is focused. The possibility of both separate and singly accented V₁-e/a V₂ sequences is comparable to the two accent patterns for Japanese V₁-te V₂ sequences discussed by Matsumoto (Chapter 6, this volume), except that accent deletion in SKK is subject to a phonological condition. Hwang (2011: 924) shows that lexical pitch accent is suppressed in the domain of an accented lexical item with focus prominence. Mantúl- ‘make’, with second syllable accent, permits deletion of the following accent if focused. Et- ‘get, receive’ is unaccented. (Unaccented verbs are realized as high, but lower their initial mora under certain conditions.) Therefore, V₂ may not lose its accent after

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

198

     

et-e ‘get-’, even when the latter has focus prominence. A pitch contour of the pattern LHL#LL pattern in (13a) is shown in in (15). (15)

200 Pitch (Hz)

compressed

0

man

tu

0

le

mek

ta 1.129

Time (s)

In contrast to V₁-e/a V₂ sequences, lexical compounds form only a single accentual phrase. With monosyllabic V₂ stems, the tendency is for V₁ to lose its accent, resulting in a LHL pattern: (16)

Lexical compound verb accentuation in South Kyŏngsang Korean a. H L HL LHL o-ta ka-ta o+ka-ta come- come- come+go- ‘go back and forth’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. L H nal-ta fly-

HL ttwi-ta jump-

LHL nal+ttwi-ta fly+jump- ‘jump about violently, attack, run amok’

The final argument for the noncompound status of V₁-e/a V₂ sequences comes from compositionality. Chung (1993: 45) points out that the two verb stems that make up the bare-root compound verb tol+po- ‘take care of ’ in (3c) may also occur as a V₁-e/a V₂ sequence. (17)

a. Meyli=ka ku kay=lul cal Mary= that dog= well ‘Mary looked after the dog well.’

tol+po-ass-ta. turn+see-- (cf. Chung 1993: 45)

b. Meyli=ka ku kay=lul tol-a Mary= that dog= turn- ‘Mary turned around and saw the dog.’

po-ass-ta. see-- (cf. Chung 1993: 45)

As Chung points out, the relationship between the meaning of the lexical compound verb tol+po- ‘take care of, look after’ and the meanings of its component verbs is “figurative,” or non-compositional, while the meaning of the V₁-e/a V₂ sequence tol-epo- ‘turn around and look’ is completely composed from the meaning

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

199

of its component verbs. This is a general difference: for example, the meaning of the component verbs ‘fly’ and ‘jump’ is only etymologically related to the meaning of the lexical compound nal+ttwi- ‘jump about violently, attack’ in (3b). Bare-root compound verbs are lexicalized in the traditional sense of the term.⁵

8.3 Functions of V₁-e/a nonthematic V₂ complexes The -e/a suffix is labeled “infinitive” by Martin (1993). The term is justified by the fact that V₁-e/a can never be tensed. Ramstedt (1939) includes -e/a in his class of ‘converbalia’. The term ‘converb’ has been revived in the current typological literature (e.g. Narrog, Rhee, and Whitman 2018), but it obscures the diverse range of syntactic functions of infinitive V₁-e/a, in particular the distinction between thematic and nonthematic V₂. Functions of the latter are summarized in Table 8.1, based on the classification in Narrog, Rhee, and Whitman (2018). Table 8.1 Nonthematic V₂ selecting infinitive V1-e/a

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Pattern Aspectual V₁-e/a issV₁-e/a kaV₁-e/a mekV₁-e/a peliV₁-e/a naV₁-e/a nayV₁-e/a nohV₁-e/a oV₁-e/a ppaciV₁-e/a tay V₁-e/a tulV₁-e/a twuBenefactive V₁-e/a cwuV₁-e/a tuliConative V₁-e/a poIntensifier V₁-e/a ttuli-

Meaning of V₂

Label and gloss

     ,   ,     ,     

Stative resultative ‘have V₁ed’ Continuative ‘go on V₁ing’ Perfective completive ‘end up V₁ing’ Perfective completive ‘end up V₁ing’ Inceptive intransitive ‘come out as V₁’ Inceptive transitive ‘bring out as V₁’ Purposive perfective ‘keep V₁ed’, stay V₁’ Perfective/Inceptive ‘come to V₁’ Perfective completive completely V₁’ Habitual/Iterative ‘keep on V₁ing’ Perfective ‘get V₁ed’ Purposive perfective ‘keep V₁ed’

  (deferential)

Benefactive ‘do V₁ for’ Deferential benefactive ‘do V₁ for’



Conative ‘try V₁ing’

I

Intensive ‘really (makes) V₁’

⁵ Chung acknowledges that there are also some “lexicalized or idiomatic S(erial) V(erb) C(instruction)s” (in our terms, V₁-e/a V₂ sequences) (1993: 46). For example, et-e mek- ‘get and eat’ in (13b) has the idiomatic meanings ‘beg for food, freeload, mooch’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

200

     

As Narrog, Rhee, and Whitman point out, the wide range of functions seen in V₁e/a nonthematic V₂ sequences is comparable to the diverse range of V₁-i/e (infinitive, ren’yōkei) nonthematic V₂ patterns in earlier Japanese (see Aoki and Frellesvig (Chapter 3, this volume)). As in earlier Japanese, infinitive V₁ is selected by benefactive verbs, cwu- ‘give’ and tuli- ‘give (deferential, goal honorific)’. In this pattern, the term “nonthematic” for benefactive V₂ is inaccurate, as benefactive V₂ adds a (usually implicit) beneficiary argument. Benefactive V₂ also affects the casemarking potential of V₁-e/a V₂. Thus, many speakers allow cwu- ‘give’ to assign accusative case to its goal argument (18b). This is not possible for the ditransitive verb kenney- ‘hand to’ (18a), but the accusative goal pattern becomes possible in the V₁-e/a cwu- benefactive pattern (18c): (18)

a. Swumi=ka chayk=ul Thaymin=eykey/*=ul Sumi= book= Taymin=/= ‘Sumi handed Taemin a book.’ b. Swumi=ka chayk=ul Thaymin=eykey/?/%=ul Sumi= book= Taymin=/= ‘Sumi gave Taemin a book.’

kenney-ss-ta. hand-- cwu-ess-ta. --

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. Swumi=ka chayk=ul Thaymin=eykey/?/%=ul kenney(-e) Sumi= book= Taymin=/?/%= hand- cwu-ess-ta. -- ‘Sumi did Taemin the favor of handing him a book.’ The existential verb iss- ‘be’ combines with achievement verbs to express stative resultatives (19a). This pattern is similar to the pre-Old Japanese pattern combining infinitive -i and existential ari ‘be, exist’ producing the OJ stative suffix -yeri, but the stative resultative V₁-e iss- in modern Korean is restricted to unaccusative achievement verbs. The Korean pattern is part of a diachronic cycle where V-e/a iss- contracted to form what is now the modern Korean past tense -e/ass- (19c). Uncontracted V-e/a iss- has a resultative meaning just for unaccusative achievements, while V-ko iss- ‘V- be’ is progressive (19b). (19)

a. Swumi=ka anc-a Sumi= sit- ‘Sumi is seated.’

iss-ta. -

b. Swumi=ka anc-ko iss-ta. Sumi= sit- - ‘Sumi is (in the process of ) sitting down.’ c. Swumi=ka anc-ass-ta. Sumi= sit-- ‘Sumi sat down.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

201

The two verbs of placement, twu- ‘place’ and noh- ‘put’ derive perfective V₂ complexes where the agent completes an action and leaves the resultant state for some future purpose (20a, b). V₁-e/a noh-, but not V₁-e/a twu-, can be used after stative intransitive V₁ to indicate that the state expressed by V₁ persists (20c). (20)

a. Swumi=ka chayksang wi=ey chayk=ul noh-a Sumi= desk top= book= put- twu-ess-ta/noh-ass-ta. --/-- ‘Sumi put the book on top of the desk (and left it there).’ b. Swumi=ka chayksang wi=ey ilum=ul ss-e Sumi= desk top= name= write- twu-ess-ta/noh-ass-ta. --/-- ‘Sumi wrote her name on top of the desk (and left it there).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. Thaymin=i yakha-y noh-ase/*twu-ese kkok kamki=ga Taemin= weak= -/- surely cold= tu-n-ta. enter-- ‘Taemin is (as always) weak so he will surely catch a cold.’ Peli- ‘discard, throw away’ and mek- ‘eat’ both derive disposal-type aspectual V₂ sequences where the action of V₁ is expressed as complete and irreversible or unrecoverable. The unrecoverability implicature of these two disposal-type light verbs is often associated with a negative evaluation of the event, which is stronger with mek- ‘eat’. As shown in (21b), V₁-e/a mek- is limited to transitive telic V₁. (21)

a. Swumi=ka Thaymin=uy ilum=ul Sumi= Taymin= name= peli-ess-ta / mek-ess-ta. --/-- ‘Sumi went and forgot Taemin’s name.’ b. Swumi=ka cip=ulo tol-a Sumi= home= return- peli-ess-ta/*mek-ess-ta. --/-- ‘Sumi ended up going home.’

ic-e forget-

k-a go-

Conative V₁-e/a po- ‘try V₁ing’ is syntactically distinct from the other nonthematic/ light V₂ patterns. This can be shown by scope of negation. Korean has both preverbal and postverbal clausal negation. In all of the patterns in Table 8.1, preverbal negation must precede V₁ (22a, b). In all of them except the conative (22c), negation must scope over both V₁ and V₂; in the conative it may scope over V₁ alone.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

202

     

(22)

a. Swumi=ka ilum=ul an cek-e Sumi= name=  write- cwu-ess-ta/twu-ess-ta/peli-ess-ta. --/--/-- ‘Sumi didn’t do the favor of writing / purposefully wrote down / went and wrote (her) name.’ Not: ‘Sumi did the favor of not writing / purposefully didn’t write down / went and didn’t write (her) name.’ b. *Swumi=ka ilum=ul cek-e an cwu-ess-ta/ Sumi= name= write-  --/ twu-ess-ta/peli-ess-ta. --/-- ‘Sumi didn’t do the favor of writing / wrote down / went and wrote (her) name.’ c. Swumi=ka ilum=ul an cek-e po-ass-ta. Sumi= name=  write- -- ‘Sumi didn’t try writing / tried not writing (her) name.’

V₁-e/a nonthematic V₂ complexes pass the tests for independent wordhood that we saw for thematic V₂ in Section 8.2. They can be split up by delimiter particles (23a), allow VP-pronominalization of V₁ (23b), and have separate accentuation of V₁ and V₂ in SKK (23c):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(23)

a. Swumi=ka ilum=ul cek-e=man Sumi= name= write-=only cwu-ess-ta/twu-ess-ta/po-ass-ta. --/--/-- ‘Sumi did the favor of / finished only writing (her) name.’ b. Swumi=ka ilum=ul cek-e Sumi= name= write- cwu-ess-ta/ twu-ess-ta /po-ass-ta. --/--/-- Minswu=to kulihay cwu-ess-ta/twu-ess-ta/po-ass-ta. Minsu=also doing.so --/--/-- ‘Sumi did the favor of/finished/ended up only writing (her) name. Minsu did the favor of/finished/ended up doing so too.’ c. L H L mantul-ta make-

HHL cwu-ess-ta --

LHL mantul-e make-

HHL cwu-ess-ta --

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  200

200

Pitch (Hz)

H

H

L

tul

0

L

cwu

ta 0.8326

Pitch (Hz)

180

man 0

0

L

tu

ess

ta

0

0.8584

H

H L

20

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

H

L

man 0

203

le

cwu Times (s)

H

ess

L

ta 1.189

Jeon (2013: 338–43) argues that the nonthematic sequences V₁-e/a na- ‘come out as V₁’, V₁-e/a nay- ‘bring out as V₁’, and V₁-e/a tul- ‘get V₁ed’ have single-word status, that is, are lexical compounds, but the arguments are unconvincing. Jeon’s first argument is that preverbal negation cannot scope over V1 alone, but must scope over both verbs (24a). But as we saw in (22), this is also true of sequences involving V₁-e/a twu- ‘keep Ved₁’, and V₁-e/a peli- ‘end up V₁ed’, even though, as we saw in (23) and Jeon recognizes, nonthematic twu- ‘put’ and peli- ‘discard’ do not form a lexical compound with V₁. Jeon’s next argument is that V₁ cannot be tensed, that is, sequences like *chac-ass-naysearch--bring.out. But this is simply a reflex of the fact that the past-tense stem V-ess-/ass-, like verb roots in Korean, is a bound morpheme. Jeon’s next argument is that na-, nay-, and tul- as V₂ cannot undergo Right Node Raising (24a). But again, this is also true of noncompound nonthematic V₂s such as twu- ‘put’ and peli- ‘discard’ (24b). (24)

pemin=ul cap-a a. *Pomwul=ul chac-a nay-ko, treasure= find- bring.out- criminal= catch- nay-ss-ta.6 .-- ‘(Pro) found the treasure and caught the criminal.’ (Jeon 2013: 340) b. *Pomwul=ul swumki-e twu-ko, pemin=ul chayphohay treasure= hide- put- criminal= arrest- twu-ess-ta. -- ‘(Pro) purposefully hid the treasure and arrested the criminal.’

⁶ Actually, speakers differ as to the judgment on (24a–c). While all speakers accept the interpretation where the V₂ nayssta ‘brought out’, twuessta ‘put’, peliessta ‘discarded’ scopes only over the second conjunct, some speakers accept the judgment where V₂ scopes over both, contra Jeon’s claim. In (24a) the first interpretation would mean something like ‘(Pro) having found the treasure, s/he caught the criminal’, where the first clause ending in gerundive -ko receives an adjunct interpretation. The second is a clausal conjunction interpretation: ‘(Pro) found the treasure and caught the criminal’. The important point here is that speakers who do not accept the second interpretation for (24a) also do not for (b–c).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

204

      c. *Pomwul=ul phokiha-y peli-ko, pemin=ul phwul-e treasure= abandon- put- criminal= release- peli-ess-ta. -- ‘(Pro) purposefully abandoned the treasure and released the criminal.’

As a final argument, Jeon claims that in some cases delimiter particles may not intervene between V₁ and na-, nay-, and tul- as V₂. But Jeon also acknowledges that in some cases particles may intervene. Below are the examples with Jeon’s judgments. (25)

a. Ai=ka cal-a =man/*=to/*=nun child= grow- =only/=even/= ‘The child has only/even/ grown.’

na-ss-ta. -- (Jeon 2013: 341)

b. Pomwul=ul chac-a=man/*=to/=nun nay-ss-ta. treasure= find-=only/=even/= .-- ‘(Pro) only/even/ found the treasure.’ (Jeon 2013: 341)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Again, not all speakers agree with Jeon’s judgments. The acceptability of delimiter particles depends on context, with =man ‘only’ generally the most acceptable, since its meaning can be inferred without prior discourse, unlike contrastive topic =nun and scalar =to ‘even/also’. This is shown by Jeon’s own example (25b), where =man ‘only’ is acknowledged as acceptable. It is easy to construct a context where all speakers accept =to ‘even/also’ breaking up V₁-e/a nay- as well: (26)

Cangso=lul al-a nay-ss-ul ppwun ani-la place= find- .-- only not.be- pomwul=ul cikcep chac-a=to nay-ss-ta. treasure= directly find-=also .-- ‘(Pro) not only located the place, s/he also found the treasure.’

To conclude this discussion of morphological status, all researchers concur that the majority of nonthematic complexes do not have compound (morphological word) status. A controversy exists over a small number of aspectual V₂s such as na- ‘come out and nay- bring out’, but we have argued that complexes headed by these do not have single-word status either. The inventory of nonthematic V₁-e/a V₂ sequences as shown in Table 8.1 partly parallels the inventory of Japanese V₁-te V₂ or Type 4 complexes (Kageyama 1993; Chapter 2, this volume) classified by Matsumoto (Chapter 6, this volume) as biclausal, but there are some differences. Korean lacks a counterpart of Japanese NP=ni V₁-te moraw- ‘have NP V₁’, while Japanese lacks a counterpart of Korean

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

205

perfective completive V₁-e/a mek- ‘eat = end up V₁ing’ with the corresponding V₂. We saw in (23) that conative ‘see=try V₁ing’, benefactive ‘give=do V₁ for’, and purposive perfective ‘place=keep V₁ed’ all allow VP pronominalization, matching one of Matsumoto’s tests for biclausality. However, as we saw in (22), only conative V₁-e/a po- ‘see=try V₁ing’ allows preverbal negation to scope only over the clause headed by V₁. This fact indicates that the “clausal” category headed by V₁ in the conative pattern must be large enough to include negation.

8.4 Other forms of V₁ in V-V sequences Examples (4)–(7) all show instances where a Japanese V-V compound is possible, but Korean must use a form of V₁ other than the infinitive in -e/a. Here we will explore briefly the conditions which block use of the infinitive, returning to this issue in Section 8.5. Sequential -e/ase in (4) is used when the events expressed by V₁..V₂ are characterized as a sequence of actions. (4) mil-e*(se) yel-ess-ta ‘pushed and opened’ constitutes a minimal pair with a V-V sequence like mil-e cechi-ess-ta ‘pushed wide open’, as shown in (27):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(27)

Swumi=nun mwukewun mwun=ul mil-e*(se) Sumi= heavy door= push- yel-ess-ta / mil-e cechi-ess-ta. open-- / push- opentr-- ‘Sumi pushed and opened the heavy door.’ (Adapted from NINJAL 2014)

Both corresponding V₂s allow V-V compounds in Japanese: (28)

Swumi=wa omoi doa=o os-i-ake-ta/ Sumi= heavy door= push--open-/ osi-hirai-ta. push--opentr- ‘Sumi pushed and opened the heavy door.’ (Adapted from NINJAL 2014)

The contrast in Korean is based on the causative nature of V₂. ThematicV₁-e/a-V₂ sequences where V₁ expresses the manner in which the action of V₂ is accomplished seem to require that V₂ be morphologically causative. (We discuss the relationship between the lexical aspect or Aktionsart of V₁ and V₂ in greater detail in Section 8.5.) Semantic causative status is not enough: yel- ‘opentransitive’ means ‘cause something to be in an open state’, but morphologically it is a simplex transitive verb. In contrast, cec-hi- in (27) is a morphological causative, formed

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

206

     

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Table 8.2 Object-sharing V₁-e/a thematic V₂ complexes with V₁ mil-/nwul- ‘push’ Acceptable with V₁-e/a V₂

Meaning

nwul-e press- mil-e push- nwul-e press- mil-e push- mil-e push- mil-e push- mil-e push- mil-e push-

‘press and kill, crush to death’

cwuk-i-ta die-- nayli-ta lower- kalaanc-hi-ta sink-- nemettuli-ta topple- nay-ta put.out- pwut-hi-ta attach-- pwus-wu-ta break-- cec-hi-ta incline--

‘push down’ ‘push and sink’ ‘push over‘ ‘push out’ ‘push on to’ ‘push and break’ ‘push open, push aside’

from the root cec- ‘incline in reverse’ and -hi-, the morphological causative suffix.⁷ This generalization is confirmed by the list of V-V sequences in Table 8.2, where V₁ is mil- ‘push’ or nwul- ‘push, press’, extracted from the NINJAL 2014 database. All but two of the examples involve the overt morphological causative morphemes -hi-/-ki-/-li/-i- or -wu-. In the case of mil-e nay- ‘push out’, causative -iderives nay- ‘put out’ from intransitive na- ‘emerge’. The two exceptions are mil-e nayli- ‘push and lower, push down’ and mil-e nemettuli-ta ‘push and topple, push over’ in (1). In the first case, V₂ nayli- ‘lower’ is one of the relatively few labile verbs in Korean, with both transitive and intransitive uses. It is possible that transitive nayli- incorporates an unpronounced causative morpheme, perhaps due to the homophony of its final syllable with causative -i-. In the second case, nemettuli- ‘topple’ is itself a V-V complex derived from V₁ nem-e ‘surpass, go over’ plus nonthematic V₂ ttuli-, which intensifies and typically transitivizes V₁, as shown in Table 8.1. Thus, all of the object-sharing examples in Table 8.2 but one have overt transitivizing or causativizing morphology on V₂. Example (5) illustrates another point of contrast between Japanese V-V compounds and Korean. In Modern Japanese, path verbs in V-V compounds occupy the V₂ position in combinations with perception verbs such as ‘see’, as shown in (29): (29)

Yama=o mi-age-ta/mi-orosi-ta. mountain= look-raise/ look-lower- ‘(Pro) looked up / looked down at the mountain.’ (Adapted from NINJAL 2014)

⁷ We owe this insight to Christina Kim (pers. comm.). Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

207

In contrast, as shown in (5), Korean path verbs precede the perception verb, but rather than appearing in the bare infinitive, they require the transferative particle =ta: (30)

San=ul olli-e*(-ta) /nayli-e*(-ta) mountain= raise-- /lower-- ‘(Pro) looked up / looked down at the mountain.’

po-ass-ta. see--

Transferative -ta can be attached to the infinitive, as in (30), the verb root (31a), or even to PPs (31b). It marks a shift in direction, position, or activity. (31)

a. Nwun=i o-ta snow= come- ‘It barely snowed.’

mal-ass-ta. stop--

b. Chayksang wi=ey-ta Table top=- ‘Please put it on the table.’

noh-a place-

cwu-sey-yo. give--

In addition to the path plus perception verb sequences in (30), transferative -ta can break up sequences of V₁ nonthematic V₂. When it does, as in (32), -ta emphasizes that the action of V₁ (borrowing) and V₂ (giving) took place in different locations:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(32)

Tosekwan=eyse ku chayk=ul pilly-e-ta library=from that book= borrow-- ‘Please pick up that book for me at the library.’

cwu-sey-yo. give--

Without =ta, (32) simply means ‘Please borrow a book from the library for me’; the speaker and hearer might both be at the library or planning to go there together. With -ta, (32) implies that the hearer is being asked to borrow the book at the library and bring it to the speaker, who is in a different location. In the case of the path–perception V-V sequences in (30), rather than implying that the two verbs denote distinct actions, -ta seems to mark the whole action (looking up, looking down) as a shift in direction from the subject’s previous stance. From this standpoint, -ta might be interpreted as a type of focus particle, with a “narrow” reading in examples like (31), and a “broad” reading in examples like (30). Note that, once again, the fact that -ta intervenes between the two verbs in (30) shows that they do not form a compound. (6)=(33) and (7)=(34) are our final examples where a V-V compound in Korean does not correspond to a V₁-e/a V₂ sequence in Korean. Japanese allows V-V compounds such as asobi-sugos- ‘play-spend = spend time playing’. In Korean, the parallel V-V sequences require gerundive -ko or conjunctive -mye:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

208

     

(33)

Na=nun tayhak ywuk nyenkan nol-*a/ko I= university six year play-/ ‘I spent six years of college playing around.’

(34)

Na nun tayhak ywuk nyenkan=ul nol-*a/mye I= college six years= play-- ‘I spent six years of college playing around.’

cinay-ess-ta. pass--

cinay-ess-ta. pass--

The syntax of gerundive -ko coordination has been extensively studied (e.g. Yi 1997; Yoon 1997), and it has been argued that -ko can be used for both clausal and VP coordination. These structures must be distinguished from the structure of both thematic and nonthematic V₁-e/a V₂ sequences, as the following contrast shows. With the exception of the scrambled pattern in (10), V₁-e/a V₂ sequences observe the same adjacency requirement as Japanese Type 4 V₁-te V₂ complexes demonstrated by Matsumoto (Chapter 6, this volume), while -ko coordination does not, as shown in (35): (35)

a. Minswu=nun chayk=ul hakkyo=ey Minswu= book= school=to ‘Minswu brought the book to school.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Minswu=nun Minswu=

chayk=ul book=

kaci-ko/-e hold-/-

kaci-ko/*-e hold-/-

hakkyo=ey school=to

ka-ss-ta. go-- ka-ss-ta. go--

There is one exception whereby V₁-e/a V₂ can be used as a coordination pattern, but this requires that the action expressed by V₁ be interpreted as complete. Thus (36) contrasts with (35b) because the accomplishment verb mantul- ‘make’ can receive a telic or completive interpretation, while the activity verb kaci- ‘hold’ cannot. (36)

Minswu=nun kwaca=lul mantul-ko/-e Minswu= sweet= make-/- ‘Minswu made sweets and went to school.’

hakkyo=ey school=to

ka-ss-ta. go--

A final difference between V₁-e/a V₂ sequences and Japanese V-V compounds is the absence in Korean of counterparts of Japanese Type 3 or syntactic V-V compounds such as V₁-i/e-hazime- ‘begin V₁ing (Kageyama 1993; Kishimoto (2006, Chapter 5, this volume). These patterns are realized in Korean as noncompound clausal complementation structures: (37)

a. Sensayngnim=i ilum=ul cek-ki(=ul) teacher= name= write-= ‘The teacher began writing her name.’

sicakha-si-ess-ta. begin---

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  b. Aki=ga wul-um=ul baby= cry-= ‘The baby stopped crying.’

209

kuchi-ess-ta. end--

The lexical complementizer -ki and the nominalizing suffix –(u)m derive clausal complements with external nominal properties (as shown by the possibility of accusative case marking in (37)) and internal clausal properties (as shown by subject honorific agreement in (37a)). Complementation with sicakha- ‘begin’ and kuchi- ‘end, finish’ and similar predicates is not possible using a V₁-e/a V₂ sequence.

8.5 Thematic V₁-e/a V₂ sequences A long tradition in syntactic analyses of thematic V₁-e/a-V₂ sequences dating back to Lee (1992) treats them as serial verb constructions (SVCs: cf. Chung 1993; Kang 1993; Zubizarreta and Oh 2007; Ko and Sohn 2015). This is based on similarities between the properties of thematic V₁-e/a V₂ and SVCs as originally identified in Twi by Stewart (1963) and described in work such as Sebba (1987). In a recent dissertation on SVCs, Cleary-Kemp (2015) reviews the properties of SVCs specified by various researchers and concludes that the defining crosslinguistic properties of SVCs boil down to four:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(38)

Defining properties of serial verb constructions (Cleary-Kemp 2015: 101) a. SVCs consist of components, each of which stand on its own as a main predicate. b. SVCs are monoclausal. c. An SVC describes something conceptualized as a single event. d. The verbs in an SVC share at least one core argument.

Korean thematic V₁-e/a V₂ sequences share these four properties. Verbs inflected with infinitive -e/a can stand alone in the so-called panmal ‘half talk’ informal pattern (39): (39)

a. Na=nun an mek-e. I=  go= ‘I won’t eat (it).’ b. Al-a. know= ‘(I) know.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

210

     

The monoclausality of thematic V₁-e/a V₂ sequences is shown by the fact that they cannot be broken up by preverbal negation of V₂, while preverbal negation before V₁ cannot scope over V₁ alone, but must scope over the VPs headed by both verbs: (40)

a. *Taymin=i Taemin=

kemok=ul big tree=

mil-e push-

an 

nemettuli-ess-ta. topple--

b. Taymin=i kemok=ul an mil-e nemettuli-ess-ta. Taemin= big tree=  push- topple-- ‘Taemin didn’t push over the big tree.’ (NOT ‘Taemin didn’t push and toppled the big tree.’) Thematic V₁-e/a V₂ sequences are interpreted like “sequential” SVCs in serializing languages: the VP headed by each verb expresses a subevent interpreted as part of a larger single event. As we saw at the end of the previous section, V₁-e/a V₂ can be used as a coordination pattern if the action expressed by V₁ be interpreted as complete or telic. Ko and Sohn (2015: 110) point out that in this case thematic V₁-e/a V₂ sequences can be broken up by time adverbials marking distinct events (Ko and Sohn analyze this pattern as involving TP adjunction, but we will continue to call it coordinate). In this pattern, in contrast to (40), preverbal negation of V₂ is possible, and preverbal negation of V₁ scopes only over the clause it heads. Note that sequential -se is also possible in this pattern:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(41)

a. Taymin=i achim=pwuthe kemok=ul mil-e(-se) halwu Taemin= morning=from big.tree= push- one.day congil mos nemettuli-ess-ta. whole.day  topple-- ‘Taemin pushed the tree from morning on and couldn’t topple it the whole day.’ b. Taymin=i achim=pwuthe kemok=ul mos mil-e(-se) onul Taemin= morning=from big.tree=  push- today nemettuli-ess-ta. topple-- ‘Taemin didn’t push over the big tree from morning on and today toppled it.’

Finally, all the examples of thematic V₁-e/a V₂ sequences we have seen share either the object argument, as in (1), or the subject argument, as in (8b). This last property is crucial for stating a condition stated by Chung (1993: 164) as “Preservation of the Highest Argument of the Head,” Lee (2003) as a requirement that V₁ and V₂ share the same semantic type, and Ko and Sohn (2015) as a stipulation that the v heads introducing each verb have the same featural property

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

211

introducing an external argument. Ko and Sohn (2015: 97) point out that their Matching Condition on Verbal Serialization in (42) is similar in effect to Kageyama’s (1993, 2016a, b) Transitivity Harmony Principle governing Japanese V-V compounds. (42)

Matching Condition on Verbal Serialization Verbs can be serialized with each other only when their v heads have the same featural property in introducing an external argument. (Ko and Sohn 2015: 97)

(42) predicts that verbs divide up into two groups. Unergatives, transitives, and causatives may all serialize with each other, because all have features introducing an external agent argument. Unaccusatives (including inchoatives) and passives form a second group of verbs that may serialize with each other, but verbs from these two groups may not serialize with one another. (1) is an example of two transitives serializing, while (8b) is an example of two unaccusatives. Ko and Sohn (2015) provide (43a) as an example of a simple transitive and a causative, (43b) as an example of an unergative and a transitive, and (43c) as an example of an unaccusative and a passive:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(43)

a. Taymin=i kaymi=lul palp-a Taemin= ant= tread- ‘Taemin trampled an ant to death.’ b. Taymin=i wulthali=lul ttwi-e Taemin= fence= jump- ‘Taemin jumped over the fence.’

cwuk-i-ess-ta. die--- nem-ess-ta. go.over--

c. So=ka cwuk-e mwut-hi-ess-ta. cow= die- bury--- ‘The cow died and was buried.’ ((43a, b, c) adapted from Ko and Sohn 2015) Thematic V₁-e/a V₂ sequences which do not match with respect to presence of an external argument result in ungrammaticality. Ko and Sohn provide (44a) as an example of a transitive and an unaccusative, (44b) as an unaccusative and a transitive, and (44c) as an unaccusative and a causative. (44)

a. *Taymin=i kaymi=lul palp-a cwuk-ess-ta. Taemin= ant= tread- die-- ‘Taemin trampled an ant to death.’ (intended) b. *Taymin=i mwul=ul kkulh-e Taemin= water= boil (intrans.)- ‘Taemin boiled and drank water.’ (intended)

masiss-ess-ta. drink--

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

212

      c. *Taymin=i kaymi=lul kwulm-e cwuk-i-ess-ta. Taemin= ant= tread- die--- ‘An ant starved and Taemin killed the ant.’ (intended) ((44a, b, c) adapted from Ko and Sohn 2015)

As Ko and Sohn point out, some of the examples in (44) are acceptable on other interpretations. For example, (44c) is marginally acceptable on the interpretation ‘Taemin caused the ant to starve and die’. Ko and Sohn account for this possibility by allowing the morphological causative affix -hi-/-ki-/-li-/-i- to attach inside the immediate projection of V₂, scoping only over it (as in (43a)), or outside the entire V₁-e/aV₂ sequence, scoping over both VPs. (44c) is marginally acceptable on the latter interpretation. On this interpretation, [starve] [die] are grouped together, satisfying the Matching Condition in (42); the causative then scopes over them both.⁸ Ko and Sohn’s basic structure for a ‘serializing’ thematic V₁-e/a V₂ sequences such as (43a) is given in (45): (45)

Structure for a ‘serializing’ thematic V₁-e/a-V₂ sequence (modified from Ko and Sohn 2015: 91) TP

Taymini=nom Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

vPcaus

vP1

ei antj=acc trample-inf

T[past]

vPcaus

ei proj die-CAUS

There are a number of details that are unclear in this structure; for instance, the identity of the category e that is the external argument of the vPs headed by ⁸ The marginality of (44c) on the interpretation ‘Taemin caused the ant to starve and die’ is due to the tendency of cwuk-i- ‘die-’ to be interpreted as a simple transitive ‘kill’. Ko and Sohn provide another example of this type that is completely acceptable: (i) 나영이 태민을 꿇어 앉혔다. (adapted from Ko and Sohn 2015) Nayeng=i Taymin=ul kkwulh-e anc-hi-ess-ta. Nayoung= Taemin= kneel- sit--- ‘Nayoung made Taemin kneel and sit.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

 - 

213

‘trample’ and ‘cause to die’, coreferent with the subject NP in the specifier of TP, Taymin.⁹ In any event, Ko and Sohn’s Matching Condition correctly predicts that verbs which require an external argument cannot group with verbs disallowing an external argument in a thematic V₁-e/a-V₂ sequence. The Matching Condition is a necessary condition for forming a noncoordinate, ‘serial’-type V₁-e/a-V₂ sequence, but it is not sufficient. Examples such as (27=4) and (33=6), that we saw in Section 8. 4, show that not all verbs which match with respect to presence of an external argument can form a verb sequence of this type. In (27) mil- ‘push’ and yel- ‘open are both transitives with external arguments, but they do not form an acceptable thematic V₁-e/a-V₂ sequence. The same applies to unergative intransitive nol- ‘play’ and transitive cinay- ‘pass’ in (33). What these examples indicate is that the Matching Condition for V₁-e/a-V₂ sequences must stipulate not just that the two verbs match with respect to presence or absence of an external argument, but that their Aktionsart or lexical aspect must match as well. For example, in terms of a Vendlerian classification of lexical aspect (Vendler 1957), mil- ‘push’ in (27) and nol- ‘play’ in (33) are activities, while the second verbs in these examples, yel- ‘opentransitive’ and cinay‘passtransitive’ are accomplishments. Both verbs in acceptable V₁-e/a-V₂ examples such as (8b) mikkuleci- ‘slip’ and tteleci- ‘fall’ are achievements, while both verbs cap- ‘grab’ and mek- ‘eat’ in (10b) are accomplishments. Ttwi- ‘run, jump’ in (43b) can have an accomplishment reading, as in han pakhwi=lul ttwi- (one rotation run) ‘run one time around’, matching the accomplishment reading of nem‘exceed, pass over’ in this example. We checked the hypothesis that thematic V₁-e/a-V₂ sequences require Aktionsart matching by examining various Korean and Japanese verb pairs in NINJAL (2014), a database of Japanese compound verbs with translations in English, Korean, and Chinese, using Japanese V-V compounds as a matrix. For instance, as an example of a typical activity V₂, ‘walk’ pairs with ten V₁s to form VV compounds in Japanese. Four of these involve transitive accomplishments, as V₁: ‘sell’, ‘bring’, ‘drink’, and ‘hold’. None of these allows V₁-e/a-V₂ sequences in Korean: ‘sell’, ‘bring/take with’, and ‘hold’ require gerundive -ko (cf. example (33)), while ‘drink’ requires conjunctive -mye (cf. (34)). As an example of a typical activity V₁, ‘play’ forms three V-V compounds in Japanese (in addition to ‘playpass time’ in (33=6)), with ‘pass the day’ (accomplishment), ‘have fun’ (possibly ⁹ This question raises a number of technical issues that are beyond the scope of this chapter. For example, the subject NP in the specifier of TP is usually considered to raise by movement from the Specifier of vP (Chomsky 1995). But if the subject is raised from the specifier of vP₂ (here, VP, considered the head vP by Ko and Sohn), and ei in the specifier of vP₁ is pro, the predicted result is a violation of weak crossover. If the subject is raised from the specifier of vP₁, the predicted result is a violation of the general prohibition on extraction from adjuncts known as the Condition on Extraction Domains (CED). The best solution may be to analyze raising of the subject to the specifier of TP as a case of across-the-board (ATB) extraction, or the kind of extraction found in parasitic gap contexts.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

214

     

an activity), and ‘get tired’ (an achievement) as V₂. None of these allows V₁-e/a-V₂ sequences in Korean. We looked at ‘die’ as an example of a typical achievement V₁. This verb pairs with five V₂s in Japanese in the NINJAL (2014) database: ‘hurry’ (activity), ‘be late for’ (achievement), ‘go extinct’ (achievement), ‘do completely’ (intensifier), and ‘part with’ (accomplishment). None of these allows V₁-e/a-V₂ sequences in Korean. In cases where a V₁-e/a-V₂ sequence might be predicted to be possible, such as ‘die + go extinct’, the expression is replaced by a Sino-Korean verb. In contrast, as a V₂, ‘die’ occurs in one Japanese V-V compound in the NINJAL (2014) database: ‘freeze-die (freeze to death)’. This is an achievement–achievement pairing, and it has an exact Korean counterpart: el-e cwuk- ‘freeze- die’. As a final check of the Aktionsart-matching requirement, we compared Japanese V-V compounds where V₂ is tuk- ‘attach to (intransitive)’ with their Korean counterparts. The NINJAL (2014) database includes fifty-one Japanese examples with this V₂. Some of them combine with clear transitive or unergative V₁s. None of these has Korean V₁-e/a-V₂ counterparts with the corresponding Korean intransitive achievement pwuth- ‘attach to (intransitive)’ as V₂. For example, Japanese kaziri+tuk- ‘bite-attach (bite forcefully)’ and kami+tuk- ‘biteattach (bite and attack)’ do not form corresponding V₁-e/a-V₂ sequences with puth- ‘attach to (intransitive)’ as V₂. The Korean counterpart in NINJAL (2014) has mwul- ‘bite’ as the main verb. Only the idiomatic Japanese expression in (46a) has a Korean V₁-e/a-V₂ counterpart with pwuth- as V₂:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(46)

a. Watasi=wa, nitiyoobi=wa terebi=ni I= Sunday= television= i-ru. (NINJAL 2014) be- ‘On Sundays, I’m glued to the television.’

kazir-i+tui-te bite-+attach-

b. Na=nun ilyoil=un thipi=ey tull-e I= Sunday= TV= approach- iss-ta. (NINJAL 2014) be- ‘On Sundays, I’m glued to the television.’

pwuth-e attach-

In (46b), both verbs ‘approach’ and ‘attach’ match in Aktionsart, as achievements. We have listed in Table 8.3 all of the Japanese V-V compounds that have corresponding Korean V₁-e/a-V₂ sequences taking pwuth- ‘attach (intransitive)’ as V₂.¹⁰ ¹⁰ The identity of the verb tall- in (4), (11–17), and (19–20) and tull- in (8) of Table 8.3 is unclear, as these do not appear as the stems of independent verbs. They are most likely intransitive counterparts of Middle Korean tol-, Modern Korean tal- ‘attach’, and its vowel-harmonic alternate. The reconstructed stem would be *tolol-.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

215

Table 8.3 V₁-e/a V₂ sequences taking pwuth- ‘attach (intransitive)’ as V₂ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

Japanese V-V compound

Korean V₁-e/a V₂ sequence

it-e + tukfreeze- attach ‘freeze over’ o-i + tukchase- attach ‘catch up to’ karamar-i tuk entangle- attach ‘get wrapped around’ kum-i + tukgrapple- attach ‘jump on to (someone)’ koor-i + tukfreeze- attach ‘freeze over/on to’ kog-e + tukbe.scorched- attach ‘be scorched on to’ kogo-e + tukfreeze- attach ‘get completely frozen’ kobir-i + tukburn- attach ‘burn and stick to pot’ sab-i + tukrust- attach ‘rust shut’ simi-i + tukseep- attach ‘seep in and stick’ syabur-i + tuksuck- attach ‘leap on and suck’ zyar-e + tukplay- attach ‘follow and play with’ su-i + tuksuck- attach ‘attach fast by sucking’ nebar-i + tukbe.sticky- attach ‘stick to’ har-i + tukstretch- attach ‘attach snugly’ hit + tukpull- attach ‘attach snugly’ hebar-i + tukbe.worn.out- attach ‘cling to’ maki-i + tukencoil- attach ‘wrap around’ matuwar-i + tukwrap.around- attach ‘get tangled with’ matowar-i + tukwrap.around- attach ‘get tangled with’ mo-e + tukburn- attach ‘catch fire’

el-e pwuthfreeze- attach ‘freeze over/ on to’ ttal-a pwuthchase- attach ‘catch up to’ engkhi-e pwuthentangle- attach ‘get wrapped around’ tall-a pwuth - ?- attach ‘adhere to’ el-e pwuthfreeze- attach ‘freeze over/on to’ nwull-e pwuthbe.scorched- attach ‘be scorched on to’ el-e pwuthfreeze- attach ‘freeze over/on to’ tull-e pwuth?/attach- attach ‘adhere to’ engki-e pwuthcongeal– attach ‘stick (e.g. rust) shut’ pay-e pwuthsoak-  attach ‘soak in and stick’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ kamki-e pwuthencoil- attach ‘wrap around’ tall-a pwuth?- attach ‘adhere to’ tall-a pwuth?- attach ‘adhere to’ nwull-e pwuthbe.scorched- attach ‘be scorched on to’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

216

     

All twenty-one V₁s in the Korean examples in Table 8.3 appear to be intransitive achievements. In cases where V₁ in the Japanese counterpart is a transitive activity, such as (11) ‘suck and attach’ and (12) ‘play and attach’, the closest Korean V₁-e/a-V₂ sequence is tall-a pwuth- ‘adhere to’, with the manner component expressed by the Japanese V₁ inferred from context. The only possible exception in these data to the generalization that V₁-e/a-V₂ sequences must match in Aktionsart is 2 in Table 8.3, ttal-a pwuth- ‘follow and catch up to’, where the independent V₁ ttalu- can be a transitive activity. In this V₁-e/aV₂ sequence, however, it seems to be coerced into the achievement reading ‘draw abreast with’. The striking absence of Korean V₁-e/a-V₂ sequences corresponding to Japanese Aktionsart mismatches such as ‘drink and walk’ or ‘bite and attach’ confirms our hypothesis that V₁-e/a-V₂ sequences are subject to a stronger Matching Condition, one that requires that V₁ and V₂ match not only with respect to presence or absence of an external argument, but with respect to verbal aspect or Aktionsart. There are a number of ways to implement this stronger condition. One would be to adopt Ko and Sohn’s (2015) approach and stipulate that the categories immediately dominating V₁ and V₂ have the same featural properties not merely with respect to introducing an external argument, but with respect to lexical aspect. A more elegant approach, we believe, is to analyze thematic V₁-e/a-V₂ sequences as true VP concatenation, dominated by light verbal heads responsible for introducing external arguments and lexical aspect. Approaches representing lexical aspect in terms of an extended light verb structure include Borer (2005) and Ramchand (2008). Under Ramchand’s “first phase syntax” approach, an initP (initiation projection, corresponding to vP) introduces external arguments; a procP (process projection, corresponding to VP) introduces internal arguments, and a resP (result projection) introduces arguments that undergo a change of state. Projections which share an argument in their specifier are coindexed. Under this approach, the verbs in a V₁-e/a-V₂ sequence such as the one in (10b) (in its unscrambled word order) cap-a mek- ‘grab and eat’ are both accomplishments, sharing subject and object and the projections that introduce them, as shown in (47):

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

217

vP (=initP)

(47)

kom=i VP1 (=procP)

bear=NOM

v

yene=lul salmon=acc resP1

cap-a

mek-

grab-inf eat resP1

salmon grab-inf

resP1

salmon eat

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

‘A bear grabbed and ate the salmon.’ The bracket notation in (47) indicates that the verbs ‘grab- ’ and ‘eat’ are realized first as the heads of resP, then again as the head of VP in accomplishments of this type. Whether this is produced by traditional verb raising or the process of ‘remerge’ discussed by Ramchand is immaterial to the present discussion. Similarly, the internal argument ‘salmon’ is the argument of both resPs, as well as of the VP. The fact that these projections share an argument is indicated by coindexing. In contrast, let us look at how the ill-formed V₁-e/a-V₂ sequence in (27=4) *mwun=ul mil-e yel- [door= push- opentr-] would be represented in this approach:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

218

     

(48)

vP (=initP)

*Swumi=ka Swumi=nom

VP (=procP)

mwun=ul

v

VP

door=acc

VPi

push-inf

resPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

door

yelopen

open

‘Swumi pushed the door open.’ (Ungrammatical as V1-e/aV2 sequence) (48) represents the attempt to conjoin an activity predicate (represented in Ramchand’s framework as a VP head by the bare verb mil- ‘push’, with no resP) and an accomplishment predicate headed by yel- ‘open’ and containing a resP with its own result predicate and an occurrence of the internal argument mun ‘door’. The resP and the accomplishment VP are coindexed, since they contain the same argument. But the predicate headed by mil- ‘push’ does not bear this index, making it effectively a different category. On the principle that only like categories can be conjoined, this approach correctly predicts (27=4) to be ill-formed. As a final note in this section, recall that at the beginning of Section 8.4 we observed that examples like the thematic V₁-e/a V₂ sequence in (1) mil-e nemettuli- ‘push-  topple’, which combines an activity verb and an accomplishment verb, are possible when V₂ is morphologically complex. As Table 8.2 shows, such examples occur when V₂ contains the morphological causative morphemes -hi-/ki-/-li/-i- or -wu-, or the intensifying stem nemettuli-. We have no definitive explanation why such examples relax the requirement that thematic V₁-e/a V₂ sequences match in Aktionsart. Within Ramchand’s framework, it is possible

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 - 

219

that these morphemes introduce an additional light verb head that is exempted from the coindexing requirement for argument sharing. We leave final resolution of this question for further research. In this section we have reviewed previous work on thematic V₁-e/a-V₂ sequences analyzing them as a type of serial verb construction. We focused on the insight in this work that Korean thematic V₁-e/a-V₂ sequences are subject to a version of Kageyama’s Harmony Principle, the Matching Condition of Ko and Sohn (2015). We then showed that thematic V₁-e/a-V₂ sequences are subject to a stricter condition, which requires that they match in lexical aspect or Aktionsart. We showed how this requirement might be implemented in Ramchand’s (2008) ‘first phase syntax’ approach.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

8.6 Conclusion In this chapter we have focused on Korean verb sequences linked by the infinitive suffix -e/a. The most basic subdivision within V₁-e/a-V₂ sequences in terms of the taxonomy introduced by Kageyama (Chapter 2, this volume) corresponds to the division between Type 1 (thematic) and Type 2 (lexical aspectual) compound verbs in Japanese. However, as we saw in Sections 8.2 and 8.3, Korean V₁-e/a-V₂ sequences are not compounds, that is, they do not have single-word status. Korean V-V compounds, as shown in Section 8.2, contrast with bare V₁-V₂ compounds in that they are highly lexicalized, with meanings frequently unpredictable from their component verbs. They are also unusual in Northeast Asia for the lack of any compound marker or linker whatsoever. Korean nonthematic (lexical aspectual) V₁-e/a-V₂ sequences show considerable similarity in terms of the aspectual categories they express with Japanese Type 4 (complex verbs) that have a similar function, e.g. conative ‘V₁ see = try V₁’ and perfective purposive ‘V₁ put = keep V₁ed’ (see Table 8.1). However, a striking difference between Korean and Japanese is the absence of a V₁-e/a-V₂ pattern corresponding to Japanese Type 3 (syntactic compound verbs). Instead, patterns like “begin the V” and “continue to V” are expressed by periphrastic constructions using clausal complementation in Korean. Finally, we saw in Section 8.5 that thematic V₁-e/a-V₂ sequences are subject to a condition similar to Kageyama’s (1993) Matching Condition. However the condition is even stricter in Korean, requiring a match in verbal aspect or Aktionsart as well. The Aktionsart-matching requirement accounts for the many discrepancies between Korean thematic V₁-e/a-V₂ sequences and Japanese Type 1 or thematic V-V compounds. The discrepancies between Korean and Japanese are all the more striking given the argument that the most deeply embedded layer of verbal inflectional morphology in Korean and Japanese contains many strong candidates for morphemes that are cognate

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

220

     

between the two languages. Although a cognate relation between proto-Korean (pK) and proto-Japanese-Ryūkyūan (pJR) is still unproven, verbal inflectional morphology provides a particularly strong argument for such a relationship (Whitman 2012). Candidate cognates include such inflectional affixes as pK adnominal *-r and pJR *-u/or and pK gerundive *-ko/u and pJR *-ku, but also two infinitive suffixes: pK adverbal *-i and pJR infinitive *-i, and pK infinitive *-ə/ a and pJR irrealis (mizenkei) -a. It is especially striking that both proto-languages had two distinct nonfinite endings for verbs. Their functional differentiation also appears to have been similar. Proto-Korean and pJR *-i were used to nominalize verbs, producing derived nominalizations such as Korean khi -e/a, which had largely displaced -i in this linker function in V-V sequences and other infinitival contexts, while in Japanese, the irrealis (mizenkei) suffix -a was lost in nonbound contexts after Old Japanese. Nevertheless, the functional differentiation between the two suffixes continued to have consequences. Perhaps because it never had a nominalizing function, Korean -e/a never developed the complementation patterns seen in Japanese Type 3 (syntactic compounds) formed with -i. The nominalizing function of -i may also have contributed to the development of this suffix in Japanese V-V compounds, as nominal coordinate compounds are pervasive in Japanese. To summarize, the development of the two infinitive verbal suffixes in different directions in the two languages accounts for the synchronic differences between thematic V-V sequences in these two languages. Despite these differences, commonalities between Japanese and Korean V-V sequences are indicative of strong areal and perhaps universal trends. The existence of a Matching Condition similar to Kageyama’s (1993) proposal for Japanese V-V compounds shows that generalizations of this sort are motivated independently of lexical or syntactic status. The large overlap in nonthematic V-V constructions attests to the prevalence of these patterns in Asian languages.

Acknowledgments The authors wish to thank Christina Kim, Hankyul Kim, and Mia Gong for their assistance. Whitman’s work was supported by the Laboratory Program for Korean Studies through the Ministry of Education of the Republic of Korea and the Korean Studies Promotion Service of the Academy of Korean Studies (AKS-2016-LAB-2250004).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

PART II

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

LANGUAGES OF SOUTH ASIA

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

9 Classification of complex verbs and the evolution of the compound verb in Marathi Prashant Pardeshi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

9.1 Introduction This chapter attempts to elucidate the functional properties of explicator/vector verbs as contrasted with auxiliary verbs with particular reference to Marathi, an Indo-Aryan language spoken mainly in the state of Maharashtra, located in western India. According to the 2011 census of India, the number of speakers is a little over 83 million. Marathi has a head-final structure with modifiers placed before modified elements and auxiliaries placed after main verbs. From a morphological point of view, Marathi verbs can be classified into two broad classes: (i) simplex verbs: those consisting of one morphological word, such as khā-ne : ‘to eat’, ye-ne : ‘to come’ (in Marathi -ne : indicates citation or dictionary form) and (ii) complex verbs: those consisting of two (or more) morphological words, such as kām kar-ne : [work (Noun) + do (Verb), N-V type] ‘to work’, kami kar-ne [less (adverb) + do (Verb), ADV-V type] ‘to reduce/lower’, and khā-un : t:āk-ne [eat (verb) + throw (verb), V-V type] ‘eat at once/eat completely.’ : Classification of morphologically ‘complex verbs’ [combinations of Noun + Verb, Adverb + Verb, and Verb₁ (nonfinite form) + Verb₂ (desinence)] is a contentious issue in Marathi grammar. Some propose putting together all types (N-V, ADV-V, and V-V) under one banner on the basis of their shared feature, i.e. morphological complexity, vis-à-vis simplex verbs, (Joshi 1925 [1889]; Kher 1899 [1895]; Navalkar 1880; Tarkhadkar 1899 [1836], to name a few), while others advocate splitting them into subgroups on the basis of semantic and/or syntactic criteria, some of which are explicitly mentioned, while others remain as tacit assumptions (Agarkar 1888; Damle 1970 [1911]; Godbole, Sathe, and Joshi 1897; Pandharipande 1993, among others). The goal of this chapter¹ is to offer a list of diagnostic criteria for the classification of Marathi complex verbs into functionally coherent subgroups by making ¹ This chapter is a revised and expanded version of Pardeshi (2003), in which a sustained discussion on the criteria for subclassification of complex verbs is offered.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

224

 

the tacit assumptions explicit and reconciling them with other criteria proposed in previous works. Of central concern is one subgroup, referred to as ‘compound verb’ (CV for short) in this chapter, which is a category shared by many languages spoken in the Indo-Turanian linguistic area.² The compound verb is a concatenation of a main/polar verb (V₁) and a vector/explicator verb (V₂). The ontogeny of V₂ (vector verbs) is also a contentious issue. Butt and her colleagues (Butt 2003, 2010; Butt and Lahiri 2002, 2013; Butt and Geuder 2003; Butt and Tantos 2004) argue that vector verbs are historically stable or eternal and do not undergo the changes that auxiliary verbs expressing tense, aspect, and mood do. In this chapter I shall attempt to shed light on the earliest attested phase of the CV in Marathi, drawing on the data from the thirteenth-century text called li:lāčaritra (assumed to be written around šake 1200 or 1278 ), and compare it with various intermittent phases up to the present with a view to describing the qualitative and quantitative changes the CV has undergone over the past 700 years. In the process of characterizing the development of the CV in Marathi, I will also compare it in other Indic languages, namely Marwari and Bangla, based on Hook (2001b) and Hook and Pardeshi (2005), to understand the synchronic differences among them in the proliferation (text frequency) of compound verbs. Statistical evidence showing gradual increase in the frequency of Marathi compound verbs over a span of 700 years suggests that vector verbs are not eternal and are subject to diachronic changes, pace Butt and her colleagues.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

9.2 Marathi complex verbs: a gallery With a view to offering a bird’s-eye view of Marathi complex verbs, I will first illustrate all types of complex verbs (N-V, ADV-V/ADJ-V, and V-V) with attested examples culled from various sources. • N-V type In the N-V type there are two subtypes: (i) the noun contained in the complex verb functions as its argument, i.e. it controls the agreement in perfective, as in (1),

² This subgroup of V-V-type complex verbs, widely known as compound verbs, is one of the hallmarks of the South Asian linguistic area (cf. Masica 1976) and has been a topic of debate in traditional as well as modern linguistic descriptions. It has been variously referred to by different scholars as modified verbal expression (Pořízka 1967–9), compound verb (Hook 1974; Singh, Subbarao, and Bandyopadhyay 1986; Singh 1998), explicator compound verb (Masica 1976; Abbi and Gopalkrishnan 1991; Gopalkrishnan and Abbi 1992), serial or compound verb (Kachru 1979; Kachru and Pandharipande 1980; Fedson 1985; Pandharipande 1990), and verbal expression (Nespital 1989, 1997c). In Indo-Aryan languages the CV is a true innovation (or a borrowing). It is absent in Sanskrit, starts showing up in the Pali writings from the early centuries of the Common Era, and then gradually expands in Modern Indo-Aryan languages (cf. Masica 1991; Hook 1991, Slade, Chapter 10, this volume).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

225

and (ii) the noun contained in the complex verb is not an argument of the verb, and thus does not control gender-number agreement in the verb of which it is a constituent, as in (2). (1)

t-yā-n-ni-ʦ ma.lā gapp rahā-n-yā-ʦ-ā : he---- I. quiet stay---- di.l-ā hot-ā give.- be.- ‘He was the one who advised me to keep quiet.’ (https://books.google.co.jp/books?id=n_HgDgAAQBAJ)

(2)

dev-ā-ne t-yā-lā ek kombdi bakšis di.l-i : god-- he-- one chicken. gift. give.- ‘God gifted him a chicken.’ (http://marathi.webdunia.com/kids-story-the-golden-egg-108020100003_1.html)

sallā advice.

• ADV-V/ADJ-V type

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(3)

a. plast:ik-ʦ-ā vāpar kami kar-uyā plastic-- use. less do-let.us ‘Let us reduce the use of plastic.’ (https://www.bobhata.com/science/plastic-eating-worms-and-insects) b. he koni kel.a? : tayār this who prepared do.-. Who prepared this?

• V-V type This type of complex verb is formally as well as semantically the most diverse, owing to the various nonfinite forms of the first verb (V₁) and the various meanings contributed by the second verb (V₂). Some representative examples are given below. (4)

to bhāt khā-t āhe/hot-ā he rice eat-ing be..3/be..3 ‘He is/was eating rice.’

(5)

to šā:l-e-t ʣā-t he school--in go-ing ‘He used to go to school.’

(6)

tu bhāt khā-t ʣā you rice eat-ing . ‘You should make it a habit to eat rice.’

ase be..3

[tense]

[habitual aspect]

[habitual aspect]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

226

 

(7)

ti gā-t rāh.i-l-i she sing-ing stay--3 ‘She kept on singing.’

(8)

te kām kar-āy-ts-a that work.. do--- ‘I have not yet done that work.’

(9)

rām bhāt khā-u lāg-l-ā Ram. rice eat- be.attached-- ‘Ram started to eat (his) rice.’

[continuative aspect]

[noninceptive/ uninitiative]

rāh.i-l-a stay--

[inceptive]

(10)

rām-lā japāni bhās: ā bol-tā ye-t-e [abilitative/ Ram- Japanese language. speak- come-- potential] ‘Ram can speak Japanese.’

(11)

mi ha:luha:lu ti-lā bol-ta ke.l-a I. slowly she- speak-  do.-. ‘Slowly I made her speak.’ (www.dainikprabhat.com)

[phrasal causative]

(12)

āi-ne šyām-lā bhāt kh-āy-lā lāv-l-ā mother- Sham-/ rice. eat-3- attach-- ‘Mother made Sham eat the rice.’

[coercive causative]

(13)

āi-ne šyām-lā bhāt mother- Sham-/ rice. ‘Mother let Sham eat the rice.’

(14)

ʦor-ā-lā šiks: ā thief-- punishment. ‘The thief was punished.’

(15)

rām-lā kām-ā-var-un kādh-n-yā-t ā.l-a [passive] : Ram- work--on-from remove--- come.- ‘Ram was sacked from the job.’

(16)

mi t-yā mul.ā-lā navin kot: kar-un di.l-ā [benefactive] I that- boy- new coat. do- give.- ‘I made a new coat for the boy.’ (Khandeker 1961: 102)

khā-u eat-

di.l-ā give--

ke.l-i do.-

ge.l-i go.-

³ Marathi has three or four allomorphs for .

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

[permissive causative]

[passive]

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

      (17)

mi tumh-ā-lā he sāŋg-u I. you-- this tell- ‘I would like to tell you that . . . ’

(18)

āvdi-niv di ādhi-ʦ sāŋg-un t:hev : : likes-dislikes beforehand- tell- . ‘Please let me/us know your likes and dislike beforehand (so that I/we can make preparations accordingly).’

(19)

čandrā-na do mit:-un : :le Chandra- eyes.. close- ‘Chandra closed her eyes.’

(20)

svatā-baddal-č-i khari māhiti to naka:lat sāŋg-un self-about--F true information. he. inadvertently tell- bas-l-ā [inadvertent/misguided] -. ‘Inadvertently he ended up giving genuine information about himself.’

(21)

gārdan-madh-il dršy pāh-un ghābar-un geli mahilā : garden-in- scene see-/ be.scared-  woman ‘Seeing the scene in the garden the woman was terrified.’ [intensivity] (https://divyamarathi.bhaskar.com/news)

(22)

dupār-paryant tān-un d.yā : noon-till stretch- . ‘Sleep till noon.’ (https://books.google.co.jp/books?id=jlcoDQAAQBAJ)

icchit-o want-

ki that

227

[desiderative]

[anticipatory action]

ghe.t-l-e [centripetal action] --. (Gokhle 1979: 176)

[idiom]

The functions of V₂ in examples (4) through (22) are varied: it expresses tense in example (4), aspectual contours in (5) through (9), ability or potential in (10), causative in (11) through (13), passive in (14) and (15), benefaction in (16), desideration in (17), anticipatory action in (18), centripetal action in (19), inadvertent/misguided action in (20), intensity in (21), and idiomatic meaning in (22). Various attempts have been made to classify the complex verbs (N-V, ADV-V/ADJ-V, and V-V types). Among them, classification of VV types (illustrated in (4) through (22)) has attracted more attention than that of N-V and ADV-V/ADJ-V types. In what follows, I will offer a critical review of previous work on complex verbs and then move on to propose formal and semantic criteria for identifying a homogenous subgroup of ‘V₁-un V₂’ type verb complexes, which are referred to as ‘compound verbs’ in this chapter.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

228

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

9.3 Previous studies of Marathi complex verbs Classification of Marathi’s complex verbs, instantiated in the previous section, is a matter of long-standing dispute among Marathi grammarians. As mentioned before, some propose putting together all types (N-V, ADV-V/ADJ-V, and V-V) under one banner on the basis of their shared feature of morphological complexity vis-à-vis simplex verbs (Joshi 1925 [1889]; Kher 1899 [1895]; Navalkar 1880; Tarkhadkar 1899 [1836], to name a few), while others advocate splitting them into sub groups on the basis of semantic and/or syntactic criteria, some of which are explicitly mentioned, while others remain as tacit assumptions (Agarkar 1888; Damle 1970 [1911]; Godbole, Sathe, and Joshi 1897; Pandharipande 1993, among others). Damle (1970 [1911]: 569–607), argues for identifying only V-V type constellations as sanyukt kriyāpad (complex verb) and also provides criteria to subclassify them. First, Damle provides a definition of V-V-type verb complexes in formal terms: concatenation of two verbs, with the first or main verb in a nonfinite form followed by the secondary verb. To this Damle adds a semantic criterion, that the first verb should be the bearer of the ‘primary’ meaning expressed by the concatenation. The secondary verb (V₂) Damle calls sahāy dhātu ‘helping verb’ and he classifies such verbs into two groups, namely arthavān (meaningful) and arthahin (meaningless), on the basis of whether or not they make a semantic contribution to the meaning of the primary/main verb. He further subclassifies the arthahin (meaningless) ones into three groups—Affirmative (as ‘be’, ho ‘become’), Negative (nas ‘not to be’, naho ‘not become’), and Voice markers (ʣā ‘go’, ye ‘come’, ho ‘become’)] and the arthavān (meaningful) group into eight groups—on the basis of the meaning they add— as summarized in Table 9.1 (the form of V₁ is supplied for clarity). According to the semantic criterion that the first verb (nonfinite) should be the bearer of ‘primary’ meaning, Damle distinguishes [V₁-un V₂]-type verb complexes

Table 9.1 Damle’s classification of arthavān (meaningful) V₂s in V₁+V₂ complexes Meaning

V₂ verbs

Example

Habitual Frequentative Potential Obligational Desiderative Permissive Help/Beginning Intensives

[V₁-t] + as ‘be’ [V₁-t] + ye ‘come’, ʣā ‘go’, ʦāl ‘walk’, bas ‘sit’, rāh ‘stay’ [V₁-tā] + ye ‘come’, [V₁-u] + šak ‘be able’, pāv ‘be able’ [V₁-āylā] + pāhije ‘need’, lāg ‘be struck’ [V₁-u] + pāh ‘want to’, čāh ‘want to’, iččhi ‘want to’ [V₁-u] + de ‘give’ [V₁-u] + ghāl ‘put on’, lāg ‘be struck’ [V₁-un] + thev ‘to put, keep’, de ‘give’, ghe ‘take’, bas ‘sit’, ʣā ‘go’, sod: ‘release’, kādh : ‘take out, draw’

(5) (6) (10)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

(17) (13) (9) (18), (19), (20), (21)

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

229

(his ‘intensives’ [examples (18) through (21)], from formally congruent serial verbs as in (23) below. (23)

dāt ghās-un kholi-t ʣā-un ʣhop teeth rub- room-in go-/ sleep. ‘Brush your teeth, go in the room, and sleep.’

[serial actions]

Damle (1970 [1911]: 569), citing the following examples, points out that, in the case of intensives, V₂ is bleached of its lexical meaning (desemanticized), while, in the case of serial verbs, V₂ retains its lexical meaning: Intensive (24)

t-yā-na he sabandh pustak he-- this complete book. ‘He wrote up this complete book.’

lih-un write-

kādh-l-a : .-.

Serial verb

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(25)

t-yā-na he āmbe dzhādā-var ʦadh-un : : he- - these mangoes. tree-on climb-/ kādh-l-e : take.out-- ‘He climbed on the tree and picked these mangoes.’

In example (25) (serial verb) primary meanings reside in both V₁ and V₂. Therefore, Damle does not recognize it as a sanyukt kriyāpad ‘complex verb’. From the foregoing overview of past research by traditional Marathi grammarians, it is clear that there are two positions on the membership of the category of complex verbs: (i) including all types (Tarkhadkar 1899 [1836]; Navalkar 1880; Kher 1899 [1895]; Joshi 1925 [1889]) and (ii) including only V-V types (Agarkar 1888; Damle 1970 [1911]; Godbole, Sathe, and Joshi 1897). In either case, the category of complex verbs is functionally heterogeneous in that it includes V₂s expressing objective grammatical meanings like tense, aspect, voice, and mood, on the one hand, and the speaker’s subjective perceptual meanings, like regret, selfbenefit, and preparative action, on the other. In modern linguistic studies like Masica (1976), Pandharipande (1990),⁴ and Hook (1974, 1991, 1993, 2001b), among others, complex verbs in South Asian languages are broadly classified into four groups: (i) Conjunct verbs (N-V, ADV-V/

⁴ It should be noted that what have been referred to as the compound verb (exx. (18) through (21) and (24)) and serial verb constructions (exx. (23), (25)) have been termed ‘serial verb construction’ and ‘conjunctive participial construction’ respectively in Pandharipande (1990).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

230

 

ADJ-V complex verbs), (ii) Main verb + auxiliary verb constellations: all subgroups identified by Damle (1970 [1911]) except ‘intensives’, (iii) Compound verbs: Damle’s intensives, and (iv) Serial verbs: Damle’s example (25).⁵ I endorse this four-way classification of complex verbs and provide explicit criteria for identifying the category of compound verb in Marathi, which expresses the speaker’s or protagonist’s subjective attitudes such as regret, self-benefit, and preparative action.

9.4 Criteria for identifying compound verbs: V₁-un V₂

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

As is clear from examples (24) and (25), serial verbs and compound verbs (Damle’s intensives)—and even some idioms, like the one in (22)—are formally identical (V₁-un V₂) showing that formal criteria alone are inadequate to identify compound verbs. It is necessary to set out precise criteria to delineate the class of compound verb and distinguish it from other subgroups. Research in the past few decades on CVs in New Indo-Aryan languages has yielded a consensus on identifying the class of compound verbs (Hook 1974, 1991, 2001; Masica 1976, 1991; Dasgupta 1977; Kachru 1979; Bhat 1979; Subbarao 1979; Nespital 1989, 1997c; Pandharipande 1990; Gopalkrishnan and Abbi 1992; Singh 1998, among others; see footnote 4 for another view). Making the tacit assumptions explicit and reconciling them with the explicit criteria proposed in past works, I provide a list of three criteria for identifying the class of CVs in Marathi.

9.4.1 The verb-verb complex should express one event The first criterion for identifying a verb-verb complex as a compound verb is that it should express a single event. Serial verb concatenations involve two (or more) verbs, each of which retains its lexical meaning and expresses an independent event. In contrast, in compound verb constellations, only V₁ retains its lexical meaning, while V₂ is ‘bleached’ of its lexical meaning. Owing to this, the compound verb constellation expresses one event: the one expressed by V₁. Note the contrast in the following pair of examples (also see the contrast between (24) and (25)). (26)

rām-na t:opi kādh-un dhut-l-i : Ram- cap. take.off-/ wash-- ‘Ram took off his cap and washed it.’

[serial verb]

⁵ In Butt (1995) and her works thereafter, the distinction between conjunct verbs and compound verbs is collapsed into one category and V₂ is called a ‘light’ verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      (27)

rām-na t:opi dhu-un kādh-l-i : Ram-  cap. wash- .-- ‘Ram washed his cap thoroughly.’

231

[compound verb]

Note that the V₁ kādh: ne : ‘take off ’ in example (26) retains its lexical meaning, while kādhne in (27) does not. The latter is homophonous with the lexical verb : : but is ‘desemanticized’ or ‘delexicalized’ to an adverbial meaning ‘to do the action expressed by V₁ thoroughly or to draw it to its extreme limit’. This distinction, however, is not always clear, and sometimes it is difficult to draw a line between CVs and serial verbs. In such a case of overlap the expression is ambiguous between a serial verb and a compound verb reading. (28)

nidān at least

bādli bucket.

bhar-un fill-

tari 

t:hev-āy-č-i {put/}---

parat again

a. ‘At least you should have refilled the bucket again and kept it ready for future use.’ [CV reading: Anticipatory action] b. ‘At least you should have refilled the bucket again and put it (there).’ [Serial verb reading: Temporally Sequential Actions] (Gokhle 1979: 64)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Using semantic bleaching of V₂ as a criterion, ‘serial verb’ type complex verbs and arguably bi-eventive causative, permissive causative, and benefactive verb sequences can be separated from the group of compound verbs, which express one event.

9.4.2 V₁ should be the sole licenser of arguments hosted by the verb-verb complex The second criterion for identifying a verb-verb complex as a compound verb is that V₁ should be the sole licenser of arguments hosted by the verb-verb complex. In other words, V₂ neither adds nor reduces arguments hosted by V₁. Note the contrast between (29) and (30). Both involve a ‘benefactive’ meaning. (29)

a. t-yā mul.ā-lā-ne ma.lā kholi sāph kar-un di.l-i that- boy.- I. room. clean do- give.- ‘That boy cleaned the room for me.’ [benefactive] b. *t-yā mul.ā-lā-ne ma.lā kholi that- boy.- I. room. ‘That boy cleaned the room for me.’

(30)

sāph clean

ke.l-i do.-

a. ti-ne sāra ma.lā samaʣ-āv-un di.l-e she- all I. understand-- .- ‘She did me the favor of explaining everything.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

232

  b. ti-ne sāra ma.lā samaʣ-āv-l-e she- all I. understand--- ‘She explained everything to me.’

In (29a), the beneficiary argument is licensed by V₂; hence, if V₂ is omitted, as in (29b), the sentence becomes ungrammatical. On the other hand, the beneficiary in (30a) is not licensed by V2; and hence omission of V₂ does not make the sentence ungrammatical, as shown in (30b). With the criterion of V₁ alone as a licenser of arguments hosted by the verb-verb complex, (30a) would qualify as a compound verb, while (29a) would not. In addition to the benefactive in (16) and (29), coercive causative, permissive causative, and passive (see examples from (12) through (15)), can also be isolated from the group of compound verbs with this criterion, since they add to or reduce the number of arguments hosted by V₁.

9.4.3 V₁ should be the semantic center of gravity of the verb-verb complex

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The third criterion for identifying a verb-verb complex as a compound verb is that V₁ should be the semantic center of gravity or, in other words, the semantic pivot. In the adverbial verb complexes illustrated in (31a) and (32a), V₂ is the semantic pivot, while V₁ is an adverbial or modifier element and hence can be dispensed with, as shown in the (b) versions. (31)

a. āmhi čāhā phuŋk-un py-āy-l-ā we. tea. blow- drink--- ‘We drank tea by cooling it enough for us to drink it by blowing it.’ b. āmhi čāhā we. tea. ‘We drank tea.’

(32)

p.yāy-l-ā drink--

a. mi phat:-i-t-un tsor-un pāh.i-l-a I. gap--in-from steal- see--. (default agreement) ‘I peeped through the gap stealthily.’ b. mi phat:-i-t-un pāh.i-l-a I. gap--in-from see-- ‘I peeped through the gap.’

In compound verbs, the semantic center of gravity is V₁, while V₂ functions as a modifier. On the other hand, in the case of adverbial V-V sequences, it is exactly the opposite—V₂ is the semantic pivot and V₁ serves as an adverbial or modifier

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

233

element. The criterion of V₁ as the semantic center of gravity helps to exclude verb-verb complexes where V₁ is not the semantic pivot from the category of compound verbs. Using this criterion, idiomatic V-V complex verbs illustrated in (33) and (34) also do not qualify as compound verbs, since V₁ is not the semantic center of gravity (in fact, they lack a semantic pivot, since their meaning is not the sum of the meanings of their constituents). (33)

a. udyog naslele lekhak dādāsāhebā-n-var tut:-un pad-l-e : business without writers Dadasaheb--on break- drop--. ‘Idle authors came down heavily on Dadasaheb.’ (Khandekar 1961: 121) b. *udyog naslele lekhak dādāsāhebā-n-var business without writers Dadasaheb--on ‘Idle authors came down heavily on Dadasaheb.’

(34)

a. mag t-yā-ne di.l-i tān-un : then he-- give.- stretch- ‘Then he slept on the lawn itself.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. *mag t-yā-ne tān-l-i : then he-- streth-- ‘Then he slept on the lawn itself.’

tut:-l-e break--.

hirva:li-vara-ʦ lawn-on-

hirva:li-vara-ʦ lawn-on-

In the foregoing discussion, three criteria were proposed for diagnosing whether a V-V concatenation qualifies as a compound verb or not. V-V complexes that satisfy all three of the criteria discussed in Sections 9.4.1 through 9.4.3 qualify for inclusion in the category of compound verbs in Marathi. Hook (2001b: 101, and footnote 2) characterizes compound verbs in South and Central Asian languages as “a sequence of two verbs AB (‘polar A’ plus ‘vector B’) that sometimes alternates with A (the ‘polar’) with little or no difference in meaning easily translatable into languages which do not have compound verbs (such as Sanskrit or English).” The criteria discussed in Sections 9.4.1 through 9.4.3 can be considered as a more concrete formulation of the ‘alternation’ between compound versus simple verbs as the key factor Hook highlights (see Section 9.5 for the discussion of the meaning contribution of V₂). The V₂s in compound verb constellations are widely referred to as ‘vector’ verbs, since many of the main verbs from which they derive have a component of directionality. Hereafter, I will call the V₂s occurring in the compound verb constellations vector verbs. Vector verbs differ from other V₂s occurring in the V-V complexes that express objective grammatical meanings such as tense, aspect, mood, and voice. Those I will label ‘auxiliary verbs’. The differences between auxiliary verbs and vector verbs (in comparison with main (lexical) verbs) are discussed in the next section.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

234

 

9.5 Vector verbs vs. auxiliary verbs vs. main (lexical) verbs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

First, both vector verbs and auxiliary verbs differ from main (lexical) verbs: main verbs constitute an open class (permitting addition of new members) and express lexical meaning. Vector verbs and auxiliary verbs, on the other hand, are a closed class (rarely permitting the addition of new members) and express nonlexical (or desemanticized) meanings. Their meanings are considered to have evolved from the lexical meanings of their homophonous main (lexical) verb counterparts via ‘semantic bleaching’, a common hallmark of grammaticalization. It is these similarities between vectors and auxiliary verbs, perhaps, that might have led traditional Marathi grammarians to treat them alike. They, however, differ significantly from each other in other ways. Auxiliary verbs and vectors can be distinguished on the basis of their distributional properties. Unlike vector verbs, auxiliaries are the outermost members of a verb phrase. Thus, a V-Aux concatenation is a closed one, while a V-Vector sequence can be extended further by an auxiliary, as exemplified below (vector verbs are shown in bold). (35)

šyām phār thak-un Sham extremely be.tired- ‘Sham was extremely tired.’

(36)

t-yā-n-ni te phād-un t:āk-l-a asel : he--- that.N tear- -- be... ‘He (honorific) might have torn it off.’ (Khandekar 1961: 250)

ge-l-ā go--

hotā be... (Khandekar 1961: 272)

Among auxiliary verbs, voice auxiliaries come right after the vector or main verb and may be followed by modal and/or tense/aspect marker auxiliaries in that order. (37)

patr phād-un t: āk-l-a ge-l-a {asāva/hota} : letter. tear- --. --. may../was.. ‘The letter {might have been/was} torn.’

It should be noted that the same lexical verb can have a ‘vector’ verb counterpart as well as an ‘auxiliary’ verb counterpart. Vector verb use and auxiliary verb use are not mutually exclusive. Verbs like GO and GIVE in Marathi are cases in point. The verb GO is used as a vector verb and also as an auxiliary of the passive. Similarly, the verb GIVE has vector verb usage and also is used as a permissive (causative) marker. As vectors, these verbs are subject to restrictions regarding which V₁s they can be hosted by, while, as auxiliary verbs, there are no such

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

235

restrictions on the V₁s they co-occur with (see Hook 2001b: 126, n. 4, for a similar situation in other languages of the Indo-Turanian area). Verb-Vector sequences are lexico-semantically constrained and are not a part of a regular contrast or paradigm in a CV-poor language like Marathi. In contrast to this, auxiliary sequences—expressing tense/aspect, mood, and voice—are in a regular paradigmatic relation with the verb with which they co-occur. The combination of verb plus auxiliary is, so to say, an open set. Almost any verb can co-occur with an auxiliary as long as it does not present what Pandharipande (1990: 181) calls ‘ontological incompatibility’. Consequently, auxiliary sequences need not be given in a dictionary as separate lexical entries. Nor do most main–vector verb combinations. Hook (2013a: 180–7) argues that not all vector verbs are grammaticalized to the same extent and proposes a spectrum of “vectorality” on which various vector verbs can be arranged according to their degree of semantic bleaching. At the less semantically bleached end of this spectrum lie those vector verbs which can be considered themselves as main (lexical) verbs. Hook calls such less grammaticalized vectors ‘factor’ verbs (or factors) and the compound verbs involving factor verbs ‘factored compound verbs’. In a factored compound verb, each constituent refers to an aspect of a single complex action, as in the concatenation of COME and ARRIVE, in which COME expresses the orientation of a movement toward a reference point, while ARRIVE expresses achievement of arrival. COME and ARRIVE, expressing two aspects of a single action, are temporally blended but conceptually distinct. The Marathi example from Hook (2013a) is reproduced here.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(38)

ti kāhi mint:ā-t-aʦ tithe ye-un pohoʦ-l-i she few minute-in- there come-/ arrive--. ‘She arrived there in just a few minutes.’ (https://books.google.co.jp/books?isbn=8171612075, accessed June 2020)

In view of the foregoing discussion, V-V type verb complexes in Marathi can be arranged on the cline of V₂’s semantic bleaching as follows. (39) Cline of semantic bleaching of V₂ Serial/adverbial verb type – Factor verb type – Compound verb type – Auxiliary verb type Less bleached

More bleached

Within the category of compound verb types as well, a continuum starting with marginally frequent vectors (e.g. ghāl-ne : ‘insert, pour’) through moderately frequent ones (e.g. t:hev-ne : PUT) to highly frequent vectors (e.g. ʣā-ne : GO) can be posited. Marginally frequent vectors or factors impose more semantic restrictions on the V₁s they collocate with. For example, in the case of ghāl-ne : ‘insert, pour’ the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

236

 

main verb should express force against resistance. In other words, there is a greater degree of semantic ‘compatibility’ between the main verb and the infrequent vector. For example, ghāl-ne : ‘insert, pour’ can collocate with only a handful of verbs, such as kond: ne : ‘lock up’ and bāndh-ne : ‘tie’, and adds a nuance of doing V₁ with force. This vector verb seems to be used in the Southern dialects of Marathi. The following example is from the Belgaum edition of a newspaper. (40)

t-yā-lā ʦor samʣ.un kāhi-n-ni khāmbā-lā he-- thief consider-ing some-- pole- bandh-un ghā.t-l-e hote tie- -- (default agreement) was ‘Considering him a thief, some people have tied him forcibly to the pole.’ (marvelousbelgaum.com/sakaal-7/, accessed June 2020)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Having given an explicit exposition of the criteria needed to identify the category of compound verb in Marathi, let us now turn to a brief discussion about the function of vector verbs (V₂) in compound verb concatenations. On the function of the V₂s in a compound verb, Masica (1976: 143) notes: the function of V₂ is varied and manifold: it connotes completion, suddenness, directionality, benefaction, intensity, violence, stubbornness, reluctance, regret, forethought, thoroughness, etc. depending on the item involved and on the circumstances . . . however, these all can be brought under one banner: greater specification (utochnenie) of features of the action already latent in the main verb itself or those that are compatible with it but have not yet been specified. It is for this reason that the combinations . . . are lexically selective; a given V₂ combines only with such V₁ as are compatible with it, or to put it another way, as have the semantic potential for it. Many unspecified V₁ have the potential for combination with a number of V₂, of course: the selection of the particular one depends on the demands of the situation as the speaker sees it or chooses to characterize it.

I agree with Masica on the function of V₂ in a compound verb sequence.

9.6 Evolution of vector verbs in Marathi: 1300–2000 As in other Indic languages, vector verbs in Marathi are homophonous with lexical verbs. A list of commonly and not so commonly used vectors, together with a rough idea of the nuances of meaning that can be ascribed to them, is given below.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      (41)

ghe de ʣā t:hev t:āk rāh pad: bas ye ān: [1] ghāl lāv lāg sod: sut: ne kādh : nigh dhar

237

TAKE GIVE GO PUT; KEEP

action for or toward oneself action for or toward others or neutral in orientation direction away; simple completion action with future use in view; firm assertion; continuation THROW speed, recklessness; psychological separation, relief STAY continuation of result of an action or event FALL unexpectedness or undesirability of action SIT unwelcomeness or misguidedness of action COME direction toward; change of state from within BRING successful completion, fullness of action INSERT completeness of action (rare) APPLY forceful completion of action (rare) TUCH reaction in response to coercive action (rare) RELEASE exhaustive (at times destructive) completion ESCAPE combines as a factor with main verbs of movement (rare) TAKE AWAY completeness of an action (limited in occurrence) PULL OUT exhaustive (at times destructive) completeness EMERGE fullness of result of an action or event HOLD continuation of implementation of an action

The V2 ān: co-occurs with very few main verbs: ghadav‘create, bring into : existence’ and ʣu:lav ‘connect; assemble’. Concatenations of each of these with ān, : for example, ghad-av-un ān: (come.into.existence-CAUS-CP BRING) and ʣu:l-av: un ān: (connect- CAUS-CP BRING) are transitive and have corresponding intransitive / inchoative counterparts: ʣu:l-un ye ‘succeed’ and ghad-un ye : (come into.existence-CP COME), as shown for the latter in (i) and (ii):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

[1]

vyavasāyā-t krānti ghad-un āli (i) yā ekā patrā-ne sanganak : : this one letter-ERG computer business-in revolution occur-CP COME ‘Due to this one letter a revolution occurred in the computer business field.’ (thebhandarkars.com/milind/archives/32) int:ernet:-jāhirātin-čyā kšetrā-t krānti ghad-av-un ān-li (ii) gugal-ne : : Google-ERG internet-adverts- field-in revolution occur-- come.- ‘Google brought a revolution into the field of Internet advertisements.’ (maharashtratimes.indiatimes) Of these nineteen V2s, the five most commonly used are the first five: ghe, de, ʣā, t:hev, and t:āk. The CV is not attested in Sanskrit, a language from which Marathi has descended. The intriguing question is: How do CVs arise in Marathi and other modern Indic

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

238

 

languages? To answer this question, one needs to trace the historical path modern Indo-Aryan languages in general have taken to reach the current stage. I will explore the historical development of the CV in Marathi through an empirical study and compare it with other Indic languages using data provided in Hook (1991, 2001b). The oldest work in Marathi prose is held to be the li:lāčaritra. This is a biography of the founder of the mahānubhāv sect, called čakradharasvāmi, and was composed after his death by two of his disciples: nāgadevāčārya and mhaimbhat: around 1278  (see Kolte 1982, prastāvanā: pp 59–66). The mahānubhāv sect produced a vast literature, and it was ciphered using a code known as saka:la lipi. Ciphering protected the language of the mahānubhāv texts from corruption and ensured that the language of the texts is of its time. These materials were deciphered in the 1960s by scholars such as V. B. Kolte and I. M. P. Raeside (see Tulpule and Feldhaus 1999 for details). The period in which li:lāčaritra was composed is called the Old Marathi period (seventh to thirteenth century ). The li:lāčaritra text consists of two parts, namely the purvārdh (Beginning part) and the uttarārdh (Concluding part). The text is 734 pages long, and it consists of 655 li:lā or episodes. For the present study I confined myself to the first 100 li:lā (episodes/anecdotes), a segment that runs to over eighty-four pages, and collected all the concatenations of verbs (V₁+V₂) occurring next to each other. These verbal concatenations share formal characteristics, namely the preceding verb (V₁) in a nonfinite form followed by another verb (V₂) carrying desinence. From the semantics of V₂ these concatenations can be classified into three broad categories: (a) combinations where V₁ is semantically dominant, (b) combinations where V₂ is semantically dominant, and (c) combinations where both V₁ and V₂ have equal semantic weight. Group (c) consists of cases of biclausal constructions conveying two temporally sequential actions, which I will refer to as ‘serial verb constructions’. Group (b) comprises ‘infinitival complement verb’ constructions, which are monoclausal and in which V₁ is subordinate to V₂. Neither (b) nor (c) is of interest to us here. Group (a) contains monoclausal structures and can be further divided into two major subgroups based on the degree of semantic beaching of V₂: (i) where V₂ is highly bleached and conveys grammatical meaning such as tense/ aspectual meaning and (ii) where V₂ is less bleached, that is, neither fully lexical nor fully grammaticalized. In the case of (i), there are almost no co-occurrence restrictions on V₁, while in (ii) collocability of V₂ with V₁ is subject to severe semantic constraints. As discussed earlier, I refer to the V₂s in the former group as ‘auxiliary’ verbs and to those in the latter group as either ‘vector’ or ‘factor’ verbs. The focus of the present inquiry is on the ‘vector’ and ‘factor’ verbs. The goal is to characterize the diachronic trajectory they have followed from the thirteenth century to the present.

9.6.1 A closer look at the verbal concatenations in the li:lāčaritra As mentioned before, I focused on the first 100 li:lās from the li:lāčaritra purvārdh, which is a text of eighty-four pages. I counted all the verb tokens in it. In the case of

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

239

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

verbal concatenations in which two verbs occur next to each other without any intervening element (V₁+V₂) I adopted the following two criteria for counting: (i) if V₂ is semantically bleached and can be deleted without causing any semantic anomaly, the concatenation is counted as one token (the V₂ in this combination is a vector or a factor verb) and (ii) if V₂ is semantically transparent, the concatenation is counted as two tokens (the V₂ in such concatenations includes aspectual auxiliaries, complement verbs, phasal verbs, and full lexical verbs). Adopting these criteria, in the pool of 2,966 verbs I found fifteen tokens in which V₂ can be potentially interpreted as a vector or a factor verb. The ratio of compound verb to simple verb is 15/2966, which comes out as 0.5 %. Thus, from the quantitative perspective, the CVs in the thirteenth-century Marathi account for only 0.5% of the total. Let us take a closer look at these fifteen tokens from a qualitative perspective. First, let us take a look at the V₂ slot in the fifteen concatenations which can be potentially treated as CVs. There are six tokens of GO, four tokens of KEEP/PUT, three tokens of GIVE, one token of SEE, and one token of TAKE. The examples are given here. (42)

te.yā.ciye āŋg-i romāva:li nigā-l-i ase he.. body- belly.hair. come.out-. be. tene te.yā.n.ce deh āčāda-uni gele ase [CV] : by that he... body cover- go. be. ‘Belly hair has come out on his body. Due to that his body has gotten covered.’ (li:lā No. 23, p. 16)

(43)

prema-sančār-mātr-e vikāru vikalpu svabhāvo devotion-perfusion-measure- passion error physical.necessity tin-hi učeda-uni gele [CV] three- dispel- went ‘Due to possession by devotion, (his) passion, error, and physical necessity, all three disappeared at once.’ (li:lā No. 29, p. 25)

(44)

hāt-i-ʦe kharga ga:la-uni hand-in- sword fall down/drop- ‘The sword in his hand fell down.’

gele went

[CV or SerV] (li:lā No. 26, p. 19)

(45)

to pau:li uda-uni gelā [CV or SerV] : that (tiger) surrounding.wall leap- went ‘That tiger leaped from the surrounding wall of the compound and went away.’ (li:lā No. 33, p. 32)

(46)

rānā pa:la-uni gelā : Rana run- go. ‘Rana fled away.’

[CV or SerV] (li:lā No. 67, p. 62)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

240

 

(47)

avaghā loku nigo-uni all people leave- ‘All the people have gone away.’

gelā go.

[CV or SerV] (li:lā No. 99, p. 83)

(48)

vidiyā kar-uni didhaliyā [CV or SerV] rolled.betel.leaf prepare- give. ‘He himself made the betel leaf roll and offered (him).’ (li:lā No. 27, p. 20 and li:lā No. 74, p. 67)

(49)

“āre ghe-ghe” mhana-uni prasādu kar-uni didhale [CV or SerV] : “Oh take-take” say-/ eatable do- give. ‘Saying “Oh, take it, take it” (the divine incarnation) tasted the edible item and gave it to him.’ (li:lā No. 20, p. 15)

(50)

mātāŋg-e pān-ā-či pāta:li kar-un t:hevili [CV or SerV] Matang- leaf-- leaf.dish make-/ keep. ‘A Matang (name of the community) person made a leaf dish and kept (it ready for something to be put on it).’ (li:lā No. 39, p. 37)

(51)

tie-vari done bhar-uni t:hevile asati [CV or SerV] : that-on leaf cup fill-/ keep. be. ‘Into one leaf cup (X) poured water and (X) placed (the cup on the leaf dish)’ (li:lā No. 39, p. 37)

(52)

eki donā udaka bhar-uni t:hevile [CV or SerV] : one.of.the leaf.cup water fill-/ keep. ‘In one of the leaf cups water was poured and kept (ready).’ (li:lā No. 39, p. 37)

(53)

suneyā-pu dhe undā kar-uni t:hevije [CV or SerV] : : : dog-before morsel.of.food make-/ keep. ‘Make a bit or a morsel of food and put it before the dog.’ (li:lā No. 76, p. 69)

(54)

mot:a sod-uni pāhili : bag untie-/ see. ‘(She) untied the bag and saw (inside).’

(55)

yā āpule daiv kādh-uni ghetale : she self fate/fortune pull- take. Lit. ‘She herself has pulled (invited) this fate/fortune.’

[SerV] (li:lā No. 94, p. 82) [idiomatic]

(li:lā No. 32, p. 31)

It should be added that in the foregoing examples from (42) to -(55) only the first two, namely (42) and (43), are clear-cut cases of vectorial CVs. These two cannot

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

241

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

be interpreted as temporally sequential actions (serial verbs) nor can they be regarded as concurrent aspects of single complex actions (factor verbs). In contrast, (44) through (53) are ambivalent, that is, they can be treated as either vectorial or factorial CVs or as cases of temporally sequential actions occurring one after another (serial verb). Example (54) is a clear case of a serial verb, while (55) sounds idiomatic. With regard to the temporally sequential actions (serial verb) exemplified in (45)–(54), the subject of V₂ does not appear overtly and is interpreted as being the same as that of V₁ since Indo-Aryan languages do not permit switch-subject verb concatenations wherein the object of V₁ is the subject of V₂. Had the subject of V₂ been expressed explicitly, it would have occurred between V₁ and V₂ and would have separated them. The occurrence of the subject of V₂ would block the CV interpretation, but since it does not occur, (45)–(54) should be treated as ambiguous between CV and SV interpretations. In this connection it is interesting to note that we found clear-cut examples of SV in the li:lāčaritra text, wherein an adverbial element intervenes between V₁ and V2, as exemplified here. (56)

hāt-i-ʦe kharga usa:la-uni parhā hand-in- sword spring.up-/ away ‘The sword in his hand sprang up and fell away.’

(57)

aise aika-uni murchā e-uni dadakari : Like.that hear-/ faint come-/ noisily padali [SerV] : fell ‘On hearing that, she fainted and fell down making a noise.’ (li:lā No. 32, p. 31)

padile [SerV] : fall. (li:lā No. 29, p. 25)

On comparing the examples of CV in twentieth-entury Marathi (exx. (35)–(36)) with those from thirteenth-century Marathi (exx. (42)–(53)), the following observations can be made: (I) The CV in the thirteenth century is still in the formative phase. Only two unambiguous tokens of vectorial CVs are attested, and both of these involve the vector GO, which is the most productive vector in the twentieth as well as the thirteenth century. (II) Factor verbs such as GO, KEEP, and GIVE are the most frequently used vector verbs in the thirteenth century. (III) The inventory of V₂ in the thirteenth century is much more restricted than that of the twentieth century. GO, KEEP, and GIVE are the only cases found in the thirteenth century, whereas in the twentieth century more than a dozen vectors are used: GO, COME, GIVE, TAKE, THROW, RELEASE, PUT/KEEP, SIT, DRAW, EMERGE, APPLY, FALL, etc. (see (41)).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

242

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The proliferation of the inventory of vector verbs seems to be directly proportional to the text frequency of the CV: 0.5% in the thirteenth century vs. 3–4 % in the twentieth century. However, at a certain stage, as overall CV flux increases, the relative prominence of the four least-marked vectors (GO, GIVE, TAKE, TOSS) also increases. Hook (1991: 60–1) offers a brief account of the diachronic development of CV in Indo-Aryan. CV is completely absent in Sanskrit, the mother language of the modern Indo-Aryan languages. Hook (ibid.) mentions that there are one or two ambiguous instances attested in late texts in Sanskrit which can be interpreted either as serial or as compound verbs. Hook also mentions that the languages that have compound verbs also have serial verb constructions, which are structurally homologous. Slade (Chapter 10, this volume) proposes that the CVs in Modern Indo-Aryan are derived from a reanalysis of structures involving a gerund combined with another verb (V₂). The data from li:lāčaritra already discussed endorse these observations. To know more about the emergence of the compound verb in Indo-Aryan languages, I compared the historical developments of CVs with other Indic languages, drawing insights from previous studies like Hook (1991, 2001b). As demonstrated by Hook (1991), the increase in flux is the same for Bangla, Marwari, and Marathi. Therefore, we can treat the characteristics of CV systems in various languages as representing successive phases along a developmental continuum. Let us take a detailed look at the diachronic development of CVs in other Indo-Aryan languages. However, the possibility of CV flux decreasing in some languages (Kashmiri?) should be kept in mind.

9.6.2 A Comparison of CVs in Marathi with other Indic languages While the Marathi compound verb seems to resemble that of Hindi-Urdu, Panjabi, Gujarati, and other Indo-Aryan languages in the choice of the grammaticalized items that serve as its auxiliary components or vectors, it differs from Table 9.2 Variation across Indo-Aryan languages in ratios of CVs to total verbs Language

Text type

Text dates

Counts

CV %

Hindi-Urdu Marwari Gujarati Bangla Bhitrauti Marathi Kashmiri

stories stories stories prose folktale stories folktales

since 1950 since 1950 since 1950 twentieth c. twentieth c. twentieth c. nineteenth–twentieth c.

7 2 2 4 1 2 3

15–20 % 13–18 % 14–16 % 10–13 % 8% 3–4 % 0.5–1.5%

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

243

Table 9.3 Variation across time in ratios of compound verbs to total verbs in Marathi texts Text

Author

Text type

Text dates

ʣāi-ʣui: prastāvanā mājhyā kathā ... pešvyānči bakhar bhāusahebānči bakhar pānipatči : bakhar dās-bodh šri-tukarāmči gāthā šridnyānešvari li:lāčaritra

Y. G. Joshi V. S. Khandekar N. S. Phadke

stories criticism essay

mid-20th c. 960 mid-20th c. 950 early 20th c. 720

Balaji G. Karkun?

chronicles

Krishnaji Shamrao? chronicles P. C. Govind? Raghunath Yadav chronicles Ramdas Svami Sant Tukaram

sermons poetry

Sant Dnyaneshvar

commentary 13th c.?? on the Gita anecdotes late 13th c.

Mhaimbhat

Total CVs

CV %

30 39 20

3.0 % 4.1% 2.8%

late 19th c.

1400 28

2.0%

late 18th c.

1200 15–18 1.3–1.5%

late 18th c.

1200 18

1.5%

17th c.?? 16th c.??

850 861

10 6

1.2% 0.7%

626

2

0.3%

1610 3

0.2%

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Table 9.4 Variation across time in ratios of compound verbs to total verbs in Bangla texts Text

Author

Text type

Text dates

Total CVs CV %

ghanadar galpa putulnačer . . . šrikānta gora bišabrkša ālaler gharer . . . annadamangal manasabijay šri-kršnakirtan : ramāyān

Premendra Mitra Bandyopadhyay Sharatchandra R. N. Tagore Bankimchandra P. C. Mitra Bharatchandra Bipradas Chandidas Krittibas

narrative narrative narrative narrative narrative narrative narrative? poetry?? poetry?? poetry

mid-20th c. mid-20th c. early 20th c. early 20th c. late 19th c. mid-19th c. mid-18th c. late 15th c. late 15th c. early 15th c.

5155 5313 5576 5103 4927 5113 4738 5127 4778 4862

599 591 707 523 257 332 77 95 99 51

11.6 % 11.1% 12.7% 10.2% 5.2% 6.5% 1.6% 1.9% 2.1% 1.0%

them on two scores: (A) It is not nearly so frequent in its occurrence and (B) it does not have so wide a range of functions. In Hindi-Urdu, for example, in connected texts the compound verb is used from four to five times more frequently than its Marathi counterpart. In Gujarati it is used twice as frequently, etc., as shown in Table 9.2 (simplified from Hook 2001b: 111).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 

244

Table 9.5 Variation across time in ratios of compound verbs to total verbs in Marwari texts Text

Author

Text type

Text dates

Total

CVs

CV %

khiraki uka:ltā āntarā mānsingh ri khyāt ajit vilās nainsi ri khyāt da:lpat vilās . . .

Rathor Moolchand anonymous anonymous M. Nainsi anonymous

narrative narrative chronicle chronicle chronicle chronicle

late 20th c. late 20th c. late 19th c. mid-18th c. late 17th c. early 17th c.

1118 1249 1125 1010 1292 1020

198 158 101 56 43 13

17.7 % 12.7% 9.0% 5.5% 3.3% 1.3%

20 18 16

Marwari Bangla Marathi

14 12 10 8 6

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

4 2 0 1300

1400

1500

1600

1700

1800

1900

2000

Figure 9.1 Increase in frequency of use of the Marathi, Marwari, and Bangla CV Source: Designed and created courtesy of Dr.Vengadesan; adapted from Hook 2001b: 109

In Indo-Aryan languages where the compound verb is more frequent, it has assumed certain functions that it does not (yet?) have in Marathi. Those functions include the expression of relative completion and the expression of perfective aspect (see Hook 2000b and Chapter 11, this volume, for details). In Marathi prose texts from about 1300  to the present there is a 1,600% increase in the ‘flux’ or text-frequency of the compound verb, as shown in Table 9.3. The data in Table 9.3 clearly shows that vector verbs are not historically stable. The gradual evolution of vector verbs is reflected in the increase in the text frequency or flux of the compound verbs of which they are a part.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

     

245

This gradual, monotonic climb in the incidence of compound verbs over seven centuries can be compared with parallel increases in the flux of compound verbs in Bangla (Zbavitel 1970) over the past 500 years (see Table 9.4) and in Marwari (Hook 1993) over the past four centuries (see Table 9.5). By plotting the points for each language on a graph and fitting a curve to each language’s points, we may obtain a clear picture of the inflation in the flux of the compound verb experienced by each of these three languages over the past several centuries (Figure 9.1). It is worth noting that while the Marathi compound verb shows a tenfold increase similar to that affecting the compound verb systems of Bangla and Marwari over a similar time period, it is more gradual; and at the beginning of the twenty-first century the Marathi compound verb remains significantly less frequent than the compound verb in Bangla and Marwari. As stated earlier, compound verbs in Marathi are not very frequent and are also restricted in function. In particular, they lack such functions as the expression of relative completion and the expression of perfective aspect (see Hook 2001b). Since Marathi compound verbs do not have these particular functions in the modern language and cannot be shown to have had them in the thirteenth century either, my primary concern in this chapter is to show the change in the kinds of compound verbs one encounters in Marathi texts from the very earliest periods as compared to the most recent ones. Comparing the compound verbs that are attested in the li:lāčaritra and the dnyānešvari with those in twentieth-century Marathi texts, the following differences may be observed. First, in the early texts, the vector components are usually “factors.” That is, they express some sub-part of the action expressed by the main verb. In (58) the vector verb gele WENT expresses a physical displacement that can be seen as an element of the overall event of the sword’s falling: (58)

hāt-i-ʦe kharg ga:la-uni hand-in- sword fall down/drop- ‘The sword in his hand fell down.’

gele went (li:lā No. 26, p. 19) (= (44))

Compare this with gelā WENT in Phadke’s essay on his own work, where physical displacement is not part of the situation expressed by the main verb: (59)

tyā-var-il t:ike-ʦā-ʦ “sāʦā” t:har-un gelā that-on- criticism-- pattern fix-  ‘The pattern of (their) criticism of it has become fixed.’ (Phadke 1962: 9)

Second, since vectors in the early texts are less grammaticalized than their modern homologues, the main verbs that they appear with tend to be restricted

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

246

 

to those expressing actions and events that are more compatible with the literal meanings of the vectors. A survey of the main verbs in the compound verbs in early texts shows that a much higher proportion of them express movement and other physical actions. In the modern texts, such verbs are still prominent as polar or main verbs in compound verbs, but their proportion of the total is markedly less. Finally, I turn to the ontogeny of vector verbs. Recently it has been asserted that the vector (or “light”) verb constituents of the compound verbs found in South, Central, and Northeast Asian languages differ in fundamental ways from auxiliaries of tense and mood (Butt and Lahiri 2002; Butt 2003; Butt and Geuder 2003; Butt and Tantos 2004; Butt 2010). A central contention (Butt 2003:15–16; Butt 2010: 19) is that while auxiliaries may evolve from full verbs over longer or shorter periods of time in ways deemed by students of grammaticalization to be universally valid pathways, the lexical sources of vector verbs do not evolve along those paths. Rather, their semantics is claimed to be sufficiently abstract to allow them to function immediately as explicators in compound verbs without the semantic bleaching or phonological ablation characteristic of the diachronic trajectory from full verb to auxiliary to affix that has been posited by “grammaticalizationists” (Heine 1993; Hopper and Traugott 1993: 108) for the development of verbal morphology. In the words of Butt (2010: 19): “as discussed in Butt and Lahiri (2003), we assume that auxiliaries are derived from the main verb, not the light verb. That is, we assume that light verbs are inert for the purposes of historical change. This idea accounts for the fact that a light verb always corresponds to a form identical main verb in the language and that light verb constructions do not give rise to auxiliaries and modals.” Counter-arguments to Butt and her colleagues’ claim that vector verbs, unlike auxiliary verbs, are immune to historical changes are offered in Hook and Pardeshi (2005) and by Slade (Chapter 10, this volume). The empirical demonstration of the steady increase (also known as inflation) over a span of 700 years in the text frequency or flux of the compound verb of which the vector verbs are a constituent implies that vector verbs are not eternal, pace Butt and her colleagues. They do evolve.

9.7 Summary In this chapter I have presented a brief review of past research on complex verbs in Marathi, pointing out that the category envisaged by traditional grammarians suffers from conflation and over inclusion. I offered a list of diagnostic criteria for the classification of Marathi complex verbs into functionally coherent subgroups by making the tacit assumptions of previous works explicit and synthesizing them with other criteria proposed in previous works. I then described the diachronic

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

     

247

development of the CV in Marathi from the thirteenth century to the present. The process shows as a steady increase in CV flux along with an increase in the number of vector verbs. In the initial stages of the CVs, the verbs occupying the V₂ slot are ambiguous between factor-verb and vector-verb interpretations. Comparison of Marathi with other Indo-Aryan languages shows that the rate of grammaticalization varies from one language to another. Finally, offering statistical evidence showing a gradual increase in the frequency of Marathi compound verbs over a span of 700 years, I have argued that vector verbs are not eternal and are subject to diachronic changes, pace Butt and her colleagues.

Acknowledgments I would like to thank Peter Hook and Taro Kageyama for their valuable comments and suggestions. The usual disclaimer applies. This chapter contains materials related to my work supported by the NINJAL project “Contrastive Studies of Japanese Prosody and Grammar” and “Development of and Linguistic Research with a Parsed Corpus of Japanese” and also by JSPS KAKENHI (Grant Number JP 15H03210).

Primary sources bhāusāhebānči bakhar. See Kanade, M. S.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Bhide, B. A., ed. and trans. (1929). sārths: ridnyānešvari. Mumbai: Keshav Bhikaji Dhavle Prakashan. Dadkar, J., P. Ganorkar, V. A. Dahake, and S. Bhatkal (eds) (1998). sankšipt marāt:hi vāŋmay-kos: [Concise Dictionary of Marathi Literature]. Mumbai: G. R. Bhatkal Foundation. dāsbodh. See Pangarkar 1968. dnyānešvari. See Bhide 1929. Gokhle, Arwind (1979). arvind gokhle yanči kathā: nivdak kathānʦā pratinidhik : sangrah [Stories by Arvind Gokhle: A Collection of representative stories written between 1945 and 1971], second edition. Pune: Continental Prakashan. Herwadkar, R. V. (ed.) (1975). pešvyānči bakhar of Balaji Ganesh Karkun (?). Pune: Venus Prakashan. Jog, Vishunubuwa Maharaj (ed. and transl.) (1999). sārth s: ritukārāmāči gāthā. Mumbai: Keshav Bhikaji Dhavle. Joshi, Y. G. (1949). ʣāi-ʣui. Pune: Continental Prakashan. Kanade, M. S. (ed.) (1993). bhāusāhebānči bakhar. Pune: Snehavardhan Publishing House.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

248

 

Khandekar, V. S. (1949). prastāvanā (introduction) to Y. G. Joshi’s dzāi-dzui. Pune: Continental Prakashan, 3–27. Khandekar, V. S. (1961). pāšānpujā [Stone worship, collection of short stories written : between 1925 and 1960]. Pune: Continental Prakashan. Kolte, V. B. (ed.) (1982). li:lāčaritra, second edition. Mumbai: Maharashtra Rajya Saahitya-Sanskruti Mandal (Maharashtra State Literature & Culture Board). li:lāčaritra. See Kolte 1982. Pangarkar, L. R. (ed. and transl.) (1924). sārth s: ridāsbodh, [1968 revised edition of Nārayanmahāraj Rāmdāsi]. Mumbai: Keshav Bhikaji Dhavle. pānipatči bakhar. See Yadav, Raghunath. : pešvyānči bakhar of Balaji Ganesh Karkun (?). See Herwadkar. Phadke, N. C. (1962) bāvankaši. Pune: Continental Prakashan. Tukaram. See Jog 1999.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Yadav, Raghunath (1997). pānipatči bakhar [Chronicles of Panipat], third edition. : R. V. Herwadkar (ed.) Pune: Venus Prakashan [Originally composed in 1763].

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

10 Development of verb-verb complexes in Indo-Aryan Benjamin Slade

10.1 Verb-verb compounds in Indo-Aryan One feature of modern Indo-Aryan (IA) languages—and South Asian languages more generally—is the employment of verb-verb compounds, which involve collocations of two verbs, where one (the “light” or “vector” verb) acts as a grammaticalized version of its full-verb counterpart. Thus, the verb GO¹ in many IA languages can appear as a light verb in verb-verb [VV] compounds, contributing various more functional semantic components, including completion.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

a. us-ne khānā (nahī ᷉ ) khāy-ā he/she.- food.. () eat._-. ‘He/she (didn’t eat)2/ate the food.’ (Hindi) b. vah khānā khā gay-ā he/she.. food.. eat. ._-. ‘He gobbled up the food.’ (Hindi)

Historically, the morphological form of the main or polar verb of the CV is an absolutive, a fixed, indeclinable form also used in earlier converb constructions, which also survive into the modern IA languages, sometimes (as in Hindi) within extended forms distinguishing them from the absolutive used in CVs. Compare (1) with (2), where the latter shows the sequential ordering of events use of the converb in (2a) or the co-eventual (Raina 2011) use of the converb in (2b).³

¹ Full capitals signals the light verb use of a verb; the gloss in capitals is the English sense of the verb when it occurs as a full/main verb. As discussed herein, the light verb senses are sometimes related transparently to their main verb meanings, e.g. Hindi GIVE, often signaling other-benefaction and/or outward-directed beneficiality; but sometimes they are related rather opaquely, as in Hindi SIT, signaling regret. ² In Hindi VV constructions are very common in perfective contexts with only special environments, like negation, triggering the use of simplex forms in non-marked uses (Hook 1974). ³ On ‘non-past’ readings of Sanskrit gerunds, see Hock (1992).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

250

 

(2)

a. vah khānā khā-ke ghar gayā he/she. food.. eat- home go._-. ‘(After) having eaten food, he went home.’ (Hindi) : : b. vah hams-ke bol-ī ‘hām’ he/she. laugh- say._-. ‘yes’ ‘Laughing(ly) she said “yes”.’ (Hindi)

Although VV compounds are a clear areal feature of South Asia (Masica 1976; Abbi 1991), at lower levels of analysis we find obvious and substantial differences in the structure, interpretation, and frequency of use of VVs across South Asia, including significant differences even between closely related languages. Hindi VVs occur in a more restricted range of contexts (see Hook 1988, 1991, 1993, Chapter 11, this volume), while VVs in Nepali and Marathi are less restricted (see Section 10.4.4). The strength of restrictions on licensed environments for VVs appears to co-vary with the frequency of occurrence of VVs overall. Thus, Hindi VVs appear to display overall a greater frequency of occurrence than do Nepali VVs. In this chapter, I discuss the historical development and properties of VV compounds in IA, with reference to VVs in Dravidian. For an overview of studies of VVs in IA, see Slade (2016); for an overview of studies of VVs in Dravidian, see Annamalai (2016); and for an overview of studies of VVs in Munda and TibetoBurman languages of South Asia, see Hock (2016). In Section 10.2 I review the history of modern IA VVs, including precursors of modern IA VVs in Section 10.2.1, VVs in Middle Indo-Aryan languages of Sri Lanka in Section 10.2.2, and VVs in early modern IA in Section 10.2.3. Section 10.3 examines later developments in modern IA VVs, including further grammaticalization into auxiliary elements of various types (including tense/ aspect-related auxiliaries). Section 10.4 focuses on the range of variation of morphosyntactic properties of IA VVs found even in closely related languages like Hindi and Nepali. Finally, Section 10.5 discusses the similarly varying inventories of light verbs used in VV constructions, not only in IA but also in neighboring Dravidian languages, and suggests that the culmination of the facts regarding variation in VV constructions throughout South Asia points to potential independently developed VV systems, with subsequent partial convergence.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

251

10.2 Historical Development of Indo-Aryan VVs 10.2.1 Prehistory of Indo-Aryan VVs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The morphological form of the main or polar verb in many modern IA VV constructions derives from the so-called Old Indo-Aryan (OIA) past gerund of prefixed verbs in -(t)ya/(t)yā (Chatterjee 1926; Hendriksen 1944; Tikkanen 1987), an element which has the “virtual value of an indeclinable participle” (Whitney 1879/1889: §989), which in general functions as a converb, meaning something like ‘having X-ed’ (though it does not always have a past tense value; as in modern IA, converbs in OIA can have the co-eventual type readings discussed by Raina 2011).⁴ Modern IA VV compounding ultimately derives from a reanalysis of structures involving a gerund combined with another verb (which for ease of reference I will refer to as V2, as it usually occurs following the gerund), so that the Sanskrit example in (3) can in a certain sense be seen as the formal precursor of Hindi (1b), repeated below as (4).⁵ (3)

annam sam-khād-ya gatah: asti food. together-eat- go._.. be..3 ‘He ate up the food and left’ (Lit., ‘Having eaten up the food, (he) left.’) (Sanskrit)

(4)

vah khānā he/she.. food.. ‘He gobbled up the food.’

khā eat.

gay-ā ._-. (Hindi)

The change from (3) to the VV construction of the type exemplified by (4) took place via the reanalysis of gerund + V2 as a single predicate, in which the semantic contribution of the gerund was taken to be primary, and the V2 as a modifying element (i.e. a vector or light verb). This only happened with a subset of verbs occupying the V2 position, specifically with verbs with broad semantic values, e.g. verbs meaning ‘go’, ‘give’, ‘take’, etc. Further, converb constructions did not disappear from modern IA; rather a sort of morphosyntactic split took place, with some gerund + V2 constructions being reinterpretable/reinterpreted as VVs. Clear examples of VV constructions of the modern IA type in (4), (5), and (6) do not appear until the modern period (see Masica 1991: 325; Slade 2013), with the exception of Sinhala (discussed in Section 10.2.2). Throughout the remainder of this

⁴ Peter Hook (pers. comm.) notes that in some IA languages (e.g. Marathi, Kashmiri) the form of the conjunctive participle is something other than a reflex of the OIA past gerund of prefixed verbs. ⁵ Example (3) and all subsequent examples from Sanskrit are shown in unsandhied form.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

252

 

chapter, whenever VVs are referred to, these are to be understood as full-verb + light-verb compound verb combinations of the prototypical modern IA-type. (5)

(6)

Hindi : a. maim yah bāt bilkul bhūl I this matter completely forget. cukā thā .._.. .. ‘I had completely forgotten about this matter.’ : b. gusse mem us-ne merā redyo : anger in he/she.- my.. radio dālā : ._.. ‘In anger, he broke my radio.’ Nepali a. āt:h baj-i-sakyo eight strike.-..3 ‘It’s already eight o’clock!’ b. anju-le bibek-lāī gupta Anju- Bibek- secret ‘Anju told Bibek the secret.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

tor: break.

kurā thing

bhan-i-hālyo speak--..3

While Butt (2010) and Butt and Lahiri (2002, 2013) argue that the modern IA system of VVs does not represent an innovation but rather continues a system of light-verb constructions found throughout all stages of IA, and that examples like (7), taken from R̥gvedic Sanskrit (the earliest attested form of IA), represent VV structures comparable to those found in modern IA, the historical developments in IA examined in this section make it clear that this is not the case, and that contra Butt (2010) and Butt and Lahiri (2013) VVs are just as susceptible to historical change as other verbal elements (see also Slade 2013; Coelho 2018; Hook and Pardeshi 2005 for additional evidence of the susceptibility of South Asian VVs to historical changes of various sorts). (7)

imé here

te yours

yé _

indra Indra.

té _

tvā+ārábhya you.+grasp.

vayám we.

cárāmasi move.1.

purus: t:uta praised-by-many. prabhūvaso rich-one.

a. ‘We here, O Indra, are yours, O one praised by many, [we] who keep holding on to you, O rich one.’ (complex predicate reading following Graßmann 1873: 437)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

253

b. ‘We here, O Indra, are yours, O one praised by many, [we] who having taken hold of you, move (around), O rich one.” (literal converb reading) (R̥gveda I.57,4 (Bandhu 1963–6); cf. Graßmann 1873: 437, Delbrück 1888: 406, Tikkanen 1987: 175) (Vedic Sanskrit) While, as Delbrück (1888: 406–7) observes, car- ‘move’ does seem to be able to bear an idiomatic sense when used with a gerund in examples like (7),⁶ there are several differences between modern IA VVs and light verb constructions in Sanskrit like (7). Firstly, unlike usage in modern IA VVs, the use of car- in (7) represents a grammatically peripheral construction, in the sense that it is not part of the central tense/aspect system of Vedic or Classical Sanskrit. Secondly, unlike modern IA VVs, which overwhelmingly impart a perfective sense, (7) and similar examples in Vedic where a combination of a gerund with another verbal form bears a noncompositional interpretation all involve a durative/continuitive sense. Rather, examples like (7) are actually similar to constructions that are grammatically more peripheral, found in modern IA languages involving combinations of present participles with a small set of light verbs: (8)

a. vah din bhar so-tā gay-ā he/she day full sleep._.. go._-. ‘He kept on sleeping all day.’ (Hindi)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. vah din bhar so-tā rah-ā he/she day full sleep._.. continue._-. ‘He kept on sleeping all day.’ (Hindi) (9)

: a. ma bhan-dai jān-chu, tam lekh-tai I. speak-_ go.-1, you. write-_ jā go. ‘I will keep dictating, and you keep writing.’ (Pokharel 1991: 194) (Nepali) b. āun-dai garnu come-_ do. ‘Keep on coming (from time to time).’ (Pokharel 1991: 194)

(Nepali)

Butt and Lahiri (2002: 23) point out the apparent similarity of Hindi constructions like (10) to a construction from Classical Sanskrit, (11).

⁶ See Hock (2008) and Slade (2013: 540–1) for further examples of VVs in Sanskrit like (7) which bear idiomatic interpretations.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

254

 

(10)

makkhī ur: fly... fly. ‘The fly flew off.’

(11)

gayī ._.. (Hindi)

tatah: maks: ikā+ ud: dīya gatā : then fly fly-up. go._.. ‘Then the fly, having flown up, left.’ (literal converb reading) (Pañcatantra 1.22, Tikkanen 1987: 176) (Classical Sanskrit)7

Notice that (11), however, can be interpreted naturally as a co-eventual use of the converb ud: dīya, i.e. “the fly left ‘flyingly’,” and does not require assuming a : modern IA VV analysis. In the next section I examine further potential early examples of IA VVs, including indubitable cases from early Sinhala.

10.2.2 Middle Indo-Aryan VVs in Sri Lanka More convincing examples of early VV of the modern IA VV variety are found in Middle IA, in Pali texts from Sri Lanka; the most persuasive of these is given in (12) (Hook 1993 also discusses these examples).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

: so tassā saddam sutvā . . . . . . he her cry. hear. . . . katvā adāsi make. give..3

: : assamapadam ānetvā aggim hermitage. bring. fire.

a. ‘He, having heard her cry, having brought her to his hermitage, made a fire (for her).’ (complex predicate reading) b. ‘He, having heard her cry, having brought her to his hermitage, having made a fire, offered it (to her).’ (literal converb reading) (Jātaka I, 296, 10; between 300 . and 400 ; cited by Hendriksen 1944: 134) (Pali) (13)

ath’ then

assa him

satthā master

: udakasāt:akam water-cloak

khipitvā throw.

adāsi give..3

a. ‘Then the master threw a bathrobe to him.’ (complex predicate reading) b. ‘Then the master gave him a bathrobe, throwing (it).’ (literal converb reading) (Dhammapad-At:t:hakathā II, 61, 10; post fifth-c. , cited by Hendriksen 1944: 134) (Pali) ⁷ Example (11) is shown in unsandhied form.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

255

Example (12) is reminiscent of Hindi constructions like: (14)

rām-ne sītā-ke liye āg Ram- Sita-for fire ‘Ram made a fire for Sita.’

jalā burn(vol).

diyā ._.. (Hindi)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

It is significant that these are found in Sri Lankan texts, as the earliest examples of undoubtable IA VVs occur in Old Sinhala, many centuries before the earliest clear examples of VVs in mainland IA. In Old Sinhala (c. eighth–tenth c. ) are found light-verb uses of GO (ya), TAKE (gan), PLACE (la), and SHOW (pa);⁸ examples of each are given in (15)–(18). (15)

Suka topa vi divi giya happiness. your. become..3 run. . ‘Your pleasure vanished.’ (Lit. ‘Your pleasure having run went.’) (Sigiri Graffito 498; 750–800 ) (Old Sinhala)

(16)

Næga gan Sihigiri rise. . Sihigiri ‘Ascend Sihigiri!’ (Lit. ‘Having risen take Sihigiri!’) (Sigiri Graffito 169; 850–900 ) (Old Sinhala)

(17)

Mitalu kiya lu gī Mitalu speak. . song.. ‘The song which has been recited by Mitalu.’ (Lit. ‘The song which having been spoken was placed by Mitalu.’) (Sigiri Graffito 498; 750–800 ) (Old Sinhala)

(18)

He:la-divi a Jabu-duvu datimi pataka Ceylon come.. Indian know..1 letter... di kese lami give. how ..1 ‘To Sri Lanka has come an Indian; I know how to give (him) a letter.” (Lit. ‘ . . . I knew how to, having given a letter (to him), place (it)’.) (Sigiri Graffito 293; 700–800 ..) (Old Sinhala)

⁸ The auxiliary verbs ya (past tense, giya) ‘to go’ and gan ‘to take’ emphasize the reflexive character of the preceding absolutive. The auxiliary verbs la ‘to place’, ‘to put’ and pa ‘to show’ either emphasize the transitive character of the principal verb or are merely periphrastic. (Paranavitana 1956: §501)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

256

 

Like the mainland IA VVs, Sinhala VVs also derive from a reanalysis of a combination of a verb in absolutive/gerund form with another verb, and such collocations are in fact in Old Sinhala still form-identical with VV collocation; see the example of a converb in (19). (19)

Mahamet-himiyā abu Nāl himiyabuyun ga liyu Lord Mahamet. wife Nāl Lady sing. written me gī this song ‘This song was sung and written down by Lady Nāl, wife of Lord Mahamet.’ (Sigiri Graffito 543; 800–900 ) (Old Sinhala)

Since clear examples of modern IA-type VVs are found in the colloquial language very early on, it is not surprising to find Sri Lankan Pali examples early on as well. Modern Sinhala still employs VVs (Paolillo 1989), though the set of light verbs is partially different from those of Old Sinhala, and includes TAKE (gannəwa), GIVE (denəwa), GO (yanəwa), COME (enəwa), and OPEN (arinəwa [volitive]; ærenəwa [involitive]).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(20)

a. gunəpālə kǣmak hadā9 gatta Gunapala meal. make.Ø . ‘Gunapala made a meal (for himself)’ (Paolillo 1989) (Modern Colloquial Sinhala) b. gunəpālə mahaeat:ə kǣmak hadə.la dunna Gunapala gentleman. meal. make. . ‘Gunapala made a meal for the boss.’ (Paolillo 1989) (Modern Colloquial Sinhala) c. gewal pol atu-wəliŋ hewili kərə-la ē houses coconut branches.. roofing do. that udətə piduru ihirō-la arinə wa : above. straw sprinkle. (vol)._ ‘Having roofed the houses with coconut branches, on top of that (they) cover it (up) with straw (i.e. sprinkle it completely with straw).’ (Paolillo 1989) (Modern Colloquial Sinhala)

Notably absent from modern Sinhala are either SHOW or PLACE, which occurred as light verbs in Old Sinhala VV constructions (as shown in (17)–(18)). SHOW (la) ⁹ Interestingly, VVs with gannəwa, unlike the others, use the (fossilized) old Sinhala absolutive form without la for the main/polar verb, glossed here and throughout as Ø. See the quotation from Paranavitana in the next paragraph on the evolution of the old light verb la to a marker of the absolutive.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

257

and PLACE (pa) as light verbs evolved to serve other functions at some point during the period between Old Sinhala and the modern language: The absolutive form of the verb la used as an auxiliary has . . . become the mark of the absolutive (in modern Sinhala); similarly the imperative form of the auxiliary verb pa (e.g. ela pan) has in the modern language been reduced to the character of a suffix indicating the imperative mood. (Paranavitana 1956: §501)

Sinhala is thus interesting both in having developed VVs very much earlier, and in that some of these early light verbs subsequently underwent further grammaticalizing developments. The next section examines the earliest uncontroversial examples of VVs in mainland IA languages.

10.2.3 Early modern Indo-Aryan VVs The first unequivocal examples of VV constructions in mainland IA languages do not appear until the early modern IA period (sixteenth–eighteenth c.).¹⁰ See the examples from early Braj Bhās: ā (a close relative of the ancestor of modern Hindi) in (21), and (22), and the “Old Shah” Nepali examples in (24) and (25).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(21)

nātaru otherwise

jau if

vacchā calf.

mari die.

jāi, ..3,

tau then

gāi cow.

cchim jāi : dāī : take_away._.. go..3 a. ‘Otherwise, if the calf dies, then the cow is taken away.’ (complex predicate reading) b. #‘Otherwise, if the calf having died goes, then the cow is taken away.’ (literal converb reading) (c. 1600 , Indrajit of Orchā’s commentary on the Nītiśataka of Bhartr̥hari, f.18b4; McGregor 1968: 57) ¹⁰ Masica (1991: 326–30) concurs on this point. It has been suggested that modern IA-type VV : constructions are to be found in Apabhramśa (see discussion in Hook 1977; Bubenik 1998). Given the : uncertainties around the interpretation of Apabhramśa (see Slade 2013: §3.4), I exclude it from consideration here. Marathi shows even earlier examples of VV constructions (Peter Hook and Prashant Pardeshi, pers.comm.): (i)

teyaaciiye aangii romaawaLi nigaali ase. he. body. belly.hair come.out be. teNe teyaance deha aachaada-uni gele ase by.that he. body cover- go. be.  ‘Belly hair has come out on his body. Due to that his body has got covered up.’ (LP, Lila No. 23, p. 16)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

258

 

(22)

bhayabhīta hvai-kari samudra kau mathivau na cchād-i : without.fear be- ocean of churning not leave- dayau ._.. a. ‘Having become fearless, (he) did not leave off the churning of the ocean.’ (complex predicate reading) b. #‘Having become fearless, (he), having not left the churning of the ocean, gave.” (literal converb reading) (c. 1600 , Indrajit of Orchā’s commentary on the Nītiśataka of Bhartr̥hari, f.17a; McGregor 1968: 54)

(23)

2 ālā motipur-kā 2 ālā vaks-i 2 field.. Motipur. 2 field.. bestow. diyā̃ chãu ._.. be..1 ‘We have given two fields of Motipur.’ (Old Shah Nepali, 1529 .; Wallace 1982: 164)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24)

(25)

(26)

: tahām kājikana vujhāi samjhāi ñahā there Kazi. understanding understand.. here pat:hā-idinu havas send-- . a. ‘Persuade the Kazi there, and send him here.’ (complex predicate reading) b. #‘Persuade the Kazi there, and having sent him here, give (him).’ (literal converb reading) (Old Shah Nepali, 1755/6   1812 V.S., Prithvinarayan Shah, letter to Pan: dit : Rājīvalocana; Pokharel B. S. 2020 (=1963 ): 196) : tyo cāmdo māgera pat:hā-i- deu : that(thing) quickly request. send-- . a. ‘Quickly request that thing and send it (to me).’ (complex predicate reading) b. #‘Quickly request that thing and having sent it, give (it) (to me).’ (literal converb reading) (1767/7   1824 V. S., Prithvinarayan Shah, letter to Haripan: dit, : Pokharel B. S. 2020 (=1963 ): 211) pachillā cit:hi-ko javāph hāmi-le les: -ilast. letter... answer we. write. rākhyā-thyãu ._..-be..1 ‘We wrote our answer to the last letter.’ (Old Shah Nepali, 1792 ; Wallace 1982: 191)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

259

(27)

timiheru-le aneka prakār-kā pāp gar-i you.- many way... sin do. rās: yā-chau ._..-be.2 ‘You have committed sins in many ways.’ (? Or, ‘You kept committing sins in various ways.’) (Old Shah Nepali, 1798 ; Wallace 1982: 191)

(28)

unai-ki maiyā hā̃ mi-le lyā-i pal-i she.-. daughter we. bring. nurture. rās: yāki chan ._.. be.3. ‘We took her daughter in and raised her.’ (? Or, ‘We took her daughter in and kept nurturing her.’) (Old Shah Nepali, 1800 ; Wallace 1982: 193)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

GIVE seems by far the most common light verb to appear in early modern IA VVs, though we also see an early instance of GO. The appearance of these two verbs is perhaps unsurprising, since these are two of the most well-established light verbs to appear in VVs not only in IA but also in Dravidian (see Section 10.5). VVs subjectively seem uncommon in Old Shah Nepali, with numerous comparable passages containing simplex verbs.¹¹ And even in modern Nepali the frequency of VVs seems significantly lower than in modern Hindi. In the next section examples of VVs undergoing further grammaticalization are examined, resulting in some cases in new tense/aspect auxiliaries or honorific auxiliaries.

10.3 The fate of IA light verbs Like other partially grammaticalized elements, light verbs sometimes experience further grammaticalization.¹² Despite suggestions to the contrary (Butt 2010; Butt and Lahiri 2013), the light verbs which participate in modern IA-type VVs can thus ¹¹ Compare: van pās: o inu tin janā-le hāmu diyāko cha forest hill these three man. we. give._.. be..3 ‘These three men have given us the forest and the hillside.’ [Old Shah Nepali, 1590 ; Wallace 1982: 164] (ii) kāt:hmādãu-le khosyā ādhā gorkhā-kan adinu : Kathmandu. open._ half Gorkha. give. ‘To give Gorkha half what is opened by Kathmandu.’ [Old Shah Nepali, 1757 ; Wallace 1982: 192] (i)

(iii) tati gharyādi-samet hā̃mi-le vaksyãu : thus house-etc.-with we. bestow..1 ‘We have given thus the house and other things.’ [Old Shah Nepali, 1767 ; Wallace 1982: 179] ¹² See also Coelho (2018) for similar observations of change in the VV system of the South Dravidian language Betta Kurumba.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

260

 

undergo independent change—including phonological change which does not affect the full-verb counterpart, as well as undergoing additional grammaticalization leading to functions more typical of elements we would call auxiliaries rather than light verbs. Further, not all light verbs do, in fact, possess full-verb counterparts.

10.3.1 Independent phonological change of light verbs The Nepali light verb baksinu provides an example of change affecting a light verb independently of its full-verb counterpart. Nepali baksinu is a light verb employed in VV constructions as an honorific when referring to Nepali royalty and other persons due great respect, e.g.: (29)

mausuph-le yo kuro ghosanā : His.Majesty- this thing proclamation gar-i- baksi-yo do-- ..-3 ‘His Majesty made this proclamation.’ (Sharma 1980: 131)

(Nepali)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Both baksinu and a phonologically reduced form are also employed in upper-class Kathmandu families (e.g. by children to parents, wife to husband, etc.), as discussed in Sharma (1980: 130–2). An example of the reduced form of baksinu is given in example (30). (30)

buwā-le bhujā khā-i- s-yo dad- rice eat-- -3 ‘Dad ate rice.’ (Sharma 1980: 132)

(Nepali)

While baksinu is employed as a main verb (31), -s- is not (32). (31)

(32)

mahārāni-le ma-lāi takmā baksi-yo queen- I- medal bestow-.3 ‘The queen bestowed a medal upon me.’ (Sharma 1980: 132)

(Nepali)

*mahārāni-le queen-

(Nepali)

ma-lāi I-

takmā medal

s-yo -.3

Thus, while baksinu as a light verb can occur in both a reduced and an unreduced phonological form, the full verb form can occur only in unreduced form.¹³

¹³ Nepali baksinu is ultimately a loanword from Persian bakhś ‘to give’, and obviously was borrowed as a full verb, not as a light verb, since, as shown by example (31), it can still be used as a full verb. See Section 10.2.3 for earlier examples of full-verb baksinu. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

261

10.3.2 Light verbs lacking full-verb counterparts Though there seems to be some tendency for light verbs to continue to have full-verb counterparts, this is not always the case. Nepali, for example, employs t:opalnu as a light verb indicating pretense, as in example (33), but t:opalnu does not exist as a full verb. (33)

u gā-it:opal-daicha he/she sing-- -_- be..3 ‘He is pretending to sing.’ (Pokharel 1991: 195)

(Nepali)

Turner (1931: 247) suggests that a full-verb counterpart may have once existed, possibly meaning ‘to cover’ (cf. Nepali t:opi ‘helmet’; cf. Hindi t:opī ‘hat’, Hindi t:opnā ‘to cover’); cf. (34). (34)

*u t:opal-daihe/she pretend-_‘He is pretending.’

cha be..3 (Nepali)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Likewise, Hindi saknā ‘to be able to’ behaves syntactically like a light verb, and also lacks a nonlight/nonauxiliary full-verb counterpart (see Slade 2013: §4.2.3 for further discussion).

10.3.3 Further grammaticalization of light verbs As discussed in Section 10.2.2, the Old Sinhala light verb uses of SHOW (pa) and PLACE (la) underwent further grammaticalization, ending up in the modern language as markers of the imperative mood and absolutive form, respectively. In the following two sections I consider two further instances of vector-type light verbs developing into tense/aspectual auxiliaries in Hindi and Nepali.

10.3.4 The development of the Hindi auxiliary rahnā The Hindi continuative auxiliary use of rahnā also originated as a light verb before being further grammaticalized as part of the core grammar of aspect in Hindi. However, in Modern Hindi, the simple present and the present continuous are clearly distinguished, as shown in (35).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

262

 

(35)

Present-day Hindi : : a. maim skūl jātā hūm I school go._.. be..1 ‘I (habitually) go to school.’ : b. maim skūl jā rahā I school go. ._.. ‘I am going to school (just now).’

: hūm be..1

: : However, Kellogg (1893: §404) cites forms like maim jātā hūm as meaning either ‘I go’ or ‘I am going’.¹⁴ Further, he categorizes rahnā as a light verb (Kellogg 1893: §428) rather than an auxiliary, suggesting that—as in the case of Modern Nepali rahanu—the use of rahnā as indicating continuous action in nineteenth-century Hindi was more peripheral and not yet integrated as part of the core grammar as a clear aspectual auxiliary; see (36) and (37), taken from Kellogg (1893: §404), retaining his translations.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(36)

(37)

19th-century Hindi : a. donom lar: ke khelte the both boy. play._.. be..3 ‘The two children were playing.’ : b. donom lar: ke khel rahe both boy. play. ._.. ‘The two children were engaged in play.’ 19th-century Hindi a. vah suntā he hear._.. ‘He hears.’ / ‘He is hearing.’

the be..3

hai be.3

b. vah sun rahā he hear. ._.. ‘He is occupied in hearing.’

hai be.3

The later auxiliary nature of Hindi rahnā represents a reanalysis which is part of larger-scale reconstruction of the Hindi verbal system (see further Slade 2013: §5.3).

¹⁴ A situation which persists marginally in modern Hindi, in much the same way as the modern English simple present can be employed with a progressive sense in certain contexts, e.g. “I am attaching a document to this email” vs. “I attach a document to this email.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

263

10.3.5 Nepali rahanu: light verb > aspectual auxiliary Nepali rahanu ‘remain, stay’ behaves morphologically like other light verbs, selecting for an absolutive participle in -i. Rahanu is also identical to other light verbs with respect to restrictions on recursion and construction-specific restrictions (on which, see Section 10.4). However, unlike other Nepali VV constructions, VVs involving rahanu (or rākhnu ‘keep’) never require agentive-marking on the subject, regardless of the transitivity of the main verb. In this, the construction including rahanu in its lightverb function appears to be a periphrastic counterpart of the Nepali synthetic imperfect construction, verb-stem + dai + inflected form of . In other words, rahanu as a light verb, (38)—with respect to agentive case-marking assignment— patterns with the synthetic imperfect in -dai, (39), not with other VVs (such as the light verb di-, as shown in (40)). (38)

Nepali a. ma mandir-mā ga-i- rah-eko chu I temple- go-- -_.. be.1. ‘I am going to the temple.’ / ‘I have been going to the temple.’ b. ma yo kām gar-i- rah-eko I this work do-- -_.. ‘I am doing this work.’ / ‘I have been doing this work.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(39)

Nepali : a. ma mandir-mā jām-dai I temple- go-_ ‘I am going to the temple.’ b. ma yo kām gar-dai I this work do-_ ‘I am doing this work.’

(40)

chu be.1.

chu be.1. chu be.1.

Nepali a. ma mandir-mā ga-i- di-eko I temple- go-- -_.. ‘I have gone to the temple (for someone).’

chu be.1.

b. mai-le yo kām gar-idi-eko I- this work do-- -_.. ‘I have done this work (for someone).’

chu be.1.

In contrast to normal VV constructions (like those in (40)), constructions in which rahanu functions as a light verb do not require agentive marking on the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

264

 

subject (even when the main verb is transitive); see (38). In this, the VV construction with rahanu closely resembles the synthetic imperfect construction, as in (39). The Nepali light verb rahanu appears thus to be in the process of becoming an auxiliary like Hindi rahnā. Both rahanu and rākhnu are further interesting in that sometimes they maintain a more contentful light-verb sense¹⁵—e.g., as in (41) and (42), respectively—another indication of their intermediate status in present-day Nepali. (41)

(42)

: bas na-āe-samma ma yahīm basi-rahan-chu bus not-came-until I here sit.--be.1 ‘I’ll keep sitting here until the bus comes.’ (Matthews 1998: 236)

(Nepali)

dāk ma-lāī aus: adhi khā-i-rākhnu bhanekā : t:ar-le doctor- I- medicine eat--. say._. thie be....H ‘The doctor had told me to keep taking the medicine.’ (Matthews 1998: 236) (Nepali)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In the following section, using data from a close examination of VVs in Hindi and Nepali, I outline various morphosyntactic parameters along which VV constructions may vary in different IA languages, some of which suggest differing degrees (or paths) of grammaticalization.

10.4 Variation in morphosyntactic properties of IA VVs The VV constructions of South Asia show considerable variation in their morphosyntactic attributes, and even within IA there are significant differences between the VV systems of different languages.

10.4.1 Interruptibility Hindi VVs are interruptible, that is, other words may occur in between the main verb and the light-verb components of the VV, as shown in (43).¹⁶

¹⁵ See also the Old Shah Nepali examples of rākh- as a light verb in (26)–(28). ¹⁶ All Hindi speakers consulted allowed for particles like to to intervene between main verb and light verb, as in (43); most speakers also allow pronouns to intervene, as in (ia); some also allow for full NPs, as in (ib), (ic).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   - (43)

mai-ne khānā khā to liy-ā, I- food eat.  ..-. ult:ī bhī ā-ī vomit also come..-. ‘I did eat food, but then I also vomited.’

lekin but

265 phir then

(Hindi)

Nepali VVs are not interruptible, even by particles like ta (equivalent to Hindi to): (44)

*mai-le bhāt khā-i- ta I- food eat--  ‘I did eat food, . . . ’

: sak-em, . . . -..

This difference is possibly reflected in writing, as Hindi VVs are written as separate words, but Nepali VVs usually as a single word.¹⁷ The inseparability of Nepali VVs suggests that they are either formed in the lexicon or else composed at a much lower level of syntax than Hindi VVs. In other words, Nepali light verbs appear to be more affixal in nature than are Hindi light verbs.

10.4.2 Recursion

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Hindi allows for only one light verb per main verb, as shown in (45).¹⁸

(i)

a. ā to vah gay-ā hai, come.  he ..-. be.3.. : sabzi nahīm lāy-ā vegetables not bring..-. ‘He has indeed come, but he didn’t bring the vegetables.’ b. %khā to khānā eat.  food ‘(I/he/she) did eat food.’

lekin . . . but [Hindi]

liy-ā ..-.

c. %liy-ā to khānā ..-.  food ‘(I/he/she) did (in fact) have food.’

[Hindi] khā eat. [Hindi]

¹⁷ This is unlikely to be just an artifact of the script, as both languages employ the same writing system (devanāgarī). ¹⁸ Peter Hook (pers. comm.) points out that there are instances in Hindi which morphosyntactically involve a converb, but semantically behave not dissimilarly to the Nepali type illustrated in (46), including V-kar rakh de- ‘(lit.) having V’ed, keep give’ and V-kar chor: de- ‘(lit.) having V’ed, leave give’; and an example of the former type is given in (i). (i) lekin kharc kī mār unke man ko mārkar but expenditure of strike his/her mind/heart  kill. rakh detī hai keep ._.. be..3 ‘But the blow of the expenditure struck his/her heart.’ [from https://books.google.com/books? isbn=8189859951]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

266

 

(45)

a. us-ne kican sāf kar he/she- kitchen clean do. ‘He/she cleaned the kitchen for me.’

dī .... (Hindi)

b. us-ne kican sāf kar dāl-ī : he/she- kitchen clean do. ..-. ‘He/she cleaned the kitchen straightaway.’ (Hindi) c. *us-ne kican sāf kar de dāl-ī : he/she- kitchen clean do. . ..-. ‘He/she cleaned the kitchen for me straightaway.’ (Hindi) In Nepali, on the other hand, VVs may involve up to two light verbs,¹⁹ as in (46), where the main verb, gari, is modified by both dii (itself in absolutive form) and hālin. (46)

un-le kican saphā gar-i- di-ihāl-in he/she.MH- kitchen clean do-- -- -.3. ‘She cleaned the kitchen for me straightaway.’ (Peterson 2002: 107) (Nepali)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In certain respects this makes Nepali VV appear more “affixal” in nature, given that main verb + light verb behaves like a simplex verb for purposes of additional operations (including the addition of other light verbs). Sinhala also allows multiple light verbs to collocate with the same main verb, as shown by example (47). (47)

meyāge bandinə wayəsə dæŋ pahu wē -gənə his/her marrying age now past become.Ø -.Ø -enəwa -. ‘Her marrying age is approaching (coming past) now.’ (Paolillo 1989, cited in Herring 1993) (Sinhala)

10.4.3 Morphosyntactic form restrictions on light verbs Nepali VVs can occur in conjunctive participles. Thus, for example, a converb in Nepali may be composed from a VV, as in (48).

¹⁹ Pokharel (1991) suggests that up to three light verbs may be used in some cases in Nepali, but my primary consultant did not accept any examples involving more than two light verbs collocated with the same main verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   - (48)

267

bhāt khā-i- sak-era u sut-na gay-o food eat-- - he/she sleep-. go.-3 ‘Having finished dinner, he went to sleep.’ (Peterson 2002: 108) (Nepali)

In contrast, Hindi converbs can only be formed from simplex verbs. Therefore, while (49b) is grammatical, (49a) is not.²⁰ (49)

a. *khānā khā le kar vah food eat.  conv he/she ‘Having eaten up the food, he left.’ b. khānā khā kar vah food eat.  he/she ‘Having eaten the food, he left.’

gay-ā go._-. (Hindi)

gay-ā go._-. (Hindi)

Hindi appears to have innovated: what is now the overt converb marker in Hindi, kar or ke, derives ultimately from the pleonastic addition of a converb form of kar ‘do’, pointing to the possibility of forming converbs of VV collocations in earlier Hindi, and suggesting that with respect to this property Nepali is conservative.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

10.4.4 Construction-specific restrictions IA languages vary with respect to the set of syntactic environments in which VV constructions are permitted or required. For example, Hindi VVs are fairly infrequent in negative contexts and “semi-negative” contexts like sirf . . . hī ‘only’, śāyad hī ‘hardly’ (Hook 1974, 1988), while the same constraint is not as strong in other languages; for example, in Marathi certain VV combinations can be easily negated (Hook 1988; Pardeshi 2001). Similarly, in Hindi, VVs are nearly obligatory whenever an event/action is perfective/completive (Hook 1974), while the same requirement is not found in other IA languages like Marathi (Hook 1988) or Nepali (Slade 2013). Hook (1988, 1993) points out also that certain contexts in Hindi strongly prefer or disprefer the use of VVs; in Hindi strongly VV-preferring environments include clauses dependent on a verb expressing fear, as in: (50)

: mujhe dar thā ki kahīm tum use cit:t:hī nā : me. fear be... that lest you he. letter  de do give. . ‘I was afraid that you might give him the letter.’ (Hook 1993: 100) (Hindi)

²⁰ Examples like (49a) are possible in Dakkhini Hindi (spoken in Hyderabad), which has undergone convergence with Dravidian Telugu; see Arora (2004).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

268

 

Hook (1988, 1993) points out that other IA languages display such preferences more weakly or not at all. Thus, in Marathi and Nepali, verbs of fear do not trigger use of VVs: (51)

cit:hī ta dienas holā tai-le bhanera ma-lāī letter part give..2. be..3 you-  me- dar lāgethyo : fear apply..1 ‘I was afraid that you might give him the letter.’ (Nepali)

The strength of restrictions of the requirement/disallowing of VVs in particular environments seems to correlate with the overall frequency of the use of VVs in a language; presumably both properties reflect the level of integration of VVs into the core grammar of tense/aspect in the language.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

10.4.5 Dominance and transitivity issues In both Hindi and Nepali, the question arises of how to reconcile clashes in transitivity between main verb and light verb in VV combinations.²¹ Before going into detail, it is important to understand the basic patterns of case-marking morphology in these two languages. Both Hindi and Nepali (more or less)²² employ a special marking for agents of transitive verbs in perfective tenses, -ne in Hindi, -le in Nepali. In Hindi, the marking of agents correlates with verbal agreement: agentive-case-marked nouns do not control predicate agreement.²³ In Nepali, (nonexperiencer) subjects control predicate agreement regardless of whether they are agentive-case-marked or not. More relevant for the purposes of VVs is the fact that in Hindi it is the light-verb member of the VV which determines whether the entire VV is treated as transitive or intransitive (and thus whether the subject receives agentive case-marking)— except in the case of intransitive main verb combined with transitive light verb (where we find variation in the assignment of transitivity to the compound for the purposes of case-assignment);²⁴ in Nepali, it is always the main verb which ²¹ Cf. Hock (1985) on issues of transitivity and case-marking in Indo-Aryan more generally. ²² Nepali also sometimes displays agentive case-marking in nonperfective contexts; the exact conditions on the use of -le in such cases are not entirely clear: see Poudel (2006) and Li (2007) for discussion. ²³ The unmarked object usually controls verb agreement; if this is blocked by the oblique postposition ko, the verb takes default masculine, singular, third-person agreement. ²⁴ Combinations of intransitive main verb and transitive light verb are very rare in Hindi, being apparently largely avoided by Hindi speakers. I have found only two examples from Nespital (1997a: 1108–9) where the case-assignment is clear (i.e. where the finite verb appears as a perfect participle):

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

269

determines the transitivity of the entire verbal compound. This is shown in the examples in (52) and (53). (52)

Hindi VVs a. Intransitive main verb + intransitive light verb = Intransitive: vah ā gay-ā he come. ..-. ‘He came.’ b. Transitive main verb + intransitive light verb = Intransitive: vah khānā khā gay-ā he food eat. ..-. ‘He ate up the food.’ c. Transitive main verb + transitive light verb = Transitive: us-ne khānā khā liy-ā he- food eat. ..-. ‘He ate up the food.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

d. Intransitive main verb + transitive light verb = (variation) (see fn. 24) vah/us-ne bait:h liy-ā he./he- sit. ._-. ‘He sat down.’

(i)

śatru senā acānak hī hamārī senā enemy army suddenly  our army lī ._.. ‘The enemy army suddenly pounced on our army.’

par on

ā come.

(ii) jab satīś apne ghar se nikl-ā when Satish his.own house from exit-_.. to uskā kuttā bhī uske pīche ho liy-ā then his dog also his behind be. -_.. ‘When Satish left his house, then his dog also followed him.’ In both cases, the subject is not marked with an agentive marker, despite the light-verb member being transitive. Remarkably, in both (i) and (ii), not only is the light verb transitive, but the VV itself appears to be semantically transitive (in the sense that both pounce and follow require complements). However, more generally, combinations of intransitive main verb and transitive light verb seem to result in variation in Hindi (probably as a result of the rarity of such combinations). That is, some speakers prefer (iii), while others prefer (iv) (yet others prefer to avoid such combinations altogether). (iii)

vah bait:h liy-ā he sit. ..-. ‘He sat down.’

(iv) us-ne bait:h he- sit. ‘He sat down.’

liy-ā ..-.

Speakers show similar variation with respect to other combinations, such as dikh diyā ‘appeared’ and ghum liyā ‘roamed’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

270

 

(53)

Nepali VVs a. Intransitive main verb + intransitive light verb = Intransitive: u mandir-mā ga-iā-eko cha he temple- go-- -... be.pres.3sg ‘He kept going to the temple.’ b. Transitive main verb + intransitive light verb = Transitive: us-le yo kām gar-iā-eko cha he- this work do-- -... be..3 ‘He has continued to do this work.’ c. Transitive main verb + transitive light verb = Transitive: us-le bhāt khā-idi-yo he- food eat-- -.3. ‘He ate up the food.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

d. Intransitive main verb + transitive light verb = Intransitive: u rām ko-lāgi bajār ga-idiy-o he Ram for-sake-of market go-- -.3. ‘He went to the market for Ram’s sake/in place of Ram.’ Thus, in Hindi the light verb usually determines the transitivity of the entire compound, for purposes of assignment (or nonassignment) of agentive marking to the subject. In Nepali, it is the main verb which determines the transitivity. Again, though it is unclear which of these represents the more conservative system, it is clear that change has taken place in the VV system of one or both languages. The uncertainty stems from the absence of extant Proto-Modern-Indo-Aryan, and the fact that for the predecessor gerund construction in earlier IA such issues did not arise: we do not find an ergative/absolute-type system in Sanskrit or Pali. In the following and final section, I examine the inventories of light verbs which participate in VV constructions more broadly, considering both various IA languages—including not only Hindi, Bengali, Marathi, and Oriya, but also Nepali, Shina, and Sinhala—as well as a range of Dravidian languages, and speculate on the implications of the great amount of variation in both lexical inventories and morphosyntactic properties of VVs not only between language families but also within them for our understanding of the historical development of VVs in South Asia.

10.5 Reflections on light verb inventories across South Asia Examining VV constructions in Dravidian languages, Annamalai (2016: 556) observes that “there are cognate and non-cognate light verbs across the languages; the meanings of the light verbs are a shared set, though not always expressed by cognate light verbs.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

271

Table 10.1 Light verbs that occur in VVs from selected Dravidian languages

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Tamil Malayalam Betta Kurumba Kannada Telugu GO SEE GIVE RELEASE/CAST/LEAVE PLACE/PUT HOLD/CONTAIN TAKE BE BECOME DROP SHOW LOSE TEAR UP PUSH DESTROY STAND SIT ACT SUITABLY BE FINISHED JOIN PLAY KEEP BUY FALL DIE

pō par kotu : vit:u vai ko:l et:u iru āku pōt:u kāt:t:u tolai kizi
ta:l:lu

pō kān, : nōkku kotu, : tar ka:lay vay ko:l et:u

pōg nōd: kod, : t̪ār kā:l

hōgu nōdu : kodu : bidu : hāku

pō cūs

wēs

ir

tulay nil iri aru:l kazi
ēku ādu : pet:t: kon pad: caw

Indeed, Table 10.1 represents a rough approximation of the range of the full-verb meanings (I have not space here to examine the different functions of Dravidian light verbs in depth) across a selection of Dravidian languages.²⁵ The range of Dravidian light verbs that occur in VVs can be compared with those found in a selection of Indo-Aryan languages, as given in Table 10.2.²⁶

²⁵ Tamil data is taken from Annamalai (1985, 2016) and Steever (2005); Malayalam data from Nayar (1979); Betta Kurumba data from Coelho (2018); Kannada data from Bhat (1979); and Telugu data from Subbarao (1979). ²⁶ Compiled from Vale (1948); Singh et al. (1986); Abbi and Gopalakrishnan (1991); Pokharel (1991); Schmidt (2004) and my own data collection. Blanks indicate a lack of evidence for the existence of the form as a light verb, while many of these may in fact occur; ‘–’ indicates a stronger belief in the actual absence of the form as a light verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

272

 

Table 10.2 Light verbs that occur in VVs from selected Indo-Aryan languages

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

GIVE GO COME FALL SIT PUT, THROW TAKE KEEP RISE MOVE DIE KILL RELEASE COME OUT TAKE AWAY SEND OPEN LEAVE STAY STAND UP BRING INSERT TOUCH APPLY ESCAPE HOLD TAKE OUT

Bengali

Hindi Marathi¹ Nepali

Oriya Kohistan² Sinhala Shina

deoja¯ oa¯ sipaṛbasphelnerakoṭhcal-

deja¯ a¯ paṛbaiṭhḍa¯ llerakhuṭhcalmarma¯ r-

deja¯ a¯ spaṛbas-

doón bójoón

nerakhuṭhca¯ lmarma¯ r-

–

nikalle ja¯ -

deja¯ yepaḍbasṭa¯ kghethev: uṭhcal-

dija¯ a¯ u(par-)³ ha¯ l– ra¯ kh-

soḍnighne-

dijoón byoón

denəwa yanəwa enəwa – – da¯ nəwa gannəwa – – –

mojoón

– –

phátoon

– arinəwa, ærenəwa –

paṭha¯ u-

ra¯ hkhaṛa¯ hoa¯ ṇghallāglāvsuṭdharkādh:

¹ Marathi based on Pardeshi (Chapter 9, this volume) ² The Shina varieties in Gilgit and Skardu do not appear to exhibit use of VVs (Peter Hook, pers. comm.). ³ Pokharel (1991: 191) lists par- as a light verb which can form a typical IA VV in Nepali, but without examples, and neither online searches nor consultation with my Nepali informant has revealed convincing instances.

Table 10.2 might even suggest more homogeneity than is in fact present in IA VV systems, for not only do these vary along various morphosyntactic parameters (as per Section 10.4), but even the semantics of the light-verb uses of verbs with similar full semantics, and even cognate verbs, varies somewhat between IA languages, which Table 10.3 gives a small sampling of. Thus, sometimes the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

  -   -

273

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Table 10.3 Selected Hindi and Nepali light verbs compared Hindi stem Hindi fullverb meaning

Light-verb sense Nepali full-verb meaning

dāl : de le de cuk bait:h

immediacy other-directed self-directed perfective completive regret

throw give take give be used up sit

Nepali stem

throw give

hāl di – give di complete / be able to sak send pat:hāu

same light-verb function is expressed using verbs with the same full-verb meaning, e.g. Hindi dāland Nepali hāl. Sometimes the forms are even cognates, as is the : case for Hindi de- and Nepali di-. But sometimes light-verb functions are expressed using verbs with distinct full-verb senses, e.g. Hindi bait:h- and Nepali pat:hau-, and in some cases a light-verb sense is expressible in some languages but not in others; for example, Nepali lacks an equivalent to the light-verb function of Hindi le-. Considering the range of variation in morphosyntactic proprieties and particular inventories of light verbs, we might wonder if these really do represent even small significant pockets of common inheritance. One of the most common light verbs in IA, and one of the ones to make clear early appearances, is GIVE. But though GIVE is present in modern Sinhala, it is not found in Old Sinhala.²⁷ That leaves only GO. And forms of GO are frequently found undergoing various types of grammaticalization, being present, for instance, in the formation of passives and futures in Hindi.²⁸ Annamalai (2016: 556) notes that “[t]he use of non-cognate light verbs for a shared meaning is not surprising because this is noticed between dialects of the same language as well. This raises a problem for comparative reconstruction: Can a grammatical phenomenon be reconstructed for the proto stage while the forms of particular light verbs that are instances of the phenomenon cannot be?” While the lack of cognates for a shared meaning between related languages is not itself necessarily an insurmountable problem,²⁹ the overall differences between even closely related Dravidian languages in terms of substantial differences in lightverb inventories may weigh against positing VVs in Proto-Dravidian. ²⁷ Of the light verbs found in Modern Colloquial Sinhala, only two of these were also found in Old Sinhala: TAKE and GO. Old Sinhala also possesses PLACE and SHOW, which have subsequently undergone further grammaticalization: see Section 10.2.2. ²⁸ And compare, outside of South Asia, the grammaticalization of GO in the English future construction be going to V. ²⁹ See Watkins (1995) on the notion of “renewal” of components of inherited formulaic phrases— essentially the phenomenon of a form within an idiomatic chunk being replaced by another form which bears the same/similar meaning—as well as Matasović (1996) and Slade (2008 [2010]), who explore additional examples of preservation of formulae involving “renewal” or lexical replacements.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 21/1/2021, SPi

274

 

Turning back to Indo-Aryan, given the significantly earlier appearance of VVs in Sinhala, and the substantial and prolonged contact Sinhala has had with South Dravidian languages, it is tempting to suggest that VVs might find their origin in Dravidian and that the late and varied appearance of VVs in IA represents Dravidian influence. However, while it would be tempting to attribute the strikingly early Sinhala development of VVs to Dravidian influence, VVs in Old Tamil are, in fact, apparently vanishingly rare: Annamalai (2016: 556) points out that “[g]iven the paucity of light verbs in Old Tamil texts before the Common Era, one may speculate that ‘delexicalization’ of full verbs into light verbs is a development that is shared, but its time and instantiation are specific to individual languages.” So one possibility this suggests is that VVs in modern South Asian languages had somewhat independent paths of development—even probably in many cases including also languages within the same language family—with some amount of later convergent development. This would also be one potential explanation for the fact that the semantic similarities between the root meanings of light verbs seem to be largely restricted to those we might independently expect to be liable to grammaticalization (e.g. GO, COME).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Acknowledgments Thanks to the editors of this volume, Taro Kageyama, Peter Hook, and Prashant Pardeshi, for the work they have put in on this collection. Thanks to my Nepali consultant Sonia Chettri. This chapter greatly benefited from extensive suggestions and comments from Peter Hook. All remaining errors and infelicities are my own, and Peter Hook’s comments also suggest a number of avenues for further exploration which time and space precluded including in the present chapter.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

11 The Hindi-Urdu compound verb and its covert semantics Births, earthquakes, meteors, and other autogenous expressions Peter E. Hook

11.1 Introduction In Hindi-Urdu the compound verb (CV) consists of the bare stem of a main verb extended by the inflected form of a “vector” (ranjak kriyā). In (1a) samajh ‘understand’ is the main verb and lete.hãĩ ‘’ is the vector + desinence:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

a. log rassī-ko sããp samajh le-te.hãĩ people rope- snake understand -. ‘People understand the rope (to be) a snake.’ (aksharabharat.blogspot.com)

The compound verb alternates with an unextended counterpart, as in (1b), and does so with little or no change in meaning that is easily translatable into languages like Sanskrit, French, or English that lack compound verbs: (1)

b. log rassī-ko . . . sããp samajh-te.hãĩ people rope- snake understand-. ‘People . . . understand the rope (to be) a snake.’ (books.google.com; isbn = 9351862232)

As indicated by the capitalized gloss “” of lete in (1a), vectors almost always have cognate forms among the basic lexical items of a language. In languages that have compound verbs the degree to which the basic lexical meanings of these cognates are missing from their vectorial counterparts varies from language to language, vector to vector, place to place, epoch to epoch, and even from speaker to speaker. In most varieties of Hindi-Urdu the grammaticalization of jā- {} has proceeded to the extent that in the compound verb form main verb ā- ‘come’ takes

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

276

 . 

the ‘bleached’ form of its basic lexical antonym jā- as its normal vector. See the compound verb form ā jāegā in (2a): (2)

a. govā-se yahãã ā jā-egā. Goa-from here come - ‘X will come here from Goa.’

As indicated by the translations of (2b) basic meaning and PAS (predicate argument structure) of the simplex āegā do not differ in significant or easily articulated ways from those of ā jāegā in (2a):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(2)

b. govā-se yahãã ā-egā. Goa-from here come- ‘X will come here from Goa.’

Finding and characterizing the number and complexity of the many syntactic, semantic, and discoursal factors that determine the speaker’s / writer’s choice between the compound verbs in (1a) and (2a) and the corresponding “simple verbs” in (1b) and (2b) is the central problem addressed in this chapter.¹ While the list of main verbs that participate in this alternation is unending, the number of vector verbs in Hindi-Urdu is limited to about two dozen, all of which have counterparts among the basic lexical verbs.² In this chapter we consider only the four most common (“semantically least marked”) vectors: jā- ‘’, le- ‘’, de- ‘’, and par: - ‘’. As a rough approximation, collocability with one or an other of these four vectors partitions the set of main verbs into general subsets definable by predicate argument structure (PAS): intransitives take jā- ‘’ or sometimes par: - ‘’; transitives take le- ‘’ or de- ‘’; semitransitives³ take jā- ‘’ or le- ‘’. Their orientation with respect to either the speaker or the protagonist crosscuts the partition by PAS: Centrifugal predicates take de‘’ or jā- ‘’; neutral or stationary ones, de- ‘ ’, jā- ‘’, or par: - ‘’; and centripetal predicates, le- ‘’ or jā- ‘’:

¹ In complexity it may be compared to the problem of identifying the multiple factors conditioning rendaku in Japanese. See Irwin (2005). ² A fuller list of Hindi-Urdu’s major and minor vector verbs (with the meanings of their basic lexical counterparts) is: ā ‘come’, ut:h ‘rise’, khar: ā ho ‘stand up’, čal ‘move’, čuk ‘already’, čhor: ‘leave behind’, čhor: -de‘quit; let go’, jā ‘go’, dāl : ‘throw; pour’, de ‘give’, dhar ‘put down’, nikal ‘exit’, nikāl ‘take out’, par: ‘fall’, pā ‘manage to’, bait:h ‘sit’, mar ‘die’, mār ‘strike’, rakh ‘put, keep’, rakh-de ‘put down’, rah ‘remain, stay’, le ‘take’, le-jā ‘take away’, and sak ‘can’. See Hook (1974: 119–44) for discussion and examples. ³ “Semitransitive” are those predicates which take direct objects but express actions oriented toward the subject: khā.nā ‘to eat’, pī.nā ‘to drink’, soč.nā ‘to think’, samajh.nā ‘to understand’, pakar: .nā ‘to catch’, čhīn.nā ‘to seize’, etc.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -   (3)

Intransitives:

nikal-nā leave- Transitives: rakh-nā put- Semitransitives: khā-nā eat-

, , , , , ,

nikal jā-nā leave - rakh de-nā put - khā le-nā eat -

, , , , , ,

277

nikal par: -nā4 = ‘to leave’ leave - rakh le-nā = ‘to put’ put - khā jā-nā = ‘to eat’ eat -

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Hindi-Urdu has other types of sequences of verb forms morphologically parallel to compound verbs. These can be distinguished from compound verbs by referring to semantic and behavioral criteria: A. Idioms and fixed sequences (čal bas.nā, Lit: move-settle = ‘to pass away, die’) are to be excluded because compound verbs alternate with corresponding noncompounds. čal bas.nā ‘die’ does not alternate with čal.nā ‘move; walk; start’. Compare V-V sequences of Type 2 in Kageyama, Chapter 2, Table 1, this volume. B. Modals (čal sak.nā, Lit: move-can = ‘to be able to move’) are excluded because, while a vector alternates with its absence with little or no easily translatable difference in meaning, English, French, Sanskrit, and most other languages of the world have ready translations for sak.nā. C. Serial verbs (phisal par: ā 100 exx; , 20 exx] (less rare: 1 in 5)

(7)

g. pappā-e . . . st:ešan-par čhod.i didh.elā bā:lako t:vit:ar-ni : father- station-at abandon - children Twitter- madad-thi mammi-sudhi pahõč.yā help-with mother-to arrived (akilanews.com) ‘With Twitter’s help kids abandoned at the station by their Dad got to their Mom.’ [search strings = , >80 exx; , >60 exx] (as past participles CVs in Gujarati are fairly common: about 3 in 7)

(8)

g. te pelũ ke:lũ čhod.i da.i-ne potā-no hāth bahār : he that banana abandon - self- hand out kādh.i šake : pull.out can ‘He can let go of the banana and get his hand out (of the trap).’ (data found online) [search strings = , >140 exx; , >160 exx] (as converbs or gerunds, CVs in Gujarati are common: about 4 in 7)

From the Gujarati data it appears that occurrence in a converb or gerund has the weakest suppressive effect and occurrence in the complements of phasal verbs, the strongest. Parallel data from the other three languages (Hindi-Urdu, Marathi, and Nepali) are needed to confirm this generalization. How to explain the inverse correlation observable in Indo-Aryan languages between the abundance of CVs in running text and their freedom of occurrence? One possibility is to regard compound verbs in CV-poor languages (Kashmiri, Nepali, Marathi) as independent lexical items on a par with, at times even competing with their noncompound counterparts. As such, they are free to occur in any of the structural environments that noncompound verbs may occur in. In contrast, in CVrich languages like Hindi-Urdu (and Panjabi) verb-compounding may be seen as a general paradigmatic feature of every or nearly every verb, a feature that functions (or is chosen by speakers) in the same way that features of tense or mood are chosen. Since Hindi-Urdu CVs occur in all the tenses and moods that noncompound verbs occur in, their function is orthogonal to categories of tense and mood. “Conceptual completeness” may be the closest general characterization of CV function in HindiUrdu, but is too vague to be of predictive power.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

283

If the term “completeness” brings to mind the notion of perfective aspect (fuzzy as that may be; see Comrie (1976)), then we might expect to find some parallels in the behavior of CVs in Hindi-Urdu and perfective forms in Russian and other Slavic languages. Indeed, such parallels exist. One, first noted by V. Pořízka over fifty years ago, is the constraint given above in (5a). The perfective (PV) forms of Russian and Czech infinitives do not (normally) occur as complements of phasal verbs such as načat’ ‘to begin’ (9a) and perestat’ ‘to stop’ (9b). That structural position is open only to nonperfective (IMPV) infinitives (Pořízka 1968: 70, 1969: 30): (9)

a. ona načala posyla-t’ (*posla-t’) den’gi materi (cf. (5a)) she began send. - (send.-) money mother. ‘She began to send money to her mother. (www.kvazar-fant.ru/story/409/text) b. v kakom godu ivan 3 perestal posyla-t’ (*posla-t’) in which year Ivan 3 stopped send.- (send.-) dan’ v ordu (cf. (5a)) tribute to Horde ‘In which year did Ivan III stop sending tribute to the [Golden] Horde?’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Recourse to a similar explanation may be had to explain the skewing of CV flux under negation. In Hindi-Urdu the presence of a negative, while not always doing so (see (10d)), in most combinations of tense, mood, and finiteness, strongly disfavors CVs. For instance, in nonnegated past tenses the CV vičār čhor: diyā ‘gave up the idea’ (10b) outnumbers its non-CV counterpart vičār čhor: ā (10a) by 4 to 1: (10)

a. jīkā mahāmārī phail.ne-par t:āt:ā mot:ars-ne jīkā brān: d: zika epidemic spead-on Tata Motors- Zika brand lā.ne-kā vičār čhor: ā bring- idea gave.up ‘With the spread of the zika virus Tata Motors gave up the idea of introducing the “Zika” brand.’ (www.khabarindiatv.com) [search string = ; 30 exx] b. is-ke.bād unhõ-ne yūrop jā.ne-kā vičār čhor: diyā this-after he- Europe going- idea give.up  ‘After this he gave up the idea of going to Europe.’ (satyagrah.scroll.in/article/101698) [search string = ; > 120 exx]

This preference for CV manifestation is dramatically reversed by the presence of a negative particle, as in (10d). Compare the ten non-CVs negated as in (10c) with the single example of a negated CV in (10d):

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

284

 . 

(10)

c. jasvīr kaur-ne apne sasur-kī pūrn: čal.sampatti Jasvir Kaur- her father.in.law- entire property ur: ā.ne-kā vičār čhor: ā nahĩĩ steal- idea gave.up  ‘Jasvir Kaur did not give up the idea of stealing all of her father-in-law’s wealth.’ (Gurudutt 2015: 102) [search strings = ; 10 exx] d. mãĩ-ne tīrth.yātrā-kā vičār čhor: nahĩĩ diyā, I- pilgrimage- idea give.up  , keval kučh kāl-ke.lie sthagit.kiyā hai only some time-for postponed have ‘I haven’t given up the idea of [making] a pilgrimage, only postponed it for a while.’ (Singh 2005: 79) [search strings = ; 1 example]

Searches show that in Hindi-Urdu the suppressive effect of negation is quite general. News reports of crimes provide hundreds of examples in which the frequency of occurrence of the CV pahũč gaī ‘arrive ’ (11b) can be compared with that of its non-CV counterpart pahũčī ‘arrived’ (11a):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(11)

a. sūčnā mil.te hī pulis mauke-par pahũčī aur information getting  police scene-on arrived and jããč šurū.kar dī investigation begin  ‘As soon as they were informed, the police arrived on the scene and began to investigate.’ (hindi.newsroompost.com) [search string = ; > 80 examples] b. sūčnā mil.te hī pulis mauke-par pahũč gaī information getting  police scene-on arrive  aur māmle-kī jããč šurū.kar dī and matter- investigation begin  ‘As soon as they were informed, the police arrived on the scene and began to investigate the matter.’ (m.dailyhunt.in) [search string = ; > 100 examples]

Searches yield no examples of negated pahũč gaī versus sixty-five examples of negated non-CV pahũčī:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -   (11)

285

c. sūčnā-ke ek ghan: t:e bād bhī pulis mauke-par information- one hour after even police scene-on nahĩĩ pahũčī  arrived ‘Even after an hour of being informed the police did not arrive on the scene.’ [search strings = , 65 exx; , 0 exx]

Such effects of negation on CV frequency may be explained by supposing that conceptual completeness is not uppermost in speakers’ and hearers’ minds when the failure of an event to occur is to be expressed. By contrast in CV-poor languages like Marathi the suppressive influence of negation is moderate:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

a. polis.ān-nā pāh.tā-ʦ to pa:lālā police- seeing- he fled ‘He ran away as soon as he saw the police.’ (loksatta.com/pune-news/crime-in-pune) [search string = , 85 exx] b. lok jamā.ho-u lāg.lyā.ʦe pāh.tā-ʦ to pa:l.un gelā people gather- beginning seeing- he run  ‘He ran away as soon as he saw a crowd had begun to gather.’ (archive.loksatta.com) [search strings = , 55 exx; , 6 exx]

Searches indicate a four-to-three preference for the positive simplex pa:lālā over the CV pa:lun gelā. At two-to-one, the corresponding ratio of negated SVs to negated CVs shows the inhibitory effect of negation to be milder than it is in Hindi-Urdu: fifty-three negated SVs in (12c) versus twenty-six negated CVs in (12d). (12)

c. mhašin-ʦā ka:lap kāhi pa:lālā nāhi buffalos- herd any ran  ‘The buffalo herd did not run away at all.’ (data found online) [search strings= , 45 exx; , 8 exx] d. mātr yā khepe-lā bibt:yā pa:lun gelā nāhi but this time- leopard run   ‘But this time the leopard did not run away.’ (www.misalpav.com) [search strings = , 16 exx; , 10 exx]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

286

 . 

The expression of conceptual completeness is not always canceled by the presence of a negative. Hindi-Urdu has two constructions that feature conceptual completeness, allow a negative element, and strongly favor verb-compounding. These are ‘until’-clauses (13a, b) and ‘fear’-clauses (14a, b): (13)

a. unhõ-ne mujh-se kahā ki jab.tak mãĩ ghar na they- me-to told that until I home  pahũč-ũũ, ban: dal na khol-ũũ : arrive- bundle  open- ‘They told me not to open the bundle until I got home.’ (navbharattimes.indiatimes.com) [search strings = ; 3 exx] b. jab.tak mãĩ pahũč nā jā-ũũ kučh mat kar-nā until I arrive  - anything don’t do-. ‘Don’t do anything until I get there.’ (khabarindiatv.com) [search strings = ; 13 exx]14

The compound verb is rather strongly favored in clauses that complement expressions of fear or anxiety, especially if they contain the particle kahĩĩ ‘lest’:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(14)

a. dar hai kahĩĩ yah sil.silā soniyā-jī tak na pahũč-e : fear is lest this affair Soniya-ji up.to  arrive- ‘There’s fear [word of] this affair might get to Soniya Ji.’ (twitter.com/atorneybharti) [search strings = ; 4 exx] b. hamẽ bas yahī dar thā ki kahĩĩ āg hamāre : us. just this fear was that lest fire our gharõ-mẽ na pahũč jā-e houses-in  arrive - ‘What we were afraid of was that the fire might get to our houses.’ (navbharattimes) [search strings = ; 12 exx]

The evidence for the CV-promoting effect of fear-clauses in Hindi-Urdu is statistical rather than categorical, and the presence on the Internet of Nepali and Marathi data is insufficient to demonstrate a contrast with Hindi-Urdu. However, given the strong preference for the perfective aspect in the complements of fear-clauses in Russian

¹⁴ In (13a, b) na and nā are allomorphs; in (13b) counts of CVs interrupted by NEG have been merged with CVs preceded by NEG. Asterisks indicate material intervenes between the subject mãĩ and the rest of the search string. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

287

(Forsyth 1970: 258–61, 297), from (14a, b) we do have additional support for recognizing a parallel between the perfective in Slavic and CV manifestation in Hindi-Urdu:¹⁵ (15)

ljudi bespokoilis’ kak.by ogon’ ne perekinulsja people were.worried lest fire  spread() [*perekidyval.sja] na ix dom [*spread()] to their house ‘People were afraid that the fire might spread to their houses.’ (www.amur.kp.ru/daily) [search strings = как бы огонь не перекинулся: 23 exx / перекидывался: 0 exx]

In addition to the one illustrated in (13a, b) Hindi-Urdu possesses another jab. tak–construction, one having a different function. The jab.tak–clause in (16) indicates the posteriority of the action or event that it expresses to the action or event of the accompanying tab.tak–clause. CVs do not occur in such clauses: (16)

a. jab-tak phāyar briged: vahãã pahũčī tab-tak sthānīy when-till fire brigade there arrived then-till local logõ-ne āg-par kābū pā liyā people- fire-on control get  ‘By the time the fire brigade got there, the local people brought the fire under control.’ [search strings = , 28 exx] b. (zero exx of CV)

[search strings =

]

By contrast a tab tak- clause expressing anteriority invariably has CVs (or completives in čuk-): (17)

jab-tak agni-šaman gār: ī pahũčī, tab-tak āg-ne when-till fire-extinguishing vehicle arrived then-till fire- vikrāl rūp le liyā monstrous form take  ‘By the time the fire truck arrived, the fire took on fearsome proportions.’ [search strings = . CV: 19 exx; non-CV: zero exx]

¹⁵ Non-CVs freely occur in this environment in CV-poor languages. (i) is a Nepali example provided by Slade (2016: 563), and (ii) is a Marathi example. For Kashmiri, see Hook and Koul (1992: 10–11). (i)

cit:hī ta dienas holā tai-le bhanera ma-lāī dar (Nepali) : letter  give...2 be..3 you-  me- fear lāgethyo apply..3 ‘I was afraid that you might have given him the letter.’ (my translation)

(ii)

kāhi vadilān-nā mulā-ʦe lād: kele tar to bighad-el aši bhiti vāt:.te tar (Marathi] : : some fathers- son- coddling did then he spoil- such fear feels whereas . . . ‘Some fathers feel that if they are affectionate with a son, then he will be spoiled, whereas . . . ’

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

288

 . 

The anti-iconic (time-inverting) phrase is se pahle hī ‘before this’ has a similar CV-promoting effect: (18)

grāmīn: āg-par kābū pāte, is-se.pahle-hī āg-ne villagers fire on control get this-before- fire- bhīs: an: rūp le liyā terrible shape take  ‘Before the villagers could get it under control, the fire assumed a horrific form.’ (data found online) [search string = ; 15 exx; one counterexample]

Replacing the hī in is se pahle hī with ki “uninverts” the temporal order of the events expressed in the linked clauses. The preference for CV (and the dispreference for non-CV) switches accordingly:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(19)

is-se.pahle ki āg vikrāl rūp letī sthānīy logõ-ne this-before that fire monstrous form take local people- āg-par kābū pā liyā fire-on control get  ‘Before the fire could take a horrific form, the local people got it under control.’ (www.jagran.com/uttarakhand/) [search string = ; 14 exx; 2 counterexamples]

In addition to negative particles other adjuncts (often adverbial) may influence CV-manifestation: (20)

a. agar us.kī ãgrejī aččhī ho to use naukarī if his English good be then him. job āsānī-se mil.tī.hai ease-with gets ‘If his English is good, then he gets a job easily.’ (http://abpnews.abplive.in/sports) [search strings = ; 9 exx] b. kah.te.hãĩ mumbaī-mẽ naukarī āsānī-se mil jātī.hai, lekin say Mumbai-in job ease-with get  but ghar nahĩĩ mil.tā house  gets ‘They say in Mumbai you can easily get a job but you can’t get a place to live.’ (navbharattimes.indiatimes.com) [search strings = ; 37 exx]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

289

c. sarkārī naukrī bar: ī muškil-se mil.tī.hai, par lag.tā.hai tumhẽ gov’t job much difficulty-with gets but seems you. is.kī tanik bhī parvāh nahĩĩ it’s little even care  ‘It’s very hard to get a government job, but it seems you don’t care about it at all.’ (data found online) [search strings = ; 18 exx] d. (zero exx of CV)

[search strings =

]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

How to explain this quadratic asymmetry? Is an easy achievement more likely to be conceived of as being complete in itself than one achieved with effort? Or does the adverbial phrase muškil-se ‘with difficulty’ come close enough to the negation inherent in the concept of ‘hard to V’ or ‘barely’ to suppress CV manifestation? Parallel effects of these two contrasting adverbial phrases on CV flux apply to other verbs: X āsānī se bik.tī (11 exx) versus āsānī se bik jātī (32 exx) as against muškil-se bik.tī (8 exx) versus muškil-se bik jātī (0 exx) ‘X sells easily’ versus ‘X sells with difficulty’. X āsānī se nikal.tī (10 exx) versus āsānī se nikal jātī (90 exx) against muškil-se nikal.tī (20 exx) versus muškil-se nikal jātī (0 exx) ‘X comes off easily’ versus ‘X comes off with difficulty’. In (21) we may observe what appears to be an instance of the influence of ‘long distance’ negation on CV occurrence. Phone calls that go through and have results may be expressed either with the CV phon kar diyā or with the non-CV phon kiyā. See (21a) and (21b): (21)

a. is-ne pulis-ko 100 nambar-par phon kiyā. pulis she- police- 100 number-at phone did police bhī turant vahãã pahũč gaī also imediately there arrive  ‘She phoned the police at [their] 100 number. The police got there right away, too.’ [search string = ; 17 exx] b. kisī-ne 100-par phon kar diyā. pulis mauke-par someone- 100-at phone do  police scene-on pahũčī aur ghāyal-ko aspatāl-mẽ bhartī.kar-āyā got and victim- hospital-in admit-caused ‘Someone phoned “100”. The police arrived and had the victim admitted to hospital.’ (vicharkhabar.com) [search strings = ; 12 exx]

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 . 

290

But phone calls that are not answered or which have no results rarely get CV expression (21d):¹⁶ (21)

c. ghāyal-ne batāyā ki us-ne 100 nambar-par phon kiyā injured- told that he- 100 number-at phone did lekin pulis mauke-par nahĩĩ pahũčī. but police scene-on  arrived ‘The victim said he phoned the 100 number, but the police did not arrive at the scene.’ (www.patrika.com) [search strings = ; 13 exx]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

d. sureš-ne 100 n.-par phon kar diyā hai. lekin Suresh- 100 number-at phone do  has but pulis ab-tak nahĩĩ āī hai police now-till  come has ‘Suresh has phoned the 100 number, but the police have yet to arrive.’ (data found online) [search strings = ; 1 ex.] How many of these factors can be explained by attributing “conceptual completeness” or perfectivity as a basic sense to the compound verb? Such an ascription may be supported by CVs’ (i) absence in complements of phasal verbs and (ii) their prevalence in expressing anxiety or feared events or (iii) limits in developments whose endpoints are defined by or presented as anterior to some intervening action or state: fear-clauses, tab-tak clauses, until-clauses. The rarity of Hindi-Urdu CVs’ occurrence in negative contexts may also be regarded as a consequence of a general perfective meaning. However, their absence in converbs and in participles, and their rarity in the expression of posteriority (16b) complicate an appeal to viewpoint aspect of the kind found in Slavic languages¹⁷. ¹⁶ The asymmetry in (21) may relate to a specific phenomenon termed ‘entailment cancellation’ (Bhatt 2006: 162). (i)

binubābū-ne jā-kar darvājā kholā, kintu darvājā nahĩĩ khulā : Binubaba- go- door opened(tr.) but door  opened(intr.) Literally: ‘Binubaba went and opened the door, but the door did not open.’ (isbn=8185134820, p. 61)

For most Hindi-Urdu speakers replacing kholā with a CV (either khol diyā or khol liyā) in (i) creates a contradiction. ¹⁷ Notice, for instance, that clauses in Russian expressing posteriority usually have verbs in the perfective aspect: (i)

k tomu vremeni, kogda požarnaja mašina prijexala na mesto, garaž to that time. when fire vehicle arrived. at place garage doma polnost’ju zgorel house. completely burned.up ‘By the time the fire truck arrived on the scene, the house’s garage was entirely consumed.’ (vk.com)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

291

11.3 Autogeny The asymmetrical sets to be examined in this section may derive from the suppressive effect of an even more abstract kind of negative, the unprepared mind. Examples (22a, b) report the falling of trees, events not unexpected in a storm. Either CV or non-CV expression may occur: (22)

a. tej ããdhī-tūphān ke.kāran: en eč-par šīšam.kā per: strong gale-storm because N. H.-on rosewood tree girā thā fallen had ‘Due to a windstorm a rosewood tree had fallen on the National Highway.’ [search string for (22a) = ; 20 exx] b. tej ããdhī āne ke.kāran: vr̥ddh.kī jhopar: ī-par bargad.kā strong gale coming because elder’s hut-on banyan per: gir gayā thā. tree fall  had ‘Due to the coming of a windstorm a banyan tree had fallen on the old man’s hut.’ [search strings for (22b) = ; 27 exx]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

However, as a completely unexpected event, the fall of a meteor very rarely gets CV expression (23b): (23)

a. 1892-mẽ bānūsūr-mẽ ākāš-se ek ulkā.pin: d: girā thā 1892-in Banusur-in sky-from one meteor fallen had ‘In 1892 in Banusur a meteorite had fallen from the sky.’ (www.facebook.com) [search strings = ; 36 exx] b. dhyān rahe ki mind [in] stay that višāl ulkā.pin: d: enormous meteor ‘Bear in mind that in Russian city.’ [search strings=

rūs-ke ek šahar-par 2013-mẽ ek Russia- one city-on 2013-in one gir gayā thā fall  had 2013 an enormous meteorite had fallen on a (news.ajitsamachar) ; 1 ex.]

The strings in (23a) and (23b) refer to “auto-creative” events. What we call a “meteor” does not exist as such prior to or outside of its fall through the atmosphere. Contrast the fall to earth of an artificial satellite:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

292

 . 

(24)

duniyā-ke vaijñānikõ-kī najar jis sait:elāit:-par t:ikī thī world- scientists- eyes which satellite-on fixed were vo dhartī-par gir gayā hai. it Earth-on fall  has ‘The satellite on which the eyes of the world’s scientists were fixed has fallen to Earth.’ (aajtak.intoday.in) [search strings = ; 6 exx]

With two recent exceptions,¹⁸ the identity of the specific objects which flash into brilliance as meteors has never been known beforehand, thus never expected or predicted by scientists. Reporting a meteor fall is reporting a creation or emergence for which the mind is unprepared. Similarly, earthquakes do not exist before they occur. Nor, prior to Columbus’s discovery of it, was ‘America’ in anyone’s mind:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(25)

a. krist:ophar kolambas-ne amerikā-kī Christopher Columbus- America- ‘Christopher Columbus discovered America.’ [search strings =

khoj kī. discovery did (aajtak.intoday.in) , 37 exx]

b. spen-ke kolambas-ne bhārat khoj.ne-ke kram-mẽ Spain- Columbus- India search- order-in amerikā-kī khoj kar dī. America- discovery do  ‘While looking for India, Spain’s Christopher Columbus discovered America.’ (upsssc.com) [search strings = ; 2 exx] The birth of each one of us as a specific person, as an “I,” is not preceded by an existence, at least not by one that most of us have access to: (26)

a. mãĩ bhārat-mẽ paidā huā aur bhārat-kī sanskr̥ti-mẽ I India-in born was and India- culture in palā-bar: hā was.raised-grew.up ‘I was born in India and was raised and grew up in India’s culture . . . ’ (rajivdixitji.com) [search string = ; 19 exx]

¹⁸ The TC3 event “was notable as the first such body to be observed and tracked prior to reaching Earth” (on October 7, 2008, in Sudan] (from Wikipedia). A second tracked meteor fell (as predicted) in Botswana on June 2, 2018.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -   b. mãĩ bhārat-mẽ paidā I India-in born ‘I was born in India.’ [search string =

ho become

293

gayā 

; 0 exx]

All these events are events that bring into being the subjects or (in the case of America) the objects referred to by their respective predicates.

11.4 Interaction of autogenous expressions with CV-promoting factors

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

There are at least five lexically specific strings known to me that suppress CV flux: mãĩ paidā ho.nā ‘I . . . to be born’, merā janm ho.nā ‘my birth . . . to occur’, ulkā.pin: d: gir.nā ‘meteor . . . to fall’, amerikā-kī khoj kar.nā ‘to discover America’, bhū.kamp ā.nā ‘earthquake . . . to occur’.¹⁹ In unemotional, constative (reportorial) statements these five resist manifestation as CVs. These five are CV-averse. In (23), (25), and (26) we have seen examples of three of these. Examples of the others: (27)

sādhāran: nagar-ke ek sādhāran: ghar-mẽ merā janm ordinary town  one ordinary home-in my birth huā / ho gayā. became / become  ‘I was born in an ordinary home in an ordinary town. (Bharat Bhushan, cited in hindivarta.com) [search strings = 80 exx / , 5 exx]

(28)

tīvr.tā-kā bhū.kamp āyā. intensity earthquake came (vs. CV = ā gayā [come ]) ‘Today in Gujarat there was an earthquake of medium strength.’ (abpnews.abplive.in) [search strings = , 28 exx / , 0 exx] gujarāt-mẽ Gujarat-in

āj today

madhyam medium

Searching for these CV-averse expressions turns up a few in which CVs do occur. A look at them reveals the identity and influence of CV-promoting factors. In (29) that factor is anteriority (cf. (16)–(19)): ¹⁹ Notice that the Russian counterparts of these expressions normally require manifestation in the perfective aspect: (i)

ja rodilsja v indii I was born. in India ‘I was born in India.’ [search string = я родился в Индии, 54 exx]

(ii) ja roždalsja v indii I was.born. in India ‘I was born in India.’ [search string = я рождался в Индии, 0 exx]

This is another feature that distinguishes the perfective in Russian from the Hindi-Urdu CV. See fn 17.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

294

 . 

(29)

čīnī khoj.kartāõ-ne kaī sāl pahle.hī amerikā-kī Chinese explorers- many years before America- khoj kar lī. discovery do  ‘Chinese explorers discovered America many years before [Columbus].’ (data found online) [search strings = X ; 2 exx]

“Prepared mind” and “unprepared mind” are concepts invoked by Bashir (1993: 2) in her account of CVs in Urdu and Kalasha: “A mind unprepared with reference to a particular event has no ‘pre-monitory consciousness’ of that event, and the event is not ‘consonant with the current state of mind of the speaker’ (Slobin and Aksu-Koc 1982: 196).” “Conversely, the prepared mind is one into which the (possibility of) the event (to be) reported is already integrated, either by virtue of its being anticipated, feared, desired, ordered . . . or by virtue of its being a natural, foreseeable consequence of a circumstance or action already realized.” If the existence of the speaker’s individual consciousness prior to being born is implied (at least as a rhetorical turn), then the CV paidā ho jā- [Lit: born become -] ‘to be born’ may occur:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(30)

a. mujhe paidā ho-nā thā, so mãĩ apne samay-se me. born become- was thus I self ’s time- paidā ho gayā. born become  ‘I did have to be born and thus in due course I was.’ (cf. (26a, b)) (from Neelam Shankar’s in www.hindisamay.com) [search string = ; 4 exx]

CVs paidā ho jā- and merā janm ho jā- may occur if speakers regret being born where or when they were. In such expressions the fact of birth is already “in mind”: (31)

a. kaī bār lag.tā.hai ki kahãã mãĩ many times seems that where I ho gayā. become  ‘Often I wonder why was I born in Haryana.’ [search string = ; 4 exx]

hariyānā-mẽ : Haryana-in

paidā born

(bhaskar.com)

b. kyā yah svapn hai athavā narak-mẽ merā janm ho  this dream is or hell-in my birth become gayā hai?  has ‘Is this a dream? Or have I been born in Hell?’ (hi.encyclopediaofjainism.com) [search string = ; 5 exx]

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

295

c. “kaī bār mãĩ kah.tā bhī hũũ ki mãĩ ġalat many times I saying even am that I wrong samay-mẽ paidā ho gayā.” time-in born become  ‘Often I may even go so far as to say that I was born at the wrong time.’ (www.bbc.com/hindi/sport-38432168) [search string = ; 4 exx] While reports of the occurrence of an actual (geophysical) earthquake invariably use the non-CV form āyā (28), use of the subordinating conjunction jaise ‘as if ’ allows the CV ā gayā freely to occur:²⁰ (32)

a. jaise.hī bhū.kamp āyā, sabhī t:eliphon lāinẽ vyast as.soon.as earthquake came all telephone lines busy ho gaĩĩ become  ‘As soon as the quake occurred, all phone lines became engaged.’ (www.jagran.com) [search strings = , 17 exx; , zero exx]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. aisā lagā jaise bhū.kamp āyā ho (Compare (32c).) so seemed as.if earthquake come had ‘It seemed as if there had been an earthquake.’ (www.wniindia.com) [search string = ; 20 exx] c. hādse-ke vakt aisā lagā jaise bhū.kamp ā gayā ho accident- time so seemed as.if earthquake come  had ‘At the time of the accident it seemed as if there had been an earthquake.’ (hindi.news24online.com) [search string = ; 25 exx] The freedom in the use of the CV form bhū.kamp ā gayā with jaise seen in (32c) is in accord with the concept of the prepared mind. Making the comparison of an accident to an earthquake presupposes that the idea of an earthquake’s occurrence is present in the speaker’s or writer’s mind.²¹

²⁰ Indeed, compared with their rare occurrence in constative contexts, CVs when used metaphorically are a common means for rhetorically vivid expression [aajtak.intoday.in/topic; from an address to Parliament]: (i) pī em bole “kal bhū.kamp āyā, ā hī gayā ākhir bhū.kamp ā hī gayā.” P.M. said yesterday earthquake came come   finally earthquake come   ‘The PM said, “Yesterday there was an earthquake. A real earthquake. Finally the earthquake came.” ’ ²¹ Compare Slobin and Aksu (1982: 196–7): “This interpretation casts light on a curious loosening . . . as the event referred to recedes in time from the moment of speech . . . . its status shifts and no longer requires being expressed as if entering an ‘unprepared mind’.” In (32c) the earthquake is now the upamān), employed to define the accident which is the “target” known, the “source,” the standard ( upamey) in a simile ( upamā). or comparee (

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

296

 . 

Similarly in (33) a public reaction is being compared to lamentations over an apocalyptic encounter with a meteor whose imagined fall is expressed with a CV: (33)

vilāp čal-rahā hai jaise pūrī mānav.tā-par koī lament proceed-ing is as.if whole humanity-on some ulkā.pin: d: gir gayā ho. meteor fall  had ‘There’s a hue and cry pouring forth as if a meteor has fallen on the whole of humanity.’ (data found online)

Amendment: Invoking “unprepared mind” as a factor favoring the nonoccurrence of a CV must allow for the use of CVs in expressing discordant (albeit not unexpectable) additional information:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(34)

tenduā rāt-ko yahãã roj pānī pī-ne ātā.hai. leopard night-at here daily water drink- comes kabhī.kabhī do.pahar-mẽ bhī ā jātā.hai sometimes afternoon-in too come  ‘The leopard comes here every night for water. Sometimes he comes in the afternoon, too.’ (from Chapter 8 of Premchand’s Godan; http://www.hindisamay.com; cf. Nespital 1997b: 48)

The first clause in (34), expressing the normal situation or state, sets up an expectation which is countered or amended by the second. This divergence from a norm is a promoter of CV manifestation. For instance the inherently CV-phobic phrase sāth denā ‘to accompany’ usually manifests in noncompound form. However, after a norm-setting clause a CV in the next clause is possible, if expressing an exception: (35)

paijāmā pīem mōdī.kī st:āil.kā mahatv.pūrn: aŋg hai, jis-mẽ pajama PM Modi’s style’s important part is which-in kabhī.kabhī jaiket: bhī sāth de detā.hai. sometimes jacket too company give  ‘A [churidar] pajama is an important part of the P.M.’s style but sometimes a (Western) jacket accompanies it.’ (hindi.oneindia.com)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -   (36)

297

kurtā-pāy.jāmā va sadrī ke.atirikt ve kabhī.kabhār sūt: bhī kurta-pajama and vest besides he occasionally suit also pahan lete.hãĩ. wear  ‘Beside his [normal] kurta, pajama, and vest, from time to time he also wears a suit.’ (https://hi.wikipedia.org/wiki)

Relevant data are too scanty for a quantitative demonstration of the CVpromoting effect of contrastive amendation.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

11.5 Concluding remarks In Section 11.2, I presented quantitative evidence in support of some earlier claims (Hook 1974) about the semantic factors that either promote or inhibit CV manifestation in Hindi-Urdu. Those same promoting or inhibiting factors are less influential in CV-poor languages such as Marathi and Nepali. The two sections that follow (11.3, 11.4) introduced CV-averse, autogenous expressions such as ‘I was born’, ‘a meteor fell’, ‘an earthquake occurred’ and explore their interaction with environments that promote CV manifestation. Among these the subordinator jaise ‘as if ’ is shown to be a strong promoter, unlike other subordinating conjunctions such as jaise hī ‘as soon as’. This capacity aligns with the effects on CV manifestation of other environments, such as “fear”-clauses, which reflect mental [vs. objective] states and events. Both “fear”-clauses and jaiseclauses feature specific meanings which are seen as deriving from the general sense of ‘prepared mind’. However, as evidenced by data on clauses expressing posteriority (15), it cannot be claimed that prepared mind is sufficient as a promoter of CV manifestation. Moreover, the dispreference for CVs in clauses expressing difficulty (as opposed to ease (18a, b)) cannot be ascribed to ‘unprepared mind’. (37) and (38) are unordered, nonexhaustive lists of factors promoting manifestation of verbal expressions in CV form and those inhibiting it. (37)

CV-promoters fear-clauses (14b) anteriority (17)–(19) ease (20a, b) until-clauses (16a, b) as-if clauses (32b, c, 33) contrastive amendation (34)–(36) regret (31b, c, d) anticipation (appendix)

(38)

CV-inhibitors complements of phasals (5b) posteriority (15a, b) difficulty (20c, d) in converbs (8b) in participles (7b) failure (21c, d) negation (10c, d, 11c) complements of modals (6b)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

298

 . 

Ordering these on scales of relative strength and investigating the effects of interacting or opposing factors must wait. It is clear, however, that searching for monocausal explanations to account for the appearance of CVs (rather than nonCVs) or non-CVs (rather than CVs) is not likely to bear fruit, though there is scope for grouping and reducing the number of factors listed in (37) and (38). Also needed is a comprehensive, ordered list of inherently CV-prone items (39) as well as an ordered list of inherently CV-averse items (40):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(39)

vectorophiles (40) čhor: (de)nā ‘to release; abandon’ bhūl(jā)nā ‘to forget’ bhããp(le/jā)nā ‘to guess’ . . . V-ne se mukar(jā)nā ‘to refuse to V’

vectorophobes bhūkamp ānā ‘earthquake to occur’ X-kā sāth denā ‘to accompany X’ V-ne lagnā ‘to begin to V’ X-ko šobhā denā ‘to behoove X’ V-nā čāhnā ‘to want to V’ ...

The items in (39) help identify and weigh CV-inhibitors; those in (40) help characterize CV-promoters. Verbs in Hindi-Urdu may be thought of as superpositions of CVs and corresponding SVs (non-CVs) that, depending on semantics and their discoursal environments, “decohere” to manifest as explicit forms. The recent rapid increase in online resources of unconsciously provided, natural (aka “found”) data now permits us to use quantitative evidence to verify (or refute) our intuitions about the semantic and discoursal conditions responsible for the use or non-use of CVs in Hindi-Urdu and opens a new way forward to satisfying the ultimate “Turing Test” for CV studies: The ability to predict the occurrence of CVs and required non-CVs (as well as optional CVs) in running texts and recorded conversations and to provide an explicit explanation for their occurrence in them.

Acknowledgments An earlier version of this chapter was presented at the International Hindi Conference in September 2016 at INALCO (Paris, France). I am indebted to Prashant Pardeshi, Taro Kageyama, Ghanshyam Sharma, Rainer Kimmig, Elena Bashir, Benjamin Slade, Rajesh Bhatt, Bettina Zeisler, Liudmila Khokhlova, Henrik Liljegren, and Sunil Bhatt for their comments, cautions, and suggestions.

Appendix Employing online data sources for quantitative studies L: Use of Google’s search engine as an “open” or uncontrolled corpus requires careful curation with attention paid to eliminate N-tuplicates, writing errors, and robo-translations. Be aware that the summary totals at the top of each search page are wildly inflated and

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 -  

299

cannot be relied on. The smaller totals given at the bottom of the last search page are also inaccurate. You should compare and count “hits” one by one by hand. P: Devise a search string that in the optimal instance will yield no more than three or four pages of search results. Too short a search string results in numbers of instances that are too large to cull for duplicates. Too long a search string yields numbers of instances that are too small to ensure statistical validity. If necessary the number of results can be increased by the use of the asterisk (*) as a wild card. It is important to examine each and every “hit” and check for misreadings (common in optical character recognition), N-tuplicates, writing errors, and robo-translations. Avoid strings that are parts of a song or poem or political slogan. E: A typical search involves four numbers. Two of them should show the frequencies of CV versus non-CV in the manifestation of an expression without the intuited or suspected factor. That ratio of CV to non-CV provides a baseline. The other two numbers introduce the factor and may show a skew. The more marked or dramatic the skew, the better. To illustrate let us take the strings (or ) and (or ) (vo din ā gayā and vo din āyā ‘that day came’) and search for the effect of and its common alternate (intazār / and in search results by intajār ‘wait’) on CV flux. We can exclude the presence of prefixing a hyphen to each of them. The first two pairs of search strings then are: a.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Google reports wildly inflated numbers for each of these strings but by clicking through to the final results page we find about 200 for (a.) and 180 tokens for (b.). Scrutinizing the dozen or so pages of results for (a.) and (b.) in a bid to find and drop duplicates and writing errors is an unmanageable burden on the short-term memory, so instead we reduce these totals by 10%. The next step is to add and to the search strings. Use of an asterisk (*) allows us to capture relevant data even though are separated from by intervening material such as ‘finally’, ‘at last’, ‘today’, etc.: a. b. String (c.) has fifteen examples. String (d.) has two. Entering all four numbers into a chisquare table: Total

Total

180

162

342

15

2

17

195

164

359

The two-tailed P value equals 0.0047. The association between rows (groups) and columns (outcomes) is considered to be very statistically significant. We may take the presence of ‘wait’ as an indication or proxy for “anticipation” as a CV-promoting factor.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

12 The matrix of verb-verb sequences in Tamil E. Annamalai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.1 Introduction The linguistic literature on V-V sequence in Indian languages employs a multitude of terms to designate this structure, of which serial verb construction (SVC) or serial verbs (Steever 1988; Pandharipande 1993) and complex predicate (CP) (Verma 1993; Butt 1997, 2010) or compound (Hook 1974; P. Dasgupta 1977; Kaul 2006) / composite (M. Dasgupta 1990) verb are two overarching names.¹ They do not, however, cover the same set of data and their internal properties differ. At the theoretical level, CP covers V-V and N-V predicates, while SVC covers only the former. But studies of the languages of the world are not unanimous about the differences between the serial verb construction (Stewert 2001; Aikhenvald 2006; Aikhenvald and Dixon 2006) and the complex predicate (Alsina, Bresnen, and Sells 1997; Hinrichs, Kathol, and Nakazawa 1998; Amberber, Baker, and Harvey 2010). Nor are they unanimous about what structures go under each of these designations. For the purpose of this chapter I will categorize nontheoretically as “V-V sequence” any sequence of two verbs in Tamil where one of the verbs does not have lexical status referentially and often structurally. Tamil belongs genetically to the Dravidian family of languages, typologically to the left-branching verb-final languages, and areally to the matrix of south Asian languages sharing specific grammatical features. Before discussing the properties and types of V-V sequence in Tamil, I shall first filter out the verb sequences in Tamil that are not considered under the V-V sequence, where V₁ is of the morphological category of verbal participle composed of verb + past tense. In these sequences also, one of the verbs is nonlexical or a light verb in the sense that it is semantically bleached, though it is often (but always in formal Tamil) phonologically indistinguishable from full verbs that occur in isolation. Examples of these are given in Table 12.1 with minimal comments for the attention of typological researchers, since these sequences may have functional and formal properties in other languages of one of the types of V-V sequences discussed in this book. (In the table, the gloss of “light verbs” is given in block letters and in abbreviated form as in the example ¹ Fedson (1981) uses “serial verb” and “compound verb” as alternate names for the same phenomenon.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

301

Table 12.1 List of light verbs and the corresponding full verbs in Tamil Light verbs

Grammatical meanings

Full verbs

pār pār kāt:t:u

TRIAL (with infinitive) ‘try to do’ TRIAL (with adverbial participle) ‘try doing’ DEMONSTRATION (with adverbial participle) ‘demonstrate doing’ EXPERIENCE (after bare verb) ‘undergo an action’

pār ‘see, look’ pār ‘see, look’ kāt:t:u ‘show’ (causative of kān, : a synonym of pār) padu : ‘experience, be touched’ vai ‘place, put’ vai ‘place, put’

padu :

CAUSE (with infinitive) ‘cause an action’ CONSEQUENCE (with adverbial participle) ‘have a consequence intentionally’ pōdu CONSEQUENCE (with adverbial participle) ‘have a pōdu : : ’drop, put’ consequence incidentally’ tole DISGUST (with adverbial participle) ‘express disgust tole ‘have something about action by self or others’ lost’ (intr.), ‘lose something’ (tr.) azu RELUCTANCE (with adverbial participle) ‘express a zu

‘cry’ reluctance to do’ kizi DISMISSAL (with adverbial participle) ‘express being kizi

‘tear’ (tr.) dismissive of other’s ability to do’ *kudu HELP (with adverbial participle) ‘do to help others’ kodu : : ‘give’ tā HELP (with adverbial participle) ‘do to help others’ tā ‘give’ vidu PERMISSION (with infinitive) ‘allow to do’ vidu : : ‘release, let go’ (tr.) (v)udu vidu : TRANSMISSION (with adverbial participle) ‘get : ‘release, let go’ (tr.) across help or a message’ (u)du COMPLETION (with adverbial participle) ‘assert an vidu : : ’release, leave’ act as done or to be done’ (intr.) pō COMPLETION (with adverbial participle) pō ‘go’ ‘completion of change from one state to another’ āgu COMPLETION (with adverbial participle) āgu ‘come to be, ‘completion of the process change from one state to become’ another stated impersonally’ kidu AFFECT (with adverbial participle) ‘do something ko:l ‘hold, have, contain’ : that affects self ’ iru STATE (with adverbial participle) (i) ‘being in a state iru ‘be, stay’ after act’, (ii) ‘stating of an act to be true in the past’, (iii) ‘saying there is evidence of a past act’ kit:uru CONTINUITY (with adv. participle) ‘action kon: diru ‘be holding : continuing over time’ (in time)’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

vay vay

* tā ‘GIVE’ is synonymous with kudu : and interchangeable with it except when the verb sequence is lexicalized.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

302

. 

sentences; each light verb is glossed in terms of its grammatical meaning, with a noun in block letters to indicate broadly its semantic function.)

12.2 Infinitive + verb Tamil uses light verbs with infinitives to express an activity (such as causation) performed on the full verb expressed in the infinitive, which activity is closer to a grammatical function. The light verbs are vay, vidu, : and pār, whose equivalent full verbs mean respectively ‘place’, ‘let go’, ‘see / look’. The light verbs express the nonreferential or grammatical meanings ‘cause’, ‘let’, and ‘try’ respectively. Note that these light verbs after the verbal participle discussed later in the chapter have different grammatical meanings. (1)

a. nī ivan-e padi.kk-a : you he- read- ‘You made/let him read.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. nī id-e padi.kk-a : you it- read- ‘You tried to read this.’

{va-cc-e --2

/ vi.t:-t:-e} / --2

pā-tt-e --2

The ‘infinitive + light verb’ structure cannot be instanced by lexical compounds because a grammatical element may be inserted between Vs, and words in the sentence may be scrambled to occur between Vs just as in a sequence of two lexical verbs. Though the morphological form of V₁ is infinitive like the compound verbs in Japanese (see Kagayema, Chapter 2, this volume), it is a syntactic compound like Type 3 of the Japanese compound verbs. But not all V₂ in this sequence are syntax-building elements.

12.2.1 Get passive The verb padu : ‘be touched, be experienced’ occurs with bare verbs and gives the light verb the meaning of ‘be subjected to’. Its grammatical equivalent in English is the so-called ‘get passive’, which is glossed as ‘get done’. (2)

kadale nal.l-ā {sudu : : / varu} pad-a-le : peanut good- roast / fry -  - ‘Peanuts did not get roasted/fried well.’

This structure is a good candidate to be a V-V sequence that is monoclausal. It is also an instance of bare V without inflection. It is monoclausal since it does not have the syntactic properties attributed to the infinitive + light verb structure mentioned above.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

303

12.2.2 Duplicated V Another case of a monoclausal V-V is duplication of the base of the finite verb in its verbal or nominal form for intensity. V₁ could be a nominal form also, which is a derived noun from the verb (as in 3c). This duplication is possible with intransitive verbs also, as in (3b), and so it is not generated in the phrase structure. (3)

a. nān avan-e oru {adi : I he- a hitting ‘I gave him a good beating.’

adi-cc-ēn : hit--1

/ midi midi-cc-ēn} / stamping stamp--1

b. nān oru ciri ciri-cc-ēn I a laugh laugh--1 ‘I had a good (scary) laugh.’ c. nān oru siri-ppu siri-cc-ēn I a laugh- laugh--1 ‘I had a good (scary) laugh.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.2.3 Two Vs in code mixing Tamil makes a V-V sequence by adding the verb pan: nu : ‘do’ to an English verb that is mixed in Tamil. pan: nu carries the inflection of the sequence such as : try-pan: nu ‘try’, marry-pa n nu ‘marry’, reserve-pa n nu ‘reserve’, etc. This looks : :: :: like a V-V sequence drawing the verbs from two languages, but it is not.² It is a morphological strategy to inflect a verb mixed from another language (Annamalai 1971: 26).

12.3 Verbal participle + verb I shall now discuss the V-V sequence in Tamil that has received most attention from linguists, which is the structure verbal participle (VbP) + light verb (often called in Tamil linguistics auxiliary verb, differentiating it from the grammatical category of auxiliary). The second verb is called a light verb here in the sense that it

² Baker and Harvey (2010: 16) claim syntactic V-V structure for a construction of V + light verb, where V is a borrowed word. The borrowed word is assimilated in their example, which is not the case in code mixing in Tamil. Butt (2010: 52) treats such mixing as instances of the N-V structure that is CP without giving any facts of their syntactic behavior in support of such a treatment. The borrowed form need not be an English verb; it could be an English predicate preposition (on-pan: nu : ‘turn on (the switch)’; it could be an English adjective and past participle when the Tamil verb is āgu ‘come to be’ (cool- āgu ‘become cool’, tired- āgu ‘become tired’) (Annamalai 1971: 20).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

304

. 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

carries a grammatical function and in the sense that it is less than a full verb.³ This sequence has functional, if not formal similarities with the types of V-V sequence in Japanese described in Chapter 2 of this volume. In a sequence of two or more verbs, all verbs other than the last one are in the morphological form of the verb inflected for past tense, also called the adverbial participle, expressing a temporal precedence relationship to the following verb. The most common meaning of this sequence is conjunction of verbs in a temporal sequence. In this structure of V₁ - V₂, both verbs could be full verbs that have referential meaning semantically, have the same lexical status, and come from an infinite set of verbs. These V-V sequences have the sequence of VbP in the form of verb + past-tense marker, which is V₁ and V₂ as a finite verb (or any verb form of the higher clause). There is no specific marker of syntactic dependency in this structure other than the past-tense marker, which simply indicates temporal sequencing of verbs. This raises the question whether these sequences of two full verbs are instances of a V-V sequence with special syntactic and lexical properties. They are not. They represent a union of clauses. A VbP clause is shown to be a subordinate clause as it allows moving out a constituent that is inside it (Annamalai 1970). Hence, the verbs in (4) will not qualify as serial verbs.⁴ Furthermore, the verbs in such a symmetrical verb sequence can have different arguments. Borrowing the book and reading it, in (4a), could take place at the same location (library) or reading could take place in another location that is unspecified or specified (e.g. house). V-V has two locatives in (4a) and two subjects in (4b). VbP+V, therefore, is biclausal, where the Vs are in a temporal sequential relationship. (4)

a. nān laibreri-le pustagam edu-ttu (vī.t:t:u-le) padi-cc-ēn : : I library- book take- (house-) read--1 ‘I borrowed a book from the library and read it there / in the house.’ b. nān laibreri-le pustagam edu-ttu avan padi-cc-ēn : : I library- book take- he read--1 ‘I borrowed a book from the library and he read it.’

This illustrative V-V sequence is a syntactic structure and so cannot be considered a verbal compound. As such, it is not a unitary predicate and does not qualify to be a V-V sequence of the types under discussion in this book.

³ The term “serial verbs” is used in Dravidian historical linguistic research to refer to a sequence of verb bases (Steever 1988), but the term includes inflected forms of the verb as well from a crosslinguistic point of view (Aikhenwald 2006: 3–4). ⁴ Stewart (2001: 270) raises the absence of tense to the level of setting a parameter for a language to have Serial Verb Construction (SVC). He generalizes the presence of bare verbs in SVC in the West African language Edo as a principle that stipulates that “No verb in the serial verb construction can bear morphological tense inflection.” But the Tamil data of the V-V sequence discussed in 12.3.1 would falsify this principle when this sequence is treated as SVC.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

305

Steever (2005: 3) gives a matrix of taxonomy of V-V sequences in Tamil through binary branching. I shall examine in this chapter the syntactic (i.e. lexical) integrity and semantic homogeneity of VbP+V forming one clause. Steever’s oneclause structure and two-clause structure are defined by these properties. In such V-V structures, V₂ could be a light verb, which is not referential, and it augments the meaning of V₁ (verb of action or verb of state) or of the whole of the sentence (i.e. proposition). It comes from a finite set of verbs. Linguists have studied the syntax of this structure (Schiffman 1969; Steever 2005) and its semantics (Lindholm 1969; Fedson 1981; Annamalai 1985). All scholars agree that the dividing line between the category of full verbs and that of light verbs is porous, and they differ in the number of verbs they designate as light verbs. Annamalai (1985) gives a set of diagnostic tests to identify light verbs after VbP, but claims that they form a cline in a scale of “verbhood” from being close to a full verb with referential meaning to a bound form with grammatical meaning and with regard to restrictions on collocation with specific full verbs (syntagmatic) as well as in relation to each other (paradigmatic).

12.3.1 VbP + semilight verb Let me examine some constructions similar to the above VbP-V sequence, but here the V₂ is not a full verb. It is not a light verb (see (5)) either.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(5)

a. nān inda pustaga.tt-e padi-ccu.p : I this book- read- ‘I tried reading this book.’

pā-tt-ēn try--1

b. nān avanu-kku inta pustaga.tt-e I he- this book- ‘I showed him reading this book.’

padi-ccu.k : read-

kā.t:-t:un-ēn. show--1

Clitics (shifters) such as the interrogative marker or emphatic marker could occur with the VbP showing that VbP+V is not lexicalized. The question remains if the sequence is monoclausal in (5). Note that kāt:t:u ‘show’ is semantically the causative of pār ‘see’, but we are not deconstructing the lexical verbs into semantic primes that are in a clausal relationship as in a generative semantic analysis. The whole VbP clause is in object relation to the finite verb pār in its semilight verb sense ‘try’. This verb requires a clausal object. (5a) would be ungrammatical without the VbP, which should translate as ‘I tried this book’ but would translate with the full verb meaning of pār, i.e. ‘I saw this book’. The same is true of kāt:t:u in its semilight verb sense of ‘try to show’. That V₂ has a clausal object and V₁ is the predicate of this clause shows that V-V in (5) is not a monoclause, though V₂ is a semilight verb semantically. The purpose of trying in (5a) is left unsaid, which

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

306

. 

could include ‘if I will be able to read, pass the exam etc., if the book is difficult, good, etc.’. The unsaid objective of trying to prove would be ‘that I was able to read, that he could see the difficulty, goodness, etc. of the book’. If any of the purpose is present in the sentence, it will be a complement clause of V-V with -nnu ‘that’. (6) is syntactically similar to (5), but mudi : ‘finish’ is a full verb having object relationship to the VbP. The VbP can be absent with this verb, giving the meaning ‘I finished this book’ to (6a). (6)

a. nān inda pustaga.tt-e (padi-ccu) : I this book- (read-) ‘I finished (reading) this book.’

mudi-cc-ēn : finish--1

mudi : does not represent the aspectual sense of completion, as the light verb vidu : does (see Section 12.3.3.1). V-V in (6) cannot be said to belong to the set of symmetrical serial verbs because it is not monoclausal and mudi : ‘finish’ has an object relationship to the VbP. (6d), which is the negative of (6a), negates only V2. The scope of negation in (6d) is the verb of finishing. V₁ and V₂ could have adverbs modifying each or one of the verbs separately, as in (6b) and (6c).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(6)

b. nān inda pustaga.tt-e āse.y-ā padi-ccu : I this book- desire- read- ‘I finished reading this book enthusiastically.’

mudi-cc-ēn : finish--1

c. nān inta pustaga.tt-e denam padi-ccu vēgam-ā mudi-cc-ēn : : I this book- daily read- speed- finish- -1 ‘I read this book daily and finished it fast.’ d. nān inda pustaga.tt-e innom padi-ccu : I this book- yet read- ‘I have not yet finished reading this book.’

mudi.kk-a-le : finish--

The V-V sequence in (6) is not a verbal compound, and V₁ is the head of a sentence which is a complement of V₂. Though Type 3 of V-V sequence in Japanese has syntactic properties of complements like this sequence in Tamil, the syntactic structure of this sequence in Tamil is not a complex predicate structure, as in Japanese. Functionally though, it is like Type 4 in Japanese, which has the attitudinal meaning of ‘try’ for the lexical verb ‘see’, as in Tamil.

12.3.2 VbP + attitudinal light verb In the following sentences, which are monoclausal, the light verb expresses an attitude of the speaker to the proposition. This sense is not one of grammatical

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

307

category, but it is semantic, as these light verbs behave as attitudinal verbs such as “be in anticipation of,” “be reluctant,” or “disbelieve.” No new verbs of attitude are to express a specific attitude, but semantically appropriate verbs permit the inference of an attitudinal sense in the context of their occurrence with VbP. Without VbP, these verbs do not express any attitudinal sense. VbP + attitudinal light verb has a monoclausal structure by virtue of the fact that the full verb and the light verb do not have separate arguments or modifiers. In fact, the light verb does not have any assigned arguments. The light verbs of attitude share the form of full verbs vay ‘place down’, pōdu : ‘put / drop down’, tole ‘get lost / lose’, azu ‘cry’, and kizi ‘tear’. Their attitudinal meanings respectively are vay ‘do deliberately for a consequence, with an eye for future usefulness’, pōdu : ‘same as vay, but doing nondeliberately’, tole ‘do against one’s will or reluctantly as a last option’, azu ‘do out of compulsion or as having no option’, and kizi ‘express lack of confidence in another person doing a thing’. In the first two examples, the attitude is deliberation (i.e. carefulness) and absence of it (i.e. casualness) respectively. (7)

a. nān inda pustaga.tt-e padi-ccu-va-cc-ēn : I this book- read----1 ‘I read this book (as it might be of use later).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. nān kadav-e {tera-ndu-va-cc-ēn / tera-ndu-pō.t:-t:-ēn} I door- open---- / open----1 ‘I kept/left the door open.’ c. nān inda pustaga.tt-e padi-ccu.t-tole-cc-ēn : I this book- read----1 ‘I read this book (as there was no other way).’ d. avan kadesi.y-ā pō-y.t-tole-nj-ān : he end- go--(intr.)--3 ‘He finally left and got lost.’ e. naan kadesi.y-ā avan-ukku pattu rūbā : I end- he- ten rupee ‘I finally gave him ten rupees reluctantly.’

kudu-ttu-azu-d-ēn. : give----1

f. nī inda pustaga.tt-e padi-ccu.k-kizi-cc-e : you this book- read----2 ‘(I don’t believe that) you will read this book.’ Some light verbs could retain the attitudinal sense even when not in combination with a full verb. The VbP could be absent in (7e) and azu could have the attitudinal sense. This is not the case with (7a), where vaccēn will have the sense ‘I put down this book’ when the VbP is absent. The exception with regard to azu could be explained

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

308

. 

as a case of metonymy, where the transferred sense by contiguity remains even when the contiguous word is absent. That the light verb does not stand alone in its lightverb sense is a defining property of light verbs. These light verbs do not have predicative properties separately from the full verb and so may be taken to be a unitary predicate, though a complex one. This overlaps with Type 4 in Japanese semantically in one of its functions, but not compositionally.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.3.3 VbP + grammatical light verbs 12.3.3.1 Assertion of occurrence Light verbs perform a grammatical function in the following sentences. The grammatical function is the expression of speaker perception about the event: whether it really took place and ended (vidu : /-du), : whether the effect of the event lingers, having relevance after the event, both of which are placed under the category of aspect in the analytical framework of the Russian verb (Schiffman 1969), whether the event’s happening is known through inference, which is categorized as evidential (iru / -ru), and whether the event has any effect on the subject (ko(:l) / kidu), : which is placed under the category of voice. The light verbs in this class of V-V sequence differ in their grammatical status. The status varies in range from being semantically opaque, being inflectional morphologically (that is, they are not selective about combining with specific verbs), and being reduced phonologically to nonlexical canonical forms, at one end of the spectrum to being semantically transparent, being syntactically a verb in assigning cases, and being selective in their combinatory choices with verbs. These light verbs thus form a cline in their freedom to combine with verbs and their status, being unambiguously categorized as light verbs (Annamalai 1985: 140). (8)

a. nān inda pustaga.tt-e I this book- ‘I read this book.’

padi-cc-ēn : read--1

b. nān inda pustaga.tt-e I this book- ‘I have read this book.’

padi-ccit:t:-ēn : read- ---1

c. nān inda pustaga.tt-e padi-ccu : I this book- read- ‘I have finished reading this book.’

mudi-ccit:t:-ēn : finish----1

d. nān inda pustaga.tt-e nālu tadave padi-ccit:t:-ēn : : I this book- four times read----1 ‘I did read this book four times.’ e. nān inda pustaga.tt-e nā:lekku padi-ccidu.v-ēn : : I this book- tomorrow read---FUT-1 ‘I will certainly read this book tomorrow.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

   -   

309

Tamil verbs are unmarked for accomplishment or successful completion. (8a) could be followed without contradiction by an equivalent of ‘but I did not finish it’. This follow-up sentence will be a contradiction in (8b).⁵ So this light verb is described as marking the successful completion of an act or closing of an event (Annamalai 1985: 80–103; Fedson 1981: 80–93; Steever 2005: 185–95). This, however, is questioned by (8c), where the light verb combines with the mudi : ‘finish’, which is not redundant. This light verb can, therefore, be said to assert the fact that the event did take place. (8d) asserts that the reading of the book took place four times; (8e) asserts that the reading of the book will take place. This assertion led linguists to assign this light verb the grammatical function of definiteness in future. But both in the past and in the future it is the assertion of the occurrence of the event or existence of a state. The light verb in (8) is in the form of -du, : which is not a canonical form of a verb in Tamil and there is no full verb of this form. Nevertheless, it relates to the light verb vidu : used in the same grammatical sense in formal Tamil. The corresponding full verb vidu : in both formal and colloquial Tamil has the intransitive sense of ‘release or leave’. (cf. kāccal / maze vit:t:uccu ‘the fever / rain stopped’). -du : behaves like a verb in that it belongs to a conjugation class of verbs and it takes all inflected forms of the verb, such as the infinitive and conditional. The meaning of this light verb is largely divorced from the meaning of its counterpart full verb. It is free to combine with any verb including stative verbs. It is a light verb that has a grammatical function and counts as a V-V sequence as discussed here. Linguists have attributed a context-specific meaning as part of the meaning of this light verb, which is the realization of the occurrence of an event that was expected to happen or not to happen (see Annamalai 1985: 91–2; Fedson 1981: 63–4; Bashir 1993: 6 for a discussion of it). (9)

a. obāmā tērdal-le jeyi-cci-t:t:-ār Obama election- win----3.. ‘Obama won the election (against the odds).’ b. nān kīze vizu-ndu-t:t:-ēn I down fall- ---1 ‘I have fallen down (unexpectedly).’ c. nī tūng-i-d-ād-e : you sleep---- ‘Don’t fall down (be careful).’

It could, however, be argued that realization of expectation or realization contrary to expectation is pragmatic and is a conversational implicature. The implicature ⁵ This is a feature of verbs in some other Asian languages also such as Japanese and Burmese, but is variously analyzed (Kato 2018).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

310

. 

arises from Grice’s principle of cooperation, which includes the maxim of quantity, which stipulates that you do not make your utterance more informative than is required (cited in Levinson 2000: 14). The sense relating to the expectation of the occurrence is implicated by the utterance, not said in it. That this sense is an implicature is borne out by the test of reinforceability (Levinson 2000: 15). This sense is not made redundant by having explicit adverbs of expectation in the sentence (9d). The sentence with the first adverb ‘as expected’ is said when the father is expected to arrive at five o’clock and his arrival is asserted to have occurred at that time; it includes the second adverb, ‘unexpectedly’, when the arrival of the father was expected at another time. Neither case is redundant to the implicature of expectation.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(9)

d. appā anju mani-kku {edirpātt-a-padi/ : : edirpā.kk-āma} father five hour- expect-- / expect- va-ndu-t:t:-ār. come----3.. ‘Father came at five o’clock as expected/unexpectedly.’

There are two other light verbs, pō ‘go’, āgu ‘come to be, become’, which are used near-synonymously as -du : in the sense of asserting occurrence. pō is used preferentially when the occurrence is of change of state and it combines with change-of-state verbs such as kedu : ‘get spoiled’; āgu is used (its agreement marker is neuter singular, taking the event itself as its subject, and it is used only as a finite verb⁶ in the light-verb sense) only when there is a presupposition that the asserted occurrence is expected to happen. Though it is linguistically appropriate to say either of the two sentences in (10), it is socially inappropriate to say (10b). Both poo and āgu could co-occur to strengthen the assertion in which two light verbs are in a sequence (10c).

⁶ The finite verb includes the indicative as well as the modal as long as the indicative is positive and the modal is obligatory. (i)

a. nān inda pustaga.tt-e padi-ccā-ccu : I this book- read---3. ‘I have read this book / I have not read this book.’

/*padi-cc-āg-a-le : / read----

b. nān inda pustaga.tt-e padi-cc-āg-anum / *padi-cc-āg-a-lām : : : I this book- read----must / read----may ‘I ought to read this book / I may certainly read this book.’ Fedson (1981: 69) notes that aagu is acceptable, though restrictively, in the negative indicative (when the sentence implies ‘not yet’) and in the relative participle used in the temporal clause (cāppit:t:āna peragu ‘after finishing eating’ [which is less attested]; *vandāna peragu ‘after finishing coming’ is unacceptable).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -    (10)

311

a. avan cettu.p-pō-n-ān he die---- ‘He did die.’ b. avan ce-tt-ā-ccu he die---+3. ‘He is dead (as expected).’ c. avan ce-ttu.p-pō-y-ā-ccu he die-----+3. ‘He is dead and gone.’

The light verb pō could co-occur with the light verb du : (10d), which is an assertion of change of state and its finality.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(10)

d. avan ce-ttu.p-pō-y-t:t:-ān he die------3. ‘He is dead beyond revival.’

12.3.3.2 Resulting state The ubiquitous verb iru ‘be’, as a light verb combined with VbP, has three senses, namely remaining in the state of a performed act to be relevant to another act or to the speech act; of an event, existing before the time of the speech act but relevant to it; extrapolating an act from the state that persists. These three senses are named respectively the perfect aspect or resulting state, existential past, and evidential. These three senses, however, are related, and the common thread is the existence of a state arising out of an act, which could be evidence for the act. This light verb is semantically transparent, as glossed in the example sentences, but in combination with VbP assumes nonreferential meaning. The sentences in (11) illustrate the first sense (i.e. resulting state), which is discussed after the following example sentences. (11)

a. avan {kudi-cc-iru.kk-ān / kudi-cc-iru-nd-ān : : he drink--stat-- / drink----3. kudi-cc-iru-pp-ān} : drink----3. ‘He is drunk / He was drunk / He will be drunk.’

/ /

b. on cat:t:e {kizi-nj-iru.kku / kizi-nj-iru-nd-udu / your shirt tear---Ø/ tear----3. / kizi-nj-iru.kk-um} tear---3. ‘Your shirt is torn / Your shirt was torn / Your shirt will be torn.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

312

.  c. avan on vīt:t:u-le {ukkā-ndu-ru.kk-ān / ukkā-ndu-ru-nd-ān / he your house- sit----3. / sit----3. / ukkā-ndu-ru-pp-ān} sit----3. ‘He is sitting in your house / He was sitting in your house / He will be sitting in your house.’ d. avan engit:t:e panam {kē.t:-t:u-ru.kk-ān / kē.t:-t:u-ru-nd-ān / : he I- money ask----3. / ask----3. / kē.t:-t:u-ru-pp-ān} ask----3. ‘He has asked me for money / He had asked me for money / He will have asked me for money.’

The sentences in (12) illustrate the second sense (i.e. past existence). (12)

a. avan munnāle oru tadave kudi-cc-iru.kk-ān : : he before one time drink---Ø-3. ‘He has drunk once before (in the past).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. on cat:t:e munnāle oru tadave kizi-nj-iru.kku : your shirt before one time tear---Ø-Ø ‘Your shirt has been torn once before (in the past).’ c. avan munnāle oru tadave on vīt:t:u-le ukkaa-ndu-ru.kk-ān : he before one time your house- sit- --Ø-3. ‘He has sat in your house once before (in the past).’ d. avan munnāle oru tadave en-git:t:e panam kē.t:-t:u-ru.kk-ān : : he before one time I- money ask---Ø-3. ‘He has asked me for money once before (in the past).’ The sentences in (13) illustrate the third sense (i.e. evidential), stating that the event was not observed directly by the speaker. They contrast with parallel sentences without iru. The source of the information about the event (i.e. drinking beer, reading the book, eating fish) in (13) in the past is an extrapolation of the speaker from the state resulting from the act observed by the speaker after the act. When avan ‘he’ is replaced with nān ‘I, the speaker’, there is no scope for extrapolation and the sentences in (13) attest the direct knowledge of the speaker as to the act as well as the resulting state. iru in the present tense and the future tense may not express the resulting state, but they state the speaker’s inference of the occurrence of the act by deduction from other facts outside the spoken sentence. This inference could cover the resulting state as well, in which case iru is repeated after the VbP, giving kudiccirundurukkān ‘has drunk beer / is drunk :

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

313

with beer’ and kudiccirunduruppān ‘would have drunk beer / would have been : drunk with beer’, but not *kudiccirundurundaan ‘had drunk beer / was drunk with : beer’. In other words, when there is no direct knowledge of an act, iru must be used as the resulting state is the linguistic means of transmission of the direct knowledge; if the resulting state is not conveying direct knowledge, iru must be repeated. The difference between iru in the present tense and in the future tense is that the future expresses a degree of probability.⁷ (13)

a. avan bīr {kudi-cc-iru-nd-ān / kudi-cc-iru.kk-ān / : : he beer drink----3. / drink---Ø-3. / kudi-cc-iru-pp-ān} : drink----3. ‘He was drunk with beer / He is drunk with beer / He may have been drunk with beer’ b. avan inda pustaga.tt-e {padi-cci-ru-nd-ān / padi-cc-iru.kk-ān/ : : he this book-acc read----3. / read---Ø- / padi-cc-iru-pp-ān} : read----3. ‘He had read this book / He has read this book / He may have read this book.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. avan mīn {sāppi.t:-t:u-ru-nd-ān / sāpp.t:-t:u-ru.kk-ān / sāppi.t:-t:u-ru-pp-ān} he fish eat----3. eat---Ø-3. eat----3. ‘He had eaten fish /He had eaten fish / He may have eaten fish.’

While VbP + iru + iru is an instance of V-V-V sequence, the question to be answered is whether iru is the same light verb that is repeated in its different senses or whether there is more than one homophonous light verb. The light verb iru does not occur with the full verb iru (14b) in the resulting state sense, but occurs with it in the senses of existential past (14c) and evidential in present and future (14d). The reason is the fact that the light verb iru in (14b) is high on a scale of verbhood and is closer to the full-verb meaning. Two full verbs⁸ or verbs closer in status are not repeated in a VbP + V construction. See, for example, *pāttuppāttān ‘(he) tried seeing’, *vanduvandān ‘(he) continued coming’.

⁷ iru in future also refers to counterfactual occurrence. (13a), for example, could also mean that his drinking beer could have taken place, but did not for a reason that is not stated. ⁸ VbP may be duplicated to express an action repeated many times. This is not a V-V sequence. va-ndu pā-tt-ān (i) avan enn-e va-ndu he I- come- come- see--3. ‘He came repeatedly to see me.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

314

. 

(14)

a. avan t:ōkyō.vu-le {iru-nd-ān / iru.kk-ān / iru-pp-ān} he Tokyo- be--3. / be-Ø-3. be--3. ‘He was in Tokyo / He is in Tokyo / He will be in Tokyo.’ b. *avan t:ōkyō.vu-le {iru-ndu-ru-nd-ān / iru-ndu-ru.kk-ān / he Tokyo- be----3. / be---Ø-3. / iru-ndu-ru-pp-ān} be---3. ‘He had been in Tokyo / He has been in Tokyo / He will have been in Tokyo.’ c. avan munnāle oru tadave t:ōkyō.vu-le iru-ndu-ru.kk-ān : he before one time Tokyo- be---Ø-3. ‘He has been in Tokyo once before.’ d. avan t:ōkyō.vu-le {*iru-ndu-ru-nd-ān / iru-ndu-ru.kk-ān / he Tokyo- be----3. / be---Ø-3. / iru-ndu-ru-pp-ān} be----3. ‘He had been in Tokyo / He has been in Tokyo / He will have been in Tokyo (I guess).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The imperative construction requires the verb that admits agency. iru ‘be’ as a full verb can be imperative and so can the light verb, as long as the bleaching of its sense of iru is slight. The light verb elsewhere does not have agency. (15)

a. nī ange iru9 you there be- ‘You be there.’ b. nī ange ukkā-ndu-ru you there sit---be. ‘Be sitting there.’ c. ? nī biir kudi-cc-iru : you beer drink--- ‘?You, be drunk with beer.’ d. ?? nī inda pustaga.tt-e padi-cc-iru : you this book- read--- ‘??Be (well) read in this book.’

⁹ The bare verb is the form of the imperative and so there is no morpheme to indicate this grammatical function. Consequently, there is an extra gloss -IMP- with no morpheme to match in the example sentence.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

315

e. *nī appā-git:t:e ad-e kē.t:-t:u-ru you father- it- ask--- ‘*You stay asked of this with father’ Note that (15 b–e) would be good if the VbP + V construction is disjunctive with the addition of -t:t:u to the VbP. This makes V the full verb iru, which is shown by the fact that it can have adverbs modifying it. (16)

a. nī ange bīr kudi-ccit:t:u : you there beer drink--- ‘You be there, having drunk the beer.’

iru be-

b. nī laybreri-le inda pustaga.tt-e padi-ccit:t:u iru : you library- this book- read--- be- ‘You be in the library, having read this book.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

c. nī kade-le.y-ē appā-git:t:e ad-e kē.t:-t:u-t:t:u iru : you store-- father- it- ask--- be- ‘Stay right in the store after having asked the father about it.’ This is true also when the verb is inflected for the durative aspect with kit:t:uru, which has the verb iru. The sentences in (16) will translate respectively as ‘keep drinking beer there’, ‘keep reading this book in the library’, and ‘keep asking the father about this right in the store’. The above facts about iru are not incompatible with the analysis of it in all its three senses as having one light verb and one lexical verb (Steever (2005: 169) considers the light verb to have the meaning of location; Lindholm (1969), Schiffman (1969: 140), Annamalai (1985: 135), an Fedson (1981: 40) consider it to have the meaning of state). The different senses of the light verb iru are explained by its place in different structural configurations (Schiffman 1969: 140; Steever 2005: 168–85) or by the variable status of iru in relation to its verbhood and semantic transparency, discussed here and in Annamalai (1985: 124–36). The advantage of variability analysis is that it allows iru to occur twice in a verb sequence. The lexical meaning of location for iru gets credence from the fact that another verb of posture keda : ‘be lying’ can occur in place of iru to give the sense of lying rather than being.¹⁰ But the light verb keda : does not have the sense of past existence and extrapolation. Its occurrence is restricted in the resulting state sense also with VbPs that do not allow the result of the act to be in the lying

¹⁰ kida, : like iru but restrictively, becomes semantically distant from its full verb meaning of ‘lie’ to give the meaning of resulting state with the connotation of casualness (similar to that found in ‘lie around’): pezeccirundaa ‘if (I) am alive’ vs. pezeccukedandaa ‘if (I) happen to be alive’. :

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

316

. 

posture. The restricted parallel with keda : shows that iru in some of its occurrences has lexical status. (17)

a. kadavu {tera-nd-iru-nd-udu /tera-ndu keda-nd-udu} : door open----3. / open--lie--3. ‘The door was open / The door lay open.’ b. nān tirumbi va-nt-a-ppa eli {ce-ttu-ru-ndu-du / I back come--rel-when rat die----3. / ce-ttu.k-keda-nd-udu} : die--lie--3. ‘When I came back, the rat was dead / was lying dead.’ c. avan inda pustaga.tt-e {padi-cc-iru-nd-ān / *padi-cci.k-ke da-nd-ān} : : : he this book- read----3. / read--lie--3. ‘He had read this book / *lay read this book.’ d. avan ange {ukkā-ndu-ru.kk-ān /*ukkā-ndu-keda.kk-ān} : he there sit---pr-3. / sit--lie-Ø-3. ‘He was sitting there / *was lying sitting there.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.3.3.3 Duration The light verb iru ‘be’ combines with the light verb kidu : ‘hold, have’ in its pasttense form (as of VbP) to give kit:t:uru. This combined light verb gives the grammatical meaning of duration of act. (18)

a. avan kudi-cci-ki. t:-t:u-ru.kk-ān : he drink-----Ø-3. ‘He is drinking.’ b. on cat:t:e kizi-nji-ki.t:-t:u-ru.kku your shirt tear-----Ø-3. ‘Your shirt is tearing.’ c. avan on vī.t:t:u-le he your house-  ‘He is sitting in your house.’

ukkā-ndu-ki.t:-t:u-ru.kk-ān sit-----Ø-3.

d. avan en-git:t:e panam kē.t:-t:u-ki.t:-t:u-ru.kk-ān : he I- money ask-----Ø-3. ‘He has been asking me for money.’ iru in the combined light verb parallels a full verb after VbP+kit:t:u, as in (19a). (18c) is ambiguous between ‘he is doing the act of sitting’ and ‘he is seated = he sat and is in that position or state’. The second meaning is similar to (11c) except that (18c) has the added sense of self-comfort, as in (20f). It could be argued that iru in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

317

the second sense of (18c) is a lexical verb meaning ‘be, stay’. It could have a verb modifier, as in (19b), which is synonymous with (19c), where the modifier sits before the combined verbal form. (19)

a. avan kudi-cci-ki. t:-t:u : he drink--- ‘He came drinking.’

va-nd-ān come--3.

b. avan on vīt:.t:u-le ukkā-ndu-ki.t:-t:u summa iru.kk-ān he your house- sit--- inactively be-Ø-3. ‘He is sitting inactive in your house.’ c. avan on vīt:.t:u-le summa he your house- inactively ‘He is sitting inactive in your house.’

ukkā-ndu-ki.t:-t:u-ru.kk-ān sit-----Ø-3.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

This discussion shows that the light verb iru is not semantically opaque like other light verbs, and that any semantic distinction between iru as a full verb and as a light verb is blurred, resulting in ambiguity.

12.3.3.4 Self-Affecting The third light verb with a grammatical function is kidu. : But there is no full verb kidu; the full verb in formal Tamil is ko l ‘hold’, whose reflex ko is used in colloquial : : Tamil in the imperative form.¹¹ The sense of this light verb is that the subject with agency is affected in some way by the act. This light verb is volitional. The inanimate subjects in (20i, j) perform by an unspecified force, natural or mechanical, which is perceived to act volitionally. Note that (20k) is odd, if not ungrammatical, where the act nil ‘stop’ is not perceived that way unless you imagine the rain as playing hide-and-go-seek with you. It may be seen that the sense of kidu : includes the grammatical sense of reflexive, where the subject and object are coreferential, but goes beyond it in its affectation to include performing the act for comfort, safety, etc. It can thus occur with intransitive verbs (unlike in Indo-Aryan languages) (Annamalai 1985: 114). (20)

a. avan tan.n-e adi-cci.k-ki. t:-t:-ān : he self- hit----3. ‘He hit himself.’ b. avan tan.a-kku-:l:le pēs-i.k-ki.t:-t:-ān he self-- talk----3. ‘He talked to himself.’

¹¹ The past form of this light verb –n: du : in Brahmin dialect is derivable from the formal Tamil past form of this verb kon: du. :

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

318

.  c. avan tan tambi.y-ōde pēs-i.k-ki.t:-t:-ān : he self brother- talk----3. ‘He chose to talk to his brother.’ d. avan tan tambi.y-e pidi-cci.k-ki. t:-t:-ān : he self brother- hold----3. ‘He held on to his brother.’ e. avan kadav-e mūdi.k-ki.t:-t:-ān he door- shut----3. ‘He shut himself behind the door.’ f. avan tan.akku-:l:le siri-cci.k-ki.t:-t:-ān he self-- laugh----3. ‘He laughed to himself.’ g. avan mara.ttu mēle ēr-i.k-ki.t:-t:-ān he tree above climb----3. ‘He climbed on to a tree (for safety).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

h. avan ammā madi-le padu-ttu.k-ki. t:-t:-ān : : he mother lap- lie----3. ‘He lay down (cozily) on his mother’s lap.’ i. kadavu mūd-i.k-ki du-ccu : : door close---+3. ‘The door closed by itself.’ j. maze pidi-cci-ki du-ccu : : rain hol---+3. ‘The rain intensified.’ k. *maze nin-nu.k-kidu-ccu : rain stop---+3. ‘The rain took a break.’ VbP + kidu : is like an inflected verb and looks like a simple predicate. It can, nevertheless, be categorized as a complex predicate. In Tamil, the bare verb base is inflected, but not the VbP base, and so VbP + kidu : cannot be an instance of inflection normally. As VbP gets light verbs attached to it, this structure is to be treated as V-V. But kidu : does not contribute to the argument structure of the predicate and does not admit any complement of its own. Its contribution to the predicate is its constitution, which is two verbs fused into a unitary predicate. In other words, its potential to make a biclausal structure is not realized and the predicate remains monoclausal. A predicate can be monoclausal with one of its constituent verbs not admitting arguments on its own, which is the case with complex predicates.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

319

The three V-V sequences with different light verbs, which are vidu : /-du, : ru/ iru, and ko(:l)/kidu and have grammatical functions (not referential meanings), : resemble Type 2 of Japanese compound verbs. They, however, are not restricted to aspectual function but include modal function.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.4 Polar opposite of light verbs The polar opposite of kidu : ‘hold’ is kudu : ‘give’, which is high in verbhood and semantic transparency, unlike kidu. This is borne out by the fact that taa ‘give’, the : synonym of kudu, can replace ku du in (21). It shares arguments with the full verb : : and merges its arguments with the arguments of the full verb. It may (21a, b) or may not (21c, d) share the arguments of the full verb depending on its meaning. The sense of this light verb is that the act is performed to be of help to the recipient in the sentence, and is expressed in the dative case. (This is not found in IndoAryan languages.) The object kār ‘car’ in (21a) is the object of both the full verb and the light verb. The object pay ‘bag’ in (21b) is the object of the light verb, but not the full verb; the locative pay ‘bag’ is the locative of the light verb, but not the full verb. Either one can be used in this sentence. In both these sentences, the indirect object ‘him’ is the recipient of the object as well as the help implied by the light verb. These two sentences will have the same indirect object as the recipient of the object without the light verb kudu : ‘give’, and the indirect object will be glossed ‘for him’. The contribution of kudu : in (21a, b) is the sense that the indirect object is also the recipient of help from the subject. In (21c, d), on the other hand, the indirect object is not the recipient of the object; the recipient is unspecified. The indirect object is the recipient of the help, which is the abstract unspecified direct object of the ditransitive light verb kudu. : (21)

a. nān avan-ukku kār vaang-i.k-kudu-tt-ēn : I he- car buy----1 ‘I bought him a car / I helped him buy a car.’ b. nān avan-ukku {pay-le / pay.y-e} paza.tt-e pō.t:-t:u.k-kudu-tt-ēn :
I he- bag- /bag- fruit- drop----1 ‘I put the fruit in the bag by way of help and handed it over.’ c. nān avan-ukku kār-e vi.t-tu.k-kudu-tt-ēn : I he- car- sell----1 ‘I helped him sell his car.’ (i.e. ‘I sold the car for him.’) d. nān avan-ukku kadidam ezut-i.k-ku du-tt-ēn : :
I he- letter write----1 ‘I wrote the letter for him (by way of help).’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

320

. 

Note that in (22) the V₁ and V₂ have a different relationship (see V₁ as explicator in Section 12.6). It is a relationship where kudu : is the main verb and the verb of VbP describes the manner or condition in which the object is given. The manner is ‘gave the meat chewed’ in (22a) and ‘*gave the meat eaten’ in (22b). Since the object is made nonexistent by eating in (28a) and there is no object for the verb kudu : ‘give’, this sentence would be ungrammatical. The same ungrammaticality holds good for the same reason when kudu : is a light verb. (22a) in light verb reading will give the meaning: ‘Mother helped the child by chewing the meat’; (22b) should mean: *‘Mother helped the child by eating the meat’, but it does not. (22)

a. ammā pi:l:le-kku kari.y-e me.n-nu mother child- meat- munch- ‘Mother gave the child meat chewing it.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. *ammā pi:l:le-kku kari.y-e tin-nu mother child- meat- eat- ‘*Mother gave the child meat eating it.’

kudu-tt-ā : give--3.

kudu-tt-ā : give--3.

While kudu : in (22) is a semantically transparent full verb requiring a set of arguments of its own, it could be semantically opaque, this enabling it to create a new word. It is a light verb in the VbP + kudu : structure that lexicalizes the monoclause when bleached semantically. tadavu ‘rub’, tat:t:u ‘tap’, and sollu ‘say’, : with kudu ‘give help’, in (23) give respectively the lexicalized meanings ‘massage’, : ‘pat’, and ‘teach’. Note that the bleached light verb kudu : does not admit an indirect object as the recipient of help in (23a, b) but has a direct object. This light verb is part of a lexical word without any predicative property of its own. The indirect object in (23c) is the one assigned by the verb sollu ‘say’. (23)

a. nān avan mudug-e tadav-i.k-ku du-tt-ēn : : I he- back- rub----1. ‘I massaged his back (in appreciation).’ b. nān nall-ā padi-cc-adu-kku avan-e : I well study--- he- ‘I patted him for having studied well.’ c. nān avan-ukku tamiz
I he- Tamil ‘I taught him Tamil.’

tat:t:-i.k-kudu-tt-ēn : tap----1.

coll-i.k-kudu-tt-ēn : say----1.

VbP + kudu : is is monoclausal and has a combined argument structure like a complex predicate. Further, it could be lexicalized, as (23) shows. In terms of their syntactic properties, the lexical verbs that are counterparts of the light verbs kidu : ‘have, hold’ and kudu : ‘give’ have a causative relationship of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

321

‘have’ and ‘cause to have’, which is similar to the lexical verbs that are counterparts of the semilight verbs pār ‘try see’ and kāt:t:u ‘try to cause to see’. Pairs of light verbs have the same lexical relation as their corresponding pairs of full verbs.

12.5 Homonymy among light verbs There is a light verb vidu : / (v)udu : ‘release, let go’ (which is homonymous with the intransitive light verb vidu : ‘release, leave’, truncated to -du, : discussed at (9)). It has a sense similar to another light verb, kudu (21), but it highlights : the fact that the help is to carry out, for another person, a task that is needed and helpful (25a), to carry out a task to relieve another person of it (25b), to arrange to carry information to another person (25c), or to make another person realize his or her folly (25d). These specific meanings may be generalized to say that this light verb gives the meaning of ‘getting across help or a message’. (24)

a. ammā pi:l:le-kku cat:t:e pō.t:-t:u-vu.t:-t:-ā12 mother child- shirt put----3. ‘Mother put on the shirt for the child.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. ammā en cat:t:e.y-e āni-leruntu edu-ttu-vu. t:-t:-ā13 : : mother my shirt- nail- take----3. ‘Mother took my shirt from the nail for me.’ c. ammā pi:l:le-kku pazam kudu-ttu-vu. t:-t:-ā : mother child- fruit give----3. ‘Mother sent fruit for the child.’ d. ammā pi:l:le.y-e nal.l-ā kē.t:-t:u-vu.t:-t:-ā mother child- good- ask----3. ‘Mother asked the child (to shame him).’ This vidu : as a light verb (unlike the assertive vidu) : could have its own argument. In (25a), ‘faucet’ is the object of tera ‘open’, but in (25b), water is the object of vidu : ‘let’. VbP + vidu structure does not have lexical integrity, as (25c) shows. :

¹² If the occurrence-asserting light verb (vi)du : is used in place of the light verb vidu : in this sentence, it would mean that the mother has done the job of putting the shirt on the child, and not what this sentence means: the mother helped the child by putting on the shirt on her, which she could not do or was not doing herself. ¹³ If the light verb kudu : is used in place of vidu : in this sentence, it would mean that the mother helped the child by taking the shirt from the nail (which she could not reach), and not what this sentence means: the mother released the shirt from the nail (on which it had become entangled) for the child, which the child could not do or was not doing herself.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

322

. 

(25)

a. ammā kozāy-e tera-ndu-vu.t:-t:-ā mother faucet- open----3. ‘Mother opened the faucet (for the water to run).’ b. ammā tan: ni.y-e tera-ndu-vu.t:-t:-ā : mother water- open----3. ‘Mother let the water run opening (the faucet).’ c. ammā kozāy-e tera-ndu tan: ni.y-e vu.t:-t:-ā : mother faucet- open- water- --3. ‘Mother opened the faucet and let the water run.’

The discussion shows that there are light verbs that correspond with one lexical verb and develop to have different grammatical meanings (vidu), and that there : are lexical verbs that relate causatively (ko:l / kodu), : and their corresponding light verbs (kidu : / kudu) : develop different unrelated grammatical meanings.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.6 V₁ as an explicator In the V-V sequences we have discussed, V₁, which takes the form of a VbP, is the full verb semantically and syntactically and V₂ is the light verb. In the following sequences, V₁, which is likewise in the form of a VbP, is a full verb semantically but not syntactically, and V₂ is a full verb (in some cases frozen) semantically and syntactically. V₁ is semantically transparent, but it does not have a temporal precedence relation with V₂, but does have modifier relationship. In other words, V₁ enriches the meaning of V₂ in the following examples just as V₂ (the light verb) enriches the meaning of V₁ in the earlier examples. Thus, in Tamil there is no single direction of verb enrichment or complexity addition in a V-V sequence. The enrichment includes the manner in which the act is performed (26a, b) and atomizing the components of the act (26c, d). This sequence is monoclausal. Its referential semantic complexity matches its lexical complexity. The sequence may take complements (26a) or each verb in the sequence may take its own complement (V₁ in (26b), V₂ in (26c)). The VbP can be duplicated in (26a, b) to express intensity of manner. (26)

a. avan romba nēram en.n-e uttu.p he much time I- be.close- ‘He stared at me for a long time.’

pā-tt-ān see--3.

b. avan nenju kulung-a vim-i azu-d-ān he chest shake- sob- cry--3. ‘He cried, sobbing so his chest shook.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

323

c. avan en.n-e kīze pidi-ccu ta:l:lu-n-ān : he I- down hold- push--3. ‘He pushed me down, holding me.’ d. avan enne mēle pō.g-āma tadu-ttu niruttu-n-ān : he I- above go- block- stop--3. ‘He stopped me from going forward by blocking me.’ Other examples of V₁ describing the manner or a component of the act include the following. These are called converbs by some in the literature. Contiguity of the verbs is not a requirement in these sentences. This fact and the fact that V₁ in some sentences takes clausal complements (e.g. (27c)) do not rule out the possibility that the sequences could be biclausal, in which case they will not be a V-V sequence of the kind discussed in this chapter. But they are brought up here because this sequence is a path for lexicalization of verbs. (27)

a. avan tirumb-i he turn- ‘He came back.’

va-nd- ān come--3.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. avan nimi-ndu he straighten-  ‘He sat up.’

ukkā-nd-ān sit--3.

c. avan tale.y-e kuni-nju pā-tt-ān he head- bend- see--3. ‘He looked down, bending his head.’ d. avan en-git:t:erundu ta:l:l-i he I- move- ‘He sat away from me.’

ukkā-nd-ān sit--3.

e. avan kīze mayang-i vizu-nd-ān he down faint- fall--3. ‘He fell down, fainting / He fainted.’ f. avan dōse.y-e rasi-ccu sāppi.t:-t:-ān he dose- enjoy- eat--3. ‘He ate the dose, enjoying it / He enjoyed the dose.’ g. avan toda-ndu peesu-ān : he continue- speak--3. ‘He spoke continuously / He continued to speak.’ Some examples of V-V sequences above that are lexicalized are these. uttu in (26a) does not occur as a verb with any other inflection such as *uttān ‘(he) was close’. But most V₁ do: vimmunān ‘(he) sobbed’, etc. This V-V sequence has

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

324

. 

commonality with lexical compound verbs in Japanese (Matsumoto 1996: 197–236), which are defined as two verbs which do not have their own arguments independently of each other when they are sequenced. (27c) does not fit with this definition, which has tale ‘head’ as the argument of V₁ but not of V2, which could have another argument tare ‘floor’. Whether this is a definitional problem needs further scrutiny: while kuninju pār ‘look down’ in tareye kuninju pāttān ‘(he) looked down at the floor’ could be treated as lexicalized, it could not be in tareye taleye kuninju pāttān ‘(he) looked at the floor head down’. Some of the Type 1 sequences in Japanese resemble this sequence in Tamil with regard to the fact that V₂ is the semantic head. When the verb sequence of the light verb kit:t:u (see 19a) (or its equivalent kon: du) : after certain full verbs is followed by deictic verbs vā ‘come’, pō ‘go’, the sequence gives rise to lexical items, as in (28). This is similar to the complex motion predicates in Japanese (Matsumoto 1996: 237–62) with some differences that need to be explored. It shares with Type 2 the feature that V₁ is the semantic head.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(28)

a. avan panam {ko.n/ pō-n-ān} : : du-va-nd-ān : he money hold--come--3. / go--3. ‘He came / went holding the money.’ = ‘He brought in / took away the money.’ b. avan pustaga.tt-e edu-ttu.k-ki. t:-t:-u {van-d-ān / pō-n-ān} : he book- take--- come--3. / go--3. ‘He came/went taking the money.’ = ‘He carried in/out the money.’ c. avan tambi.y-e kū.t:-t:i.k-ki.t:-t:u {va-nd=ān / he brother- accompany--- come--3. / pō-n-ān} go--3. ‘He came / went accompanying his brother.’ = He brought in / took away his brother.’ d. avan tambi.y-e aze-ccu.k-ki.t:-t:u {va-nd-ān / pō-n-ān} he brother- invite--- come--3. / go--3. ‘He came in / went out inviting his brother with him.’ = ‘He took in / took out his brother.’ e. avan tambi.y-e tūkk-i.k-ki.t:-t:u {va-nd-ān / pō-n-ān} he brother- lift--- come--3. / go--3. ‘He came in / went out carrying his brother.’

There is no space here to discuss in detail the possibilities of sequencing two or more light verbs and the possibilities of light verbs occurring in nonfinite verb

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

325

forms. The general fact of the sequence of light verbs is that there is a hierarchical order among the light verbs and that the functionally grammaticalized light verbs follow the light verbs that are not so, as semilight verbs. With regard to the V-V sequence in nonfinite verb forms, the functionally grammaticalized light verbs acquire a different, but related sense. This is the case when a light verb is in VbP form after a full verb and is followed by a full verb. The light verbs that are not functionally grammaticalized do not have different semantics whether they occur in finite or nonfinite forms of verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

12.7 Conclusion V-V sequences in Tamil, as in Japanese or in any language, can be classified into types based on their grammatical functions and morphological/syntactic structure. Some specific cases of these sequences create composite lexical items, as evidenced by their unique referential function, which is distinct from the referential function of each V. Crosslinguistically, the grammatical functions and the structural features of sequence categories differ from language to language. The interesting point about the taxonomy of V-V sequences across languages is that specific members of each category or type of the sequence may be distributed differently across the matrix. In other words, an instance of a V-V sequence of Type 1 in language X may belong to Type 3 in language Y. Let me illustrate this by comparing Tamil with Japanese. The benefactive (II) and probative (III) senses of Type 4 V-V sequence in Japanese (see Kageyama, Chapter 2, this volume) have parallels in Tamil to the point of having the cognate light verbs kudu : ‘give ()’ and pār ‘see ()’ respectively ((5a) and (21)). They are, however, different from the Japanese case, structurally and categorically, as pointed out in the body of the chapter. This suggests that across languages V-V sequences may belong to different types, in spite of shared grammatical and semantic functions. While the grammatical function of ‘completion ()’ belongs to Type 2 in Japanese (ami-ageru [knit-accomplish] ‘finish knitting’) and Tamil (8b–e), the grammatical function of ‘continuity’, which also belongs to the same Type in Japanese (hurisikiru [fall-repeat] ‘rain continually’), belongs to Type 1 in Tamil (27g). These facts point out that in a typological comparison, types may be comparable across languages, but the individual instances in those types may not be common. This is true between Indian languages. Structurally, V-V sequences in Tamil have a range from being flexible in the contiguity of their component verbs to being uninterruptible morphological forms. This variable flexibility does not seem to be the case in Japanese. This property of a cline of variability can be said to vary across languages. This chapter, however, shows that it is found within the same language, namely Tamil. The light verbs have a range in a scale of verbal status and semantic transparency. This

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

326

. 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

shows the dynamic nature of the light verbs in Tamil. The dividing line between full verbs and light verbs may be fuzzy in Tamil as a class and in their specific sequencing questioning the validity of setting up a separate grammatical category of light verbs in Tamil (Butt 2010: 59). With regard to the theoretical concepts Serial Verb Construction and Complex Predicate, Tamil data brings out variability in their feature matrix. As for the Complex Predicate, its definition must be relaxed from the requirement that both verbs must have the same arguments, or may have different arguments that are merged into one argument structure of a unitary predicate. Relaxation of the definition is necessary because when such a unitary predicate is a lexical composite, one of its two verbs may not assign any argument at all. This expands the definition of a serial verb construction to include the V-V sequence where V₁ specifies the manner of performing an act or the semantic components of an act. It may be noted that the grammatical function of the two verbs in this sequence is reversed. Semantically transparent V₁ in the V-V sequence may play a secondary role syntactically similar to the semantically bleached light verbs that are V2, as does the Japanese data classified as Type 1. Tamil data of V-V sequences show that a taxonomic typology between languages may not have exact matches and identical theoretical constructs. Tamil data, in addition, show that the light verbs may be graded in a scale of lexicality and they exhibit some of lexical relations such as synonymy, antonymy, polysemy, and transitivity, just like full verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

13 Verb + verb sequences in Dravidian Sanford B. Steever

13.1 Introduction The Dravidian language family has many constructions that may be characterized as V + V sequences. They serve various purposes within the grammar, including the formation of complex sentences and compound verbs. This chapter focuses on their use as compound verbs. Dravidian grammars may have three basic kinds of compound verb: auxiliary compounds, lexical compounds, and reduplicating compounds. The first provides a means of conveying those verbal categories not encoded in the simple morphology of a language; the second, a means of expressing those lexical meanings not encoded in a single lexeme of the language; and the third, a way to express notions such as distribution and quantification.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

13.1.1 The Dravidian languages The Dravidian language family consists of twenty-four languages spoken in South India and adjacent regions. It has four subgroups: South Dravidian: Badaga, Irula, Kannada, Kodagu, Kota, Kurumba, Malayalam, Tamil, Toda, and Tulu; South-Central Dravidian: Gondi, Konda, Kui, Kuvi, Manda, Pengo, and Telugu; Central Dravidian: Konekor Gadaba/Ollari, Kolami, Naiki, and Parji; and North Dravidian: Brahui, Kurux, and Malto. At least three languages, Kannada, Tamil, and Telugu, distinguish modern from premodern varieties; the premodern forms differ sufficiently from the modern to be considered separate languages. Reports of other languages have appeared, but without adequate descriptions we cannot say whether these are new languages or dialects of ones already known. All of them descend from a common ancestor, Proto-Dravidian, for which a partial reconstruction is available. See Steever (1998) and Krishnamurti (2003) for details. The Dravidian languages are geographically concentrated in the Indian states of Andhra Pradesh (Telugu), Karnataka (Kannada), Kerala (Malayalam), Tamil Nadu (Tamil), and Telangana (Telugu), where they are majority languages. They are also spoken in Chhattisgarh, Jharkand, Maharashtra, Orissa, and West Bengal, where they are minority languages. Dravidian languages are also spoken in

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

328

 . 

Nepal (Kurux, under the name Dhangar), Pakistan (Brahui), and Sri Lanka (Tamil). A number are endangered, e.g. Kota, Pengo, and Toda. The Dravidian languages share the South Asian subcontinent with three other language families: the Indo-Aryan branch of Indo-European, the Munda branch of Austroasiatic, and Tibeto-Burman, as well as some isolates, e.g. Burushaski. Interaction over the millennia has given rise to the South Asian linguistic area, where grammatical features in one family are thought to have influenced the development of similar features in the others. While undeniably true at a general level, vigilance is required in specific instances to distinguish cases of grammatical borrowing between families from cases of independent innovation within a single family.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

13.1.2 Characteristics of Dravidian grammar The Dravidian languages have relatively transparent agglutinating morphologies, mixed with some fusional traits. Dravidian morphology is primarily suffixing; prefixes, where they occur, are borrowed. Dravidian word-formation rules typically generate words consisting of three parts: a lexical base, optional derivational suffix, and inflectional suffix. Languages such as Kui, Pengo, and Tulu have historically evolved more complicated verb morphologies, particularly between the lexical base and following suffixes. As one proceeds from the lexical base outwards through the layers of suffixes, the morphology moves from fusional to agglutinating and, in the case of postpositions for nouns or auxiliaries for verbs, analytic. All Dravidian languages distinguish between free forms and clitics. Dravidian clitics, postclitics marked here by =, typically involve discourse cohesion, emphasis, and quantification. Free forms include the two basic parts of speech, noun and verb. Both are characterized by a general word-formation rule: a word consists of a lexical base (+derivational suffix) + inflectional suffix. A noun typically marks case, number, and, in pronouns, (natural) gender. A verb form typically encodes mood, tense, and polarity, but other verbal categories appear, especially in compound verb constructions. For many Dravidian languages, the set of verb bases is closed, prompting the use of compounding among other devices to generate new expressions. Some grammars add a category of adjectives on functional grounds; formally, however, these appear to be defective nouns. Beyond this, consensus on the number and nature of parts of speech rapidly declines. Some languages, such as Kannada and Malto, have innovated sets of formally distinct conjunctions through the grammaticalization of inherited forms or borrowings from Indo-Aryan languages. The restriction on the number of parts of speech to the bare minimum of two, noun and verb, means that Dravidian grammars must rely on these two word classes, along with their inflected and compound forms, to carry out grammatical functions for which other languages would use adverbs, conjunctions, etc.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-   

329

For example, the lack of adverbs, including anything corresponding to English not, necessitates the expression of negation through the verbal category of polarity in basic conjugations and compound verbs. Dravidian syntax is characteristically SOV, or head-final. The expected constituent-order harmonics hold: subordinate clauses precede matrix clauses, relative clauses and genitives precede the nouns they modify, postpositions are used instead of prepositions, and main verbs precede auxiliaries. This gives rise to a syntax with robust left-branching properties. Some languages along the Dravidian-Indo-Aryan frontier have borrowed function words, such as Hindi ki ‘that’ and Oriya je ‘that’ which impose a rightward orientation on their complements, in contrast to inherited Dravidian function words, which impose a leftward orientation. Thus, in languages such as Brahui, Malto, and Parji, a modified order has evolved that superimposes some SVO features on the fundamental SOV template. The details of these modified systems continue to be worked out, particularly in interclausal relations, but within individual clauses the constituent order remains SOV. For the purposes of this chapter, this means that main verbs always precede their auxiliaries no matter whether the language exhibits a robust SOV or a modified SOV constituent order. SOV typology creates many opportunities for two verbs to stand together in a sequence. In one example, sentential object complement structures often take the shape S0[NP0 S1[NP₁ V₁]V₀], in which V₁ and V₀, though in different clauses, are adjacent. In another, coordinate structures with the shape S0[S1[NP₁ V₁] S2[V₂]] may also directly juxtapose two verbs. This is not restricted to instances where V₂ is intransitive: the obviation of arguments permits transitive and ditransitive verbs to occupy this position as well.

13.2 Analyses of V + V sequences In the Dravidian languages, a sequence of two verbs, . . . V₁ . . . V₂ . . . , may be analyzed in several ways. In the first and limiting case, the two verbs belong to different structures, so the sequence is not a construction at all. In the second, more interesting case, the sequence does constitute a construction. These constructions include two major subsets. In the first, the sequence is a compound clause structure, i.e. a structure with multiple clauses, such as a matrixcomplement or a coordinate construction. In the second, the sequence forms a compound verb which, despite the multiplicity of verb forms, refers to a single event/situation, has a single set of arguments and modifiers, and appears within a single clause. This second subset is our topic here. The Dravidian languages require inflectional suffixes to serve as a morphological intermediary between the members of a V₁ + V₂ sequence, whether that sequence is a compound clause or a compound verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

330

 . 

(1)

Tamil a. S0[S2[paiyaṉ kat:ittu.c] cettuppōṉāṉ]]. S1[pāmpu boy- snake- bit- die--3. ‘The snake bit the boy and he died.’ b. S0[S1[pāmpu kat:ittu.p] S2[paiyaṉ cettuppōṉāṉ]]. snake- bite- boy- die-pst-3. ‘The snake bit the boy and he died.’

(2)

Tamil colli] kēt:ka.v-illai.]] S0[avaṉ S1[nāṉ he-- I- say- listen-. ‘He didn’t listen to what I was saying.’

Example (1a) contains the sequence, kat:ittu.c cettuppōṉāṉ ‘(snake) biting . . . (he) died’, a coordinate structure. Each verb belongs to a different clause: cettuppōṉāṉ ‘died’ to the main clause and kat:ittu ‘biting’ to the dependent clause. Each has its own subject, as clarified by the paraphrase in (1b). Similarly, (2) contains the sequence colli kēt:kavillai ‘saying . . . didn’t listen’ in a matrix-complement structure. Here again, each verb belongs to a different clause; each has its own subject. Such sequences come under the study of complementation (see Steever 1998). Note, however, that the conjunctive suffixes in (1) and (2), which mediate between the two clauses, are the very same suffixes that also appear in compound verbs. This means that we cannot analyze a sequence as a compound verb or compound clause solely by reference to the suffixes it incorporates; further, syntactic criteria are required. It also calls into question whether we can treat these suffixes merely as phonological adjustments of one verb to another in a sequence. We see in the course of the chapter that the choice of suffix in any construction is governed by the semantic distinction between modal and indicative (Steever 2005: 82–7). The criteria used to establish the four types of V + V sequence in Kageyama (Chapter 2, this volume) appear to reflect certain features of Japanese morphology, such as the distributions of the -te perfective form (Type 4) and the infinitive (Types 1–3). As comparison shows, these criteria do not neatly coincide with the morphological and syntactic criteria used for Dravidian compound verbs. For example, the Japanese -te perfective appears only in Type 4 compound verbs, while its closest Dravidian counterpart, the conjunctive form, occurs in Types 1, 2, and 4 compound verbs. Conversely, the Japanese infinitive is excluded from Type 4 compound verbs in that language, while its closest Dravidian counterpart regularly occurs in Type 4 compounds. (What are characterized as Type 3 V + V compounds in Japanese appear in Dravidian to correspond to certain matrixcomplement structures; each verb in the Dravidian structure may host an adverb independently of the other.) The cluster of morphosyntactic criteria that define compound verbs in Dravidian thus differs somewhat from those used in Japanese.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-   

331

Note also that the two elements of a Type 1, 2, or 4 Dravidian compound verb may be separated by the insertion of a clitic particle (Steever 2005: 132–4), which is possible in Type 4 but not in Types 1–3 Japanese compounds. Further, unlike their Types 1 and 2 Japanese counterparts, Dravidian lexical compound verbs permit the postposing of V₁ to the right of V₂ in a sequence (Steever 2005: 90). This flexibility reflects, I believe, crosslinguistic differences in morphology and syntax which determine for each language how tightly the two constituents of a compound verb may be bound together, some permitting discontinuous compounds and tmesis, others not (Steever 2005: 21). Even so, it does not compromise the status of the compound verb as a linguistic unit that functions as a single verb in the larger grammatical context and with respect to grammatical rules. In Tamil, for example, the rules of Scrambling and Clefting treat a compound as a single verb in a single clause, not as two separate verbs (Steever 2005: 111–20). Despite such differences, there is an apparent kinship between compoundverb sequences in Tamil and their counterparts elsewhere. To achieve a degree of comparability, we rely on functional and syntactic criteria, since morphological criteria are more likely to vary across languages. A compound verb is a linguistic expression of two or more verb forms that is treated as a single verb by the rules of the grammar. Regardless of the multiple forms, the compound verb refers to a single event/situation, has one set of arguments and modifiers, and occurs within a single clause. An auxiliary compound verb (ACV) is a periphrastic verb form that conveys those verbal categories or combination of categories which are not expressed in the simple verb forms of the language. A lexical compound verb (LCV), by contrast, is a periphrastic verb form that conveys those lexical meanings or combinations thereof that are not expressed in any single lexeme of the language. The ACV expands the grammatical resources of the language; the LCV, its lexical resources (see Steever 1993, 2005). A reduplicating compound verb (RCV), while sharing certain surface characteristics of other compounds, requires a separate analysis, noted below. Any meaningful comparison among languages requires an empirical basis; to that end, we present a catalog and brief analysis of compound verbs in Dravidian. Section 13.3 takes up ACVs. They distinguish primarily between sequences in which V₁ is inflected for the infinitive and those in which it is inflected for the conjunctive; other minority patterns are also discussed. Section 13.4 introduces LCVs, which include a broad variety of subsets. There we discuss the properties that not only differentiate LCVs from ACVs, but distinguish among subsets of LCVs, as well. Section 13.5 introduces two kinds of RCVs: while exhibiting some traits of other compound verbs, they result from processes that apply equally to nouns and verbs. The primary focus is synchronic, but diachronic observations are also made.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

332

 . 

13.3 Auxiliary compound verbs 13.3.1 Morphology and syntax An auxiliary compound verb¹ extends the grammatical resources of a language by providing a complex morphosyntactic vehicle to convey those verbal categories, or combination of categories for which the language does not supply a verbal inflection. As a rule of thumb, the verb in the left slot (V₁), the so-called main verb, is modified by the verb in the right slot (V₂), the auxiliary, which conveys a semantically more specific verbal category than is conveyed by any basic verb inflection. Careful segmentation of the ACV reveals that the compound’s denotational and inflectional functions are distributed over the main and auxiliary verbs. The base of the main verb lexically identifies the compound, indicating what sort of situation or activity it denotes, while the auxiliary bears the verbal inflections the grammatical context imposes on the compound. Factoring out these two components leaves us with the grammatical inflection of the main verb and the lexical base of the auxiliary: the combination of these two elements conveys the verbal category conventionally associated with a particular AVC, e.g. perfect tense, benefactive voice (Benveniste 1965). Consider these examples; for ease of identification, the compound verbs are enclosed by braces:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(3)

Tamil a. avaṉ ceṉṉai.k-ku.c [ce.l-la.p pō-kiṟ-āṉ]. he-- Chennai- go- go--3. ‘He is going to go to Chennai.’ b. ava:l [camai.t-tu iru-kkiṟ-ā:l]. she-- cook- be--3. ‘She has cooked.’

Example (3a) contains the ACV cella.p pōkiṟāṉ ‘he is going to go’. In keeping with its denotational function, the main verb, cel-l.a ‘go’, incorporates the verb base cel‘go’. The auxiliary pō-kiṟ-āṉ ‘he is going’ bears the present-tense marker -kiṟ- and the personal ending -āṉ, inflections imposed by the grammatical context. What remains are the infinitive marker of the main verb, -a, and the verb base pō- of the auxiliary: together they signal prospective tense, i.e. ‘is going to V’, a meaning not encoded in the basic verb inflections of Tamil. Combined with the auxiliary’s present-tense marking, the whole ACV signals present prospective tense. In the ACV camaittu irukkiṟā:l ‘she has cooked’ in (3b), the main verb incorporates the ¹ Compound verbs with three or more components are treated as a series of nested binary constructions.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

333

lexical root camai- ‘cook’, while the auxiliary bears the present-tense suffix -(k)kiṟand the personal ending -ā:l marking subject-verb agreement (SVA). The two remaining elements of the compound, the conjunctive suffix -t.tu of the main verb and the verb base iru- ‘be’ of the auxiliary, jointly convey perfect tense, another category not found among the basic verbal inflections of Tamil.² The entire AVC is thus a present–perfect-tense form.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

13.3.2 Classifying ACVs The main verb of the ACV in (3a) incorporates the infinitive; its counterpart in (3b), the conjunctive. The morphological distinction between these two suffixes directly correlates with a basic opposition in the verbal category of mood (see Steever 2005: 85–7). Mood expresses the ontological evaluation of the process denoted by a given verb. It opposes modal and indicative forms: modal forms express modality, contingency, possibility, etc., whereas indicative forms express actuality, reality, and noncontingency. The infinitive is the least-marked modal verb form in Tamil; the conjunctive, the least-marked indicative form. This opposition between the infinitive and the conjunctive is projected to the ACVs that incorporate them: hence, (3a) is a modal ACV, (3b) an indicative ACV. ACVs are further classified according to the specific verbal categories they convey, e.g. aspect, attitude, and tense. As the unmarked member of the opposition, the infinitive has a broader distribution, occurring in contexts where modality is neutralized, e.g. the Tamil periphrastic passive (Steever 2005: 77, 87). The distinction between modal and indicative ACVs rests not just on the morphological opposition between the infinitive and conjunctive forms of the verb, but on other grammatical phenomena. For example, modal auxiliaries may be reduplicated to signal emphasis, a possibility not open to ACVs (Steever 2005: 132). In response to a question with a modal ACV, the answer may omit the main verb, repeating just the auxiliary; this possibility is also closed to indicative ACVs (Steever 2005: 140) We now introduce an important subclass of indicative ACVs: the attitudinal auxiliary compound. (4)

Tamil contakkāraṉ va.n-tu tolai-nt-āṉ. relative- come- -+-3. ‘(My) relative arrived, damn it!’

² By contrast, Pengo does mark perfect tense in its basic verb morphology.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

334

 . 

The indicative ACV vantu tolaintāṉ ‘(he) came/arrived, damn it!’ in (4) has the same basic properties as the ACV in (3b): despite having two verb forms, the compound makes one predication and has just one set of arguments. The main verb vantu ‘coming, arriving’ incorporates the lexical base va- ‘come’, satisfying the compound’s denotational function. The auxiliary tolaintāṉ ‘(he) got lost’ bears inflections for the past tense, -nt-, and SVA, -āṉ, as required by context. The residue, the main verb’s conjunctive suffix -n.tu and the auxiliary’s lexical base tolai-, taken together, signal the speaker’s antipathy toward the event denoted by the main verb, roughly translated by the epithet damn it. This ACV encodes the verbal category of attitude: the speaker’s subjective evaluation of the narrated event (see Steever 2005: 239–89). Although the literature often overlooks this category and these compounds, grammatical tests support the conclusion that attitudinal ACVs are a subset of indicative ACVs (Steever 2005).

13.3.3 Less common variants of the ACV While ACVs based on an infinitive or conjunctive form of V₁ predominate in Dravidian, other kinds occur, with differences in fine structure.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(5) a. Parji (Burrow and Bhattacharya 1953: 77) kindik vey [vend cī] pēpa en-d-a : little rice cook- - aunt- say--3 ‘(The monkey) said, “Cook a little rice for me, auntie.” ’

ge. 

b. Parji (Burrow and Bhattacharya 1953: 129) ok manj-en [andki-t-er mẽ-d-ar]. : one man- kill--3 --3 ‘They have killed a man.’ c. Konda (Krishnamurti 1969: 308) veyu [kāk-t-a ma-n-ar]. mouth open-- --3 ‘They have opened their mouths.’ d. Pengo (Burrow and Bhattacharya 1970: 66–7) hur: -taŋn-aŋ ( ‘put on horseback and took along’ or ‘took along on horseback’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

367

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In all such cases, I had earlier assumed that we are dealing with compound expressions, because the informants rejected the interpretation in terms of a biclausal chained construction. I revised my position, not least because some informants rejected a compound reading, not only in these more problematic cases, but also in those cases where the type verb is aimed at the same goal as the path verb. When given the option that the first verb may modify the second like an adverb, almost all informants instantly agreed. This means that we are dealing with a semantically complex single event, but syntactically with an embedded adverbial clause and a superordinate main verb. (8) 

aʒaŋ Domkhar-na Le-a, r̥t˖eka ʒon-e, joŋs. uncle Domkhar- Leh- horse˖: ride- come. ‘The uncle came from Domkhar to Leh, riding on a horse.’ Perhaps also: ‘ . . . came riding . . . ’, but most probably not: ?‘rode hither from Domkhar to Leh.’

(9) 

rinboʧhe, ʧhips-ika ma-ʧhips-pa, skjot. chief.priest hon.horse-: 2-hon.ride- hon.come. ‘The chief priest came/ went without riding a horse.’ Perhaps also: ‘ . . . did not come/ go riding’, but most probably not: ?‘did not ride (hither/ away).’

(10) 

ɲopa-ŋun-is bagma, r̥t˖eka skjon-e, khers. bridesmen-- bride horse˖: let.ride- take.away. ‘The bridesmen put the bride on the horse and took her along.’ OR: ‘ . . . took the bride along on horseback.’

A similar problem is observed with the transitive expression rʤes ʧat ‘follow, search’ (lit. ‘cut the trace’) combined with intransitive path-motion verbs, as in example (11), and with the combination of the transitive expression lam bstan ‘show the way’ with transitive path-movement verbs as found in the LLV: lam bstan·te hkhyer·te ‘showed (him) the way and took (him) along’ or ‘led (him) along the way’. : (11) 

khoŋ-ø, ~ khoŋ-is jaγ-i rʤes ʧat-e, joŋs. they- they- yak- trace cut- come. ‘They followed the track of the yak and came (to us).’ OR: ‘ . . . came by following the track of the yak.’ Probably NOT: ?‘ . . . followed the track of the yak toward us.’

Again, the situation is not as straightforward as it might appear. Whenever such combinations are used themselves as modifying adverbs, they tend to develop more compositional properties, comparable to those of the intensifying combinations to be described in Section 14.3.7. It seems that, in such cases, the directional meaning of the path verbs is weakened.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

 

368

(12) a. Dorʤe-s Rigdzin-i kaŋrʤes ʧat-en,  Dorje- Rigdzin- foot.print cut- ri-a dzaks. mountain- climb. ‘Dorje followed Rigdzin’s footprints and climbed the mountain.’ b. Dorʤe-s Rigdzin-i kaŋrʤes ʧat-en-soŋ-se, Dorje- Rigdzin- foot.print cut--go- ri-a dzaks. mountain- climb. ‘Dorje rigidly followed Rigdzin’s footprints and so climbed the mountain.’ According to the informant, the complex construction indicates that the person had to rely solely on the footprints and would have been completely lost had there been none. The simple verb indicates that s/he might also have had other cues.

14.3.3 The use of give I: expression of a beneficiary or recipient

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In Ladakhi and Balti, the verb taŋ ‘give’ is frequently used in accordance with its basic meaning to indicate a beneficiary or maleficiary, that is, when the agent aims at the benefit or detriment of another person. In these cases, one can argue for a compound reading or for an embedded modification reading, as well as for a chained sequential reading: (13) 

goba-s julpa sami tshaŋma˖(ː) sakjat rere skal-e-teaŋs. goba- villager farmer all˖ land each allot--give. ‘The goba (village chief) allotted a piece of land each to all the farming villagers [i.e. the individual households] (compound reading). OR: ‘ . . . gave each of the . . . villagers a piece of land by allotting it [to them]’ (embedded adverbial reading). OR: ‘ . . . allotted a piece of land each to the . . . villagers and (then) gave it [to them]’ (sequential reading).

(14) kho˖s . . . žag·daŋ liŋs-la soŋ-ste |  s/he˖ every.day hunt- go.- skyin-daŋ broŋ maŋ·po khyoŋ-ste | ibex- wild.yak many bring- a·po a·bi gñis·ka-la ša maŋ·po khyoŋ-ste-btaŋs | grandfather grandmother both-all meat much bring--give. ‘He . . . went hunting every day and brought a lot of ibex and wild yak and brought the old couple a lot of meat’ (compound reading). OR: ‘ . . . and brought a lot of meat to the couple and gave [it to them]’ (sequential reading).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

369

Schmidt and Kohistani (2008: 245–6, exx 10 and 11) present similar combinations in Indus Kohistan Shina, which they treat as biclausal sequential expressions: (15) sé-si góos: san-eé d-áa-o  he-. house build- give--3.. ‘He built the house and gave it (to someone).’ (16) sé-si makái uc̣h-eé d-áa-o  he-. maize bring- give--3.. ‘He brought the maize and gave it (to someone).’

14.3.4 Give, take, and bring: mostly commercial transactions

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In Ladakhi, the verb give and its directional counterparts take away or bring hither are also used to specify the direction of a (commercial) transaction when the transaction verb itself is unspecific: lend/ borrow, exchange, return. The direction of the transaction can be inferred from the case marking of the second argument (the person from whom something is borrowed or to whom something is lent). Nevertheless, the speakers prefer the complex construction; cf. (17). The common construction is with a clause-chaining marker (Type B), but in the Domkhar dialect, the serial verb construction with the bare stem (Type A) is also frequently found. Independent of the construction type, the interpretation depends very much on the implicit context. All three logical possibilities have been attested: • a sequence of two events: first a contract or agreement on the transaction type, then, after a temporal gap, the actual transaction • modification of the second verb: e.g. taking by borrowing, not by stealing— the combination, however, usually appears without any apparent contrast • a compound reading focusing on the resulting state: e.g. a loan is yet to be paid back While the interpretation usually depends on the situation the informant has in his/her mind, I could also observe that it might depend on how one formulates the question. (17) a. ŋa-s kho-a pene maŋbo jars-pin. / &jar-e-teaŋs-pin.  I- she- money much lend.- &lend--give.- ‘I lent him/her a lot of money. / I (agreed to) lend him/her a lot of money and then gave [it to him/her]’ (sequential reading). OR: ‘ . . . / I gave him/her a lot of money by way of lending’ (embedded adverbial reading). OR: ‘ . . . / I have lent him/her a lot of money (and s/he has not yet returned it).’ (compound resultative reading).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

370

 

b. ŋa-s kho-ikana pene maŋbo jars-pin. / &jar-e-khjoŋs-pin. I- she-: money much borrow.- &borrow--take.- ‘I borrowed a lot of money from him/her. / I (asked [him/her] to) borrow a lot of money from him/her and then took [it from him/her]’ (sequential reading). OR: ‘ . . . / I took a lot of money from him/her by way of borrowing’ (embedded adverbial reading). OR: ‘ . . . / I have borrowed a lot of money from him/her (and have yet to return it)’ (compound resultative reading).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

14.3.5 The use of give II: expression of heightened intentionality or force In Ladakhi and Balti (occasionally also in other varieties; see Hein 2007 for Spiti), the verb give (or throw in Balti) highlights an actor’s intentionality, often with a negative connotation of bad temper, destructive intentions, or performance against norms, expectations, or benevolent advice. The combination may also signal the application of force, that is, more than usual or necessary. On a positive note, the construction is frequently used in commands, signaling that the addressee should just perform the task and not be shy. In this function, give has no directional counterpart. The common construction type for all Ladakhi dialects is with the bare stem II (Type A, serial verb construction). This indicates a closer unity between the two verbs and a narrowing down, if not bleaching of the semantic content of the second verb. Conversely, one may then say that the Type B converb constructions of the previous sections, which are used to express directionality or beneficial readings, should perhaps not be viewed as complex monoclausal predicates. (18) 

kha-s laptse lip skon-teaŋ˖sok. snow- sheaf onom dress-give.˖. ‘Unexpectedly, the snow covered the sheaves completely.’

(19) 

riri-o l ̥ʧus-teaŋs-pa, ʒik. radio- turn.-give.- break.down ‘When [I] turned the radio on and off again and again (more than necessary), [it] broke down.’

In Indus Kohistan Shina, the parallel compound construction with give does not differ formally from the biclausal beneficiary construction of examples (15) and (16); cf. example (20) (Schmidt and Kohistani 2008: 246, ex. 12a). The combination has a slightly different semantics and indicates the “completion of a task performed for another or for the common good” (ibid.; cf. also their example 13).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

371

It may thus also have a beneficiary connotation, and it may not always be easy to discriminate between the two constructions. The negative connotation of the Ladakhi combination is expressed by the verb ‘leave, release’, as in example (21) (Schmidt and Kohistani 2008: 248, ex. 18). (20) sé-si dóoli bah-ií d-áa-o  he-. field plow- give--3.. ‘He plowed the field (and the job is done).’ (21) 

sé-si makái dah-ií phát-aa-o he-. maize burn- leave--3.. ‘He completely ruined the maize (by neglecting to care for it).’

14.3.6 The use of give II: ambiguous cases As in Indus Kohistan Shina, several usages in Ladakhi allow both the reading of heightened intentionality or force and a more literal reading of the verb give, even with the Type A serial verb construction, at least in those dialects that allow serial verb constructions more generally, such as Domkhar.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(22) 

aba-s ŋa˖(ː) baik-ʧik ɲos. / ɲos-teaŋs. father- I˖ bike- buy. buy.-give. ‘[Our] father bought a bike for me. / bought a bike and gave [it] to me.’ OR: ‘ . . . / [Our] father just bought a bike for me (against our mother’s advice or against my own expectations).’

Compare also the following example, where the informant, despite the Type B converb construction, at least initially did not accept the meaning ‘and gave’, and compare this also with example (15), from Indus Kohistan Shina, which is analyzed by Schmidt and Kohistani as a biclausal construction: (23) 

kho-i miŋbo-s kho-a naŋ-ʧik ʧos-e-teaŋs. s/he- brother- she- house- construct.--give. ‘Her brother just constructed a house for her.’ OR: ?‘ . . . constructed a house and gave it to her.’

14.3.7 Intensifying constructions With unwanted events both the intransitive and the transitive path verbs can have an intensifying function, emphasizing the immediate danger, the suddenness, or the completeness and irreversibility of the event. Type A serial combinations of

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

372

 

negatively connotated verbs (mostly intransitive and inagentive verbs, particularly also adjectivals) with stem I ʧha ‘go’ are used in Ladakhi to indicate the chance or even immediate danger that an unwanted event or change of state might happen, as in example (24). When the same verb type is combined with stem II soŋ ‘went’, the connotation may be of completeness, irreversibility, additionally of regret or surprise, as in examples (25) and (26). Some of these connotations are, of course, also aspectual. The combination with soŋ seems to be common among many modern Tibetic languages; see also Hein (2007) for a similar construction in Spiti Tibetan. In many Tibetic varieties, including some of the Ladakhi Kenhat dialects, soŋ has further developed into an evidential past-tense marker, indicating personal observation plus movement away from the speaker. (In some of the Kenhat dialects, this evidential usage blocks the intensifying function of soŋ.)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24) Standzin-la spera maŋbo zer-na, rʤet-ʧha-ruk.  Standzin- speech much say- forget-go-. ‘If you tell Standzin (too) many things, there is danger that [he] forgets [half/ most/ all of it].’ NOT: *‘ . . . he is going (away), forgetting [it].’ (25) 

tshan-la Aŋmo-a l ̥ʧaŋm˖e zdoŋbo mi-a thoŋ-se, night- Aŋmo- tree˖ trunk man- see- ɖoks lo. / ɖoks-soŋ lo. be.scared.  be.scared-go.  ‘In the night, Aŋmo took a tree trunk for a man, and was scared, / and was totally terrified, [she] said.’ NOT: *‘ . . . and went in a frightened manner . . . ’

(26) 

daŋ kho-i guspu-a me duks, me˖(ː)aŋ gjel-soŋ-se. yesterday s/he- hair- fire burn. fire˖: fall-go.- ‘Yesterday, his/her hair caught fire, (namely) as s/he fell into the fire.’ NOT: *‘ . . . as s/he went by falling into the fire.’

Type B converb constructions are also quite common. In some cases, when a directional component is still feasible, the path verbs form directional counterparts, as in (27); in others only one of the pair seems possible; cf. (28) to (30). In many cases, the path verbs no longer describe a real movement. Intransitive-type verbs may combine either with intransitive or with transitive path verbs. In the latter case, the intensifying connotation is even stronger. In such intransitivetransitive combinations, case marking is triggered by the first verb; cf. (28), third alternative, and (30).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

373

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(27) ŋa-s bal khal-tsana, snanma phaŋ-ikana  I- wool spin-when thread spindle-: luks. / luk-se-joŋs.4 / luk-se-soŋ. fall.off. fall.off--come. fall.off--go. ‘While I was spinning, a bit of the thread came loose.’ / ‘ . . . the whole thread came loose (toward me / away from me).’ (28) 

Aŋmo-a ɲit thips-enak lo. / Aŋmo- sleep get.dull-.  thips-e-joŋ-enak / thips-e-kjoŋ-enak lo. get.dull--come-. get.dull--bring-.  ‘Aŋmo is getting sleepy, [she] says. / feels extremely sleepy [and cannot stay awake], / feels overwhelmed by sleep, [she] says.’

(29) 

aʧo Smanla-s gonp˖eaŋ elder.brother Smanla- monastery˖: l ̥abris-i rimo ʧats. / ʧad-in-khjoŋs. mural- drawing cut. cut--bring. r̥tiŋna r̥pon Skarma-s r̥tsi teaŋ-in-khjoŋs. after master Skarma- color- give--bring. ‘The elder brother Smanla drew the outlines for the wall painting in the monastery. / drew all the outlines . . . completely. Thereafter, master Skarma applied all the colors completely.’

(30) 

baŋ teaŋ-se-r̥tiŋna, kho tshun. / tshun-e-khjoŋs. running give--after she- pant. pant--bring. ‘After running, s/he was panting.’ / ‘ . . . s/he was panting heavily.’

In yet a few other cases, the combination is about to lexicalize or has lexicalized, that is, the simple verb is hardly used or can no longer be used (different speakers may have different intuitions about the applicability of the simple verb): (31) 

khakhi skjuk-se-joŋs plow.bolt get.loose--come. ‘The bolt of the plow came off.’

(32) a. ŋ˖i  I˖

lakpa hand-

(morok) (onom)

/

?skjuks. ?get.loose.

l̥ʧus-kheːs. twist.-carry.

/

*l̥ʧus. *twist.

⁴ The serial form lugz-joŋs (get.off.-come.) is also available with no shift of meaning.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

374



  b. ŋ˖i lakpa (morok) I˖ hand- (onom) ‘I twisted my hand badly.’

&l ̥ʧu-se-kheːs. &twist--carry.

/

%l ̥ʧus. %twist.

According to the Domkhar speaker, the compound form in example (32) is obligatory, at least in the past tense. For the Teya speaker the combination has an intensifying function, but is clearly preferred to the simple verb. The various combinations may yield very subtle differences in meaning, as in the following example with Type A serial verb constructions: (33) a. ama-s ser-i zaŋzbu r̥kunm˖elagna thams.  mother- gold- pot thief˖: seize. ‘The mother seized/tried to seize the golden pot from the thief.’ b. ama-s ser-i zaŋzbu r̥kunm˖elagna thams-teaŋs. mother- gold- pot thief˖: seize-give. ‘The mother tried to seize the golden pot from the thief with intention/ force (but could not get it back).’ c. ama-s ser-i zaŋzbu r̥kunm˖elagna thams-bors. mother- gold- pot thief˖: seize-keep. ‘The mother seized the golden pot from the thief (and both held on).’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

d. ama-s ser-i zaŋzbu r̥kunm˖elagna thams-khjoŋs. mother- gold- pot thief˖: seize-bring. ‘The mother successfully seized the golden pot from the thief.’ Combinations with the intransitive path-motion verb go are also attested in Indus Kohistan Shina, where they emphasize the completion and irreversibility of the result, (Schmidt and Kohistani 2008: 240–1, with ex. 2b): (34) zúun: mayãã´ ro bilj-ií bój-aa-n-o,  Zúun: Markhor melt- go---3.. ‘Hey, Zúun: Markhor is melting away!’

waá emph

14.3.8 Aspectuals: expression of duration In Ladakhi and Balti, as in many other Tibetic varieties, verbs with the meaning sit, stay and put, keep are used to describe an ongoing situation or ongoing resulting state. The intransitive verb ‘stay’ with the form duk in Shamskat and Leh and dat in Kenhat is used when the ongoing situation pertains to the subject or agent; the transitive verb bor ‘keep’ is used when the ongoing situation pertains to the patient. In some of the examples the bleached character of the second element is quite evident (cf. (38) and (39)), but in others one could always opt for a

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

375

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

full lexical meaning (cf. (35) to (37) and (40)). In a few cases, the combination leads to a slight shift in meaning; cf. l̥ta ‘look’, but l̥ta-se-duk ‘stare’. (35) 

kho ɖonpo-ŋun-la ʈhel-ba, ip-se-duks. she guest-- feel.shy- hide(intrs)--stay. ‘She felt shy before the guests and thus hid away.’ OR: ‘ . . . stayed hidden.’

(36) 

Waŋgjal-is ab˖ekana pene zba-se-bors. Waŋgyal- father˖: money hide(trns)--keep. ‘Waŋgyal hid [his] money from [his] father.’ OR: ‘ . . . kept [his] money hidden away . . . ’

(37) 

Gjapa ʧo-se mi-ɦun nān-de-bor-aɦo-kanak. Gyapa ruler- people- oppress--keep-- ‘The lord of Gya must have oppressed the people (all the time).’ OR: ‘ . . . must have kept the people oppressed.’

(38) 

/ ul-e-duks. . . . ŋa˖(ː) penʃin-po-aŋ ul. . . . I˖ pension-- loose. loose--stay. ‘[If I use the pension to buy a car, and if the car gets destroyed,] the pension also will be lost (lit. will have been lost) for me. / will be lost for me forever.’ NOT: *‘ . . . will have stayed lost.’

The continuous/ iterative form with -in is frequently found with the verb stay, less frequently also with the verb keep: (39) 

phrugu-ŋun baŋ teaŋ-in-duk-se, . . . child- running give--stay- ‘The children are/ were continuously running [on the roof] . . . ’

(40) 

ʈūgu ʧūːn-a, pī-a khap tā-fa, thok-te, child small- hip- injection give.- have.pain- kh˖e am˖e lālok pē˖ːn-bor-uk. she˖ mother˖ flipflop turn˖-keep-. Since the small child, when given an injection into the hip, got some pain [in the hip], his/her mother is/ was turning him/her around repeatedly.’ OR: ‘ . . . keeps/ kept turning [him/her] for a while.’

(41) 

khoŋ-is rgun-ifia griːnhaus ʧo-se-boːs. they- winter- greenhouse construct--keep. ‘They constructed a greenhouse for the winter (and kept it so).’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

376

 

With respect to the last example, the informant stated that the combination with bor ‘put, keep’ is commonly used when one constructs or makes something not for immediate but for future use or consumption. The verb stay mostly follows intransitive verbs, but it may also follow a transitive verb when the event is reflexive or subject-related in the widest sense, as in (42). The choice of the case marker for the “subject” depends on a compound or embedded reading, that is, when a compound reading is intended, the “subject” is in the ergative, as in (42a); when an embedded reading is intended, it is in the absolutive, as in (42b): (42) a. kho-s ŋaŋos-la dun stan-e-duks.  she- I.direction- front show--stay. ‘S/he faced me/looked in my direction for some time.’ b. kho, ŋaŋos-la dun stan-e, duks. she I.direction- front show- stay. ‘S/he stood there, looking in my direction.’ c. ama-s non-e rdoŋ-po ɲima˖(ː) stan-e-bors. mother- son- face- sun˖ show--keep. ‘The mother turned her little son’s face into the sun.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

If the resulting state has a more negative connotation, the verb lus ~ li̱ ‘remain, stay behind, be left behind’ is used instead of duk or dat ‘stay’. The occurrence of this combination is naturally less frequent. (43) 

ŋa̠˖(ː) ʒaktaŋ ʈūgu-a go khor-de-li-arak. I˖ every.day child- head rotate--be.left-. ‘I am always getting lost in thoughts about the child.’ ~ ‘My mind is always wandering about [what to do for] my child.’

Combinations with a verb sit or stay are typologically quite common. In Indus Kohistan Shina, the verb sit appears with both intransitive and transitive verbs, indicating a lasting result (Schmidt and Kohistani 2008: 242–4). The Ladakhi combination with bor ‘keep’ does not have a direct parallel in Indus Kohistan Shina. While the combination with the verb ‘leave’ yields a similar meaning in Hindi/Urdu (cf. mār chor: nā ‘to kill and leave it (for a later use)’ (Nespital 1989: 186)), the corresponding combination in Indus Kohistan Shina expresses a reprehensible or regrettable action that cannot be undone (Schmidt and Kohistani 2008: 247). The combinations with the verb duk ‘stay’ in Shamskat and Leh are formally and semantically close to the fully grammaticalized present perfect construction. However, in the perfect construction the verb duk ‘stay’ has become an evidential

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

377

auxiliary indicating visual knowledge, and does not inflect anymore. In the semantically related verb verb combination, by contrast, the second verb can still take all finite and nonfinite morphemes, including, of course, the evidential auxiliaries. In the Kenhat dialects, the difference between the two constructions is more apparent, as two different verbs are used: duk as the experiential auxiliary for the perfect construction and dat ‘stay’ as the aspectual “vector” verb.

14.3.9 The perfect construction expressing complete disappearance The present perfect construction consists of the verb stem plus the lhag.bcas morpheme plus an evidential auxiliary. However, with verbs expressing the annihilation or disappearance of items, the negated existential auxiliaries have a double function: they may indicate either that the event did not take place or, quite in the opposite, that the event did take place and the item in question is no longer there or is completely or already gone. The latter usage appears to be more frequent. The intended meaning is usually obvious from the context, but the informants admit that they might get confused; cf. example (44). Again, in some cases, the notion of nonexistence is not yet fully bleached; cf. example (46).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(44) 

kho ʃi-se-met. ~ ʃi-se-met. she die--not.exist=. ~ die--not.exist=.. ‘S/he has died (and is no longer there).’ ~ ‘S/he has not [yet] died.’

(45) ŋa˖(ː) pene rdzok-se-met. / -mi-nuk. / -mi-nak.  I˖ money finish--not.exist. / -not-exist- / -not-exist- ‘I happen to have (my) money spent (lit. finished) completely (I know and take responsibility / I find out upon seeing / I find out upon groping in my pocket).’ (46) 

bom jes-tsana, ŋa ʧha-tshar-e-met-pin. bomb explode-when I go-end--not.exist.- ‘When the bomb exploded, I had already gone/ left (and was no longer there).’

14.3.10 Communication verbs The case of communication verbs is similar to that of the motion and movement verbs: there are quite a few type verbs, but only one semantically rather empty verb suitable for the end or the introduction of a quotation.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

378

 

The quotation verb is adjacent to the proposition, and the type verbs come on the outer periphery. That is, in speech introductions, we find the combination of type verb and quotation verb. At the end of quoted speech, we find the combination of quotation verb and type verb. The latter construction is commonly used in place of an indirect or embedded proposition. When closing a quotation or proposition, the second element may be of a comparatively complex nature (light verb constructions or collocations) and additional arguments or adjuncts may be inserted. All in all, this combination seems to be the least likely candidate for univerbation. However, since the construction is used in order to avoid the incorporation of indirect propositions into the main sentence, as in (47a), it shows a certain tendency toward grammaticalization. (47) a. gergan-is ʈhugu-un-la ma-sil-khan-ifia jat ʂoks.  teacher- child-- -study-- memory frighten. ‘Yesterday, the teacher scolded the children badly for their not having studied.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. gergan-is ʈhugu-un-la, «sil-ma-sil-ba!» zer-e, teacher- child-- study--study-emph say- jat ʂoks. memory frighten. ‘Yesterday, the teacher scolded the children badly, saying: «[You] did not study at all!»’

14.4 Criteria for a compound reading of semantically related verb verb combinations (SVVCs) It will not be possible to discuss all the details, but none of the following criteria yields a reasonable result for the compound status of the semantically related verb verb combinations: conceptual unity, frequency or obligatoriness, accentuation and tonal features, other signs of phonetic fusion, scope of negation, and questions. As shown in Section 14.2.5, the overall frequencies are not very high, and it seems that the most frequent combinations, namely the directional combinations, are those where the semantic unity is least developed. As far as I can hear, there are no differences in accentuation between seemingly compound constructions and ordinary clause chaining or embedding. Maybe one could detect some differences with high-tech recording, but that is beyond my possibilities.

14.4.1 Phonetic reduction and fusion Quite similarly, it seems that in the tonal dialects of Upper Ladakh the tone of the second verb is not affected in a significant way. The tonal distinction is

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

      

379

usually neutralized in non-first syllables: high-tone and neutral syllables remain unaffected, whereas low-tone syllables become neutral or even high. Syllables with a voiced oral stop initial are low-tone by virtue of being voiced, but they have no tonal counterpart. Since the low tone is a concomitant feature of voicedness, it does not change significantly in non-first syllables. Only the nasals, the palatal, and the lateral glide have high-tone (unvoiced) counterparts. Syllables with a voiceless aspirated onset and voiceless fricatives are neutral. Syllables with a voiceless nonaspirated nonfricative onset are either high- or low-tone. By chance, most of the verbs that are used as second verbs in the semantically related verb verb combinations fall outside the tonal opposition: the verbs with neutral tone: ʧha, soŋ ‘go’, khjoŋ ‘bring’, and kher ‘take away’, and the verbs with an inherent, nonphonemic low tone: ɦoŋ ‘come’, dat ‘sit, stay’, and bor ‘put, keep’. In the case of combination with the phonemic low-tone verb li̱ ‘be left behind, remain’, which is somewhat less frequent and arguably functionally more marked than the more typical second verbs, I often heard a low tone, while the consulted speaker was hesitant, thinking that the low tone was neutralized. The latter feature would correspond to a certain tendency for the less frequent and thus functionally marked constructions to be more likely compound expressions than the most frequent, and hence functionally less marked constructions (see also the final discussion in Section 14.5). However, the fact that I was able to perceive a low tone, although this feature often escapes my attention, may speak against tonal neutralization and thus against intonational unity. Phonological unity, however, does not necessarily indicate a conceptual unity, not to speak of a syntactic unity, as can be shown with respect to assimilation features and phonological reduction. Type A serial verb constructions in Domkhar could be taken as phonologically reduced variants of Type B converb constructions. One might expect that they were more advanced in the process of univerbation. In the serial verb constructions, the two verbs may also show assimilation features, and thus a certain tendency to phonetically fuse at the lexeme boundary; cf. examples (48) and (50), where final s and t of the first verb are assimilated to z and d, owing to the following voiced initial of the second verb joŋ ‘come’. Assimilation features in the Tibetic languages are usually restricted to word-internal lexeme or morpheme boundaries and can thus be used to define the unit of a phonological word. One might thus expect that any kind of assimilation occurring between the two verbs would indicate a certain degree of syntactic fusion. However, I came across several examples where this cannot be the case, because the two verbs are semantically unrelated and there is no feasible connection between the two verbs that would allow us to treat them as a single compound:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

380 (48) 

  rardzi-a goatherd-

rama goat

ʧikʧik single

ri-a mountain-

luz˖ leave˖

joŋ˖sok. come˖.

[−ctr]

‘The goatherd apparently happened to leave [ ctr] one goat in the mountains but came (back).’ Roughly equal to: ‘ . . . apparently came back with one goat missing’. (49) a. bespa˖(ː) ʤola naŋ.p˖eaŋ lus-e-(na) / lus traveler˖ bag house-˖: leave--() / leave[−ctr]

soŋ-sok. go.-. ‘The traveler apparently happened to leave [ ctr] his/her bag(s) in the house and went.’ Roughly equal to: ‘ . . . apparently went without his/her bag’. b. bespa-s ʤola naŋ-p˖eaŋ bor-e-(na) / bor traveler- bag house-˖: put--() / put[+ctr]

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

soŋ+sok. go.+. ‘The traveler apparently left [+ctr] his/her bag(s) in the house and went.’ Roughly equal to: ‘ . . . apparently went without his/her bag’. With respect to (49), the informant insisted on an embedded modification reading for both constructions, with a rough translation as ‘went without his bag’. Only the morphologically heavier constructions with the additional ablative marker lusena or borena would yield a sequential reading as ‘left the bag and went’. The informant’s interpretation certainly reflects the everyday situation of an undesired combined result: the goatherd being back, although one goat is missing, or a traveler being gone, while the bags are still here. However, this interpretation is against the logic of the case assignment, which emphasizes the accidental or intentional leaving of the bag: in an embedded modifying construction, the foregrounded motion verb would have governed absolutive case marking. Unfortunately, the speaker’s stance, as reflected in his or her translation, may not always be helpful for the syntactic analysis. Similar examples are (50) and (51). In (51), the phonetic fusion is even more apparent, since according to a word-internal sound law, the final s of l̥tas combines with the following ʧ- of the verb ʧha ‘go’ to become ʃ. (50) 

nama-naŋ ma-thun-ba, makpa ʧad˖joŋs. wife- -harmonize- husband cut˖come. ‘As [he] did not harmonize with [his] wife, the husband [who had married into the girl’s family] came back [to the parents], after divorcing (lit. cutting).’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

       (51) 

381

kho Khalatse l ̥tas-soŋ. / l̥ta˖ʃ˖enuk. s/he Khalatse look.-go. look.˖go˖. ‘S/he went / is going while facing Khalatse.’ Roughly equivalent to: ‘S/he went / is going off directly toward Khalatse.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

14.4.2 Scope of negation Haspelmath (2016: 299–301) takes single negatability with wide scope over all verbs as a defining feature of a single clause. Negated verb verb combinations are infrequent in Ladakhi. Among the 1,354 instances of apparent complex constructions in my database of mostly elicited sentences,⁵ 88 necessarily contain a negated second verb (the perfect construction expressing complete disappearance) and cannot further be negated. Among the remaining 1,266 examples only 82 (6,48%) are negated⁶—and it may well be the case that most, if not all of them have to be dismissed as biclausal constructions. This result is not surprising, as negated verb verb combinations are not wellmotivated. Directional vector verb constructions, for example, describe two aspects of a single event, the path and the type, one of which is more accidental. But when a movement does not take place, there is not much need to describe the accidental part. Depending on which aspect is focused upon and negated, the other one is not accessible. If somebody did not come or go (or did not bring or take something), then there is usually no need to state the manner in which the person did not come or go. Similarly, if somebody did not run or jump (or did not steal or buy something), then we can hardly specify the direction. Negation is needed only when we want to contrast the way the person came or went (or moved or transferred something). However, in such cases, the construction must be taken as biclausal with a positive path-motion or path-movement verb in the main clause and a negative modification in the embedded clause in the case of narrow-scope or, alternatively, in the case of wide-scope constructions, with a negated path-motion or path-movement verb in the main clause and a positive modification in the embedded clause; cf. example (58), second alternative. It would be similarly odd to highlight the intention or force used in a nonevent. On the other hand, negation is motivated when talking about the duration of a negative state or the nonduration of a positive state. Accordingly, most of the negated examples in my database are combinations with the verbs stay (thirty items), followed by combinations with bring (seventeen items), ⁵ A valency dictionary of Ladakhi verbs (work in progress; numbers as per July 2015). ⁶ While this ratio within the elicited data is certainly arbitrary, the low frequency of negation is corroborated by the even lower ratio of negated verb verb combinations in the LLV. Here, among potentially 567 negatable combinations (585 potentially complex constructions minus 18 perfectof-complete-disappearance constructions) only 11 (1,94%) are negated, most of them (7) concerning the intensifying happened-completely construction with soŋ ‘went’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

382

 

go (thirteen items), and keep (eleven items). Only one instance of negation has been observed with the give-II construction of heightened intentionality. In this case, the second verb should perhaps be interpreted in the sense of ‘did not grant’: (52) 

gjapo-s r̥kunma-ŋun tson-iaŋna phiŋs-ma-teaŋs. king- thief- jail-: release.-2-give. ‘The king did not at all release the thieves from jail.’ OR: ‘The king did not grant the release of the thieves from jail.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In some combinations, negation can co-occur with either verb. When the first verb is negated, the scope is narrow: V2 happens without V1 or V1 does not happen and then V2. This type is mostly found when the duration of a nonevent is indicated (seventeen items). In such aspectual cases, a compound reading is most likely: V1 does not happen for a long time. In most combinations, however, the second verb is negated. As it turns out, the negation on the second verb typically has wide scope over the first verb (eleven items of negated duration with the verbs stay or keep, thirty-six items of negated manner with path-motion or path-movement verbs). In a few cases, negation has been attested with either the first verb or the second verb without any apparent difference in meaning. In the following, I shall give a few examples, starting with those examples where a compound reading seems to be most likely, proceeding to those where a compound reading is least likely. (53) 

mi rombo zgo-a thoks-e, biŋs-ʧha˖(ː)-ma-ɲan. man stout door- get.stuck- go.out.-go˖-2-able. ‘The stout person got stuck in the door and could not pass through.’ (Negated modal verb with wide scope over both preceding verbs) Rather unlikely: ?‘ . . . and could not go by getting through.’

(54) 

ʒiŋ-un phaŋ-se-ma-mbor-ba gjet! field- give.up--2-keep- cultivate. ‘Do not let the fields lie fallow (lit. Do not give up the fields), cultivate [them]!’ Perhaps: ‘Do not keep the fields abandoned, . . . ’ (Wide scope) Most probably not: ?‘Do not keep the fields after having given them up, . . . ’

(55) 

te̱zane am-e «kha zot!» zer-zane, then mother- mouth keep.quiet. say-when ŋa̱ kha ma̱-zor-a-dat-pen. / I mouth 2-keep.quiet--stay.- / &zot-e-ma-dat-pen. &keep.quiet--2-stay.- ‘Then, when [our] mother said: “Shut up!” I did not stop talking for some time (lit. I stayed with my mouth not shut, narrow scope). / &I did not keep quiet (lit. I did not , wide scope).’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

      

383

(56) 

snaŋa ma-mba˖(ː)-duks-pa, rardzi-a ʃaŋku ʃor. attention 2-do˖-stay.- goatherd- wolf happen. ‘Because [s/he] did not pay attention for a while, the goatherd happened to have a wolf [kill] the goats.’ (Narrow scope) Less likely: ‘Because [s/he] stayed without paying attention, . . . ’

(57) 

Aŋm-e phar-e, Nāmgyal-e ɲīŋ ʧhak. Aŋmo- abandon- Namgyal- heart break. ʧīa zer-na, tīŋna ʧī-aŋ ma̱-za-ama̱-thu-a-dat. why say- after what- 2-eat- 2-drink--stay. ‘Having been abandoned by Aŋmo, Namgyal was completely discouraged. That is, after that [experience] he did not eat and drink any longer.’ OR: ‘ . . . he stayed without eating and drinking.’ (Narrow scope)

With respect to example (55), I had been tempted to classify the first alternative, ma̱zora-datpen with narrow scope, as an aspectual construction, whereas I classified the second one, zote-madatpen with wide scope, as a biclausal embedded construction. Even though the analysis based on the literal translation might yield a slight difference in meaning, as indicated in the example translation, the consulted speakers do not see any difference, except that the second construction sounds better. In the following example, (58), which is clearly a biclausal construction, the speakers I consulted do not make any difference between the two ways of negating a complex situation. The example was confirmed by speakers from several dialects. (58) 

kho zos-e ma-soŋ. = ma-za˖(ː) soŋ. s/he eat.- 2-go. = 2-eat˖ go. ‘S/he did not .’ = ‘S/he went without eating.’

In this case, the first construction with the wide scope is the preferred one. The theoretically possible reading Having eaten, she did not go or She ate and then did not go was rejected by all the speakers consulted. But as in so many other cases, this reading may be acceptable if a suitable context is found. Double negation is sometimes possible (but may be blocked for semantic reasons) when we deal with an embedded adverbial construction of related or not so related events, e.g. kho, las maʧoa, manduks ‘s/he did not sit around without working’, roughly corresponding to kho, las ʧose/ ʧoen, duks ‘s/he stayed after/ by doing some work’. Note the absolutive construction in contrast to the compound expression khos las ʧoseduks ‘she worked for a while’ with the ergative. With true compound constructions, as in the case of intensifying compounds, double negation is not possible for obvious reasons. The

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

384

 

possibility of double negation with many of the directional complex verb constructions is a strong indicator that these constructions are syntactically not compounds. It is, however, not a definite proof, since the doubly negated constructions and their positive counterparts are not fully mutually equivalent, and the doubly negated constructions are clearly semantically marked and more emphatic. The possibility of negation seems to be a weak indicator of compositionality. Fully compound constructions, such as grammatical constructions, are negatable, typically with a wide-scope negation on the auxiliary. Aspectual combinations likewise occur with negation, with both wide and narrow scope of negation. Narrow scope of negation, however, seems to be more frequent. By contrast, wide scope appears to be preferred in arbitrary combinations, such as (58). Wide or narrow scope is thus not indicative of compositionality in Ladakhi.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

14.4.3 Questions When the addressee wants to recheck the content of a statement, s/he might focus on the first element of a combination, the last element, or the full combination. The respective choices could provide some indications with respect to the compositional character. However, when a question refers to a particular constituent of a preceding utterance, speakers usually employ only the corresponding question word and do not repeat the full sentence; cf. (59), first alternative. I have thus only a few examples in my database, and I am not completely sure whether these follow a consistent pattern. (59) ʧipa ʧikʧik ur-te-soŋ. – karu-a? / karu soŋ?  little.bird single fly--go. where- where go. ‘A little bird flew away.’—‘Where to? / Where did it go?’ (60) 

aba-s phrug-i kitap-(p)o l̥tsaŋs-ikana sɲeps-khjoŋs. – father- child- book- shelf-: reach.out-bring. ʧi khjoŋs? / gana khjoŋs? what bring. whence bring. ‘The father reached down the child’s book from the shelf.’—‘What did [he] bring? / Where did [he] bring [it] from?’

(61) 

[kho-s] rdo-eka r̥kos-taŋ-duk. – ʧi r̥kos-taŋ-duk? s/he- stone-: carve.-give-. what carve.-give-. ‘[He] carves into the stone(s).’—‘What does [he] carve?’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

385

(62) a. daŋ aʒaŋ-is ŋa˖(ː) galɖi˖k ɲos-teaŋs. –  yesterday uncle- I˖ car˖ buy.-give. ʧi ɲos? / &ʧi ɲos-teaŋs? what buy. &what buy.-give. ‘Yesterday, [my] uncle bought me a car.’—‘What did [he] buy?’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. daŋ aʒaŋ-is ŋa˖(ː) galɖi-k ɲo-se teaŋs. –  yesterday uncle- I˖ car- buy- give. / &ʧi teaŋs? ʧi ɲo-se teaŋs? what buy- give. &what give. ‘Yesterday, [my] uncle bought a car and gave it to me.’—‘What did [he] buy and give? / &What did [he] give?’ With motion verbs, the main focus appears to be on the directional path-motion verb. One would not ask where the bird flew to, but where it went, as in (59). Similarly, in example (60), the question is what the father brought, not what he reached down. In the case of the volitional give-II construction, by contrast, one would use either the compound construction, as in (61), or simply the first verb, as in (62a, first alternative), but the compound construction would be preferred (62a, second alternative). By contrast, if the verb give is used with a beneficiary meaning or as a full lexical verb, one could use either the complex expression or the second verb, with the latter being preferred, (62b, second alternative). One could thus conclude that if the first verb or the verb verb combination stands in the focus of the question, this would indicate that we are dealing with a real compound. However, both the first verb and the verb verb combination can be in the focus of the question when we deal with two unrelated verbs in a phonetically merged serial combination, as in examples (63) and (64a). The first verb may also be used in the question when we deal with a converbial construction, as in example (64b). (63) 

rardzi-a rama ʧikʧik ri-a luz˖ joŋ˖sok. – goatherd- goat single mountain- be.left˖ come˖. ʧi luz˖ joŋ˖sok? / ʧi lus˖ok? what be.left˖ come˖. what be.left˖. ‘The goatherd happened to come back with one goat apparently being left in the mountains.’—‘With what being left did [he] come back? / What had been left?’

(64) a. ama-s dares ʧhotpa ʃamz˖ joŋs. –  mother- earlier offering hum.throw.˖ come. garu ʃamz˖ joŋs? – khenɖag-a. where hum.throw.˖ come. roof.top- ‘[Our] mother threw out the offering and came [back].’—‘Where did [she] throw [it] out and then come [back]?’—‘On the roof.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

386

  b. ama-s dares ʧhotpa ʃam-se, joŋs. – mother- earlier offering hum.throw- come. garu ʃams? – khenɖag-a. where hum.throw. – roof.top- ‘[Our] mother threw out the offering and came [back].’—‘Where did [she] throw [it] out?’—‘On the roof.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

There are also cases where the focus on the second verb is impossible owing to semantic considerations. The combination of skor ‘take around’ and khjoŋ ‘bring’ indicates that a person was taken on a tour (skore) and then brought back home (khjoŋs). It is clear that the question garu khjoŋs ‘where was [s/he] brought to’ does not make sense; hence only the first verb or the combination can appear in the question. It follows, thus, that a focus on the first verb or on the complex expression is no proof of a compound structure. The evidence is rather indirect: when a compound reading is intended, a focus on the second verb is ruled out, and when the second verb is focused upon, a compound reading is ruled out. This is nicely illustrated in the following example. Here the directional path-movement verb kher ‘take away’ is used with an intensifying function. Intensifying usage leads to a compound-like structure. Accordingly, the question in (65a) focuses on the first verb. If the speaker focuses on the second verb, s/he shows that s/he did not fully understand the preceding sentence. The answer thus first contradicts the wrong interpretation, before explaining what kind of work was done so fast (65b). (65) a. deriŋ khoŋ-is r̥napsa hjuk-se-kheːs. – ʧi hjuks? today they- harvest work.fast--take.away. what work.fast. ‘Today they did the harvest very fast.’—‘What did they do fast?’ b. ʧi kheːs? – men, ʧi-aŋ ma-kheːs, what take.away. no what- 2-take.away. r̥napsa hjuk-se-kheːs, zeːs-pa. harvest work.fast--take.away. say.- ‘What did they take away?’—‘No, [they] did not take anything away; [they] did the harvest very fast, as [I] said.’ Like examples (48) and (49), examples (63) and (64a) formally suggest a sequential reading, according to their case-marking pattern. The informants, however, unanimously claim that the first verb is an embedded adverbial and that the person comes or goes by doing X or happening to have X. Note that the inferential markers in examples (63) and (48) above have wide scope and primarily refer to the first verb, not to the second one, as the speaker has observed the coming, but can only infer what happened before. One may say, however, that the inference

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

387

then covers the whole situation of coming back without the goat, of which not all parts have been observed. This kind of compound perspective is clearly the case in (49), where the inferential marker may directly refer to the foreigner’s going away as well as to his/her leaving behind of the bag. It may be this wide-scope usage of the inferentials that leads the speakers to conceive of such combinations as nonsequential ones.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

14.4.4 Case marking: combination of transitive and intransitive verbs When verbs of different valency are combined, case alignment with either the intransitive verb (absolutive) or the transitive verb (ergative) should indicate whether we are dealing with a biclausal construction or a compound. However, only the resultative and/or durative construction with the intransitive verb duk ‘sit, stay’, the intensifying volitional construction with the transitive verb taŋ ‘give’, and the intensifying constructions with kher take away’ and khjoŋ ‘bring’ yield unambiguous results: with all these combinations, case marking is always triggered by the first verb. Cf. examples (35) and (42a) for the combination of V2 duk with an intransitive and a transitive V1 respectively, examples (29) and (30) for the combination of an intensifying transitive V2 with a transitive and an intransitive V1 respectively, example (18) for the combination of V2 give-II with a transitive V1, and the following example, (66), for the combination with an intransitive V1. In example (42a), the combination of the transitive V1 (stan ‘show’) with the intransitive V2 (duk ‘stay’) yields the same ergative marking as the simple verb would have done; in examples (30) and (66) the combination of an intransitive V1 (tshun ‘pant’ and duk ‘stay’) with a transitive V2 (khjoŋ ‘bring’ and taŋ ‘give’), yields the same absolutive marking as the simple verb. In the case of combinations of a formally transitive, but semantically intransitive path-motion verb, such as gom ‘step on, over’, with an intransitive directional motion verb, case marking is ambiguous in Type B converb constructions, that is, it may be triggered either by the transitive first verb or the intransitive second verb, as in example (67d). However, in the Domkhar Type A serial verb constructions, case marking is triggered by the intransitive motion verb, as in (67e). This again shows that a construction that, on the formal, phonological level, seems to be a compound form is not necessarily one on the semantic or syntactic level. (66) 

daŋ ŋa khoŋ-ika duks-teaŋs-pin. yesterday I- they-: stay.-give.- ‘Yesterday, I simply stayed (overnight) in their house.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

388

 

(67) a. Dorʤe-s peathaŋ rgjaps.  Dorje- sand.plain cross. ‘Dorje crossed the sandy plain / desert.’ peathaŋ sand.plain

b. Dorʤe-s Dorje-

rgjap-se, cross-

ø

( . . . ) joŋs. come.

(ERG)

(ABS)

‘Dorje crossed the sandy plain/ desert and ø came (to . . . ).’ (Sequential clause-chaining construction) c. Dorʤe, Dorje-

ø

peathaŋ sand.plain

rgjap-se, cross-

(...)

joŋs. come.

(ERG) (ABS)

‘Dorje came (to . . . ) by ø crossing the sandy plain/ desert.’ (Embedded modifying construction)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

d. Dorʤe / Dorʤe-s peathaŋ rgjap-se-joŋs. Dorje- Dorje- sand.plain cross--come. ‘Dorje crossed the sandy plain/ desert in my direction.’ OR: ‘ . . . came by crossing the sandy plain/ desert.’ (Type B converb) e. Dorʤe / *Dorʤe-s peathaŋ rgjabz-joŋs. Dorje- *Dorje- sand.plain cross.-come. ‘Dorje crossed the sandy plain/ desert in my direction.’ OR: ‘ . . . came by crossing the sandy plain/ desert.’ (Type A serial verb)

14.5 Conclusion Semantically related verb verb combinations in Tibetic languages are inherently ambivalent. In some instances, they can or must be analyzed as representing two sequential events (e.g. the contract situation). In other cases, they can or must be analyzed as representing a simple event, possibly associated with connotations of suddenness, regret, surprise, completion, or remaining obligations. More often than not, they represent a complex event semantically, but must be analyzed syntactically as a biclausal construction consisting of an adverbial type verb and a superordinate path verb. My approach toward these constructions has changed over the years. Initially, I was overly enthusiastic, including verb pairs in the category of compounds that I better had not included, such as mount a horse + go/come > go/come by horse; cf. (8)– (11). Presently, however, I wonder whether we are dealing with complex predicates at all. My doubts have only been reinforced in view of the data from the other languages presented at the International Symposium Mysteries of Verb-Verb Complexes in Asian Languages, Tokyo, December 14–15, 2013; cf. the corresponding contributions in this volume.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

      

389

Aikhenvald (2006: 18) treats adverbial or modifying verb verb combinations as semantic units. This seems to be problematic, but could be motivated, if serial verb constructions stand in contrast to converbial or other morphologically marked constructions. In Ladakhi (and all Tibetic languages with converb constructions), such opposition does not exist, and there is no obvious formal feature such as intonation or morphological reduction that could distinguish the semantically related verb verb constructions from ordinary biclausal constructions with omitted arguments. Morphological and phonological reduction or merging has been shown to be misleading, as it may also happen with semantically completely unrelated verbs in coincidental adjacency. Most other tests likewise remained inconclusive. Speaker intuition, double negation, the focus in a checking question, and case-marking behavior indicate that we are dealing with a fuzzy continuum between constructions that show only some features of compositionality and those that may be treated as compounds. In a recent workshop on Multiverb Constructions (Humboldtuniversität Berlin, December 7–8, 2017), Kilu von Prince, as an invited speaker, posed the question whether there is a continuum across the various multiverb constructions described for different languages. From the perspective of the Tibetic languages, one can only say, “Of of course, there is.” One can find such a continuum, even within a single language or dialect and even synchronically, between arbitrary adjacency, biclausal sequential or adverbial converb constructions and monoclausal or (almost) compound constructions. While some combinations, such as those in examples (18) and (19) in Section 14.3.5 and (24) to (32) in Section 14.3.7 are clearly compounds and cannot be analyzed as biclausal sequential or adverbial constructions, other combinations based on clause chaining are rather ambivalent. In sorting the semantically related verb verb combinations along this continuum, one could set up the following hierarchy, from the most developed to the least developed combinations, even though the different factors appear to be somewhat contradictory (semantic restriction vs. syntactic merging): • most advanced and monoclausal • give-II for heightened intentionality (phonetically reduced: Type A serial verb construction; semantically unrestricted; alignment with V1; no semantic counterpart; negation attested only once) • perfect with negated existential verb (semantically restricted, but semigrammaticalized; alignment with V1; negation impossible) • intensifying usages (semantically restricted to unwanted events; alignment with V1; a directional counterpart is sometimes possible, often not; negation does not seem to be feasible, except in the ‘happen-’ combination with go; double negation can be ruled out)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

390

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

• intermediate (monoclausal?) • aspectual constructions (restricted to intransitive or transitive verbs respectively; moving toward grammaticalization; alignment with V1; negation: on V2 with wide scope or on V1 with narrow scope; double negation yields a biclausal, embedded adverbial reading with different case marking) • modal verb and causative constructions (not discussed here, but cf. Zeisler 2014; syntactically intertwined; negation on modal or causative V2 with wide scope) • least advanced, possibly still biclausal • directional vector verbs (restricted to intransitive motion and transitive movement verbs; alignment mostly invisible, ambivalent in Type B converb constructions, with V2 in Type A serial verb constructions) • beneficiary give (restricted to transitive verbs; alignment invisible; negation not attested; possibly ambiguous and biclausal) The highly frequent combinations with directional vector verbs must be analyzed as biclausal embedded modifying constructions—in the process of becoming compound constructions. However, in 1300 years of language development, these constructions show little progress in compounding or grammaticalization. The reason might be that both components of the pair are usually semantically well motivated. One might still argue with Aikhenvald that, since the directional combinations usually refer to one single, albeit complex event, they form a conceptual unit, and hence correspond to compounds. However, they can be contrasted with the more advanced combinations, such as the intensifying constructions, which clearly refer to a single noncomplex event. The latter certainly have more properties of a compound than the directional combinations. Because of this contrast, conceptual unity cannot be the sole criterion. Furthermore, some semantically nonrelated combinations, such as leave something behind plus go/ come or go/ come by leaving something behind show the same type of adverbial modification or situational complexity, and even the same type of phonological reductions and fusions, but one would not want to treat them as compounds. To some extent, the data corroborates Hook’s (Chapter 11, this volume) findings with respect to Indo-Aryan languages that a low frequency of compound verbs goes along with fewer structural restrictions, which may also be taken as a sign that the construction has not grammaticalized. As he states in an e-mail communication (April 16, 2017): Below a certain threshold CVs behave like non-CVs in their privileges of occurrence. Above that threshold vectors behave more like morphemes of tense, aspect, and mood and alternation of CVs with non-CVs becomes a normal

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

low

high

movement, transfer  / medium

low

 motion  

transitive

high

motion  

directionals

give benefactive

medium

medium

intransitive

none

medium

transitive

modal verb

medium

high

intransitive reflexive

go intensifying

aspectuals

annihilation

undesired, typically [–ctr]

special perfect

medium

low

agentive events

give intentional V1 V1 V1

– () –

invisible invisible

– (+)

V2

V1 or V2

– (+)

+

–

V1+V2

V1

– –

+ ()

V1

alignment

syntactic:

+

relative frequency phonetic. reduced

other intensifying.

semantic restriction

type

Table 14.5 Locating the verb-verb constructions on the univerbation path

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

n. a.

V2 wide V1 narrow (biclausal?)

V2 wide

V2 wide V1 narrow (biclausal?)

not feasible

V1 narrow

impossible

rare, V2 wide

negation scope

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

392

 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

paradigmatic feature of every or nearly every lexical verb. . . . My impression is that the ratio of CVs to non-CVs in Ladakhi and Kohistani Shina is not high enough to meet that threshold. That is, non-CVs in those languages cannot be said to alternate with their corresponding CV counterparts. Hence, the inhibiting effects of factors [other than negation] will not be noticeable.

However, the picture in Ladakhi is somewhat more complicated. When we look at the different types of verb verb combinations, we can see that frequencies and compositionality do not match. While the semantically restricted directional constructions with the highest relative frequency (combination of type-motion with path-motion verbs in relation to simple type-motion verbs) show little signs of compositionality, not to speak of grammaticalization, one can observe syntactic developments that indicate compositionality in the highly frequent, semantically unrestricted intentionality-give construction, in the likewise highly frequent, but semantically restricted happen-go construction, and in the comparatively infrequent intensifying constructions. We find even signs of grammaticalization both in the semantically nonrestricted durative construction of medium frequency and in the semantically restricted perfect-of-annihilation construction of comparatively low frequency. Frequency, therefore, cannot be taken as the sole or main indicator of semantic bleaching or grammaticalization. It turns out that most of the semantically related verb verb combinations of Ladakhi and, more generally, the Tibetic languages, are not complex predicates in a strict sense (not to speak of compound verbs). Are they at least complex predicates in a rather loose sense? Or are we dealing with coincidental adjacency, which only accidentally looks semantically motivated enough to be taken for a complex expression? In a way, the white blotch on Masica’s map, or at least his question mark, is more than justified.

Acknowledgments My heartfelt thank goes to the main informants—all deeper insight into the Ladakhi language is due to their patience with me and my stupid questions. The following speakers have contributed to this chapter: late meme Tondup Tshering from Khalatse, Tshewang Tharchin, Trhinlas Chosphel, Jigmet Angcuk, Tshering Tshomo, and others from Domkhar, Tshering Dolkar from Teya, Phuntsok Dolma from Saspol, Thrinlas Wangmo from Leh, and Mengyur Tshomo and Jigmet Yangdrol from Gya-Mīru. I should also like to thank the organizers of the International symposium Mysteries of Verb-Verb Complexes in Asian Languages, at NINJAL, Tokyo, December 14–15, 2013, Prof. Taro Kageyama (NINJAL), Prof. Peter Edwin Hook (Universities of Michigan and Virginia/NINJAL), and Prof. Prashant Pardeshi (NINJAL) for their invitation to the Symposium. Many thanks in particular to Prof. Peter Edwin Hook for his comments on

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

      

393

the draft version of this chapter. I should further like to thank the anonymous taxpayer who, via the Deutsche Forschungsgemeinschaft, enabled me over many years to pursue my quite unprofitable interests (2002–8: Project on Semantic roles, case relations, and crossclausal reference in Tibetan, 2010–14: Project for A valency dictionary of Ladakhi verbs, and 2016–18: Project on Evidentiality, epistemic modality, and speaker attitude in Ladakhi).

Orthographic conventions

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

- separates segmentable morphemes ˖ separates nonsegmentable morphemes · (x·x) separates written graphemes . (x.x) separates compound and inherent elements = functional equivalence (not a clitic!) & preferred usage or interpretation % less preferred usage or interpretation ? questionable usage or interpretation * ungrammatical

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

PART III

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

L A N G U A G E S OF CE N T R A L A N D NORTHWEST ASIA

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

15 Verb–verb complexes in Turkic languages Interaction of lexical and delexicalized verbs Andrey Shluinsky

15.1 Overview of Turkic V–V complexes

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15.1.1 Introduction The goal of this chapter is to show the ways in which lexical verbs and delexicalized verbs interact in V–V complexes of Turkic languages. Based on the data of four Turkic languages, I demonstrate that some auxiliaries may be combined only with lexical verbs of a specific semantic group (e.g. with dynamic verbs); other auxiliaries have different meanings depending on the semantic features of the lexical verb. As is well known, Turkic languages are a sub-branch of the Altaic language family and form a group of about forty closely related languages spoken mainly (though not exclusively) in Russia. Their internal classification is disputable. Turkic languages are agglutinative, SOV, accusative, and prefer nonfinite structures in clause-combining. In this chapter I present case studies of V–V complexes in Karachay-Balkar, Chuvash, and Tubalar Altai based on field data collected in collective field trips organized by Moscow State University. Karachay-Balkar is spoken by 300,000 people in the Kabardino-Balkar and Karachay-Cherkess Republics of Russia, being one of the official languages of both republics. Karachay-Balkar data (Cherek dialect of Balkar variety) were collected in the village of Verkhnjaja Balkaria, Kabardino-Balkaria, in 2002, and the results were published in Lyutikova et al. (2006). Chuvash is one of the official languages of the Chuvash republic of Russia and has about one million speakers. Chuvash data (Anatri / Lower dialect) were collected in the village of Shimkusy, Chuvashia, in 2001; the results were published in Shluinsky (2006a). Tubalar Altai (Tuba, Tuba-Kuzhi) is a highly endangered language spoken by about 400 people in the Altai Republic of Russia; Tubalar Altai used to be classified as a dialect of Altai, but now is typically regarded as a separate language. Tubalar Altai data were collected in the village of Kebezen, Altai, in 2006; the results were published in Shluinsky (2009a). Additionally, I use examples from Mishar Tatar that come from a Moscow State University collective field trip (Tatar Yeltan, Tatarstan, 1999–2000; cf. Pazelskaya

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

398

 

and Shluinsky 2007), examples from Sagay Khakas that come from a Russian State University for Humanities collective field trip (Kazanovka, Khakassia, 2002; cf. Shluinsky 2006b), and examples from other Turkic languages that come from published grammars. The structure of the chapter is the following. In Sections 15.1.2–15.1.3, I give an overview of the Turkic V–V complexes. In Section 15.2 two case studies of absolutely productive actional delexicalized verbs are presented: tur- ‘< stand’ in Karachay-Balkar in Section 15.2.2 and il- ‘< take’ in Anatri Chuvash in Section 15.2.3. In Section 15.3, I present a more complex case study of the productive delexicalized verbs in Tubalar Altai. Section 15.4 concludes the chapter.

15.1.2 Structure of Turkic V–V complexes

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Turkic V–V complexes belong to Type 4 in Kageyama’s classification of Japanese (Chapter 2, this volume) and have the following structure. A V–V complex consists of (a) a lexical verb in a converb form and (b) a delexicalized verb in any form depending on the syntactic and semantic context. The verbs in a V–V complex are juxtaposed. The delexicalized verb does not retain its lexical meaning. There are no “compound verbs” (Types 1–3 in Kageyama’s classification]) in Turkic languages. A typical example of a Turkic V–V complex is the Bashkir sentence (1). The lexical verb ešl- ‘work’ takes a converb marker -əp; the delexicalized verb jat- ‘< lie’ takes a finite present-tense form and keeps none of the meaning of ‘lying’; the delexicalized verb immediately follows the lexical verb. Bashkir (1) šəp kenə ešl-əp rather well work- ‘They work rather well.’

jat-a-lar --3 (Juldašev 1977: 122)

15.1.3 Converbs used in V–V complexes Turkic languages have a considerable number of converbs. Following the terminology of Nedjalkov (1995), one can divide them into a bunch of specialized converbs vs. two contextual converbs. A specialized converb has a specific circumstantial meaning; for example, (2) contains an example of an anterior converb that is limited to expressing anteriority. Bashkir (2) haldat-ka al-ɨn-ɣas hər vakɨt ber-gə bul-dɨ-k soldier- take--. all time one- be--1 ‘When we became soldiers, we were together all the time.’ (Juldašev 1977: 74)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

399

A contextual converb has a more general circumstantial meaning and is interpreted by context as expressing anteriority or simultaneity, manner, reason, etc. Turkic languages have only two contextual converbs. One of them, exemplified in (3), takes the marker -p; Chuvash and Yakut (also known as Sakha) are the only two Turkic languages that lack the -p converb, but Chuvash -sa and Yakut -n are their functional analogs. The second one takes the marker -V / -j and may be used in its simple form, as in (4). Bashkir (3) ul armija saf-ɨ-nan kajt-ɨp nəšriət-tə s/he army rank-3- return- publishing_house- ešlə-j bašla-nɨ work-. start- ‘Having returned from the army, he started working in a publishing house.’ (Juldašev 1977: 31) Bashkir

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(4) ɣəjšə ašɨk-ma-j ɣɨna Gajsha hurry--. only ‘Gajsha got undressed unhurriedly.’

sisen-de undress- (Juldašev 1977: 156)

Turkic contextual converbs have a number of structural differences from specialized converbs; for example, contextual converbs are normally same-subject, while specialized converbs may be same-subject or different-subject; contextual converbs may be transparent for the scope of mood; contextual converbs may be used for clause-chaining, i.e. expressing a series of events that have more or less the same status in the structure of the discourse. Such uses are not possible for specialized converbs. Only contextual converbs may be used when, together with the main verb, they express tightly related subevents of a macroevent; cf. two instances of such in (5). Mishar Tatar (5) šul čak jugarɤ-dan mɤ-nɤŋ kul-ɤ-na ber that time height- this- hand-3- one kil-ep teš-ä dä tägärä-p kit-ep come- fall- and roll- leave- ‘At that time a clew fell down from above into her hands further.’

jɤmgak clew bar-a go- and rolled away

“Macroevent” uses of contextual converbs like those in (5) are the functional and semantic base for the development of V–V complexes. Only contextual converbs

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

400

 

are used in Turkic V–V complexes. Most typically, the -p converb (-sa in Chuvash, -n in Yakut) is used in V–V complexes, as in (6a), but the -V / -j converb is also present, as in (6b). Tubalar Altai (6) a. toš qajɨl-ɨp par-dɨ ice melt- - ‘The ice melted’. b. kem araɣɨ urla-j who alcohol steal-. ‘Who stole the alcohol?’

par-dɨ ? -

V–V complexes are different in a number of features that distinguish them from other converb constructions. Firstly, inflected delexicalized verbs used in Turkic V–V complexes are conventionalized and form a closed class, although some of them are used more widely than others. In contrast, for macroevent uses there is no closed class of verbs. Secondly, delexicalized verbs in V–V complexes do not keep their lexical meaning. In contrast, all the verbs in macroevent uses keep their lexical meanings, although these meanings always interact in an idiosyncratic way.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15.1.4 Delexicalized verbs used in Turkic V–V complexes There is no systematic study summarizing all the V–V complexes reported for all the Turkic languages or, consequently, all the verbs that may be used as delexicalized verbs in Turkic languages in general. The sets of delexicalized verbs are, of course, different in different Turkic languages, although some of them are very productive in Turkic languages of different groups and areas, such as al- ‘< take’, ber- ‘< give’, and tur- ‘< stand’. Meanings expressed by delexicalized verbs in Turkic V–V complexes form the following three general semantic domains. A. Deictic meanings expressed by verbs of motion indicate the direction of the event in the space; cf. (7)–(8). B. Actional and Aktionsart meanings expressed by verbs of different semantic groups introduce a phase into the event structure or emphasize an existing phase or specify the way an event takes place; cf. (9)–(10). C. Valency-changing benefactive meanings expressed by the verbs ‘take’ and ‘give’; cf. (11)–(12).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

401

Anatri Chuvash (7) karap puxta-ran věšʲ-se tux-r-ě ship bay- swim- _--3 ‘The ship went out from the bay.’ Bashkir (8) ul irtük kajt-ɨp kit-te s/he early return- - ‘He left home early in the morning.’

(Juldašev 1977: 127)

Mishar Tatar (9) ä a-nɤŋ kaja kit-kän-e-n pränkä rafig-e kür-ep kal-a and s/he-GEN where leave--3- Prjanik Rafik-3 see- - ‘And Rafik Prjanik saw where she went.’ Bashkir (10) min bɨl kitap-tɨ uk-ɨp I this book- read- ‘I read this book completely.’

sɨk-tɨ-m --1 (Juldašev 1977: 129)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Sagay Khakas (11) ulax mäč tut pir-gen boy ball catch. - ‘The boy caught the ball for someone.’ Uzbek (12) bu lazzat-li taom-ni this taste- food- ‘I ate myself this tasty food.’

je-b eat-

ol-di-m --1 (Kononov 1960: 265)

The same delexicalized verb may express meanings of different semantic domains, also in the same language. For example, (13) is polysemous, since the delexicalized verb ber- ‘< give’ has in Tubalar Altai both an actional inchoative meaning, as in (13a), and a valency-changing other-benefactive meaning, as in (13b). Tubalar Altai (13) wasʲa ežik ač-ɨp ber-di Vasja door open- - a. ‘Vasja started opening the door.’ b. ‘Vasja opened the door to someone.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

402

 

Having different frequency, different delexicalized verbs have different degrees of grammaticalization. For example, in Sagay Khakas sal- ‘< put’ morphologically is still an independent verb, as in (14a), and *-is ‘< send’ is, together with the converb marker, reanalyzed as a derivational suffix, as in (14b). The suffix -ɨbɨs- (i) is more regular than auxiliaries, (ii) is convergent to the vowel harmony, and (iii) morphologically fills the slot of the derivational suffixes, for example, precedes the negation marker. V–V complexes sensu stricto are, therefore, to be distinguished from morphological markers that result from them via a grammaticalization process. Sagay Khakas (14) a. ajdo pičik-ti pas Ajdo paper- write. ‘Ajdo wrote the letter.’ {ab}

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. ajdo Ajdo

pičik-ti paper-

sal-ɣan -

paz-ɨbɨs-xan write-PFV-

(< *paz-ɨp ɨs-xan) write- -

As I have already said, in contrast to standard converb constructions, in a V–V complex the converb of a lexical verb cannot be linearly separated from the delexicalized verb. (15a) exemplifies a V–V complex consisting of a converb followed by a delexicalized verb, and (15b), where the converb is linearly separated from the delexicalized verb, is ungrammatical. In (16a), we deal with a V–V complex, but (16b), where the converb is separated from the delexicalized verb by an adverb, is grammatical only in the reading of a standard converb construction, not that of a V–V complex. In contrast to a finite verb in a standard converb construction, a delexicalized verb cannot have its own complements or adjuncts. Constructed sentences where a finite verb has its own complements or adjuncts, such as (16b) and (16c) in contrast to (16a), are not always ungrammatical, but contain standard converb constructions, not V–V complexes. Anatri Chuvash (15) a. ivan vil-se Ivan die- ‘Ivan died.’

kaj-r-ě --3

b. *vil-se die-

kaj-r-ě --3

ivan Ivan

Karachay-Balkar (16) a. sabij-le zɨj-ɨl-ɨp tur-a-dɨ-la child- gather-- --3- ‘The children have already gathered.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

403

b. sabij-le zɨj-ɨl-ɨp entda tur-a-dɨ-la child- gather-- again --3- i. ‘The children, having gathered, are standing again.’ ii. *‘The children have already gathered again.’ c. sabij-le zɨj-ɨl-ɨp tur-a-dɨ-la baxca-da child- gather-- --3- garden- i. ‘The children, having gathered, are standing in the garden.’ ii. *‘The children have gathered in the garden.’

15.2 V–V complexes as productive actional delexicalized verbs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15.2.1 Describing the parameter of actionality In this section I present two case studies of productive V–V complexes that function as actional delexicalized verbs, i.e. change the actional structure of the verb. One of them is a case study of the V–V complexes with the delexicalized verb tur- ‘< stand’ in Karachay-Balkar, the other is a case study of the V–V complexes with the delexicalized verb il- ‘< take’ in Chuvash. In both cases I have chosen the most productive V–V complex in the language in its most productive form. In both cases, studies of the actional semantics of a V–V complex are based on the actional features of lexical verbs in their independent use. For identifying the actionality of a verb, a method proposed by Tatevosov (2002) is used in the study. Actionality of a verb is defined by actional meanings of (a) its perfective and (b) its imperfective form; five actional meanings are distinguished:  (S);  (P);  s (MP);     (ES); and     (EP). S, P, and MP are accessible both for perfective and for imperfective forms, and ES and EP are accessible only for perfective forms. Any form may have one or more actional meanings, i.e. may be monosemous or polysemous in the domain of actionality. The following notation is used for an actional class or for actional characteristics of a given verb: actional meanings of a perfective form are listed before a semicolon; actional meanings of an imperfective form are listed after a semicolon; if a form has more than one actional meaning, its actional meanings are separated with a comma; all the record is placed in angle brackets. This notation is illustrated by Chuvash examples (17) and (18). The verb tavrǎna ‘return’ is telic and belongs to the actional class ; cf. (17a) with its perfective form that denotes an entry into a state, while (17b) with its imperfective form that denotes a process. The verb čirle ‘be ill’ is weak inceptive and belongs to the actional class ; cf. (18a) with its perfective form that is ambiguous and may denote both an entry into a state and a state, and (18b) with its imperfective form that denotes a state.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

404

 

Anatri Chuvash (17) a. pětěr tavrǎn-č-ě Peter return--3 ‘Peter returned (in time X / * for time X).’ b. pětěr tavrǎna-tʲ Peter return-.3 ‘Peter is returning.’ Anatri Chuvash (18) a. ivan cirle-r-ě Ivan be_ill--3 i. ‘Ivan fell ill.’ ii. ‘Ivan was ill (for time X).’ b. ivan čirle-tʲ Ivan be_ill-.3 ‘Ivan is ill.’ Karachay-Balkar and Tubalar Altai actionality is described in these terms in more detail in Lyutikova et al. (2006) and Pazelskaya and Rybinceva (2009). For Chuvash the same method is used.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15.2.2 V–V complexes with tur- ‘< stand’ in Karachay-Balkar The most productive V–V complexes in Karachay-Balkar use the delexicalized verb tur- ‘< stand’. Both contextual converbs are used with this delexicalized verb. The construction -V tur- that includes the -V / -j converb tends to denote an ongoing process, as in (19a), while the construction -p tur- that includes the -p converb tends to denote a state, as in (19b). Karachay-Balkar (19) a. fatima awru-j Fatima be_ill-. ‘Fatima is falling ill.’ b. fatima awru-p Fatima be_ill- ‘Fatima is ill.’

tur-a-dɨ --3

tur-a-dɨ --3

The case study I present here is limited to the construction -p tur- with the -p converb and by the imperfective form of the delexicalized verb. At first glance, the Karachay-Balkar construction -p tur- has a resultative meaning in the sense of Nedjalkov and Jaxontov (1983/1988), i.e. refers to the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

405

resulting state that takes place when a telic event reaches its natural endpoint. The sentence (20a) is a typical example of this construction; although it is translated into English with a perfective form of Present Perfect, in fact in Karachay-Balkar it denotes the state of ‘having cooked the soup’ that continues at the reference point.¹ A habitual interpretation in (20b) makes the resultative meaning more explicit, since this sentence expresses exactly a regular result, in contrast to (20c), where the same lexical verb with no delexicalized verb expresses a regular ongoing state. Karachay-Balkar (20) a. ɨnna šöndü šorpa bišir-ip grandmother now soup cook- ‘Grandmother has cooked the soup.’

tur-a-dɨ --3

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. min üj-ge kel-gen-de, ɨnna xar kün-den I home- come-- grandmother every day- šorpa bišir-ip tur-a-dɨ soup cook- --3 ‘When I come home, grandmother has already cooked the soup every day.’ c. min üj-ge kel-gen-de, ɨnna xar kün-den I home- come-- grandmother every day- šorpa bišir-e-di soup cook--3 ‘When I come home, grandmother cooks the soup every day.’ However, as was shown by Nedjalkov and Nedjalkov (1987), in fact the construction -p tur- can express different phases of the event expressed by the lexical verb. The phase that is expressed depends on the lexical verb’s actionality. With strong telic (21) and punctual (22) lexical verbs, the construction -p tur- always has the meaning of the resulting state and indeed introduces the resultative stative phase into the event structure. Karachay-Balkar (21) alim öl-üp Alim die- ‘Alim is dead.’

tur-a-dɨ --3

¹ I use the same convention in translations of all the other examples with the resultative meaning since English has no productive means for expressing it.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

406

 

Karachay-Balkar (22) alim kitap-nɨ tab-ɨp Alim book- find- ‘Alim has found the book.’

tur-a-dɨ --3

Weak telic lexical verbs fall into two subclasses of lexical verbs, corresponding to result vs. manner verbs in terms of Rappaport Hovav and Levin (1998). Result verbs lexically specify the state that takes place after the culmination point, whereas manner verbs specify the process that may finish with a resulting state. With result verbs, the construction -p tur- has only the meaning of the resulting state, as in (23). With manner verbs, the construction -p tur- has two meanings: the meaning of the resulting state, as in (24a), and the meaning of the ongoing process, as in (24b). Karachay-Balkar (23) fatima qartoš sat-ɨp tur-a-dɨ Fatima potatoes sell- --3 ‘Fatima has sold the potatoes.’ Karachay-Balkar

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24) fatima qofta eš-ip tur-a-dɨ Fatima jersey knit- --3 i. ‘Fatima has knitted the jersey.’ ii. ‘Fatima is knitting the jersey.’ With (weak) inceptive-stative lexical verbs, the construction -p tur- has the meaning of the lexical state and is synonymous with the lexical verb in its imperfective form, as in (25) and (26). However, since the verbs of this actional class have an inceptive meaning, the lexical state can be also analyzed as the resulting state. Karachay-Balkar (25) kerim šöndü fatima-nɨ kör-üp Kerim now Fatima- see- ‘Kerim is seeing Fatima now.’

tur-a-dɨ --3

Karachay-Balkar (26) kerim asijat-nɨ süj-üp Kerim Asijat- love- ‘Kerim loves Asijat.’

tur-a-dɨ --3

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

407

With stative-process lexical verbs² ((27)) and ((28)), the construction -p tur- has the meaning of the lexical state, but not that of the lexical process; cf. olturadɨ ‘is sitting / is sitting down’. Karachay-Balkar (27) kerim oltur-up tur-a-dɨ Kerim sit- --3 ‘Kerim is sitting / *is sitting down.’ Karachay-Balkar (28) kerim zat-ɨp tur-a-dɨ Kerim lie- --3 ‘Kerim is lying / *is lying down.’ With telic-stative lexical verbs ((29)) and ((30)), the construction -p tur- has the meaning of the resulting state. In contrast to telic and punctual verbs, telic-stative verbs have the resulting phase in their event structure outside the construction -p tur-, since this phase can also be expressed by a perfective (but not imperfective) form of a lexical verb; cf. batxandɨ ‘drowned (for time X) / drowned (in time X) / was under the water (for time X)’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Karachay-Balkar (29) kerim bat-ɨp tur-a-dɨ Kerim drown- --3 ‘Kerim has drowned.’ Karachay-Balkar (30) alim ustaz bol-up Alim teacher become- ‘Alim has become a teacher.’

tur-a-dɨ --3

With (weak) ingressive-process lexical verbs, the construction -p turhas the meaning of the lexical process, as in (31), and is synonymous with the lexical verb in its imperfective form. Since the verbs of this actional class have an ingressive meaning, the lexical process can be also analyzed as the resulting process.

² Mainly posture verbs belong to these two actional classes.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

408

 

Karachay-Balkar (31) alim kül-üp tur-a-dɨ Alim laugh- --3 ‘Alim is laughing (now).’ With two-endpoint telic ³ lexical verbs, the construction -p tur- has the meaning of the resulting state, as in (32), and cannot have the meaning of the lexical process. Karachay-Balkar (32) üj zan-ɨp tur-a-dɨ house burn- --3 ‘The house has burnt down / *is burning.’ With multiplicative ((33)) and multiplicative-process ((34)) lexical verbs, the construction -p tur- has the meaning of the resulting state of a single quantum of the multiplicative process, as in (33a); the meaning of the ongoing lexical multiplicative process is not possible, as in (33b). Karachay-Balkar

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(33) asijat qol-u bla bulʁa-p tur-a-dɨ Asijat hand-3 with wave- --3 i. ‘Asijat has waved with her hand.’ ii. *‘Asijat is waving with her hand.’ Karachay-Balkar (34) kerim top-nu at-ɨp tur-a-dɨ Kerim ball- throw- --3 i. ‘Kerim has thrown the ball.’ ii. *‘Kerim is throwing the ball.’ With stative lexical verbs, the construction -p tur- has the meaning of the lexical state and is synonymous with the lexical verb in its imperfective form, as in (35). Karachay-Balkar (35) kerim nalčik-ta zaša-p Kerim Nalchik- live- ‘Kerim lives in Nalchik.’

tur-a-dɨ --3

³ Mainly, movement verbs belong to this actional class.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

409

Table 15.1 Meaning of the construction -p tur- in Karachay-Balkar depending on the actional class of the lexical verb actional class of the lexical verb

meaning of the construction -p tur-

strong telic

resulting state

punctual

resulting state

weak telic , result verbs

resulting state

weak telic , manner verbs

resulting state, lexical process

(weak) inceptive-stative

lexical state

stative-process lexical verbs ,

lexical state

telic-stative lexical verbs ,

resulting state

(weak) ingressive-process

lexical process

two-endpoint telic

resulting state

multiplicative

resulting state of a single quantum

multiplicative-process

resulting state of a single quantum

stative

lexical state

process

no episodic uses

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

With process lexical verbs, the construction -p tur- has no episodic uses denoting a single event, as in (36b), and may be used only in habitual contexts as in (36a). Karachay-Balkar (36) kerim kitap-nɨ izle-p tur-a-dɨ Kerim book- look_for- --3 a. ‘Kerim (regularly) looks for the book.’ b. *‘Kerim is looking for the book (now).’ Table 15.1 summarizes the interaction of the delexicalized verb tur- ‘< stand’ in Karachay-Balkar with the actionality of a lexical verb. The construction -p tur- in Karachay-Balkar has the nuclear meaning of the resultative state that follows the endpoint of a process. This meaning interacts with the actional structure of the lexical verb in the following way. If the actional structure of the lexical verbs contains such an endpoint, the subsequent stative phase is expressed by the construction -p tur- both if this phase is introduced by this construction and if it is the same as its lexical durational phase. If the actional structure of the verb contains no entry into a state, the construction -p tur- can express a lexical state (keeping the general meaning of a state) and a resulting process (keeping the general resultative meaning). The endpoint of the manner verbs is ignored.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

410

 

15.2.3 V–V complexes with il- ‘< take’ in Anatri Chuvash V–V complexes in Chuvash involve the contextual -se converb that is functionally analogous to the -p converb in other Turkic languages. il- ‘< take’ is the most productive delexicalized verb. The construction -sa il- has two groups of meaning. The first, of low productivity in Chuvash, is the valency-changing self-benefactive meaning, as in (44). I do not discuss such uses of this construction. Anatri Chuvash (37) aršɨn ača kanra-na tɨt-sa man child rope- grasp- ‘The boy took away the rope.’

il-č-ě --3

The second group of meanings of the construction -sa il- in Chuvash is related to limiting the event in time; this group is absolutely productive. There are three meanings of this group: (a) limitative: an event is restricted in time, as in (38), (b) canceled result: an event reaches its endpoint, but this result is canceled, as in (39), (c) semelfactive: one quantum of an event takes place, as in (40).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Anatri Chuvash (38) pětěr jurla-sa il-č-ě Peter sing- --3 ‘Peter sang for a while.’ Anatri Chuvash (39) šʲuta sün-se il-č-ě light go_out- --3 ‘The light went out and lit again.’ Anatri Chuvash (40) šɨv tumla-sa il-č-ě water drip- --3 ‘The water dripped once.’ The choice of a meaning is defined by actionality. With strong telic ((41)) and punctual ((49)) lexical verbs, the construction -sa il- always has the meaning of canceled result.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

411

Anatri Chuvash (41) pětěr tavrǎn-sa il-č-ě Peter return- --3 ‘Peter returned and left again.’ Anatri Chuvash (42) ača šʲuxal-sa il-č-ě child get_lost- --3 ‘The child got lost and was found again.’ As in Karachay-Balkar, weak telic lexical verbs fall into two subclasses of lexical verbs, corresponding to result vs. manner verbs in terms of Rappaport Hovav and Levin (1998). With result verbs, the construction -sa il- has the meaning of canceled result, as in (43). With manner verbs, it has the limitative meaning, as in (44). Anatri Chuvash (43) ivan kil-se il-č-ě Ivan come- --3 ‘Ivan came and left again.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Anatri Chuvash (44) xer kofta-na šɨx-sa il-č-ě girl jersey- knit- --3 ‘The girl knitted the jersey for a while.’ With inceptive-stative ((45)), weak inceptive-stative , ((46)), and stative ((47)), lexical verbs, the construction -sa il- has the limitative meaning restricting in time the lexical state. Anatri Chuvash (45) ača pravila-na ǎnlan-sa il-č-ě child rule- understand- --3 ‘The child understood the rule for a while.’ Anatri Chuvash (46) ivan čirle-se il-č-ě Ivan be_ill- --3 ‘Ivan was ill for a while.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

412

 

Anatri Chuvash (47) ivan šɨvǎr-sa il-č-ě Ivan sleep- --3 ‘Ivan slept for a while.’ With (weak ingressive-process ((48)), and process ((49)), lexical verbs, the construction -sa il- has the limitative meaning restricting in time the lexical process. Anatri Chuvash (48) šɨv věrě-se il-č-ě water boil- --3 ‘The water boiled for a while.’ Anatri Chuvash (49) kupǎsta šer-se il-č-e cabbage rot- --3 ‘The cabbage rotted for a while.’ With stative-process lexical verbs⁴ ((50)), the construction -sa il- has the limitative meaning restricting in time the lexical state, but not the limitative meaning restricting in time the lexical process; cf. vɨrtatʲ ‘is lying / is lying down’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Anatri Chuvash (50) pětěr vɨrt-sa il-č-ě Peter lie- --3 ‘Peter was lying for a while.’ With multiplicative ((51)) lexical verbs, the construction -sa ilhas the limitative meaning restricting the lexical multiplicative process, as in (51a), or the meaning of a single quantum of the multiplicative process, as in (51b). Anatri Chuvash (51) ivan avtamat-ran per-se il-č-ě Ivan tommy_gun- shoot- --3 i. ‘Ivan shot for a while with the tommy gun.’ ii. ‘Ivan shot once with the tommy gun.’ Table 15.2 summarizes the interaction of the delexicalized verb il- ‘< take’ in Anatri Chuvash with the actionality of a lexical verb. This construction has the ⁴ Mainly posture verbs belong to these two actional classes. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

413

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Table 15.2 Meaning of the construction -sa il- in Anatri Chuvash depending on the actional class of the lexical verb actional class of the lexical verb

meaning of the construction -sa il-

strong telic

canceled result

punctual

canceled result

weak telic , result verbs

canceled result

weak telic , manner verbs

limitative

inceptive-stative

limitative

weak inceptive-stative

limitative

stative

limitative

weak ingressive-process

limitative

process

limitative

stative-process lexical verbs

limitative (restricting lexical state)

multiplicative

limitative quantum

limitative meaning with atelic (including inceptive / ingressive atelic) verbs and the meaning of canceled result with telic verbs; the semelfactive meaning is one of the possibilities for the multiplicative verbs. The endpoint of manner verbs is ignored, so that in interaction with this construction they behave as atelic verbs. Both Karachay-Balkar and Anatri Chuvash case studies show that the specific meaning of an aspectual V–V complex may be predicted from the actional class of the lexical verb. Although the meanings of the constructions in focus are significantly different, the main difference is found in the behavior of telic vs. atelic verbs, i.e. verbs with the actional structure finishing by an endpoint vs. verbs with the actional structure finishing by a continuous phase. The only exception is the same in both cases: weak telic verbs do not from a uniform actional class but fall into two subclasses—weak telic manner verbs and weak telic result verbs. In both cases weak telic manner verbs behave like atelic verbs: the endpoint is ignored.

15.3 A case study of a system of Turkic V–V complexes and its lexical restrictions: Tubalar Altai 15.3.1 An overview of the Tubalar Altai V–V complex system In this section I provide a case study of a whole system of delexicalized verbs in V–V complexes, their lexical restrictions and meanings. The most productive delexicalized verbs in Tubalar Altai are described here. Still, not all the constructions in focus are as productive as the -p tur- construction

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

414

 

in Karachay-Balkar and the -sa il- construction in Anatri Chuvash. As a consequence, we are sometimes dealing with an interaction of the meaning of a construction with the meaning of a lexical verb, but sometimes we are dealing with lexical restrictions. Many constructions are possible only with some lexical verbs, but this class of lexical verbs may be defined in terms of its semantic features. With aspectual delexicalized verbs, again as in the Karachay-Balkar and Anatri Chuvash constructions discussed earlier, the main clue to the distribution of meanings and to the lexical restrictions is the actional class of the lexical verb. With auxiliaries that have other meanings, the restrictions may be defined by other lexical features. Baskakov (1966: 46–7) provides a list of Tubalar Altai delexicalized verbs in V–V complexes. Still, based on my field data I have worked out another list of the delexicalized verbs that are used regularly. The following Tubalar Altai delexicalized verbs are discussed in this section: • • • • • •

al- ‘< take’ (limitative and self-benefactive -p al- constructions) ber- ‘< give’ (inchoative and -benefactive -p ber- constructions) kör- ‘< see’ (conative -p kör- construction) tur- ‘< stand’ (imperfective -p tur- construction) oŋo- ‘< know’ (capacitive -p oŋo- construction) ij- ‘< send’ (punctual -p ij- construction)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

15.3.2 al- ‘< take’: limitative and self-benefactive constructions The delexicalized verb al- ‘< take’ in Tubalar Altai (as well as in other Turkic languages) is involved in two different V–V complexes. One is the limitative construction with the aspectual meaning of limiting the event in time. This grammaticalization path may be explained with the metaphor of taking some part in an event. The other is the self-benefactive construction with a valencychanging meaning. This meaning is related to the self-oriented lexical meaning of the verb ‘take’ as well. The limitative construction -p al- ((52)) is not very productive. It is regular with atelic verbs, i.e. with the verbs of the following actional classes: , , . (Weak) telic lexical verbs normally are not compatible with the limitative construction -p al; cf. (53). The limitative construction is occasionally attested with such verbs only if they have an incremental cumulative Theme (in the terminology of Dowty 1991 and Krifka 1992), as in (54). Tubalar Altai (52) wasʲa qattɨr-ɨp al-dɨ Vasja laugh- - ‘Vasja laughed for a while.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

415

Tubalar Altai (53) wasʲa ežik ač-ɨp al-dɨ Vasja door open- - *‘Vasja opened the door for a while.’ Tubalar Altai (54) maša ajaq-qožɨq nün-üp al-dɨ Masha cup-spoon wash- - ‘Masha washed the dishes for a while.’ The self-benefactive -p al- construction introduces a new participant into the event structure of the lexical verb. This participant has the semantic role of a Beneficiary and is coreferential with the subject, as in (55). Noteworthy is the fact that the Beneficiary introduced by the V–V complex on the semantic level is not obligatorily the same as the Beneficiary that may be optionally expressed by an NP in Dative; cf. (64a,b). Tubalar Altai (55) maša učuq-tɨ üz-üp al-dɨ Masha thread- tear- - ‘Masha tore the thread for herself.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Tubalar Altai (56) a. wasʲa u ed-ip al-dɨ Vasja house make- - Vasja built a house for himself.’ b. wasʲa ana-zɨ-na u ed-ip al-dɨ Vasja mother-3- house make- - ‘Vasja built a house for his mother with a profit for himself.’ Introducing a Beneficiary influences the actional structure of the lexical verb as well. An atelic event frequently is reinterpreted as a telic one, since the ‘profit’ of the Beneficiary presupposes a kind of resulting state. For instance, the sentences (57a) and (58a) with the self-benefactive -p al- construction are telic, although the lexical verbs themselves are stative and therefore atelic ((57bi, 58bi)) and have no endpoint in their actional structure that is present in (57a) and (58a); cf. (57bii, 58bii). However, the effect of a reinterpretation of an atelic event as a telic one is not obligatory; cf. (59), which is parallel to (57a) and (58a), but still is atelic.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

416

 

Tubalar Altai (57) a. wasʲa oč-up al-dɨ Vasja sit- - ‘Vasja got something for himself via sitting.’ {e.g. one sits in offices of different institutions} b. wasʲa oč-tɨ Vasja sit- i. ‘Vasja was sitting.’ ii. *‘Vasja got something via sitting.’ Tubalar Altai (58) a. wasʲa uxta-p al-dɨ Vasja sleep- - ‘Vasja slept his fill.’ b. wasʲa uxta-dɨ Vasja sleep- i. ‘Vasja slept.’ ii. *‘Vasja slept his fill.’ Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(59) wasʲa ište-p al-dɨ Vasja work- - ‘Vasja worked for himself.’ Generally, there is a clear lexical restriction for the self-benefactive -p al- construction: it is grammatical with volitive / controlled lexical verbs. Otherwise the construction is always interpreted by the speakers as a pun; cf. (60). Tubalar Altai (60) toš qajɨl-ɨp al-dɨ ice melt- - ‘The ice melted on purpose for itself.’ However, there is an exceptional group of nonvolitive / uncontrolled lexical verbs that are used with the delexicalized verb al-, so that the V – V complex as a whole has a general telicizing meaning, as in (61) and (62). One may assume that these are cases of a lexicalization of the self-benefactive construction. Tubalar Altai (61) wasʲa maša-nɨ süj-üp Vasja Masha- love- ‘Vasja fell in love with Masha.’

al-dɨ -

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

417

Tubalar Altai (62) wasʲa tül-üp al-dɨ Vasja fall- - ‘Vasja fell down.’

15.3.3 ber- ‘< give’: inchoative and benefactive constructions The delexicalized verb ber- ‘< give’ in Tubalar Altai is also involved in two different V–V complexes. One is the inchoative construction with an aspectual meaning; the other is the benefactive construction with a valency-changing meaning. In the inchoative construction, ber- functions as an actional delexicalized verb that introduces a beginning point ((63a)) and a preliminary process ((63b)) of the event expressed by the lexical verb. In the case of telic verbs, the beginning point and preliminary process are primarily absent in the actional structure and that of the lexical verb. Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(63) a. wasʲa pičik šʲi-ip ber-di Vasja paper write- - ‘Vasja began writing a letter.’ b. wasʲa pičik šʲi-ip ber-bit Vasja paper write- - ‘Vasja is beginning writing a letter.’ The inchoative -p ber- is possible with all the actional classes of lexical verbs, except for punctual lexical verbs. (64) shows that this delexicalized verb is possible with a punctual verb only in a benefactive V–V complex, as in (64i), not in an inchoative V–V complex, as in (64ii). This exception may be easily explained: the actional structure of punctual verbs contains no process that could have the preliminary beginning point and process. Tubalar Altai (64) wasʲa klüč-ti tab-ɨp ber-di Vasja key- find- - i. ‘Vasja found the key for someone.’ ii. *‘Vasja started finding the key.’ With (weak) inceptive-stative ((65)) and ingressive-process ((66)) lexical verbs, the construction -p ber- refers to the culmination point that is already present in the actional structure of the lexical verb. For instance, (65a) and

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

418

 

(66a) express a beginning point of a state or process; (65b) and (66b) as well may express this beginning point; cf. (65bi) and (66bi). Tubalar Altai (65) a. wasʲa maša-nɨ süj-üp Vasja Masha- love- ‘Vasja fell in love with Masha.’

ber-di -

b. wasʲa maša-nɨ süj-di Vasja Masha- love- i. ‘Vasja fell in love with Masha.’ ii. ‘Vasja loved Masha.’ Tubalar Altai (66) a. wasʲa qattɨr-ɨp ber-di Vasja laugh- - ‘Vasja began to laugh.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. wasʲa qattɨr-dɨ Vasja laugh- i. ‘Vasja began to laugh.’ ii. ‘Vasja laughed.’ The benefactive -p ber- construction introduces a new participant into the event structure of the lexical verb. This participant has the semantic role of a Beneficiary and is not coreferential with the subject, as in (67). Tubalar Altai (67) wasʲa qožɨq-ti sal-ɨp ber-di Vasja spoon- put- - ‘Vasja put the spoon down for someone.’ The main semantic lexical restriction for the benefactive construction -p ber- is the same as the restriction for the self-benefactive construction -p al-: it is grammatical only with volitive / controlled lexical verbs. (68) and (69) illustrate that a combination of a nonvolitive verb with this construction is not accepted by speakers. Tubalar Altai (68) wasʲa maša-nɨ süj-üp ber-di Vasja Masha- love- - *‘Vasja loved Masha for someone.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

419

Tubalar Altai (69) toš qajɨl-ɨp ber-di ice melt- give- *‘The ice melted for someone.’ In addition to the semantic restriction on volitivity, the benefactive construction -p ber- has pragmatic limitations that are not so strict and depend on the context. Speakers tend to invent a possible interpretation for the event expressed by the lexical verb, so that it would have a beneficiary different from the subject. (70) and (71) provide examples of benefactive constructions accepted by speakers in spite of being pragmatically bizarre. Still, for some lexical verbs such interpretations could not be found; so, e.g. (72) was not accepted. Tubalar Altai (70) wasʲa uxta-p ber-di Vasja sleep- - ‘Vasja slept for someone.’ {for example, Vasja did not want to disturb him} Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(71) wasʲa töš-üp ber-di Vasja cough- - Vasja coughed for someone.’ {e.g. the doctor asked him to cough on purpose} Tubalar Altai (72) wasʲa te-p ber-di Vasja eat- - ?? ‘Vasja ate for someone.’

15.3.4 kör- ‘< see’: conative construction The conative construction -p kör- is used to express an attempt to complete the effectuation of an event expressed by the lexical verb. What exactly is done during such an attempt is defined neither by the semantics of the lexical verb nor by the semantics of the construction. With telic lexical verbs the conative construction is used to express an attempt to reach the endpoint of the event. (73)–(75) are typical examples of this construction.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

420

 

Tubalar Altai (73) wasʲa ežik ač-ɨp kör-di Vasja door open- - ‘Vasja tried to open the door.’ Tubalar Altai (74) wasʲa qožɨq-tɨ stol-tɨŋ al-ɨp kör-di Vasja spoon- table- take- - ‘Vasja attempted to take the spoon from the table.’ {e.g. he is ill} Tubalar Altai (75) maša čaška nün-üp kör-di Masha cup wash- - ‘Masha tried to wash the cup.’ With inceptive-stative ((76)) and ingressive-process ((77)) lexical verbs, the construction -p kör- is used to express an attempt to reach the starting point of the state or the process. Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(76) wasʲa maša-nɨ kör-üp Vasja Masha- see- ‘Vasja tried to see Masha.’

kör-di -

Tubalar Altai (77) it ür-üp kör-di dog bark- - ‘The dog tried to bark’. {e.g. a puppy} With atelic lexical verbs —i.e. stative ((78)) and process ((79)) verbs—the construction -p kör- is used to express the state or process that takes place for a relatively short time and pragmatically has a low estimate. Tubalar Altai (78) wasʲa kebezen-de tʲad-ɨp Vasja Kebezen- live- ‘Vasja tried to live in Kebezen.’

kör-di -

Tubalar Altai (79) wasʲa ište-p kör-di Vasja work- - ‘Vasja tried to work.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

421

As a strong tendency, the construction -p kör- presupposes that the participant expressed by the subject controls the event. Therefore, if a lexical verb has a volitive / controlled reading (even marginal) and may be used to express a controlled event, the construction is not acceptable with a subject that has no controlling potential, as in (80b). A controlled reading is applied where possible, as in (80a) and (81). Tubalar Altai (80) a. wasʲa tül-üp kör-di Vasja fall- - i. ‘Vasja tried to fall down (on purpose).’ ii. ?? ‘Vasja almost fell down.’ b.

čaška tül-üp kör-di cup fall- - ‘The cup almost fell down.’ ??

Tubalar Altai (81) wasʲa ül-üp kör-di Vasja die- - i. ‘Vasja tried to die (on purpose).’ {e.g. he attempted suicide} ii. *‘Vasja almost died.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Still, if a lexical verb always expresses a nonvolitive / uncontrolled event, the construction -p kör- has a proximative meaning: the event expressed by the lexical verb is close to its endpoint, but does not reach it; cf. (82) and (83). Tubalar Altai (82) toš qajɨl-ɨp kör-di ice melt- - ‘The ice almost melted.’ Tubalar Altai (83) čaj qajna-p kör-di tea boil- - ‘The tea almost boiled.’ With ‘ingestive’ lexical verbs (‘eat’, ‘drink’, ‘smoke’, ‘have sex’), the construction -p kör- has both the standard meaning of an attempt, as in (84ii), and a lexicalized meaning of tasting, as in (84i).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

422

 

Tubalar Altai (84) pala jablaka te-p kör-di child apple eat- - i. ‘The child tasted the apple.’ ii. ‘The child tried eating the apple.’

15.3.5 tur- ‘< stand’: imperfective construction In contrast to the Karachay-Balkar V–V complex discussed in Section 15.2.1, in Tubalar Altai the construction -p tur- has no regular resultative meaning. Its meaning is imperfective, following the standard terms of the viewpoint aspect: it presents an event as ongoing and refers to its internal temporal structure (cf. Comrie 1976). For example, (85) has an imperfective meaning, as in (85i), while a resultative meaning is not acceptable, as in (85ii). A resultative meaning is attested only with a couple of telic verbs as a marginal secondary meaning; for example, (86ii) is felicitous as well as (86i). Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(85) wasʲa eski u-nɨ od-ɨp tur-dɨ Vasja old house- break- - i. ‘(When I came,) Vasja was breaking the old house.’ ii. * ‘(When I came,) Vasja had the old house broken.’ Tubalar Altai (86) wasʲa ajaq-tɨ al-ɨp tur-dɨ Vasja cup- take- - i. ‘(When I came,) Vasja was taking the cup.’ ii. ‘(When I came,) Vasja had already taken the cup.’ There are no lexical restrictions on the imperfective construction. With lexical verbs of actional classes with a single durated phase, this phase is presented as ongoing. While with telic verbs it is the phase that precedes the natural endpoint, as in (85) and (86i), with inceptive stative verbs it is the phase that follows the beginning point; cf. (87). Tubalar Altai (87) wasʲa maša-nɨ kör-üp tur-dɨ Vasja Masha- see- - ‘(When I came,) Vasja was seeing Masha.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

423

With stative-process lexical verbs , the construction -p tur- can refer only to the stative phase, as in (88i), not to the phase of the process, as in (88ii). Tubalar Altai (88) wasʲa jablaka tut-up tur-dɨ Vasja apple hold- - i. ‘(When I came,) Vasja was holding an apple.’ ii. * (When I came,) Vasja was catching an apple.’ With atelic lexical verbs—both stative ((89)) and process ((90))— the construction -p tur- has an additional semantic component of an extended duration. (89a) and (90a) contain no information about duration of the expressed events, while (89b) and (90b) present the same events as persisting longer than expected. Tubalar Altai (89) a. wasʲa kebezen-de tʲat-tɨ Vasja Kebezen- live- ‘Vasja lived in Kebezen.’ b. wasʲa kebezen-de tʲad-ɨp tur-dɨ Vasja Kebezen- live- - ‘Vasja lived in Kebezen for a long time.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Tubalar Altai (90) a. wasʲa qattɨr-dɨ Vasja laugh- ‘Vasja laughed.’ b. wasʲa qattɨr-ɨp tur-dɨ Vasja laugh- - ‘Vasja laughed for a long time.’

15.3.6 oŋo- ‘< know’: capacitive construction The capacitive construction -p oŋo-⁵ is used to express the meaning of capacity to fulfill an event expressed by the lexical verb. The meaning of this construction is related to the semantic domain of modality, which is untypical for other Turkic V–V complexes.

⁵ A synonymous construction with the same delexicalized verb oŋo- and the infinitive lexical verb is also present in Tubalar Altai. I do not discuss it, since in general infinitive constructions in Turkic languages have a different structure from V–V complexes. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

424

 

Shluinsky (2009b) has shown that crosslinguistically inherent vs. acquired capacitives are distinguished. The construction -p oŋo- is used only for the acquired capacitive, i.e. it may express only a capacity that an animate Agent has acquired during his life. (91)–(93) are typical examples of the capacitive construction. Tubalar Altai (91) maša čočko-nɨ taqšɨn Masha piglet- outdoors Masha can drive out a piglet.’

čer-ip drive_out-

oŋo-bɨt -

Tubalar Altai (92) it ür-üp oŋo-bɨt dog bark- - ‘The dog can bark.’ {e.g. a puppy} Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(93) wasʲa klüč-ti tab-ɨp oŋo-bɨt Vasja key- find- - ‘Vasja can find the key.’ {i.e. he knows where to look for it} Lexical restrictions on the capacitive construction require the subject to be an Agent (and, therefore, animate). Therefore, lexical verbs that express states, as in (94) and (95), and patientive atelic or telic processes, as in (96) and (97), are ungrammatical with the construction -p oŋo-. Tubalar Altai (94) *wasʲa tʲad-ɨp oŋo-bɨt Vasja live- - intended ‘Vasja can live.’ Tubalar Altai (95) *wasʲa maša-nɨ kör-üp oŋo-bɨt Vasja Masha- see- - intended ‘Vasja can see Masha.’ Tubalar Altai (96) *suu qajna-p oŋo-bɨt water boil- - intended ‘The water can boil.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

425

Tubalar Altai (97) *wasʲa tül-üp oŋo-bɨt Vasja fall- - intended ‘Vasja can fall down.’

15.3.7 ij- ‘< send’: punctual construction The punctual construction -p ij- fixes and emphasizes a culmination point in the actional structure of the lexical verb. This culmination point may be unexpected, may happen earlier than expected, or may take more effort than expected. Example (98) shows a number of possible interpretations of the punctual construction that depend on pragmatics. Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(98) maša učuq-tɨ üz-üp ij-di Masha thread- tear- - ‘Masha suddenly / already / hardly tore the thread.’ The V–V complex with the construction -p ij- as a whole behaves like a punctual verb. The construction -p ij- has no episodic uses in Present: (99a) has a Past form like (98) and expresses a single culmination point, while (99b) has a Present form and may have only a habitual interpretation, as in (99bi), but no durated episodic interpretation, as in (99bii). That the Past form of the construction -p ij- cannot express a durated phase may by shown via the Vendlerian test: (100a) with an adverbial ‘in time X’ is felicitous, but (100b) with an adverbial ‘for time X’ is ungrammatical. Tubalar Altai (99) a. wasʲa u ed-ip ij-di Vasja house make- - ‘Vasja has hardly / quickly / already built the house.’ b. wasʲa u ed-ip ij-bit Vasja house make- - i. ‘Vasja builds houses quickly / with difficulty.’ ii. *Vasja is building a house quickly / with difficulty now.’ Tubalar Altai (100) a. wasʲa eki minut-xa ajaq-tɨ stol-a sal-ɨp ij-di Vasja two minute- cup- table- put- - ‘Vasja has already / quickly / with difficulty put the cup on the table in two minutes.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

426

  b. *wasʲa eki minut ajaq-tɨ stol-a sal-ɨp ij-di Vasja two minute cup- table- put- - ‘Vasja quickly / hardly put the cup on the table for two minutes.’

There are no lexical restrictions on the punctual construction. The emphasized culmination point may be either taken from the actional structure of a lexical verb or introduced into the actional structure. With (weak) telic lexical verbs, most typically the lexical inceptive point is taken, as in (101). For some verbs of this actional class, a beginning point also may be introduced as a secondary meaning; cf. (102a), which has two meanings with the construction -p ij-, and (102b), which cannot express the beginning point on its own with the same lexical verb. Tubalar Altai (101) wasʲa öl-üp ij-di Vasja die- - ‘Vasja already / suddenly died.’ Tubalar Altai

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(102) a. wasʲa eski u-nɨ od-ɨp ij-di Vasja old house- break- - i. ‘Vasja already / suddenly / hardly broke the old house.’ ii. ‘Vasja already / suddenly started breaking the old house.’ b. wasʲa eski u-nɨ ot-tɨ Vasja old house- break- i. ‘Vasja broke the old house.’ ii. *‘Vasja started breaking the old house.’ With (weak) inceptive-stative lexical verbs, the construction -p ijexpresses the lexical inceptive point, as in (103). Tubalar Altai (103) wasʲa maša-nɨ kör-üp ij-di Vasja Masha- see- - ‘Vasja suddenly / already saw Masha.’ With (weak) ingressive-process lexical verbs, the construction -p ijeither expresses the lexical ingressive point, as in (104i), or introduces a pragmatically defined endpoint, as in (104ii).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

427

Tubalar Altai (104) wasʲa qattɨr-ɨp ij-di Vasja laugh- - i. ‘Vasja already / finally began to laugh.’ ii. ‘Vasja already / finally laughed enough.’ With stative lexical verbs, an inceptive point is introduced, as in (105a); (105b) shows that such a lexical verb on its own cannot have an inceptive reading. Tubalar Altai (105) a. wasʲa kebezen-de tʲad-ɨp ij-di Vasja Kebezen- live- - ‘Vasja suddenly / already settled in Kebezen.’ b. wasʲa kebezen-de tʲat-tɨ Vasja Kebezen- live- i. ‘Vasja lived in Kebezen.’ ii. *‘Vasja settled in Kebezen.’ With process lexical verbs, the construction -p ij- introduces an ingress point, as in (106i), or an endpoint, as in (106ii).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Tubalar Altai (106) wasʲa ište-p ij-di Vasja work- - i. ‘Vasja already / finally started working.’ ii. ‘Vasja already / finally worked enough / did his work.’ With multiplicative lexical verbs, the construction -p ij- introduces an ingress point, as in (107). Tubalar Altai (107) wasʲa töš-üp ij-di Vasja cough- - ‘Vasja suddenly / abruptly had a fit of coughing.’ With stative-process lexical verbs, the construction -p ij- either expresses the lexical inceptive point, as in (108i), or introduces an endpoint, as in (108ii) (in the latter case the stative phase is reanalyzed as a dynamic one).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

428

 

Table 15.3 Restrictions on lexical verbs with V–V complexes in Tubalar Altai V–V complex

restrictions on the lexical verb

al- ‘< take’, limitative

regular only with atelic lexical verbs , , ; occasionally attested with weak telic lexical verbs with an incremental cumulative Theme

al- ‘< take’, selfbenefactive

regular only with volitive lexical verbs; lexicalized with a general telicizing meaning with a closed group of nonvolitive lexical verbs

ber- ‘< give’, inchoative

no restrictions except for punctual lexical verbs

ber- ‘< give’, benefactive

only with volitive lexical verbs; unstable pragmatic restrictions

kör- ‘< see’, conative no restrictions; volitive interpretation if applicable, proximative interpretation otherwise tur- ‘< stand’, imperfective

no restrictions

oŋo- ‘< know’, capacitive

only with Agent subjects

ij- ‘< send’, punctual no restrictions

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Tubalar Altai (108) wasʲa sumka-nɨ tud-up ij-di Vasja bag- hold- - i. ‘Vasja suddenly / hardly / already started holding the bag.’ ii. ‘Vasja already finished holding the bag (the necessary time).’

Table 15.3 summarizes restrictions on different lexical verbs in V–V complexes of Tubalar Altai. As may be seen, some complexes do not have lexical restrictions and may take any lexical verb, but other complexes are strictly restricted by transparent factors. Factors restricting lexical verbs include actional class, volitionality, and agentivity. More strict pragmatic factors also are in play.

15.4 Conclusion In this chapter I have presented three case studies of the interaction of Turkic V–V complexes with lexical verbs. Two of them were case studies of specific highly productive V–V complexes in Karachay-Balkar and Anatri Chuvash, and the third was a more complex case study of the system of V–V complexes in Tubalar Altai. In all the V–V complexes considered, as well as in Turkic V–V complexes

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    

429

in general, the delexicalized verbs are inflected and do not retain their lexical meaning but express a more abstract meaning; the lexical verbs are used in the form of a contextual converb and retain their lexical meaning. A delexicalized verb modifies the actional structure or the argument structure of a lexical verb or, less commonly, has a modal meaning. I have shown that the meaning of a delexicalized verb is not uniform but interacts with the meaning of the lexical verbs. Absolutely productive delexicalized verbs have different interpretations depending on the semantics of the lexical verb; less productive delexicalized verbs also have lexical restrictions, being ungrammatical with some classes of lexical verbs. The grammaticality of a delexicalized verb and/or its specific meaning in most cases may be predicted from the lexical features of the lexical verb. If a delexicalized verb changes the actional structure of the lexical verb, its specific meaning may be predicted from the actional class of the verb; another relevant distinction is the distinction between manner vs. result weak telic verbs. If a delexicalized verb changes the argument structure of the lexical verb or has a modal meaning, the most relevant features are volitivity and agentivity of the lexical verb.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Acknowledgments The research was conducted as part of project No. 16-18-02081 supported by the Russian Science Foundation. I am grateful to all my consultants from collective Turkic field trips in 1999–2006 dealing with Mishar Tatar, Anatri Chuvash, Sagay Khakas, Cherek KarachayBalkar, and Tubalar Altai and to my colleagues with whom I collaborated in studying Turkic data, especially Anna Pazelskaya, Sergei Tatevosov, Ekaterina Lyutikova, Nina Sumbatova, Inna Kulakova, Vera Maltseva, and Vera Kovalskaya. Responsibility for this chapter is entirely my own.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

16 Verb-verb complexes in Central and Eastern Turkic languages Noriko Ohsaki and Fuyuki Ebata

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

16.1 Introduction Turkic languages, spoken in a vast area spanning the Eurasian continent, include more than thirty dialects that are generally classified into several groups according to their linguistic features. Since these groupings roughly correspond to their geographical distribution, each group is sometimes named after its region: the Southwestern (SW) group, which includes Turkish, Azerbaijanian, and Turkmen; the Southeastern (SE) group, which includes Uzbek and Uyghur; the Northwestern (NW) group, which includes Tatar, Bashkir, Kazakh, and Kyrgyz; and the Northeastern (NE) group, which includes Sakha (also known as “Yakut”), Tuvan (also spelled “Tyvan”), and Khakas (cf. Johanson 1998b: 82–3). This chapter discusses verb-verb complexes in Turkic, mainly focusing on Kyrgyz in the NW group and Sakha and Tuvan in the NE group, along with some Turkic languages in the SE and NE groups. When we refer to “Central Turkic,” it includes the languages of the NW and SE groups, while “Eastern Turkic” includes the languages of the NE group. Although these are not common designations, they are useful and convenient terms when discussing verb-verb complexes in Turkic. Turkic languages, which are agglutinating languages with SOV constituent order, have developed rich systems of verb-verb complexes (henceforth VVCs). VVCs, which are defined as two contiguous verbs that form a single predicate, often involve more than two verbs in Turkic, but they have been mostly excluded from the present discussion. The predominant type of VVCs in Central and Eastern Turkic, where few or no morphological compounds are observed (see Section 16.4), corresponds to the syntactic complex verbs of Type 4 in the taxonomy of Japanese V-V complexes (Chapter 2, this volume). VVCs in Turkic include constructions that have mostly been called “auxiliary verb constructions” or “postverbal constructions” in the literature. In auxiliary verb constructions (AVCs), the first verb (V1) of a VVC typically appears in a converbal form and bears an essential part of the meaning of the VVC, while the second verb (V2), which can behave as a finite verb, loses its

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

431

lexical meaning and very often its argument structure as well, supplying additional or modifying meaning such as aspectual, directional, benefactive, and motion manner meaning, as shown by the examples in (1) and (2). Kyrgyz (1)

apa-m-a ayt-ïp mother-.1- tell- ‘I suddenly told my mother.’

ǰiber-di-m --1

Sakha

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(2)

bettex kel-e tart this.way come- ..2 ‘Come here quickly!’

In (1), the V2 ǰiber-¹ ‘to send’ loses its lexical meaning and expresses the quickness, definiteness, decidedness, or completion of the action denoted by the V1 (Kudaybergenov 1987: 229). In (2) also, the V2 tart- ‘to pull’ loses its lexical meaning and expresses the quickness of the action denoted by the V1. Central and Eastern Turkic languages have about twenty or more delexicalized V2s.² In Central Turkic, Kyrgyz uses twenty-one delexicalized V2 verbs, while Uzbek uses twenty-seven (Hojiev 1973, qtd in Ibrahim 1995: 12), and Uyghur uses nineteen (Hahn 1991) or twenty (Ibrahim 1995). In Eastern Turkic, Sakha uses twenty-four V2 verbs, and Tuvan uses twenty-six, depending on the authors’ view.³ The functions of these delexicalized V2s can be roughly classified into four classes: (A) Aspectual, (B) Attitude or Manner of action along with Aspectual meaning, (C) Benefactive, and (D) Modal. (See Table 16.2 in the Appendix for a list of delexicalized V2s, i.e. V2s in AVCs, in Kyrgyz and Sakha and a description of their classifications.) Although the AVCs have been the principal focus of many of the studies on VVCs in Turkic (cf. Yuldašev 1965; Johanson 1998a; Anderson 2004), Turkic languages actually contain various VVCs other than AVCs, which will be closely examined in Section 16.3.

¹ In this chapter a hyphen is used to indicate the basic form of a verb according to the tradition in Turkic linguistics. ² Ščerbak (1981: 135) notes that one Turkic language with less V2 variation in AVCs uses fifteen V2s, while another language with a larger variation uses nearly thirty V2 verbs. ³ In the studies on Sakha AVCs, Xaritonov (1960) lists fifteen verbs and Ubrjatova (1972: 595) lists thirteen verbs as aspectual auxiliaries. Anderson and Harrison (1999: 40) state “a large number of auxiliary verbs (20–25) are used to index a wide range of temporal, aspectual, and modal distinctions” in Tuvan.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

432

    

We will first discuss the grammatical and semantic properties of VVCs in Turkic, taking examples from Sakha and Tuvan (Section 16.2). In Section 16.3, some complex verb predicates other than AVCs are examined to consider whether they can be considered VVCs. We will then discuss which type(s) of the general taxonomy of VVCs Turkic VVCs fit into. Although VVCs in Turkic are rarely qualified as morphological compounds, some suspicious forms still need to be examined (Section 16.4).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

16.2 A sketch of VVCs in Sakha and Tuvan This section gives an overview of Turkic VVCs, primarily AVCs, providing examples from two Eastern Turkic languages, Sakha and Tuvan. Sakha is the easternmost Turkic language, spoken in the Sakha Republic, an area that is isolated from other Turkic dialects. Tuvan is one of the so-called South Siberian Turkic languages, spoken in the Tyva Republic. Sakha and Tuvan verb forms are divided into three groups: finite verbs, participles, and converbs.⁴ A verb stem must take either a finite suffix, a participle suffix, or a converb suffix. In Turkic languages, the number of converb forms varies from language to language, but they have in common simultaneous and sequential converbs, both of which are highly recurrent as the V1 of VVCs. Tuvan has the simultaneous and sequential converb suffixes -a and -(i)p, while Sakha has -a and -(a)n.⁵ In the following subsections, we examine Aspectual, Manner, Modal, and Benefactive V2s in Sakha and Tuvan.

16.2.1 Four productive aspectual V2s common to Turkic Most Turkic languages have four basic V2s frequently used to mark the continuous aspect. In Sakha also, four cognate verbs are extensively used in VVCs: sït- ‘to lie’, tur- ‘to stand’, olor- ‘to sit’, and sïrït- ‘to visit, to exist’. The first two can take both animate and inanimate agents, while the second two mostly co-occur with animate agents. These four V2s of Sakha can be combined with V1s in three types of converbs: simultaneous, sequential, and purposive converbs. These three types of converbs bring different aspectual meanings, which we examine using olor- ‘to sit’ as an example of V2. When the V1 is in simultaneous

⁴ Other Turkic languages have verbal nouns (nominalized verb forms) as well. Sakha and Tuvan also employ a participle for nominalized verbs. ⁵ Note that they have several allomorphs and the Sakha simultaneous converb may be formed through stem-final vowel alternation, for example ahïï ‘eat()’ from ahaa- ‘to eat’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

433

converb form, the VVC has a continuous meaning, as in ahïï olor- [eat() ] ‘to be eating’. When the V1 takes a sequential converb suffix, the VVC has a resultative meaning, as in üör-en olor- [rejoice- ] ‘to be glad’. When the V1 takes a purposive converb suffix, the VVC has a prospective meaning, as in suruy-aarï olor- [write- ] ‘to be going to write’.⁶ All three types of V1s can be negated by using the corresponding negative converb forms. The negative simultaneous and the negative sequential denote ‘without doing’, as in ustu-bakka olor- [put.off-. ] ‘sitting without taking off (one’s clothes)’. The negative purposive V1 is also possible, but combines only with the V2 tur- ‘to stand’: kiir-im-eeri tur- [enter-- ] ‘standing without entering’. Tuvan also has four continuous aspect V2s: čït- ‘to lie’, tur- ‘to stand’, olur- ‘to sit’, and čor(u)-, ‘to go’. Unlike those of Sakha, these four verbs are combined only with sequential V1s: ag-ïp čït- [run- ] ‘[river] to be running’, and there is only one negative converb form: ažïlda-vayn tur- [work- ] ‘being without working’ (ažïlda- ‘to work’). Although both Sakha and Tuvan use these continuous aspect V2s, they do not employ the V2s in similar ways. Tuvan continuous V2s are almost always used when the predicate has a continuous meaning, as in (3).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(3)

doŋ-main olur=sen=be freeze- () =2 = ‘Don’t you feel cold now?’

On the other hand, Sakha does not necessarily use a VVC when the inflected lexical verb itself has a continuous connotation. Example (4) below depicts the same situation as (3). (4)

toŋ-mok-kun=duo freeze-:-2 = ‘Don’t you feel cold now?’

16.2.2 Less productive aspectual V2 In addition to the four abovementioned V2s, Sakha has another progressive V2 is‘to go, to approach’ for the continuous aspect: küühür-en is- [get.strong- ] ‘to be getting stronger’. Sakha and Tuvan employ VVCs also for inceptive, completive, and attemptive (or incomplete) aspects. ⁶ The prospective aspect may also be expressed with another V2 gïn- ‘to do’, as in suruy-aarï gïn[write- do] ‘to be going to write’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

    

434

Both Sakha and Tuvan have several inceptive V2s. Sakha has bar- ‘to go’ and kel- ‘to come’, while Tuvan has kir- ‘to enter’, ün- ‘to exit, to ascend’, and ber- ‘to give’. Sakha xarağ-a sïrdaa-n kel-le (5) emeexsin old.woman eye-.3 become.bright- -.3 ‘The eyes of the old woman started to become bright.’ Tuvan avtobus šimče-y (6) am=na now = bus move- ‘At last the bus started to run.’

ber-di -

Both languages also have several completive V2s. Sakha has xaal- ‘to remain’, kebis- ‘to quit, to throw away’, and uur- ‘to put’, while Tuvan has kal- ‘to remain’, kag- ‘to put’, and bar- ‘to go’. Sakha

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(7)

muosta-ğa sït-an utuy-an xaal-bïp-pïn floor- lie- sleep- --1 ‘I (unintentionally) slept, lying on the floor.’

Tuvan samdarar-ïp (8) ïnčangaš asfal’tï-žï therefore asphalt-.3 get.old- ‘Therefore, the asphalt got old.’

kal-gan -

Although the two languages have completive V2s with similar functions, Tuvan seems to use VVCs more frequently. Sakha often manages with an inflected verb, as in (9), where Tuvan employs a VVC, as in (10). Sakha (9)

sïlay-dï-ŋ=duo be.tired--2= ‘Have you gotten tired?’

Tuvan (10)

šïla-p kal-dï-ŋ=be be.tired- --2= ‘Have you gotten tired?’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

435

Sakha has two incomplete aspect V2s: tüs- ‘to fall’ and sïs- ‘to fail’. They refer to an attempted and imperfective action. We will examine the V2 sïs- ‘to fail’ in Section 16.2.6.

16.2.3 Manner V2 Turkic VVCs may express the manner of action. In Sakha and Tuvan, there are several manner V2s whose functions range from suddenness and quickness to completeness to unexpectedness. Sakha has such Manner V2s as tüs- ‘to fall down’, oğus- ‘to hit’, and tart- ‘to pull’, whereas Tuvan has šap- ‘to gallop, to attack’, düš- ‘to fall down’, sok- ‘to hit’, tïrt- ‘to pull’, xalï- ‘to jump’, and xon- ‘to lodge’. In the following examples, the V2 oğus- ‘to hit’ in (11) expresses suddenness and the V2 düš- ‘to fall down’ in (12) completeness. Sakha (11)

ïksal-ïnan taŋn-a oğus-tu-m hurry- wear- --1 ‘I got dressed in a hurry.’

Tuvan

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

sagïž-im-ga bodal mind-.1- thought ‘An idea came to my mind.’

kir-e enter-

düš-tü -

16.2.4 Modal V2 VVCs in Turkic include modality expressions such as tentativeness, ability, and possibility. Among them, tentative modality is expressed by the V2 kör- ‘see’ in both Sakha and Tuvan.⁷ These two languages, however, display a sharp contrast in the expression of ability, diverging in the choice and position of the verbs in VVC expressing ability. In Tuvan, the V2 bil- ‘know’ is used for denoting ability, as in (13).

⁷ Interestingly, other Turkic languages also use the verb ‘to see, look, or watch’ for tentative modality, but not all of them are cognate verbs: Kyrghyz and Kazakh use kör- but Uyghur uses bakand Tatar uses kara-.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

436

    

Tuvan (13)

ešt-ip bil-ir=sen=be swim- -=2= ‘Can you swim?’

In Sakha, on the other hand, sataa- ‘to be able’ appears in the V1 position in a converb form and a lexical verb follows it, as shown in (14) Sakha (14)

saxalïï sataa-n saŋar-ar in.Sakha .- speak-.3 ‘(S)he can speak Sakha.’

The placement of the verb ‘can’ in the V1 position is found only in Sakha among Turkic languages; if sataa- ‘be able to’ is placed in the V2 position, as in (15), it comes to indicate a tentative action, not ability. (15)

at-ï-n ayanna-t-a horse-.3- make.trip-- ‘He tried to let his horse run.’

sataa-bït .-.3

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

16.2.5 Benefactive V2 Turkic languages have V2s that mark benefactive notions. Sakha employs V2 ïl‘to take’ to indicate self-benefactive and bier- ‘to give’ for other-benefactive verbs.⁸ Sakha (16)

isti-bik-ki-n bütünnüü suru-n-an hear--2- wholly write-- ‘Write down all that you hear!’

ïl ..2

(17)

büötür tug-u kör-büt-ü-n barï-tï-n Peter what- see--3- all-.3- kepsee-n bier-bit-e speak- --3 ‘Peter told me in detail all that he saw.’

sihilii in.detail

⁸ Anderson (2004) uses the terms “subject version” and “object version” for self- and otherbenefactives.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

437

Tuvan also uses the V2 al- ‘to take’ and ber- ‘to give’ for self- and otherbenefactives. The two V2s are quite often used along with the V1 that has a transportational meaning, such as ‘to give’, ‘to bring’, ‘to sell’, ‘to tell’, and ‘to ask’. Tuvan (18)

aŋaa nom-nu damčïd-ïp 3: book- convey- ‘Please give the book to him/her.’

ber-iŋer -.2

16.2.6 Participle as V1 Although investigations on the auxiliary verbs in Turkic languages tend to focus on the “converb plus auxiliary” structure (e.g. Anderson 2004), there are a considerable number of examples involving the “participle plus auxiliary” structure as well. For example, Sakha uses buol- ‘to become’ for the inceptive aspect, and Tuvan employs appar- ‘to take away’ for necessive modality (Anderson and Harrison 1999: 63). Sakha

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(19)

nehilieg-im žon-un üks-ü-n village-.1 people-.3 most.part-.3- bil-er buol-lu-m know-. --1 ‘I got to know most of the people in my village.’

Tuvan (20)

bis čoru-ur 1 go-. ‘We had to leave.’

appar-gan =bis .- =1

Tuvan uses čas- ‘fail’ for the incompletive aspect, that is, for expressing ‘to nearly do something’. (21)

aŋdar-l-ïp ka-ar turn.over-- put-. ‘I nearly fell to the ground.’

čas-tï-m --1

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

438

    

Sakha employs sïs- ‘fail’, which is a cognate of the Tuvan čas-, to express a similar incompletive aspect.⁹ (22)

deriebine uok-ka bïlžat-a sïs-pït-a village fire- be.lost- --3 ‘The village was nearly lost in the fire.’

Comparison of the two examples in (21) and (22) suggests that a participle (in Tuvan) and a converb (in Sakha) carry out a similar function to a V1 in the respective VVCs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

16.2.7 Integration between the two verbs Sakha and Tuvan exhibit a notable difference regarding the integration between the two verbs in VVCs. In Sakha VVCs, the two elements are merely juxtaposed, where there are no phonetic or phonemic signs that indicate they serve as a single predicate. In Tuvan, in contrast, there are several phonological features that indicate the two components configure a VVC. First, metathesis occurs between the two verbs. For instance, in the VVC tön-üp kal- (finish- ) ‘to finish’, the final /p/ of the V1 and the initial /k/ of the V2 usually switch and yield the surface form tön-ük pal-. Second, the final /p/ of the V1 is quite often fricativized when followed by a vowel-initial V2: biž-iv aayn (write- ..1) ‘Let me write!’ (< biž-ip al-ïyn). Third, the converb suffix is completely omitted in certain combinations of two verbs: kel čït- (come ) ‘to be coming’. In Section 16.4, we will undertake a closer examination of whether the two component verbs of VVC are qualified as morphological compounds.

16.2.8 Summary This section has provided an overview of Turkic VVCs, based on linguistic materials from two Eastern Turkic languages, Sakha and Tuvan. The V2s of Turkic VVCs cover a wide range of Aspectual, Manner, Modal, and Benefactive

⁹ Tofa, a genetically and geographically similar language to Tuvan, also uses the verb yas- ‘to fail’ for the incompletive aspect with a V1 in a converb form (Anderson 2004: 145). Dictionaries of Sakha, including Slepcov (1972: 362) and Slepcov et al. (2010: 477), explain that the V2 sïs- is an auxiliary usage of the verb sïs- ‘to break’. Although the stem-final /s/ of this verb sïs- shows an irregular alternation into /ňň/ after certain suffixes, the V2 never shows this irregularity. Therefore, it must be regarded as an auxiliary usage of the verb sïïs- ‘to fail’, appearing with vowel shortening.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

439

functions. Among them, both Sakha and Tuvan have four basic V2s frequently used to mark the continuous aspect in common, whereas they show a sharp contrast in expressing ability. Since the sequential and simultaneous converbs are highly recurrent as V1s in Turkic VVCs, previous studies tended to confine their discussion to the two types of converbs. However, Sakha employs the purposive converb, too. In addition, not only converbs but also participles are candidates for the V1 of the Sakha and Tuvan VVC. The two components of Sakha VVCs are phonologically independent, and the V2 merely gives an additional value to the lexical meaning of the V1. Contrastingly, sometimes the two components of Tuvan VVCs show several phonological signs indicating that they serve as a single predicate, and the V2 often has a supplementary role with the lexical meaning of the V1.¹⁰

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

16.3 VVCs in Central and Eastern Turkic and related constructions Turkic languages contain various VVCs other than AVCs. This section examines some of the complex verb predicates other than AVCs and considers their constructional status in order to decide whether they are VVCs. The first issue concerns verb phrases meaning ‘to bring’ in Central Turkic, which may lie on the dividing line between converb constructions and VVCs. Second, the constructional status of verb phrases meaning ‘to buy’ in Central Turkic will also be discussed. Third, we will discuss the case where V2s develop to become suffixlike elements and consider their entry to VVCs.

16.3.1 The verb phrases meaning ‘bring’ in Central Turkic In Central Turkic, the notion of ‘to bring’ is expressed by the combination of two verbs, ‘to take’ and ‘to come’, as in al-ïp kel- in Kazakh and Kyrgyz, el-ip käl- in Uyghur, and ol-ib kel- in Uzbek. The second verb kel- ‘to come’ may appear in the AVC, as in the Kyrgyz example in (23), where the verb kel- ‘come’ loses its lexical meaning and serves as an auxiliary verb:

¹⁰ Johanson (1998b: 114) also remarks that South Siberian Turkic, including Tuvan, “seldom use[s] simple verb forms.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

440

    

Kyrgyz (23)

Menin kïz-ïm 10 ǰïl-day el aralïk my daughter-.1 10 year-about nation between uyum-dar-da ište-p kel-di. organization-- work- -.3 ‘My daughter has worked at the international organizations for about ten years.’

When the verb kel- ‘to come’ follows the converb form of the verb al- ‘to take’, as in (24), however, it holds its lexical meaning and is involved in the choice of the argument, thus ensuring its nonauxiliary status:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(24)

Üy-gö bala-m-du home- child-.1- ‘I brought my child home.’

al-ïp take-

kel-di-m. come--1

It is not easy to determine whether this combination of two verbs denotes a single event or two. We will confirm the VVC status of the Kyrgyz verb phrase al-ïp kel‘to bring’ by examining two syntactic criteria: word order and passivization. First, the alterable word order of a clause containing the verb phrase al-ïp kel‘to bring’ can certify the syntactic status of the verb phrase. The Kyrgyz example in (24) may be regarded as being composed of two clauses, not as a single predicate, as follows: Üy-gö [bala-m-du al-ïp] kel-di-m (home- child-.1- take come--1) ‘I came home, bringing my child’. The clause bala-m-du al-ïp can be a subordinate clause embedded in the superordinate clause, üy-go keldi-m.¹¹ In contrast, no subordinate clause is found in the clause with an alternative word order for (24), as shown in (25), where the verb al- ‘to take’ and its object noun bala-m ‘my child’ are discontinuous, and thus cannot be regarded as subordinate: (25)

Bala-m-du üy-gö child-.1- home- ‘I brought my child home.’

al-ïp take-

kel-di-m. come--1

If we consider that the verb phrase al-ïp kel- in (25) is a single predicate and is thus a VVC, and that both noun phrases are the arguments of the VVC al-ïp kel- ‘to bring’, the possible alternative word orders shown in (24) and (25) can be naturally understood.

¹¹ Haspelmath (1995: 12–17) provides several criteria for subordination, one of which is the alterable constituent order of a converbal clause inside its superordinate clause.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

441

The second criterion adopted for evaluating the syntactic status of a verb phrase is passivization. Central Turkic languages employ three strategies for passivizing complex verb phrases: (i) V1- + V2-: Passive suffixes are attached to both the first and second verb, as in (26). This strategy is typically used in Kyrgyz. (ii) V1 + V2-: Passive suffixes are attached only to the second verb of complex verb phrases, as in (27). This is a common strategy in Central Turkic, including Kazakh, Uzbek, and Kyrgyz. (iii) V1- + V2: Passive suffixes are attached only to the first verb of the complex verb phrases, as found in (28). This is common not only in Central Turkic but also in Sakha and Tuvan.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Kyrgyz kïrgïzstandïk kïz Türkiya-dan (26) ǰeti ǰaštagï 7 years.old Kyrgyzstanian girl Turkey- al-ïn-ïp kel-in-di take-- come--.3 ‘A seven-year-old Kyrgyz girl was brought back from Turkey.’ (27)

Kïrgïzstan-ga al-ïp kel-in-gen tamak-aš-tar Kyrgyzstan- take- come-- meal-food- közömöl-dön öt-pö-y-t. supervision- pass---3 ‘Groceries brought into Kyrgyzstan pass through with no supervision.’

(28)

Moskva-da karoosuz kal-gan üč bala Kïrgïzstan-ga Moscow- unsupervised remain- 3 child Kyrgyzstan- al-ïn-ïp kel-di. take-- come-.3 ‘Three children who stayed in Moscow without supervision were brought to Kyrgyzstan.’

Strategy (iii), in which only the first verb is passivized and is then followed by the second verb, seems to yield another VVC. We consider strategies (i) and (ii) as the criteria to define the degree of the integration of two verbs. It should be noted that complex verb phrases can be passivized only when the first verb is a transitive verb, as illustrated in (29). If the first verb is transitive, however, passivization is possible regardless of whether the second verb is transitive ((30a)) or intransitive ((30b)).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

442

    

Kyrgyz (29)

Kayt-ïp kel-di return- come-.3 ‘(s/he) came back.’ *Kayt-ïl-ïp return-- *Kayt-ïp return-

kel-in-di come--.3

*Kayt-ïl-ïp return-- (30)

kel-in-di come--.3

kel-di. come-.3

a. Rossiya federaciya-sï-nïn aymag-ï-nan čïgar-ïl-ïp Russia federation-.3- area-.3- bring.out-- sal-ïn-a-t. .---3 ‘S/he will be expelled quickly from the area of the Russian Federation.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. šašpastan kayra-kayra kara-p čïg-ïl-ïp . . . unhurriedly again.and.again watch- .-- ‘being re-examined unhurriedly [the constitution]’ (Krippes 1998: 577) As the second verb čïk- ‘to go out’ in (30b) is intransitive, it is not usually passivized;¹² however, it can be passivized when it is the constituent of complex verb phrases, as observed in (30b). The passive suffixes formally attached to the second verb in (30b) can be considered to be attached not to a single verb but to the verb phrase or to the sequence of V1 and V2 as a whole as follows: [kara-p čïg]ïl-ïp (watch- .--). This verb sequence, kara-p čïg-, exhibits further integration when passivized by the strategy (ii), because no element can be inserted between the two verbs when the passive suffix is attached only to the second verb, as shown in (31). Kyrgyz (31)

a. kara-p da čïg-a-t watch- also .--3 ‘(s/he) also examines.’

¹² Intransitive verbs have passive forms in the impersonal passive construction, such as the following: Kečee kino-go bar-ïl-ba-y yesterday movie- go--- ‘Yesterday, no one went to the movie.’

kal-dï. -.3 (Yudaxin 1965: 113)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-        b. *kara-p watch-

da also

443

čïg-ïl-a-t .---3

c. kara-l-ïp da watch-- also ‘(It) is also examined.’

čïg-ïl-a-t .---3

The capability of being passivized by strategy (ii) (V1 + V2-) can assure the positive integrity of V1 and V2. In addition, strategy (i) (V1- + V2-), which is often adopted in Kyrgyz, is also possible only when the verb phrases are integrated as a syntactic unit, because otherwise intransitive V2s would not be passivized. Therefore, the Kyrgyz verb phrase al-ïp kel- ‘to bring’, which can be passivized by strategies (i) and (ii), is regarded as an integrated verb phrase as a predicate, and thus a VVC. In addition to the verb phrase al-ïp kel- ‘to bring’, al-ïp bar- [take- go] and al-ïp ket- [take- leave], both of which also mean ‘to bring, take away’, can be regarded as VVCs because they fulfill the criteria in the same way as al-ïp kel-. Furthermore, the contracted forms of these three verb phrases, apkel- ( apar‘to bring’ > äpkel‘to bring’

Kazakh (36)

al-ïp take- al-ïp take-

(Musaev 2008: 265)

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-       

447

The form alpar- in Kyrgyz seems to be a compound of the two verbs al- ‘to take’ and bar- ‘to go’, but it is only a contracted form of the syntactically formed VVC, al-ïp bar- ‘to bring’, as already mentioned in section 16.3.1. The devoicing of the initial consonant of V2 bar- is attributed to the voicelessness of the final consonant of the converb suffix of the preceding verb. The other three forms in (35) and (36), however, preserve the consonant of the converb suffix with the elimination of the consonants of V1 stems. All instances in (35) and (36) are primarily syntactic products and then contracted, resulting in reduced forms. Although the Kazakh form äpkel- ‘to bring’ is even accompanied by vowel harmony, these contracted forms should be distinguished from primary morphological compounds because they semantically show no difference and syntactically behave in the same way when compared to phrasal verbs in full converb form. Additionally, in Khakas, the converb suffix of V1 is often left out, as in (37), as if V1 stems were directly combined with V2 to form a word. Khakas

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(37)

pasčatxan

(< paz-ïp write-

+ čat-xan) - ‘writing’

(Karpov 1975: 239)

This combined form in Khakas is also regarded as a contracted form of syntactically formed verbal complex, not as a morphological compound, because the omission of converb suffixes of V1 is phonologically conditioned: The omission may occur only when the converb suffix follows the stem-final consonant, and V1s appear in full converb form when the verb stem ends with a vowel, as in čooxta-p sal [say- ..2] ‘Speak!’ (Baskakov and Inkiǰekova-Grekul 1953: 424). Moreover, even when the converb suffix is omitted, V1 and V2 that contain different classes of vowels from each other occur as two words, as in ügren turbïn [learn -1] ‘I am learning’. Therefore, the form in (37) is also the secondary product derived from syntactically formed verbal complexes, which should be distinguished from primary morphological compounds.

16.5 Summary This chapter has described VVCs in Turkic, providing a comprehensive survey of VVCs in two Eastern Turkic languages, Sakha and Tuvan. Both languages share major features of VVCs, but differ from each other in the choice of V2 verbs and the situation in which the VVCs are used (Section 16.2). Since the Turkic VVCs are typically the combination of converbal V1 and subsequent finite V2, the distinction between VVCs and the converb constructions could be an issue. We propose two syntactic criteria, word order and passivization, in order to determine

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

448

    

the VVC status of the complex verb predicates. Furthermore, we examine V2s that have developed into suffixes, or are in the developmental stage of suffixes, taking instances from Kyrgyz, Uyghur, and Uzbek (Section 16.3). Lastly, we examine some Turkic VVCs to determine whether they are morphological compounds or not, finding there seem to be no verb-verb compounds in Central and Eastern Turkic (Section 16.4). Compared to Japanese, which develops compounding in verb-verb predicates, Turkic languages have no verb-verb morphological compounds, solely employing syntactic strategies to form VVCs. However, some of the V2s have been grammaticalized and have developed into enclitics or suffixes. Further setting of subclasses of syntactic verb concatenations may be possible and useful in Turkic languages.

Acknowledgments

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Part of this work was supported by MEXT/JSPS KAKENHI Grant number 18H03578, 26704004, and 17H04773, and a research grant from Mishima Kaiun Memorial Foundation. We are grateful to two anonymous reviewers for comments on the first draft.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

3. ‘to sit (down)’

2. ‘to stand’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.















otur-: progressive Kat ǰaz-ïp otur-a-m. ‘I am writing a letter.’

tur-: progressive or repeated action Al kel-genče kitep oku-p tur-du-m. ‘Until he came along, I was reading a book.’ tur-: prospective Uul-um uya-dan uč-kanï tur-a-t. ‘My son is about to fly the nest.’

ǰat-: progressive Kayak-ka bar-a ǰat-a-sïŋ? ‘Where are you going?’ ǰat-: progressive Baldar oyno-p ǰat-a-t. ‘Children are playing.’ ǰat-: prospective Aškana-ga bar-ganï ǰat-a-bïz. ‘We are about to go to a cafeteria.’ tur-: continuous for a short time Küt-ö tur-uŋuz! ‘Wait for a while.’

1. ‘to lie down’

(A) Aspectual



Kyrgyz

Converb or Original meanings as participle forms lexical verbs

Appendix

V2 verbs

Table 16.2 The list of V2s in AVCs in Kyrgyz and Sakha

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

sït-: progressive Ol kihi utuy-a sït-ar. ‘That man is sleeping.’ sït-: resultative state Taas tüh-en sït-ar. ‘A stone has fallen down.’ sït-: prospective Oğo utuy-aarï sït-ar. ‘The child is about to sleep.’ tur-: (1) progressive Olox-put ularïy-a tur-ar. ‘Our life is changing.’ (2) inceptive Mağahïïn dieki xaam-a tur-da. ‘He started to walk to the shop.’ tur-: resultative state Xatïŋ üün-en tur-ar. ‘A birch has grown up.’ tur-: prospective Suol alžan-aarï tur-ar. ‘The road is about to collapse.’ olor-: progressive Bihigi kepset-e olor-o-but. ‘We are talking.’ olor-: (1) progressive Salalta ïksaa-n olor-or. ‘The director is hurrying.’

Sakha

Continued

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

V2 verbs

 

6. ‘to go’







5. ‘to come’

4. ‘to move’ or ‘to visit’



Original Converb or meanings as participle forms lexical verbs

Table 16.2 Continued

otur-: prospective On ǰïldïgïn belgile-geni otur-a-bïz. ‘We are about to mark the tenth anniversary.’ ǰür- ‘to move’: continuous for a long term Meni este-y ǰür. ‘Remember me for a long time.’ ǰür- ‘to move’: continuous for a long term Biškekte oku-p ǰür-ö-t. ‘He has been studying in Bishkek.’ ǰür- ‘to move’: prospective Baarïn ǰok kïl-ganï ǰür-üš-ö-t. ‘They are about to lose everything.’ kel-: gradual progressive Üč ǰïl ište-p kel-di-m. ‘I have been working for three years.’ bar-: gradual progressive Okuučular-dïn sanï azay-ïp bar-a-t. ‘The number of students is decreasing.’

Kyrgyz

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

sïrït- ‘to visit’: progressive Bu oğo nuuččalïï saŋa-nï ist-e sïld-ar. ‘This child is hearing Russian speech.’ sïrït- ‘to visit’: resultative state Ol alta uon oğo-tton kini ere xaal-an sïlž-ar. ‘Out of those sixty children only he remains.’ sïrït- ‘to visit’: prospective Bu uol oskuola-ğa kiir-eeri sïlž-ar. ‘This boy is about to enter the school.’ kel-: gradual progressive Xarağ-ïm sïrdaa-n kel-le. ‘My eyes are becoming bright.’ bar-: gradual progressive Maayïs ïtaa-n bar-da. ‘Maayis started to cry.’ is-: (gradual) progressive Kïtïï-tïgar čugahaa-n is-pit. ‘He was approaching the wayside.’

(2) resultative state Atah-ïm kel-en olor-or. ‘My friend has come.’ olor-: prospective Piibe-ni ih-eeri olor-or. ‘He is about to drink beer.’

Sakha

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

(B) Attitude or Manner of action along with Aspectual meaning

koy-: quick and completed action At-tï al-a koy-du. ‘He took the horse quickly.’ koy-: completion, preparation Arzan sat-ïp koy-du-m. ‘I sold out at a cheap price.’ ket-: incidental or unexpected action Biz-ge kel-e ket-iŋiz! ‘Drop by our place.’

11. ‘to put on’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

12. ‘to leave’ 





büt-: terminative Kitep ǰaz-ïp büt-tü-m. ‘I finished writing a book.’



9. ‘to become’

kir-: inceptive Al ǰaman sözdördü ayt-ïp kir-di. ‘He began saying the bad words.’

bašta-: inceptive Kïrgïz tilin üyrön-ö bašta-y-bïz. ‘We will start learning the Kyrgyz language.’ bašta-: inceptive Erteŋ kat ǰaz-ïp bašta-y-t. ‘Tomorrow she starts writing a letter.’

10. ‘to finish’ 







8. ‘to enter’

7. ‘to begin’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Continued

sağalaa- ‘to begin (v.t.)’: inceptive Kini ülelee-n sağalïï-r. ‘He starts to work.’ sağalan- ‘to begin (v.i.)’: inceptive Xamsaahïn umay-an sağalan-na. ‘A (social) movement started to develop.’ kiir-: inceptive Tolkuydaa-n kiir-en bar-ar. ‘He starts to think things over.’ Note: Used only with the verb bar- ‘to go’, in the combination kiir-en bar-. buol-: inceptive Oğo aağ-ar buol-la. ‘The child began to read.’ büter- ‘to finish (v.t.)’: terminative Dakïlaat-ï suruy-an büter-bit. ‘He finished writing a thesis.’ büt- ‘to end (v.i.)’: terminative Futbollaa-n büt-te. ‘He finished playing soccer.’

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

V2 verbs

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.











17. ‘to throw’ 

16. ‘to bend’ 

15. ‘to send’

14. ‘to remain’

13. ‘to see’



Original Converb or meanings as participle forms lexical verbs

Table 16.2 Continued

ket-: quick or sudden completed action Uyku-m kel-ip ket-ti. ‘I got sleepy.’ kör-: confirmation of doing an action, always used in imperative Bul kat-tï ber-e kör! ‘Don’t forget to give (her) this letter.’ kör-: (1) tentative Süröt tart-ïp kör-dü. ‘He tried to draw a picture.’ (2) experience Ooru-p kör-sö kerek. ‘He must have contracted an illness.’ kal-: sudden or quick action or change It-i ǰat-a kal-dï. ‘His dog lay down quickly.’ kal-: unexpected action or change Koy-u öl-üp kal-dï. ‘His sheep died suddenly.’ ǰiber-: sudden or quick, and completed action Suu-nu ič-ip ǰiber-di. ‘He drank the water quickly.’ iy-: sudden or quick completed action Biz kïykïr-ïp iy-di-k. ‘We shouted suddenly.’ tašta-: sudden or quick completed action Apama ayt-ïp tašta-dï-m. ‘I talked to my mother quickly.’

Kyrgyz

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

kebis-: thoroughness or completion Onu umn-an kebis-ti-m. ‘I forgot that.’

xaal-: thoroughness or completion Utuy-an xaal-bïp-pïn. ‘I (unintentionally) slipped into sleep.’

kör-: tentative Süg-en kör-ör da kïay-bat. ‘He tries to lift (it) up but can’t.’

(bar- ‘to go, to leave’)

Sakha OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi



  

2 0. ‘to fall’

21. ‘to pull’

22. ‘to hit’

23. ‘to fail’



19. ‘to put in’ 

18. ‘to go out’ 

čïk-: thoroughness or completion of the action Bul kitep-ti oku-p čïk-tï-m. ‘I read this book (completely).’ sal-: sudden or quick completed action Maga ayt-a sal! ‘Tell me quickly!’ sal-: thoroughness or completion of the action Süt-tü tög-üp sal-dï-m. ‘I threw away the milk.’ tüš-: (1) sudden change or action Mas bol-o tüš-tü-m. ‘I suddenly became drunk.’ (2) continuation of the action (a little while or more) Bac-a tüš-üŋüz! ‘Keep walking a little more.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Continued

tart-: sudden or quick completed action Tur-a tart! ‘Get up right now!’ oğus-: quick and sudden Olus ïksal-ïnan taŋn-a oğus-tu-m. ‘I put on my clothes very quickly.’ sïs-: incompleteness Xomuoh-u-n ïhïkt-a sïs-pït. ‘He almost dropped his ladle from his hand.’

tüs-: temporarily or slightly At-ï-n toxtot-on kör-ö tüs-püte. ‘He stopped his horse and gave a glance.’

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

25. ‘to give’

26. ‘to learn’ 

(C2) Benefactive or Aspectual

(D) Modal

27. ‘can’

al-: possibility or ability Bul at tez čurka-y al-a-t. ‘This horse can run fast.’ al-: self-benefactive Baarbïz-dïn nomeribiz-di ǰaz-ïp al-dï. ‘He wrote down the number of all of us (for himself).’

24. ‘to take’

(C1) Benefactive or Possibility

ber-: other-benefactive Maga töšök-tü sal-ïp ber-iš-ti. ‘They laid the mattress for me.’ (Note that benefactive ber- holds its own argument.)





ber-: continuation of the action Kamïr-dï ǰuuru-y ber-di. ‘He kept kneading the dough.’







Kyrgyz

Original Converb or meanings as participle forms lexical verbs

V2 verbs

Table 16.2 Continued

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

üören-: voluntative Saxalïï kepset-e üören-er. ‘He wants to talk in Sakha.’ sataa-: voluntative Oğo-lor töröö-büt nehiliek-teri-ger kel-e satïï-l-lar. ‘Children want to come to their native villages.’

bier-: unexpected and sudden Bil-bet kihi-tin xahïï-tï-ttan uhukt-a bier-de. ‘He suddenly woke up because of a shout from an unknown man.’ bier-: (1) other-benefactive Barï-tï-n cuolkaydïk et-en bier-de. ‘He clearly explained all things (for me).’ (2) gradual progressive Xallaan sïlïy-an bier-de. ‘The weather came to be warm.’

ïl-: self-benefactive (unproductive) massïïna-larï-gar utuy-an ïl-ar žon. ‘people who sleep in their cars’

Sakha

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

17 Turkish and Uyghur verb-verb complexes in contrast Yuu Kuribayashi

17.1 Introduction This chapter discusses the verb-verb complexes in Turkish with occasional reference to Uyghur and some other surrounding languages. The Turkic languages are a language family of at least thirty-five languages spoken by Turkish peoples mainly from the Middle East to Siberia, and Turkish is located in the Southwestern group in the geographic classification of Turkic languages below.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

i. ii. iii. iv.

Southeastern group: (Karluk) Uzbek, New Uyghur, Old Turkic, etc. Northeastern group: (Siberian) Sakha, Dolgan, Tuvan, etc. Northwestern group: (Kipchak) Kazakh, Nogay, Tatar, Kirghiz, etc. Southwestern group: (Oghuz) Turkish, Azerbaijanian, Turkmen, Kashkay, Gagauz v. Others: Chuvash, Khalaj, etc.

Like other Turkic languages, the linguistic properties of Turkish are characterized by such typological features as vowel harmony between stems and suffixes, agglutinative morphology, and head-final structure. Turkish V-V complexes can be divided into two types, the gerundive-type and the converb-type, depending on the conjunctive morphemes/endings attached to the main verb in V1: • Gerundive-type: main verb (V1) in the gerundive form + auxiliary verb (V2) • Converb-type: main verb (V1) in the converb form + auxiliary verb (V2) The choice between the two types is lexically determined by individual auxiliary verbs in V2. Before we go into a discussion of Turkish, the V-V complexes in Uyghur will be introduced for comparison.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

456

 

17.2 V-V complexes in Uyghur We begin with New Uyghur (hereafter referred to as Uyghur), spoken primarily by the Uyghur people of the Xinjiang Uyghur Autonomous Region of Western China, as it has a somewhat richer system of V-V complexes than Turkish.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

17.2.1 Form and meaning of Uyghur delexicalized verbs Uyghur has a total of nineteen delexicalized auxiliary verbs (called “postverbs” or “descriptive verbs”) that function as V2 in V1–V2 complexes formed from a gerundive or converbial V1 (cf. Takeuçi 1990: 298). Of these, only two, ’al‘take’ and bär- ‘give’, can take main verbs in the gerundive. The gerundive form of the verb is made by adding the conjugational suffix A (realized as /ä/ or /a/ according to vowel harmony) to a verb stem to make intraterminal items such as present or imperfect (Johanson 1998a: 47). As shown in Table 17.1, the auxiliary verb ’al- ‘take’ in käl-ä ’almadim undergoes phonological reduction, resulting in keläl-, which in turn is fused with the negative suffix -ma- to give rise to a negative suffix of modality designating the absence of dynamic or epistemic potentiality. The auxiliary verb bär-, which lexically means ‘give’, also has undergone a phonological change in the initial consonant (b>w), forming one morphological unit. The phonological change makes it a progressive suffix (-wär-). Grammaticalized suffixes such as -äl- and -wär- do not impose strict semantic restrictions on the main verbs in V1. The converb-type of V-V complex is characterized by the converb marker -Ip added to the stems of V1 verbs (the vowel I in Ip alternates between /i/, /u/, and /ü/ according to vowel harmony). The converb marker -Ip occurs at the rightmost edge of a simplex verb to close off the morphological domain of the verb and carries out two functions in syntactic structure: as a clause coordinator and as a verbal adverb. The converb used as V1 in V-V complexes is considered an extension from the adverbial usage. The important characteristic of Uyghur V-V complexes that differentiates them from Turkish V-V complexes (Section 17.3) is a relative Table 17.1 Delexicalized V2 verbs in gerundive-type V-V complexes in Uyghur V2 Original verbs lexical meanings

Bleached meanings as V2

Examples

’al-



potential

bär-



continuous, progressive

käl-ä ’al-ma-di-m [come- ---1] > kel-äl-mi-di-m [come----1] ‘I could not come’ yaz-a bär- (write- -) > yez-i-wär [write-] ‘continue to write’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

   -   

457

abundance of auxiliary verbs in the converb-type. Table 17.2 shows seventeen bound auxiliaries (including three grammaticalized suffixes) that are combined with converb verbs in V1 with relatively high productivity (Takeuçi 1990). Note that the two auxiliary verbs, ’al- and bär-, can participate in the converb-type complexes as well as in the gerundive-type. The auxiliary verbs in the converb-type convey a variety of aspectual and modal meanings, of which only the representative ones are shown in Table 17.2 (see Ibrahim 1995 for more details). As is generally the case with “auxiliary verb constructions” (Anderson 2006), grammatical aspect is the most accessible source for grammaticalization in Uyghur. Particularly in the converb-type, aspectual meanings are denoted by -Ip ’al- (resultative), -Ip bar- (continuous), -Ip čiq- (resultative), -Ip qal- (continuous result), -Ip käl- (continuous), -Ip ’oltur(progressive), -Ip tur- (continuous), -Ip yat- (continuous), and -Ip yür- (repetitive). However, the meanings expressed by -Ip baq- (experimental; ‘try -ing’), -Ip qoy(sudden, intentional), -Ip kör- (experimental), -Ip tašla-(intensive), -Ip ’ät (intensive, sudden), as well as the benefactive meanings of -Ip ’al- and -Ip bär-, are not easily subsumed under the standard conceptions of temporal aspect and deontic/epistemic modality. Based on his large database of auxiliary-verb constructions in the world’s languages, Anderson (2006: 369–72) enumerates thirty-seven semantic notions which tend to be grammaticalized in the “content > functional semantic shift.” Some of the semantic shifts shown in Table 17.2, such as , , , and  for progressive/continuative, indeed find their counterparts in Anderson’s list, but others such as  for terminative and  for repetitive are not included in his catalog. More importantly, the following notions are missing from his list: , , and  for directional,  and  for intensive, and  for suddenness and intentionality. These notions, which go beyond the repertoire of conventional aspectual categories, have much in common with the nonaspectual meanings such as suddenness, directionality, benefaction, intensity, and volitionality that are characteristically associated with explicators/vectors in South Asian languages (Masica 1976; Abbi and Gopalakrishnan 1991; Slade 2016). However, the pragmatic overtone of ‘regret’ found in some Indian languages has no equivalent in Uyghur.

17.2.2 Morphosyntactic properties of Uyghur V-V complexes In both gerundive- and converb-types, a V-V complex as a whole functions as the predicate of a clause, no matter whether it is a morphological word or a syntactic concatenation. This is most readily shown by the impossibility or unnaturalness of putting a pause between the two member verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

458

 

Table 17.2 Delexicalized V2 verbs in converb-type V-V complexes in Uyghur V2 verbs

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-Ip ’al-

Ip bär(Turkish ver-) -Ip bar(Turkish var-) -Ip baq(Turkish bak-) -Ip bol(Turkish ol-) -Ip čiq(Turkish çık-) -Ip qal(Turkish kal-) -Ip qoy(Turkish koy-) -Ip käl(Turkish gel-) -Ip kät(Turkish git-) -Ip kör(Turkish gör-) -Ip ’oltur(Turkish otur-) -Ip tašla-Ip tur(Turkish dur-) -Ip yat-

-Ip yür-Ip ’ät(Turkish et-)

Original lexical meanings 

Bleached meanings as V2

Examples

yaz-ip al- [write- -] > yez-i-wal [write--] ‘write for oneself ’  ’oqu-p bär-di-m [read- -1] ‘read someone a book’  ket-ip bar-idu [leave- ] ‘continue to go’ ,   yez-ip baq! [write-  ] ‘try writing!’ ,  terminative ’oqu-p bol! [read- ] ‘finish reading!’  ,   resultative, ’oqu-p čiq-ti-m [read-  thoroughness --1] ‘I read thoroughly’  achievement, toxta-p qal-di [stop- continuous result ] ‘remain stopped’  sudden, resultative bildür-üp qoy [inform- ]‘prepare to inform’  intentional, ’išlä-p käl-di [work- continuous ] ‘continued to work’  directional kün pet-ip kät-ti [sun set in-conv leave-pst] ‘the sun went down’  experimental ’oyla-p kör-äy [consider- ] ‘let me try considering’  progressive kitab ’oqu-p ’oltur-idu [book read -] ‘s/he remained reading the book’  intensive ’yiğla-p tašli-di [(cry- -) ‘burst into tears’ , , continuous saqla-p tur-du [wait-   ] ‘S/he continued to wait’ ,   continuous yaz-ip yat- [write- -] > yez-i-wat- [write- -] ‘S/he is writing’ ,  repetitive Qora-da ’oyna-p yür-idu [park- play- -] ‘They are playing in the park’  intensive, sudden kül-üp-’ät- [laugh-- ] > kül-ü-w’ät-ti [laugh--] ‘S/he burst out laughing’ directional, benefactive, resultative directional, benefactive directional, continuous experimental

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -    (1)

*Män ganggira-p # I freeze- ‘I (unexpectedly) froze.’

qal-di-m. --1

459

(# stands for a pause.) (Dwyer et al. 2013)

The most common way of evaluating the morphological integrity of verbal complexes is syntactic noninterruptibility. Uyghur V-V complexes generally resist the interruption of their internal structure by particle insertion, showing that they are morphologically integrated at least at a surface level. Morphological integration is overtly seen if V2 auxiliary verbs are bound suffixes, as in the case of the progressive suffix -wal, which derives from the erosion of V2 al- ‘take’ and its coalescence with the converb marker, as in yaz-ip al- > yezip al > yeziwal ‘writing’. In marked contrast to the gerundive-type verbal complexes, which do not allow either dA-insertion or mu-insertion, as in (2a, b), the converb-type verbal complexes turn out to permit insertion of an emphatic particle mu ‘even’, which originates from an interrogative particle, as shown in (3b), although they preclude dA-insertion, as in (3a). Uyghur gerundive-type (2) a. yez-i-(*de)-wal-mi-di-m. write--() ---1 ‘I was not continuing to write.’

(dA-insertion)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. kül-ü-(*mu)-w’ät-ti. (mu-insertion) laugh--()-- ‘S/he even burst out laughing.’ Uyghur converb-type (*da) bär-di-m. (3) a. ’oqu-p read- also --1 ‘I also read a book for someone.’ b. ’oqu-p (mu) bol-di. read- even - ‘S/he finished even reading.’ Although it is not known exactly what is responsible for the disparity between da in (3a) and mu in (3b), the palpable contrast between (2b) and (3b) suggests that the morphological structure of converb-type V-V complexes is less tightly integrated than that of gerundive-type complexes. The high degree of morphological integration in gerundive V-V complexes, however, does not rule out the possibility of syntactic rules like passive voice applying to the main verb in V1, as illustrated in (4).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

460

 

Passive in gerundive-type V-V complexes maqalä ikki yil yez-i-wär-il-di. (4) a. Bu this paper two year write---- ‘This article has been written continuously for two years.’ b. Bu maqalä ikki yil yez-il-i-wär-di. this paper two year write---- ‘This article was constantly being written for two years.’ The sentences in (4a) and (4b) differ in the location of the passive morpheme -il-. In (4a), the passive marker is attached to the whole of the main+auxiliary complex yez-i-wär, as is expected from its integrated morphology. In (4b), on the other hand, the passive appears on the main verb yez-. Given the syntactic nature of the passive morpheme, it follows that the gerundive V-V complex originates from a biclausal structure in syntax, and that the auxiliary verbs are affixed (or cliticized) to the gerundive main verbs at a superficial level. The converb-type exhibits the same behavior with regard to passivization. In (5), passive may apply either to the auxiliary verb in (5a) or to the main verb in (5b) (One of the informants rejected both (5a) and (5b)).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Passive in converb-type V-V complexes bär-il-di. (5) a. Bu maqalä yez-ip this paper write- -- ‘This article has been written continuously.’ b. Bu maqalä yez-il-ip bär-di. this paper write-- - ‘This article was constantly being written.’ Interestingly, both of these sentences are accepted only in the aspectual meaning of continuation; they are unacceptable in the benefactive meaning of yez-ip bär‘write for someone’ for some reason that is unclear to me. Uyghur also presents a curious double passive construction in converbial V-V complexes, where both main and auxiliary verbs in a single sentence are manifested in passive forms while representing a single passive meaning, as in (6). (6)

Yeti yašlik qiz Türkiye-din el-l-ip seven year girl Turkey-ABL hand-- ‘A seven-year-old girl was brought back from Turkey.’

kel-in-di. come--

Such double passives are not allowed in Turkish. The converb-type V-V complexes in Uyghur have other syntactic privileges as well. For example, they permit a negative suffix to be attached to either the main

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

461

verb or the auxiliary verb, as illustrated by the examples in (7a) and (7b) involving one and the same auxiliary verb qal- ‘remain’ conveying an adverbial meaning of accidentality. (7)

a. U sowghat-ni kör-mä-y qal-di. s/he gift- see-- - ‘Accidentally, s/he did not see the gift.’ b. U sowghat-ni kör-üp qal-mi-di. s/he gift- see- -- ‘Not accidentally, s/he did not see the gift.’

The positional difference reflects a semantic difference. In (7a), negation takes scope over the main verb, resulting in the negation of the propositional content. In (7b), by contrast, the scope of negation includes both aspectual auxiliary verb and main verb (see Ibrahim 1995: 98–9 for similar examples). Another syntactic privilege of converb-type complexes is that they allow “noun + light verb” constructions like yardäm qil- (‘help do’) in the V1 position. This construction is prohibited from occurring in gerundive-type complexes.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(8)

a. Converb-type [yardäm qil-ip] bärhelp do- ‘to help for someone’

b. Gerundive-type complex *[yardäm qil-e] bär-/qili-wärhelp do- -/-‘to help for someone’

Here, only the converb-type (8a) is acceptable despite the fact that one and the same auxiliary verb bär- is employed in both examples. This observation supports our assumption that converb-type complexes are more loosely connected internally than gerundive-type complexes.

17.3 V-V complexes in Turkish Following the survey of Uyghur V-V complexes, we now turn to Turkish, where a rather different picture emerges, although the V-V complexes in both languages are partitioned into the same two types: gerundive-type and converb-type.

17.3.1 Form and meaning of Turkish delexicalized verbs The gerundive form in Turkish is made by adding the conjugational ending A (realized as /a/ or /e/ according to vowel harmony) or I (realized as /i/, /ı/, /u/, and /ü/) to a verb stem. The gerundive has the general property of indicating the

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

462

 

Table 17.3 Auxiliaries in gerundive-type V-V complexes in Turkish V2 verbs -A bil-I ver-

Original lexical meanings  

-A gel- 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-A dur- ,  -A kal-  -A yaz-    (obs.)

Bleached meanings as V2 potential (suffix) suddenness and swiftness habitual or customary continuous continuous propinquitive

Examples yi-yebil-ir ‘be able to eat’ gir-i ver! ‘enter suddenly!’ kutla-ya geldi ‘S/he has been celebrating’ yürü-ye dur! ‘keep on walking!’ don-a kaldım ‘I was frozen.’ düš-e yazdım ‘I nearly fell to the ground.’

manner in which the action of the main verb is performed, typically emphasizing the repeated or continuous nature of the action in subordinate clauses. Six Turkish auxiliary verbs can participate in gerundive-type V-V complexes, as contrasted with only two in Uyghur. Compare, thus, Table 17.1 with Table 17.3. The auxiliaries in Table 17.3 represent the development of V2 verbs to grammatical suffixes of modality or aspect along the lines of Hopper and Traugott’s (2003) cline of grammaticalization: content item > grammatical word > clitic > inflectional affix. In the first example, the verb bil-, originally meaning ‘know’, is delexicalized to the modal meaning of the subject’s ability in yi-yebil- ‘be able to eat’, thus resulting in the potential suffix -abil- (e.g. yaz-a bil- (write- -) > yaz-abil- (write-) ‘be able to write/will write’). The potential suffix -abil-, which no longer functions as an independent verb, retains the syntactic property of tense inflection as a verb. Likewise in the second, third, and fourth examples, the auxiliary verbs have shifted to aspectual or aspect-related meanings such as swiftness and suddenness (gir-i ver- ‘suddenly enter’ < ver- orig. ‘give’), habitualness (kutla-ya gel- ‘have been celebrating’ < gel orig. ‘come’), and continuation (yürü-ye dur- ‘keep on walking’ < dur orig. ‘stop’). Although they have not reached the stage of suffixes in grammaticalization, they are morphologically “bound auxiliaries” (Göksel and Kerslake 2005: 157), thus forming a tight unit with the gerundive verbs in V1. The last two bound auxiliaries are lexically restricted to combinations with specific V1s. For example, V2 -a kal (- ) is combined with a few activity verbs such as bak ‘watch’, as in bak-a kal- ‘keep on staring (in amazement)’. Likewise, the propinquitive yaz- (düš-e yaz- ‘nearly fall to the ground’) generally co-occurs with V1 verbs that represent nonvolitional events/ actions. Note that Turkish dur- and kal- are not phonologically reduced, unlike in other Turkic languages (in Azerbaijanian, for example, converbial al-ıp gel- ‘bring’ is reduced to apar-). In a nutshell, the combinations of gerundive V1 and auxiliary V2 have lexical restrictions and are not fully productive. In addition to the verbs listed

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

463

Table 17.4 Auxiliary verbs in converb-type V-V complexes in Turkish V2 verbs Original lexical meanings -Ip dur- ,  -Ip kal- 

Bleached meanings as V2 continuous result

Examples gid-ip dur- ‘keep on going’ otur-up kal- ‘remain seated’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

in Table 17.3, there is an obsolete verb yorï-, meaning ‘walk’, which is grammaticalized to a progressive suffix -yor, as in yaz-ıyor (write-) ‘be writing’. In marked contrast to Uyghur, the Turkish system of converb-type V-V complexes is highly impoverished: there are only two auxiliaries, dur- ‘stop’ and kal- ‘remain’, that productively combine with V1 verbs in the converb form, as shown in Table 17.4. Semantically, these auxiliary verbs denote aspectual notions in relation to V1’s events. While the aspectual V2 of gerundive-type V-V complexes have lexical restrictions, mostly used with stative or motion verbs such as ‘go’ or ‘walk’, often in the imperative mood, the aspectual V2 of converb-type V-V complexes have more freedom in combinations with V1 verbs. Morphologically, the converb-type V-V complexes are characterized by the converb marker -Ip added to the stems of V1 verbs (the vowel I in Ip alternates between /i/, /ı/, /u/, and /ü/, according to vowel harmony). The converb originally has a clause-combining function and closes off an adverbial clause with the meaning of post-terminality (e.g. resultative, perfect, constative). (9) is such an example, where not only the verb in the converb form but also the verb following it conveys a literal meaning. (9)

Ali kitap al-ıp (da), Ali book take- also ‘Ali took a book (also) and came.’

gel-di. come-

Since the converb in the conjunctive use makes up a separate clause, it is possible to put a particle like da ‘also’ after it. The clause-combining function of the converb marker can be regarded as the source of its verb-linking function in V-V complexes. In this usage, particle insertion is no longer possible, as illustrated in (10a, b). (10)

a. Ali gid-ip (*da) dur-uyor. Ali go- (*also) - ‘Ali keeps on going, too.’ b. Ali otur-up (*da) kal-ıyor. Ali sit- (*also) - ‘Ali remains seated, too.’

The impossibility of da-insertion in (10a, b) suggests that the converb in V1 and the auxiliary in V2 are morphologically integrated into a complex predicate. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

464

 

Noninterruptibility is also a hallmark of gerundive-type V-V complexes. An exception has to do with the potential suffix -ebil- (cf. Table 17.3), which permits insertion of focus/emphatic particles like de (‘also’) in front of it. In this case, however, special emphatic stress is placed on the gerundive verb, as in (11). (11)

O gél-me-ye-de-bil-ir-di. s/he come------ ‘S/he will not come.’

(Öztürk 2004: 61)

The two types of V-V complexes in Turkish exhibit different behavior in passivization. The examples in (12a, b) involving the passive morpheme in different positions of a gerundive-type V-V complex are both grammatical, though with different meanings. This contrasts with the examples of converb-type complexes in (13). Passive in Turkish gerundive-type V-V complexes yap-ıl-ı-ver-di. (12) a. Köprü 3 gün-de bridge 3 day- do---- ‘Immediately, the bridge was repaired in three days.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. Köprü 3 gün-de yap-ı-ver-il-di. bridge 3 day- do---- ‘The bridge was immediately repaired in three days.’ Passive in Turkish converb-type V-V complexes yaz-ıl-ıp dur-du. (13) a. Bu roman hakkında aylarca this novel about many months write-- - ‘This novel has been written about for many months.’ b. *Roman yaz-ıp dur-ul-du. novel write- -- Intended: ‘A novel was still being written.’ Particularly interesting is the contrast in acceptability between (13a) and (13b), which is asymmetrical to the corresponding passive sentences in Uyghur, shown in (14). In the converb-type complexes, Turkish allows the pattern of “V1- V2”, while Uyghur allows “V1 V2-”; conversely, Turkish prohibits “V1 V2” and Uyghur “V1- V2.”¹

¹ Note that there is some variation as to whether the passive morpheme attaches to V1 or V2 (Sugar 2017: 5). As shown earlier in (5), both “V1- V2” and “V1 V2-” are sometimes allowed in Uyghur.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

465

Uygur: Passive applies to V2 in converb-type qoy-di. (14) a. *Roman yez-il-ip novel write-- - ‘A novel was written up.’ b. Roman yez-ip qoy-ul-di. novel write- -- ‘A novel was written up.’ Turkish: Passive applies to V1 in converb-type hakkında roman yaz-ıl-ıp (15) a. Bu köy this village about novel write-- ‘A novel about this village continued to be written.’

(Sugar 2018)

dur-du. -

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. *Bu köy hakkında roman yaz-ıp dur-ul-du. this village about novel write- -- *‘Novels/ A novel about this village were/was continued to write.’ Many different scenarios can be posited to account for the asymmetry of the acceptability patterns between the two languages. One possibility is to resort to the biclausal vs. monoclausal distinction in syntactic structure. This approach, however, is precluded because we already know that a V-V complex of the converb-type makes up a morphological or functional unit in both languages because of noninterruptibility. Another possible account would be that the Turkish dur- ‘stop’, which rejects the passive, does not have the syntactic properties that are generally expected of lexical verbs, such as the assignment of case. Since passivization calls for (accusative) case, its absence will naturally lead to an ungrammatical passive sentence like (15b). The lack of case might in turn be attributed to the fact that the converb-type V-V complexes in Turkish were copied from South Asian languages through language contact (see Section 17.5). The Uyghur auxiliary verbs, on the other hand, are considered to have accusative case on their own. Earlier, in (6), it was pointed out that Uyghur allows double passive in converbtype V-V complexes. Since Turkish auxiliaries do not passivize, it is naturally predicted that double passive is not allowed in Turkish. This prediction is borne out by the ungrammaticality of (16) (Kuribayashi 2018: 239). (16)

*Köprü 3 gün-de işçi-ler tarafından yap-ıl-ıp bridge 3 day- worker- by do-- ver-il-di. -- ‘The bridge was repaired immediately by workers in three days.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

466

 

Turkish and Uyghur V-V complexes exhibit different behavior in other syntactic constructions like the occurrence of negative polarity items as well. They may also differ in prosody. High pitch on the first verb in Turkish disambiguates an auxiliary reading from a conjunctive reading: çálış-ıp dur-du ‘keep on working’ vs. çalış-íp durdú ‘worked and (then) stopped’ (Csató 2018). In Uyghur, the first member retains its primary stress on its last syllable, as in yez-íp tur-idu ‘keep on cleaning’. On the other hand, there are also similarities between Turkish and Uyghur. One such case is the occurrence of light-verb constructions in the V1 position of converb-type complexes as opposed to their exclusion from the gerundive-type complexes (cf. Kuribayashi 2018). Compare, thus, the contrast between (17a) and (17b) in Turkish with that between (8a) and (8b) in Uyghur. (17)

a. Converb-type [telefon ed-ip] durtelephone do- ] ‘keep on calling’

b. Gerundive-type *[telefon ed-e] durtelephone do-  ‘keep on calling’

Since one and the same auxiliary verb, dur-, is used in (17a) and (17b), it follows that the availability of light-verb constructions in the V1 position is solely attributed to the morphological property of the converb inflection as against the gerundive ending.

The preceding sections have revealed some salient differences between Turkish and Uyghur. Most notable is the asymmetry in the numbers of available auxiliary verbs in the gerundive-type and the converb-type. In Uyghur, the -Ip converb-type is predominant and the gerundive-type impoverished (Tables 17.1 and 17.2), whereas in Turkish, the opposite is the case (Tables 17.3 and 17.4). This asymmetry could be generalized to other languages in the Southeastern group and the Southwestern group. (18) Asymmetry in the predominant types of V-V complexes converb-type more predominant ➔ gerundive-type more predominant 2 (Northwestern group) Kazakh (Southwestern group) (Southeastern group) Uyghur Azerbaijanian, Orkhon Turkic, Turkish (Southwestern group) Turkmen3 ➔

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

17.4 Explaining differences between Turkish and Uyghur

² Muhamedowa (2016) identifies eleven converb-type and ten gerundive-type auxiliary verbs in Kazakh. ³ Hanser (1977) identifies eleven types of V-V complexes, most of which are formed from the converb-type, in Turkmen, spoken in the eastern edge of the Southwestern group. This fact suggests that areal diffusion rather than genetic relation plays an important role in explaining the predominant use of the converb-type in Central Asia.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

467

Table 17.5 Auxiliary verbs in Orkhon Turkic V-V complexes

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

V2 verbs Original lexical meanings -a/-p bar-  -a ber -a ïd -a elt -a qal -a käl -a kör -a u  -a yorï- 

Bleached meanings as V2 completely finished directional intensive intensive continuation continuation intensive ability continuation

Examples täzip bar- ‘run away’ basa ver- ‘come over’ irtürü id- ‘make arrive’ yana elt- ‘disperse’ yatu kal- ‘lie down’ udu käl- ‘keep following’ yälü kör! ‘ride fast!’ artatï u- ‘be able to destroy’ udu yorï- ‘keep pursuing’

The predominance of the gerundive-type in Turkish, Azerbaijanian, and other languages in the Southwestern group can be linked to the fact that they are the direct descendants of Orkhon Turkic, the oldest recorded Turkic language.⁴ Tekin (1968) identifies nine auxiliary verbs in the gerundive-type V-V complexes in this language, as shown in Table 17.5. By contrast, the increased use of the -Ip converb-type V-V complexes in Uyghur apparently reflects the areal linguistic feature of explicator/vector verbs abundantly found in Indo-Aryan languages (Masica 1976). As pointed out in Section 17.2, the semantic functions of V2 verbs in the Uyghur converb-type complexes largely correspond to explicators or vectors in South Asian languages, which are characterized semantically by such notions as completion, suddenness, directionality, benefication, intensity, regret, and thoroughness, and syntactically by their status as finite quasi-auxiliary verbs. Because Uyghur geographically belongs to the Southeastern group of the Turkic language family, it would be plausible to hypothesize that the relative predominance of the converb-type V-V complexes in this language is influenced and enhanced by contact with South Asian languages. However, it must be noted that not all of the converb-type V-V complexes in Uyghur are copied from South Asia, because there were explicator auxiliaries formed from an already existent converbial V1 at least in Uyghur, as opposed to Indian explicator constructions based on a nonconverbial V1, which is formed from a stem or participial form (Abbi and Gopalakrishnan 1991: 162). Why is the range of V1-V2 complexes in Turkish so limited compared to that of other Turkic languages such as Uyghur, Kazakh, or Turkmen in Central Asia? For example, while Turkish has only four to eight auxiliary verbs of V-V complexes, ⁴ The Orkhon script (Turkish runes) is the oldest record in a Turkic language available at present. It was found in the Orkhon valley of Mongolia in the nineteenth century in an inscription dating back to the eighth century (Ergin 1988; Tekin 2010). Its language is called Old Turkic or Orkhon Turkic in Turkological (taking a philological approach to Turkic languages) literature.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

468

 

Uyghur has more than nineteen auxiliary verbs. Masica (1976: 150) mentions Anatolian Turkish as a language with a somewhat aberrant type of explicator compound verb. Although we are not prepared to account for the aberrant type in Anatolian Turkish, the gradual decay or paucity of V-V complexes in Turkish leads us to speculate that the answer might be found in the high frequency of the analytic strategy of using lexically derived adverbials in Turkish and other southwestern Turkic languages, as contrasted with the right-to-left modification in the explicator-type V-V complexes found in Uyghur and other Turkic languages of Central Asia. With regard to frequent use of complex verbs in spoken Uyghur and Uzbek, Ibrahim (1995: 40–1) claims that Turkic languages originally lacked adverbs of manner, degree, duration, and repetition as well as modal verbs, and that at a later stage in the history of Turkic languages, adverbs of manner and degree were calqued from Arabic or Persian. In Turkish, which probably has the least usage of V-V complex verbs among the Turkic languages, lexically derived adverbs are frequently used without recourse to morphosyntactic strategies copied from foreign languages. In contrast to the converb-type V-V complexes, the gerundive-type of V-V complexes is indigenous to Turkic languages, as they are attested in Orkhon Turkic. The gerundive-type, however, gradually waned in the inventory of auxiliary verbs in southwestern Turkic languages, including Azerbaijanian (Rüstəmov 2007 identifies nine auxiliary verbs of this type in the Azerbaijanian spoken in the Republic of Azerbaijan). Subsequently, gerundive-type V-V complexes partially emerged in Anatolian Turkish as Turkish-speaking people moved to Anatolia. Uyghur more frequently uses converb-type complex predicates than Turkish and Azerbaijanian, because the linguistic feature of explicators diffused from South Asia into Central Asia, ranging from the Southeastern to the Southwestern group of Turkic languages. As a result, while grammaticalization of converbial complexes is often found in, as in the progressive suffix wat- originating from yazip yat- (write- lie.down) > yeziwat (be writing), there is no instance of a grammaticalized bound suffix originating from a converbial complex in Turkish.

17.5 Conclusion: Language contact and linguistic change Comparing the form and meaning of V-V complexes in Uyghur and Turkish, this chapter has illuminated an asymmetry between the two languages in the use of the gerundive-type and the converb-type and suggested an explanation from the viewpoint of language contact.⁵ By way of conclusion, we will touch on two ⁵ For further elaboration, historical perspective must be definitely taken into consideration. The literary language of Chaghatay (fifteenth–seventeenth centuries.), which is a common predecessor of Uzbek and Uyghur, already has eighteen descriptive verbs (three converb-type, seven gerundive-type, and eight converb/gerundive-type), divided into seven semantic groups: ability, duration, sudden, directional, propinquity, inchoative, and terminative (Meltem and Uçar 2016). It underwent strong

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   -   

469

other cases of language contact that resulted in a shift in syntactic structures. The first involves Kashkay (Qashqā’ī), an Oghuz Turkic language spoken in the southwestern part of the Islamic Republic of Iran. The Kashkay are a seminomadic people, forming the second largest Turkic group in the country after the Azerbaijanis. Estimates of their population range from 141,000 to 530,000 (cf. Soper 1996). According to Stilo (2004), Kashkay has typical OV characteristics with head-final word orders in object + verb, noun + postposition, genitive + noun, and adjective + noun. However, rapid syntactic changes in the language have been reported in recent investigations (cf. Csató 2004; Kuribayashi 2009). Among the syntactic changes are the use of preverbal modal auxiliaries, the occurrence of postverbal dative nominals (also found in Azeri, IranianAzerbaijanian), and the absence of sentential passives. These changes are considered to be a result of the language’s contact with contemporary Persian (Johanson 2002: 94). Specifically, the loss of verb serialization by -Ip converbials is the most remarkable change in Kashkay due to language contact with prestigious Persian. As shown by the following example, the V1 + converb construction, still in use in Azerbaijanian, has been lost and the perfect form has taken its place in clause conjoining in Kashkay. (19)

a. Ged-di dukan-a (vä) bir qarpïz al-lï. go- store- and a watermelon buy- ‘He went to the store and bought a watermelon.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

b. *Ged-ip dukan-a bir qarpïz al-lï. go- store- a watermelon buy- (Soper 1996: 278) The second case involves the Turkic Khalaj of Iran, a group of people who live about 200 km south of Tehran, the capital city of Iran, who speak an endangered Turkic language. The population has been recently estimated at about 42,107. Because of the preservation of archaic features, some scholars have speculated that the Khalaj are the descendants of Arghu Turks (Turkicized Sogdians; cf. Doerfer 1971), and their language is generally classified as an isolated non-Oghuz group surrounded by Oghuz Turkic groups. The Khalaj language, like Kashkay, shows heavy linguistic influences from Persian and has lost its converb form (Minorsky 1940), as did Kashkay. (20)

Seher at-a min-miš get-e-k daġ bašï-na. morning horse- ride.on- go--1 mountain side- ‘In the morning, let us go to the countryside riding a horse.’ (Minorsky 1940: 91)

linguistic influence from Arabic and Persian during the historical period. One possible prospect is that an adoption of South Asian linguistic features might be through Classical Persian, which was widely used as a transregional lingua franca in Central Asia. This possibility needs further investigation. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

470

 

Although Khalaj has a few gerundive constructions that are inherent to Turkic languages, such as the grammaticalized potential suffix -abil-, the grammaticalized present suffix -i-yor, the imperative use of -V1-A/I + V2 (al ‘take,’ kal ‘remain,’ bar ‘go’, etc.), the adverbial use of V1 in the V1+V2 complex has become uncommon because of the influence of Persian syntax. Example (21) below shows that the gerundive or adverbial use of V1 kil- ‘laugh’ does not occur when the two finite verbs are placed linearly. (21)

män kil-di-m iēšik-dän käl-di-m. I laugh--1 door- come--1 ‘I laughed when I came in through the door.’

(Doerfer 1998: 130)

The decline or loss of V-V complexes found in Kashkay and that found Khalaj are essentially the same in nature. Contact with languages that do not have the constructions can easily undermine the traditional strategy of using V-V complexes in languages of minority groups, as Anderson (2004) notes regarding the loss of converbs in Khakas due to influence from Russian. The loss of converbs in similar situations is also reported by Rentzch (2008: 168) for Altay Turkic and by Ohsaki and Ebata (Chapter 16, this volume) for Khakas.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Acknowledgments I wish to express my deepest gratitude to Taro Kageyama for helpful suggestions and generous comments in revising drafts of the present chapter. Thanks are also due to anonymous reviewers for detailed comments. I am also grateful to John Haig for checking and commenting on the final draft. Needless to say, all mistakes and inadequacies are my own responsibility. This research was supported in part by JSPS KAKENHI Grant NumberJP16K02676, JP18H03578.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

18 Verb-verb complexes in Avar Hisanari Yamada

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

18.1 Introduction This chapter investigates the V-V complexes in Avar, a Daghestanian language spoken in the Northeast Caucasus, especially in the Republic of Daghestan in the Russian Federation. Avar has an absolutive-ergative case-marking system. Thus, in simplex verb constructions, the S of an intransitive verb and the O of a transitive verb are marked with the absolutive case, and the A of a transitive verb occurs in the ergative case. In addition, Avar has verbs whose experiencer argument appears in the dative or the superlocative case. I will refer to verbs with a dative experiencer argument as “dative verbs,” and those with a superlocative experiencer argument as “locative verbs.” Avar does not have morphological (synthetic) V-V compound verbs but uses syntactic (periphrastic) V-V complexes (Type 4 in Chapter 2). These syntactic V-V complexes comprise a semantically main verb, which is referred to as first verb or V1, and a syntactically main verb, which is termed as second verb or V2. To be more precise, the first verb of a complex verb defines the lexical meaning of the complex verb, and the second verb is the head of the complex consisting of the second verb and the clause or phrase headed by the first verb. Avar has verbs allowing an argument slot to be occupied by either a noun or nonfinite verb like AM-oł’-ize ‘to like, to want’.¹,² Such verbs are not regarded as second verbs in this chapter. The following examples illustrate Avar V-V complexes.³ (1)

cogida-l soldat-az mun other- soldier-. you. ‘Other soldiers will kill you.’

č’wa-ze kill-

j-igo. -. (DJu-A: 234)

¹ Avar uses the Cyrillic alphabet. In this chapter, they are transliterated into Latin characters as: а=a, б=b, в=w, г=g, гъ=ǧ, гь=h, гI=ʕ, д=d, е=e, ж=ž, з=z, и=i, й=j, к=k, къ=q’, кь=ł’, кI=k’, л=l, лъ=ł, м=m, н=n, о=o, п=p, р=r, с=s, т=t, тI=t’, у=u, ф=f, х=x, хъ=q, хь=ç, хI=ħ, ц=c, цI=c’, ч=č, чI=č’, ш = š, щ = šš, э=è, ю=ju, я=ja, ъ=’. ² In this chapter, verbs are cited in the infinitive form. ³ All the example sentences are taken from novels and other books written in Standard Avar and listed as ‘sources of examples’ at the end of this chapter. See the abbreviations of sources in this list. The data of Avar in this chapter are mainly drawn from published texts written in Standard Avar. They include forty-two books, an issue of the magazine Maʕarulaj (1996), some issues of the newspaper Ħaq’iq’at (1997, 2002, 2009, 2010), and the Islamic newspaper Assalam (2007–10). The data on the meaning (or semantic functions) of second verbs come from my Avar native speaker consultants, including three main informants. They are nonlinguists. I am very grateful to all my consultants.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

472

 

(2)

Dica he-b I. that-. ‘I am reading it.’

(3)

Dica tamašałi I. surprising. ‘I was surprised.’

c’al-ule-b read-.-

b-ugo -. (ShG-K: 395)

ha-b-un do--

b-uk’-ana -- (SM2-T: 9)

In (1), (2), and (3), the existential verb AM-uk’-ine ‘to be’ is combined with verbs in the infinitive, general tense participle, and perfective converb form, encoding the future tense, progressive aspect, and resultative aspect respectively.⁴, ⁵ The resultative aspect expresses the state resulting from the action denoted by a first verb. When indicating the future tense, the second verb AM-uk’-ine ‘to be’ only occurs in present-tense form. In contrast, when encoding the progressive or resultative aspect, AM-uk’-ine ‘to be’ can appear in any tense form. The tense of complexes denoting the progressive or resultative aspect is determined by tense forms of the second verb AM-uk’-ine ‘to be’. The combination of the present form of AM-uk’-ine ‘to be’ with the perfective converb form of a verb is ambiguous, being also used to indicate the nonevidential past tense. Avar uses periphrastic tense forms only for the future and nonevidential past tenses and only employs the second verb AM-uk’-ine ‘to be’ for this purpose. Sentences (4) and (5) are also examples of Avar V-V complexes.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(4)

(5)

do-sie mun=gi j-oł’-ule-j that-. you=too -like-.- ‘Probably he likes you, too.’

j-at-ila... -- (GG1-G: 131)

Gulla=cin w-ox-un w-uk’-un w-ati-la-ro hedin Gulla.=even -be come.happy- -- --- so ‘Probably even Gulla was not so happy.’ (ShM1-C: 11)

Examples (4) and (5) use the verb AM-at-ize ‘to find’ in the synthetic future form as second verbs combined with first verbs in the general-tense participle and perfective converb form. The second verb AM-at-ize ‘to find’ in the synthetic future-tense form is usually employed with a first verb in the infinitive, generaltense participle, or perfective converb form, but it can also occur with a first verb in the past- or synthetic future-tense participle form (Mallaeva 2000: 172–3;

⁴ Avar verbs have the past-, general-, and future-tense forms. In addition to the three tense forms, only the verb AM-uk’-ine ‘to be’ has the present-tense form. Following the Avar grammar tradition, I will use the terms “general tense” and “present tense”, though it seems to be better to refer to them as the habitual present and momentary (or punctual) present tenses. ⁵ Avar grammarians often use the term “past-tense converb” rather than “perfective converb.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-   

473

Bashirova 2008: 60). In whichever inflectional form a first verb occurs, the second verb AM-at-ize ‘to find’ in the synthetic future-tense form always expresses ‘probably’. With the second verb AM-at-ize ‘to find’, inflectional forms of first verbs indicate the tense of complexes. The general tense is encoded by the generaltense participle form, the past tense by the perfective converb or past-tense participle form, and the future tense by the infinitive or synthetic future-tense participle form. The second verb AM-at-ize ‘to find’ representing probability also occurs in the infinitive form combined with the second verb AM-uk’-ine ‘to be’ forming the periphrastic future form. The synthetic and periphrastic future forms express the same degree of probability. Moreover, embedding the second verb AM-at-ize ‘to find’ in the infinitive form under the verbs kk-eze ‘to happen, etc.’ and AM-eh-ize ‘to make something reach’ used as second verbs indicates higher and lower probability than the synthetic or periphrastic future form. Therefore, Avar has three types of V-V complexes: periphrastic tense forms, V-V complexes whose tense is indicated by the second verb, and ones whose tense is expressed by the first verb. The verbs raʕ-ize ‘to hear’ and AM-iç-ize ‘to see’ in the general-tense form also occur as second verbs of the third type, and they express probability, like AM-at-ize ‘to find’ (Shakhova 1998: 62–3; Abdusalamova 2006: 37–42, 126–9). The verb raʕ-ize ‘to hear’ as a second verb indicates the speaker’s judgment based on information given by other people, while the second verb AM-iç-ize ‘to see’ expresses the speaker’s judgment based on visual information. The second verbs raʕ-ize ‘to hear’ and AM-iç-ize ‘to see’ allow their first verbs to appear in the infinitive, general tense participle, or perfective converb form, but disallow their occurrence in the past- or future-tense participle form, unlike AM-at-ize ‘to find’. Avar also uses other verbs as second verbs. They are of the second type noted in that they indicate the tense of V-V complexes. This type of second verb also occurs in various forms without a tense distinction: imperative, infinitive, converb, and the form called “Masdar” with a case distinction. Moreover, it is possible to embed nonfinite second verbs of this type into other second verbs of the same type multiple times. In what follows I will discuss only this type of second verb. From a typological perspective (Bybee and Dahl 1989, etc.), it is important to know in what inflectional forms first verbs occur, and what verbs serve as second verbs to express what meaning in Avar and other languages. Languages mark first verbs of V-V complexes with various inflectional forms. With Avar second verbs whose tense determines the tense of V-V complexes, first verbs occur in the infinitive, general-tense participle, or perfective converb form. Hereafter I will refer to second verbs combined with first verbs in the infinitive, general-tense participle, and in perfective converb form as i-second verbs, gp-second verbs, and pc-second verbs respectively. Avar has eleven i-second verbs, three gp-second verbs, and fourteen pc-second verbs in all. Here, I will show an example of each type.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

474

 

(6)

t’eç c’al-ize łuh-ana book. read- - ‘She began to read a book.’

(7)

ʕod-ule-w č’-ole-w cry-.- -.- ‘My son will not stop crying.’

(8)

He-s keč’ aħ-un that-. song. sing- ‘Till he finished singing a song’

he-i. that-. (MP2-K: 247) heč’o ..

dir I.

was. son. (ShG-K: 78)

b-aq-ine-ʕan -.--till (GG1-G: 30)

In (6), łuh-ine ‘to happen, etc.’ serves as an i-second verb, expressing ‘to begin’ (Bokarev 1949: 147). In (7), č’-eze ‘to stay, to stop’ acts as a gp-second verb and represents ‘to stop’ (Bokarev 1949: 136). In (8), AM-aq-ine ‘to stand up, etc.’ functions as a pc-second verb and denotes ‘to finish’ (Saidov 1967: 62). Second verbs are not restricted to the finite form. The second verb in (6) is in the finite form, while the second verbs in (7) and (8) are nonfinite verbs. The pc-second verb AM-aq-ine ‘to stand up, etc.’ in (8) occurs in the infinitive form. Moreover, the gp-second verb č’-eze ‘to stay, to stop’ in (7) stands in the general-tense participle form and is embedded under the gp-second verb AM-uk’-ine ‘to be’. Avar has many converb forms (Samedov 1996; Pajzulaeva 2009). Among them, perfective and simultaneous converb forms are frequently used.⁶ The perfective converb form is mainly utilized to express that the event denoted by the main clause follows the event described by the subordinate clause whose head verb occurs in the perfective converb form. Some languages have V-V complexes with a first verb in the simultaneous converb form. Avar, however, does not make use of such V-V complexes. This volume mainly compares V-V complexes in Japanese and other Asian languages. Comparing Asian and European languages, it can be seen that Asian languages often employ V-V complexes with first verbs in the converb form, while European languages often use V-V complexes with infinitive first verbs, but hardly employ ones with first verbs in the converb form. Moreover, there seem to be very few languages using V-V complexes with first verbs in the participle form. Therefore, this chapter will discuss Avar pc-second verbs more carefully and concretely than i-second verbs and gp-second verbs. Some second verbs in languages do not change the number of arguments of a first verb, others increase it, and yet others decrease it. Avar has no second verbs which decrease the number of arguments. In Avar, five second verbs add an ⁶ The literature on Avar usually uses the term “present time converb” instead of “simultaneous converb.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

475

argument to the argument structure of a first verb. These five second verbs are causative in the broad sense. The i-second verb ha-AM-ize ‘to do, to make’ is a normal causative verb. The i-second verb t’am-ize ‘to put’ is a coercive causative verb. The i-second verbs t-eze ‘to leave’ and AM-ičča-ze ‘to release’ are permissive causative verbs translatable as ‘to let something/somebody do something’. The pc-second verb t-eze ‘to leave’ is a permissive causative verb that does not prevent the continuance of the result of the event depicted by the first verb, as exemplified by (9).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(9)

mun=go w-aq’-un t-ole-b you= -be hungry- -.- ‘The people leaving you hungry’

xalq’ people. (GG1-G: 144)

The discussion of causative verbs does not seem to be important for the purpose of this book. Because space does not permit me to consider all aspects of Avar second verbs, I will limit my further discussion to Avar second verbs that do not change the number of arguments. This chapter will discuss Avar second verbs, focusing on how they are used in published Standard Avar texts. Section 18.2 will present an inventory of Avar second verbs. Most Avar second verbs originate from lexical verbs. I will roughly describe the meaning of second verbs and explain the meaning of the original lexical-verb use of these second verbs. Section 18.3 will be devoted to the case marking in V-V complexes. Section 18.4 will present additional information on semantic and syntactic characteristics of some second verbs. Section 18.5 will address itself to the linear adjacency between a first and second verb, showing that Avar V-V complexes are syntactically created. In Section 18.6, I will discuss agreement patterns in V-V complexes.

18.2 Inventory of second verbs that do not change the number of arguments Some languages have many verbs that occur only in the V2 position. Avar, however, has only one such verb: k’w-eze ‘to be able to do something’. The other Avar second verbs are used also as lexical verbs. The verb k’w-eze ‘to be able to do something’ is a locative verb demanding a superlocative nominal argument functioning as an experiencer and an infinitive argument acting as a stimulus. Tables 18.1, 18.2, and 18.3 list i-second verbs, gp-second verbs, and pc-second verbs, excluding the i-second verb k’w-eze ‘to be able to do something’ already noted.⁷ The three tables present the meaning of verbs as lexical and as second verbs. ⁷ The verb ine in Table 18.3 has no verbal stem (or has a phonetically null stem), and ine is an infinitive suffix.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

476

 

Additional information on the V2 meaning of some verbs, however, will be provided in Section 18.4. In the three tables, the transitivity of each lexical verb is indicated by (i), (t), or (l) following each meaning, and (i), (t), and (l) are abbreviations of “intransitive,” “transitive,” and “locative” (verbs). The three tables present the possibility of the change in cases of arguments of semantically main verbs. “No” means that adding the second verb to a semantically main verb does not change the case of any argument of the semantically main verb. “Yes” indicates that the addition of the second verb may change the case of some arguments. The details are provided in Section 18.3. The meaning of some lexical verbs needs further explanation because they are used in multiple senses. The lexical verb łuh-ine means (i) ‘to happen, to appear’, (ii) ‘to get somewhere’, (iii) ‘to become’, etc. The lexical verb kk-eze expresses (i) ‘to happen’, (ii) ‘to get somewhere ([often] unintentionally or accidentally)’, (iii) ‘to become ([often] unintentionally or accidentally)’, etc. The lexical verb AM-aq-ine is employed in several senses: (i) ‘to stand up’, (ii) ‘to start somewhere, to leave’, (iii) ‘to become’, etc. The lexical verb AM-a-ze has many senses: (i) ‘to sow, to sprinkle something solid (e.g. grain, sand)’, (ii) ‘to hang’, (iii) ‘to build’, (iv) ‘to hit’ (e.g. zar b-a-ze ‘to strike one’s fist somewhere’), (v) ‘to spend (time)’, etc. The lexical verb AM-eh-ize ‘to make something reach’ is used only as a transitive verb in Standard Avar, but some dialects employ this lexical verb also as an intransitive verb meaning ‘to reach, to come’. Standard Avar may also have intransitively used this verb, and the second-verb use may originate from the intransitive use. Though Bokarev (1949: 136) describes the i-second verb č’-eze ‘to stay, to stop’, its use is very rare. I have found only seven occurrences of this i-second verb. Avar grammarians have paid very little attention to V-V complexes. Thus, there are very few previous studies on them. It would have not been set out that ššw-eze ‘to reach’ in Table 18.1 and AM-iç-ize ‘to see’ and łuh-ine ‘to happen, etc.’ in Table 18.3 serve as second verbs. The fact that Avar uses č’-eze ‘to stay, to stop’ Table 18.1 i-second verbs in Avar meaning as a lexical verb and transitivity

meaning as a second verb

change in cases

1. bajbiç-ize

‘to begin’ (i), ‘to begin something’ (t)

‘to begin’

Yes

2. łuh-ine

‘to happen, etc.’ (i)

‘to begin’

Yes

‘to begin’

Yes

3. žu-AM-a-ze ‘to join’ (i) 4. kk-eze

‘to happen, etc.’ (i)

‘to have to do’

No

5. AM-eh-ize

‘to make something reach’ (t)

‘may’

No

6. ššw-eze

‘to reach’ (i)

‘to succeed in doing something’

Yes

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

477

(Mallaeva 2007: 334), rex-ize ‘to throw’ (Hook 2013b: 1), kk-eze ‘to happen, etc.’ (Bokarev 1949: 136), and AM-a-ze ‘to sow, etc.’ (Hook, 2013b: 2) as pc-second verbs is pointed out, but what these pc-second verbs express is not mentioned. Kageyama (Chapter 2, this volume) classifies Japanese syntactic V-V complexes with first verbs in the gerund form into aspectual, attitudinal, and benefactive. The Japanese gerund is functionally analogous to the Avar perfective converb. Avar has no benefactive second verbs. The second-verb use of AM-iç-ize ‘to see’ has a similar meaning to the second-verb use of Japanese mi-ru ‘to see, to look’ and Table 18.2 gp-second verbs meaning as a lexical verb meaning as a second verb and transitivity 1. AM-uk’-ine ‘to be’ (i)

change in cases

the progressive aspect

Yes

2. xut’-ize

‘to remain’ (i)

‘to remain doing something, to continue to do something’

Yes

3. č’-eze

‘to stay, to stop’ (i)

‘to stop doing something’

Yes

Table 18.3 pc-second verbs meaning as a lexical verb and meaning as a second verb transitivity

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

1. AM-uk’-ine ‘to be’ (i)

the resultative aspect

change in cases Yes

2. xut’-ize

‘to remain’ (i)

the continuance of the result Yes of an action

3. č’-eze

‘to stay, to stop’ (i)

the continuance of the result Yes of an action

4. łuʕ-ize

‘to finish’ (i), ‘to finish something’ (t)

‘to finish’

No

5. AM-aq-ine

‘to stand up, etc.’ (i)

‘to finish’ or no specific meaning

No

6. t-eze

‘to leave’ (t)

the completeness of an action No

7. rex-ize

‘to throw’ (t)

the completeness of an action No

8. AM-ač’-ine

‘to come’ (i), ‘to bring’ (t)

the completeness of an action No

9. ine

‘to go’ (i)

the completeness of an action No

10. kk-eze

‘to happen, etc.’ (i)

an unintentional action

11. AM-a-ze

‘to sow something solid, etc.’ (t) repetition

No

12. AM-iç-ize

‘to see’ (l)

‘to try V-ing’

No

13. łuh-ine

‘to happen, etc.’ (i)

no specific meaning

Yes

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Yes

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

478

 

expresses ‘to try doing’. Thus, this second verb is attitudinal. Avar has no other attitudinal second verbs. Table 18.3 shows that Avar has many aspectual second verbs. The pc-second verb use of kk-eze ‘to happen, etc.’ indicates the lack of intention and would demand another new category. Moreover, i-second verbs are aspectual or modal, and all gp-second verbs are aspectual. Some second verbs have high productivity in the sense that they are used with semantically and syntactically varied first verbs, and others show low productivity in the sense that they occur only with a restricted number of first verbs from a semantic and syntactic point of view. The second verbs with low productivity include the pc-second verbs t-eze ‘to leave’, rex-ize ‘to throw’, AM-ač’-ine ‘to come’, ine ‘to go’, kk-eze ‘to happen, etc.’, AM-a-ze ‘to sow, etc.’, AM-iç-ize ‘to see’, and łuh-ine ‘to happen, etc.’. In Section 18.4, what verbs often make complex verbs with these second verbs will be described.

18.3 Case marking

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

All Avar nouns and pronouns are inflected in an absolutive-ergative pattern in simplex verb constructions. Each simplex verb construction contains maximally one absolutive argument and maximally one ergative argument. Adding a second verb to a semantically main verb sometimes changes the case of an argument of the semantically main verb from the ergative or oblique to the absolutive case, resulting in a bi-absolutive construction, namely, one involving two absolutive arguments.⁸ (10)

t’eç c’al-ule-w book. read-.- ‘He was reading a book.’

w-ugoan -.

he-w. that-. (RG2-U: 155)

In (10) with the gp-second verb AM-uk’-ine ‘to be’, as well as the O argument, the A argument is marked by the absolutive case. Avar also has another type of biabsolutive construction, illustrated by (11). (11)

He-w ʕemera-b mex-ał that-. many- time- ‘He was sad for a long time.’

rak’ heart.

baq’wa-n dry-

w-uk’-ana. -- (BZ-T: 98)

In (11) involving the pc-second verb AM-uk’-ine ‘to be’, in addition to the S argument rak’ ‘heart’ of the first verb, its possessor He-w ‘he’ occurs in the absolutive case. In simplex verb constructions, the possessors of S arguments ⁸ In this chapter, oblique cases are cases other than absolutive and ergative ones.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

479

are marked with the genitive case rather than the absolutive case. The use of this type of bi-absolutive construction and analogous bi-nominative constructions in languages with a nominative-accusative case-marking system would be typologically rare. For two absolutive NPs in bi-absolutive constructions, in what follows I will refer to the absolutive argument that is absolutive in simplex verb constructions as the first absolutive NP and to the absolutive NP that is not absolutive in simplex verb constructions as the second absolutive NP. The tables in Section 18.2 show that some second verbs provide no change in the case of any argument of semantically main verbs. These second verbs include the i-second verbs kk-eze ‘to happen, etc.’ AM-eh-ize ‘to make something reach’, the pc-second verbs łuʕ-ize ‘to finish’, AM-aq-ine ‘to stand up, etc.’, t-eze ‘to leave’, rex-ize ‘to throw’, AM-ač’-ine ‘to come’, ine ‘to go’, kk-eze ‘to happen, etc.’, AM-a-ze ‘to sow, etc.’, and AM-iç-ize ‘to see’. These second verbs never occur in bi-absolutive constructions, and A arguments always occur in the ergative case, as in (12) with the i-second verb kk-eze ‘to happen, etc.’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(12)

He-s die kumek that-. I. help. ‘He has to help me.’

ha-b-ize do--

kk-ola. - (ShM1-C: 53)

The i-second verbs ššw-eze ‘to reach’ and bajbiç-ize ‘to begin’ also employ no biabsolutive constructions. The i-second verb ššw-eze ‘to reach’ always requires a dative argument, and this dative argument is regarded as functioning as a controller of a phonetically null element (PRO) in the embedded clause. In (13), the S argument is a PRO and controlled by the dative argument of the i-second verb. (13)

Die t’ur-ize ššw-ana. I. escape- - ‘I succeeded in escaping.’

(ShM1-C: 140)

The i-second verb bajbiç-ize ‘to begin’ optionally marks animate S arguments of some intransitive verbs either with the absolutive or ergative case. Example (14) involves an ergative S argument. (14)

ħalt-ize bajbiç-ana work - ‘We began to work.’

nižeca we. (ShM2-Q: 66)

The other arguments occur in the same case as in simplex constructions. Thus, the A always appears in the ergative case (Bokarev 1949: 147, etc.). There are ten second verbs left: the i-second verbs łuh-ine ‘to happen, etc.’, žuAM-aze ‘to join’, the gp-second verbs AM-uk’-ine ‘to be’, xut’-ize ‘to remain’, and

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

480

 

č’-eze ‘to stay, to stop’, and the pc-second verbs AM-uk’-ine ‘to be’, xut’-ize ‘to remain’, č’-eze ‘to stay, to stop’, kk-eze ‘to happen, etc.’, and łuh-ine ‘to happen, etc.’ These second verbs sometimes occur in bi-absolutive constructions. Most biabsolutive constructions contain a transitive verb, an absolutive O, and an absolutive A, as in (10) with the gp-second verb AM-uk’-ine ‘to be’ and in (15) involving the pc-second verb AM-uk’-ine ‘to be’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(15)

C’aq’ ʕažaibłi ha-b-un very surprising. do-- ‘He was very surprised.’

w-ugoan -.

he-w. that-. (GG2-A: 140)

Most absolutive A arguments are animate. Though I have found several examples of inanimate absolutive A arguments, all of these contain the pc-second verb AMuk’-ine ‘to be’. Inanimate A arguments have a strong tendency to occur in the ergative case for all second verbs, judging from my text data. The gp-second verb AM-uk’-ine ‘to be’ (Bokarev 1949: 108–17, etc.) and the pc-second verb AM-uk’ine ‘to be’ (P. A. Magomedova 2006) often employ both the ergative and absolutive cases for animate A arguments though the use of the ergative case exceeds 80% for both second verbs.⁹ In (2) and (3), these two second verbs occur with ergative animate A arguments. The other eight second verbs use the absolutive case for their animate A arguments in all or most examples from my text data (Bokarev 1949: 147 for the i-second verbs łuh-ine ‘to happen, etc.’ and žu-AM-aze ‘to join’).¹⁰ The pc-second verbs AM-uk’-ine ‘to be’ and łuh-ine ‘to happen, etc.’ sometimes change the case of oblique arguments into the absolutive case differently from the other eight second verbs permitting bi-absolutive constructions. Such oblique arguments include dative and superlocative experiencer arguments of dative and locative verbs and goal arguments of some intransitive verbs. (16)

Eldar-il raq kkw-eze b-oł’-un łuh-ana Karim. Eldar- side. hold- -want- - Karim. ‘Karim came to want to support Eldar.’ (RG2-U: 56)

(17)

Uruzmag=gi rix-un Uruzmag.=too hate- ‘He hated Uruzmag, too.’

w-uk’-ana --

he-w, that-. (SM1-N: 47)

⁹ Periphrastic progressive aspect forms of transitive verbs in other Daghestanian languages also optionally occur in bi-absolutive constructions with absolutive A arguments (Forker 2012; Gagliardi et al. 2014). ¹⁰ It is necessary to say that some native speakers do not accept some case-marking patterns that are observed in texts or that other native speakers accept.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    (18)

(19)

kep šš-un satisfaction. reach- ‘He was satisfied.’

łuh-ana -

Kep šš-un=išš satisfaction. reach-= ‘Are [you] satisfied?’

481

he-w. that-. (MP1-S: 31)

w-uge-w, -..- (RG2-U: 386)

Sentence (16) is an example of the pc-second verb łuh-ine ‘to happen, etc.’ with a dative first verb. In (17), containing the pc-second verb AM-uk’-ine ‘to be’, the first verb is rix-ine ‘to hate’. This verb is used either as a dative verb or as a locative verb. In (18), the goal argument of ššw-eze ‘to reach’ occurs in the absolutive case with the pc-second verb łuh-ine ‘to happen, etc.’. Example (19) also involves the intransitive verb ššw-eze ‘to reach’. In (19), the second absolutive NP does not overtly occur. The agreement marker for the singular human male in the stem of the second verb, however, shows the existence of an omitted second absolutive NP, though I will discuss agreement patterns in complex verb constructions in Section 18.6. The goal argument of the verb ššw-eze ‘to reach’ with kep ‘satisfaction’ as the S argument occurs in the dative case in simplex verb constructions, as displayed by (20).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(20)

Čakar-ie kep Chakar- satisfaction. ‘Chakar was satisfied.’

ššw-ana, reach- (MP2-K: 200)

It is noted here that Avar uses bi-absolutive constructions where the S argument and its possessor occur in the absolutive case. Such bi-absolutive constructions are only observed with the pc-second verbs AM-uk’-ine ‘to be’ (Yamada 2010: 105–6), as in (11) above, łuh-ine ‘to happen, etc.’, and kk-eze ‘to happen, etc.’. An example of the pc-second verb łuh-ine ‘to happen, etc.’ will be presented in Section 18.4. With the pc-second verbs AM-uk’-ine ‘to be’ and łuh-ine ‘to happen, etc.’, a variety of verbs serve as their first verbs in this type of bi-absolutive constructions. By contrast, the pc-second verb kk-eze ‘to happen, etc.’ only uses the verb q’os-ine ‘to lose one’s way’ as the first verb, and nux ‘road’ is the only noun employed as the S argument of this verb. There are also found bi-absolutive constructions where the possessor of the O argument of a transitive verb made up by causativizing an intransitive verb. Such examples are very rare. These examples involve the pc-second verb AM-uk’ine ‘to be’ or łuh-ine ‘to happen, etc.’. Example (21) has the pc-second verb AMuk’-ine ‘to be’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

482 (21)

  he-sul raʕabaz rak’=gi that-. word.. heart.=and xinł-iz-a-b-un w-uk’-ana become.hot---- -- ‘His words had warmed Sultan’s heart.’

Sultan. Sultan. (RG2-U: 67)

18.4 Further semantic and syntactic characteristics of some second verbs Section 18.2 roughly described the meaning of Avar second verbs. Some second verbs require further explanation. Moreover, some Avar second verbs are employed with a restricted number of first verbs. I will discuss what first verbs such second verbs often occur with in my text data.

18.4.1 xut’-ize ‘to remain’ as a gp-second and pc-second verb and č’-eze ‘to stay, to stop’ as a pc-second verb

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

As mentioned in Section 18.1, the verb AM-uk’-ine ‘to be’ serves as a gp-second verb encoding the progressive aspect and as a pc-second verb representing the resultative aspect. The verb xut’-ize ‘to remain’ also acts as a gp-second verb, as in (22), and as a pc-second verb, as in (23). (22)

(23)

k’ut’bi č’am-ule-w xut’-ana lip.. bite-.- - ‘Abakar remained biting his lips.’ he-w èxet-un that-. stand up- ‘He remained standing.’

Abakar Abarkar. (RG1-G: 300)

xut’-ana - (RG1-G: 274)

The gp-second verb xut’-ize ‘to remain’ expresses the continuance of the action denoted by a first verb and can be translated as ‘to remain doing something, to continue to do something’ (Yamada, 2013: 165), and pc-second verb xut’-ize ‘to remain’ indicates the continuance of the result of the action expressed by a first verb (Bashirova 2008: 78). The pc-second verb č’-eze ‘to stay, to stop’ represents that the result of the action depicted by a first verb lasts for some time. (24)

dun k’odo-da cewe I. grandmother- in.front.of ‘I stood in front of the grandmother.’

èxet-un stand.up-

č’-ana - (SM1-N: 92)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

483

The gp-second verb and pc-second verb uses of AM-uk’-ine ‘to be’ and xut’-ize ‘to remain’ do not limit the semantic type of their first verbs. In contrast, verbs occurring as the first verbs of č’-eze ‘to stay, to stop’ are semantically somewhat restricted. The first verbs of č’-eze ‘to stay, to stop’ tend to express physical actions performed by human beings or animals. Verbs expressing a change of psychological state are not used with the pc-second verb č’-eze ‘to stay, to stop’. This may show that č’-eze ‘to stay, to stop’ is grammaticalized more weakly than AM-uk’-ine ‘to be’ and xut’-ize ‘to remain’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

18.4.2 AM-aq-ine ‘to stand up, etc.’ The verb AM-aq-ine ‘to stand up, etc.’ is employed as a pc-second verb to indicate ‘to finish doing something’ (Saidov 1967: 62). An example is given as (6). The pc-second verb AM-aq-ine ‘to stand up, etc.’ appears with various first verbs. However, this pc-second verb does not always express ‘to finish doing something’. In my text data, there are many instances of AM-aq-ine ‘to stand up, etc.’ combined with baʕarł-ize ‘to become red’ or qaħł-ize ‘to become white’. In such combinations, AM-aq-ine ‘to stand up, etc.’ does not indicate ‘to finish doing something’. The complex verbs baʕarł-un AM-aq-ine and qaħł-un AM-aq-ine simply mean ‘to become red’ and ‘to become white’. That is, in these complexes, the pc-second verb expresses nothing. The negation of the pc-second verb AM-aq-ine ‘to stand up, etc.’ shows the difference between the two usages of this pc-second verb. Examples (25) and (26) have the negation of the pc-second verb AM-aq-ine ‘to stand up, etc.’. (25)

(26)

Qwa-n b-aq-inč’e-b write- -.-..- ‘A story that [he] did not finish writing’ Baʕarł-un=cin j-aq-inč’o, become.red-=even -.-. ‘[She] even did not turn red.’

xabar story. (ShM2-Q: 12)

(ShM2-Q: 228)

In (25), AM-aq-ine ‘to stand up, etc.’ indicates ‘to finish doing something’. This example entails that [he] started to write this story. In contrast, (26) does not entail that [she] started to become red and simply expresses that [she] did not become red. It is not yet clear whether only baʕarł-ize ‘to become red’ and qaħł-ize ‘to become white’ are exceptional.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

484

 

18.4.3 t-eze ‘to leave’, rex-ize ‘to throw’, AM-ač’-ine ‘to come’ and ine ‘to go’ Several previous studies (Saidov 1967: 476; Mallaeva 2000: 161, etc.) note that the pc-second verb t-eze ‘to leave’ expresses the completeness of the action denoted by the first verb. Isaev (1975: 134–5) says that the pc-second verbs AM-ač’-ine ‘to come’ and ine ‘to go’ represent the completeness of an action. D. A. Magomedova (2010: 131) states that the pc-second verb AM-ač’-ine ‘to come’ always indicates the completeness of an action. Although I do not think that the pc-second verbs t-eze ‘to leave’, rex-ize ‘to throw’, AM-ač’-ine ‘to come’, and ine ‘to go’ always express the completeness of an action, the main semantic effect of these four pc-second verbs is to emphasize such completeness. It is, however, necessary to add that, according to my native-speaker consultants, the semantic effect of these pc-second verbs is very small. Without these pc-second verbs, verbs commonly used with these pc-second verbs listed below usually express the completion of an action.

18.4.3.1 t-eze ‘to leave’ The pc-second verb t-eze ‘to leave’ mostly occurs with transitive verbs. The verb that is most often used with this pc-second verb is, however, the locative verb k’oč-ene ‘to forget’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(27)

ʕaššaqatun-ida he-w Ashakatun- that-. ‘Ashakatun forgot him.’

k’oč-on forget-

t-ana - (GG1-G: 56)

My main text data use t-eze ‘to leave’ as a pc-second verb 1,118 times, and this pc-second verb occurs with the verb k’oč-ene ‘to forget’ 469 times.¹¹ Thus, 42% of the occurrences of the pc-second verb t-eze ‘to leave’ have k’oč-ene ‘to forget’ as the first verb. Moreover, k’oč-ene ‘to forget’ prefers being accompanied by the pc-second verb t-eze ‘to leave’. In my main text data, the verb k’oč-ene ‘to forget’ occurs 690 times. Therefore, 68% of the occurrences of k’oč-ene ‘to forget’ are supported by the pc-second verb t-eze ‘to leave’. Verbs commonly occurring with this pc-second verb are very restricted, including the following verbs.¹²

¹¹ My main text data consist of nineteen books. ¹² In verbal lists shown in this chapter, the figure after each verb indicates the frequency of the verb in my main text data. If a verb occurs with the second verb in more than 10% of all occurrences of this verb, the percentage is added after the figure showing the frequency.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    (28)

485

k’oč-ene ‘to forget’ (469 [68%]), AM ičča-ze ‘to release’ (132), ab-ize ‘to say’ (38), t’am-ize ‘to place’ (19), AM axč-ize ‘to hide’ (22), ł-eze ‘to put’ (13), c’un-ize ‘to save, to keep’ (11), ł’-eze ‘to give’ (10)

List (28) shows that this pc-second verb is often used with verbs expressing a movement or displacement. Except for k’oč-ene ‘to forget’, they are not frequently accompanied by the pc-second verb t-eze ‘to leave’.

18.4.3.2 rex-ize ‘to throw’ The pc-second verb rex-ize ‘to throw’ always occurs with a transitive verb, as in (29). (29)

dica he-l I. that-. ‘I killed them.’

č’wa-n kill-

rex-ana - (SM2-T: 101)

Verbs functioning as the first verbs of this pc-second verb are very limited and exemplified by list (30).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(30)

č’wa-ze ‘to kill’ (46), q’ot’-ize ‘to cut’ (24), AM-aq-ize ‘to take off, to remove’ (11), t’-eze ‘to take off, to remove’ (8), AM-iq-ize ‘to tear’ (4), gukk-ize ‘to deceive’ (3), AM-orł-ize ‘to perforate’ (3), k’k’wa-ze ‘to shave’ (2), qw-eze ‘to cut’ (2)

List (30) shows that this pc-second verb often occurs with verbs expressing partial or full separation. When this pc-second verb is combined with the verb q’ot’-ize ‘to cut’, the complex verb q’ot’-un rex-ize clearly has a different meaning from the simplex verb q’ot’-ize. This complex verb means ‘to cut off ’ and entails that an object is divided into two (or more) parts. It is, however, not always the case that this pc-second verb shows a clear semantic function. This pc-second verb is most frequently employed with the verb č’wa-ze ‘to kill’. The complex verb č’wa-n rex-ize has almost the same meaning as the simplex verb č’wa-ze, though my consultants seem to feel that č’wa-n rex-ize expresses an energetic or speedy action.

18.4.3.3 AM-ač’-ine ‘to come’ and ine ‘to go’ In (31) and (32), the verbs AM-ač’-ine ‘to come’ and ine ‘to go’ are employed as pc-second verbs. (31)

baʕarł-un b-ač’-ana become.red- -- ‘Hamzat’s face became red.’

Ħamzat-il Hamzat-

humer face. (HH-I: 223)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

486

 

(32) C’oro-n ana Gulla-l become.cold- . Gulla- ‘Gulla’s body became cold.’

čerx. body. (ShM1-C: 65)

These two verbs are not broadly used as pc-second verbs. Lists (33a–c) show what first verbs the pc-second verbs AM-ač’-ine ‘to come’ and ine ‘to go’ frequently occur with in my main text data.¹³ List (33a) exemplifies verbs that are only employed with the pc-second verb AM-ač’-ine ‘to come’, while the verbs in list (33b) are examples of verbs that are only combined with the pc-second verb ine ‘to go’. List (33c) includes verbs occurring with either AM-ač’-ine ‘to come’ or ine ‘to go’.¹⁴

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(33)

a. AM-akk-ize ‘to appear’ (113 [15%]), ʕ-eze ‘to grow’ (64), AM-iž-ize ‘to grow’ (26), bur-ize ‘to spring, to gush’ (9 [36%]), baʕarł-ize ‘to become red’ (8), č’agoł-ize ‘to become lively’ (7 [18%]), gurginł-ize ‘to become round’ (5 [45%]), c’ił-ize ‘to become new’ (5 [11%]), t’eha-ze ‘to bloom’ (5) b. t’erç-ine ‘to disappear, to get out’ (68 [29%]), ssw-ine ‘(for fire) to go out’ (48 [18%]), t’aʕ-ine ‘to disappear’ (28), AM-ixx-ize ‘to be destroyed’ (27), AM-uk’k’-ine ‘to be wrinkled’ (19 [12%]), AM-i-ine ‘to melt’ (18 [17%]), łuʕ-ize ‘to finish’ (18), ššušša-ze ‘to be destroyed’ (17 [17%]), c’oro-ze ‘to become cold’ (9), q’-ine ‘to run dry’ (8), hit’inł-ize ‘to become small, young’ (6 [25%]), sas-ine ‘to become quiet, to subside’ (6 [10%]), AM-at’ał-ize ‘to separate’ (5) c. rah-ize ‘to open’ (20 : 11), xinł-ize ‘to become warm, hot’ (6 : 2), qaħł-ize ‘to become white’ (2 : 3), AM-ek-ize ‘to break’ (2 : 3), AM-uħ-ize ‘to burn, to become warm’ (2 : 6), AM-iq-ize ‘to tear’ (2 : 7), ł’erx-ine ‘to become pale’ (1 : 5), tamaxł-ize ‘to become soft’ (1 : 5 [16%])

Lists (33a) and (33b) seem to demonstrate that verbs expressing an appearance, growth, and similar actions tend to be used with the pc-second verb AM-ač’-ine ‘to come’, while verbs indicating a disappearance, decline, and similar actions have a tendency to be employed with the pc-second verb ine ‘to go’. This tendency may be related to the fact that the lexical verbs AM-ač’-ine ‘to come’ and ine ‘to go’ are

¹³ It is difficult to distinguish between the lexical verb and second verb uses of AM-ač’-ine and ine. The verbs AM-ač’-ine and ine in the combination of some verbs in (33) may be regarded as lexical verbs rather than as second verbs. ¹⁴ In list (33c), two figures follow each verb. The former is for AM-ač’-ine and the latter is for ine.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

-   

487

metaphorically used to express ‘to appear’ and ‘to disappear’, respectively, for abstract nouns. Previous studies (Isaev 1975: 134; Bashirova 2008: 73; D. A. Magomedova 2010: 131) state that the pc-second verb AM-ač’-ine ‘to come’ sometimes expresses the suddenness or unexpectedness of an action. The two second verbs also have the following partially syntactic and partially semantic effect. Avar has many labile verbs functioning as either intransitive or transitive, e.g. AM-ek-ize ‘to break’, AM-uss-ine ‘to return’. In Avar, the ergative A of a transitive verb frequently does not explicitly occur. Some languages utilize a passive construction to background the A of a transitive verb. For this purpose, Avar simply omits the ergative A. Furthermore, Avar allows arguments to be omitted if their reference is clear from the context. Thus, when a labile verb is used without an ergative A, it is impossible to distinguish whether this verb is used intransitively or transitively. Combining a labile verb with the pc-second verb AM-ač’-ine ‘to come’ or ine ‘to go’ indicates that an action happens without an animate causer and makes it possible to tell that the labile verb is used as an intransitive verb. The pc-second verbs AM-ač’-ine ‘to come’ and ine ‘to go’ have a further semantic effect. Avar has many verbs that can be interpreted either as changeof-state verbs or as activity/state verbs indicating the continuation of the result of a change of state.¹⁵ For example, the verb AM-ox-ize acts either as a change-of-state verb expressing ‘to become happy’ or as an activity/state verb meaning ‘to be happy’. By contrast, the complex verb AM-ox-ун AM-ač’-ine is only interpreted as a change of state verb with the meaning ‘to become happy’. Thus, the pc-second verbs AM-ač’-ine ‘to come’ and ine ‘to go’ can exclude the possibility of being interpreted as activity/state verbs from first verbs that can be used either as change-of-state verbs or as activity/state verbs in simplex verb constructions. Moreover, when native speakers compare complex verb constructions involving AM-ač’-ine ‘to come’ or ine ‘to go’ and simplex verb constructions without a context, there may be the following tendency. Examples (31) and (32) can be metaphorically interpreted as indicating that Hamzat’s face became red because he was ashamed of something and that Gulla’s body became cold because he was afraid of something. Many other verbs can also be used metaphorically. My informants prefer to interpret complex verb constructions such as (31) and (32) in a metaphorical sense. In contrast, they prefer to interpret simplex verb constructions in a direct sense. In most examples of complex verb constructions involving the pc-second verb AM-ač’-ine ‘to come’ or ine ‘to go’, intransitive verbs serve as first verbs. ¹⁵ The terms “activity” and “state” are those used in Vendler (1969), Dowty (1979), etc. I am not ready to test the distinction between activity and state verbs for Avar verbs. Thus, I will use the notation activity/state verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

488

 

Interestingly, some textual examples, however, have transitive verbs created by causativizing intransitive verbs as first verbs. This is exemplified by (34) with the pc-second verb AM-ač’-ine ‘to come’. (34)

He-sul raʕabaz toxtir that-. word.. doctor. w-ač’-ana -- ‘His words surprised the doctor.’

t’urk’-iz-a-w-un be surprised----

(RG1-G: 203)

18.4.4 kk-eze ‘to happen, etc.’ as a pc-second verb The pc-second verb kk-eze ‘to happen, etc.’ denotes the lack of intention, accidentalness, or uncontrollability of the action represented by the first verb. (35)

dun ł’iž-un I. fall.asleep- ‘I fell asleep.’

kk-ana - (GG1-G: 69)

Verbs commonly used with this pc-second verb are not so many and include the following verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(36)

ł’iž-ize ‘to fall asleep, to sleep’ (63 [18%]), req’-eze ‘to fit’ (45), xis-ize ‘to change’ (28), q’os-ine ‘to lose one’s way’ (19 [10%]), AM-aq-ine ‘(for voice) to get out, etc.’ (15), ǧapulł-ize ‘to doze off, to feel asleep’ (14 [34%]), kwat’-ize ‘to be late’ (14), AM-orč’-ize ‘(for voice) to get out, etc.’ (9), AM-ixx-ize ‘to be confused’ (8), ruhunł-ize ‘to become accustomed’ (7)

Moreover, this pc-second verb has the same semantic effect as the pc-second verbs AM-ač’-ine ‘to come’ and ine ‘to go’ noted above. When the pc-second verb kk-eze ‘to happen, etc.’ is used with a first verb that can serve either as a change-of-state verb or as an activity/state verb, the resulting complex only has the change-of-state interpretation.

18.4.5 AM-a-ze ‘to sow something solid, etc.’ The pc-second verb AM-a-ze ‘to sow, etc.’ represents the thorough repetition of the action denoted by the first verb.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-    (37)

489

Gulla-ca co laħzat-alda t’ut’u-n r-a-na Gulla- one moment- tear- -- q’eda č’wa-ra-l kaǧt-al wall. put up-.- paper-. ‘Gulla tore the papers put up on the wall for a moment.’ (ShM1-C: 32)

The use of this second verb is rare, and this second verb is employed only with transitive verbs such as those in (38). (38)

t’ut’u-ze ‘to tear many times’ (17 [47%]), ǧur-ize ‘to crush, to destroy’ (9), AM-iq-AM-iq-ize ‘to tear many times’ (10 [43%]), su-su-ze ‘to cut many times’ (4 [57%]), AM-iqirq-ize ‘to tear many times’ (3 [42%]), q’ot’-q’ot’-ize ‘to cut many times’ (3 [21%]), ǧur-ǧur-ize ‘to crush, to destroy many times’ (3 [33%]), qweq-ize ‘to cut’ (3 [23%]), qur-ize ‘to pound’ (2)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

List (38) shows that this pc-second verb is mainly employed with first verbs expressing cutting or destruction. Moreover, this list shows that the pc-second verb AM-a-ze ‘to sow, etc.’ often has a first verb with a repeated stem. Avar productively employs the duplication of verbal stems to express the repetition of actions. When a verb with non-repeated stem is used as a first verb, a plural NP usually occurs in the O position.

18.4.6 AM-iç-ize ‘to see’ The pc-second verb AM-iç-ize ‘to see’ expresses ‘to try doing something’. (39)

urǧ-un w-iç-e think- -- ‘Try thinking.’

(ShM2-Q: 91)

This pc-second verb is characterized by frequent use in the imperative form. My main text data contain fifty-four instances of this pc-second verb. In fifty-one of the fifty-four instances, this pc-second verb occurs in the imperative form. Both intransitive and transitive verbs are used with this pc-second verb. The verbs in list (40) are examples of verbs used with this pc-second verb. Such verbs are restricted. The figures after the verbs show the frequency of occurrences in the imperative form.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

490 (40)

  urǧ-ize ‘to think’ (15 [42%]), AM-alah-ize ‘to look’ (10), pikru ha-b-ize ‘to think’ (5 [23%]), ab-ize ‘to say’ (3), ħisab ha-b-ize ‘to take account of ’ (3 [10%])

List (40) demonstrates that in imperative sentences, urǧ-ize ‘to think’, pikru hab-ize ‘to think’, and ħisab ha-b-ize ‘to take account of ’ prefer being accompanied by the pc-second verb AM-iç-ize ‘to see’.

18.4.7 łuh-ine ‘to happen, etc.’ as a pc-second verb Example (41) uses the pc-second verb łuh-ine ‘to happen, etc.’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(41)

w-ox-un łuh-ana -become.happy- - ‘Saxrat became happy.’

Saxrat. Saxrat. (RG1-G: 231)

Adding this pc-second verb to a semantically main verb does not provide any change of the meaning of the semantically main verb. The only semantic effect of this pc-second verb is that this pc-second verb restricts the interpretation of the action denoted by the first verb to a change of state rather than an activity/state. The significant characteristic of the pc-second verb łuh-ine ‘to happen, etc.’ is that this pc-second verb is frequently used in bi-absolutive constructions where the S argument of an intransitive verb and the possessor of the S argument occur in the absolutive case, as in (42). (42)

rak’ b-ek-un łuh-inč’o heart. -break- -. ‘The young man’s heart did not break.’

ʕoloqanči. young man. (DG-G: 447)

The pc-second verb łuh-ine ‘to happen, etc.’ often uses this type of bi-absolutive construction. In my main text data, there are 198 instances of the pc-second verb łuh-ine ‘to happen, etc.’, and 64 of the 198 instances have this type of bi-absolutive construction. In such constructions, the following nouns are commonly used as the heads of the S arguments: body parts (e.g. rak’ ‘heart’, humer ‘face’, and bet’er ‘head’), raħat ‘rest’, ħal ‘condition’, ʕamal ‘behavior’, ł’er ‘color’, etc. Nouns denoting body parts are used with a variety of intransitive verbs. In contrast, raħat ‘rest’ is used only with the verb xw-eze ‘to die’, and ħal ‘condition’ is predominantly employed with the verb xis-ize ‘to change’. The pc-second verb łuh-ine ‘to happen, etc.’ also allows the possessors of nouns denoting body parts to appear in the genitive case, as exemplified by (43).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(43)

dir rak’ ʕodoc-un I. heart. calm- ‘My heart calmed down.’

491

łuh-ana - (DG-G: 422)

There is, however, a significant difference in frequency of occurrence between biabsolutive constructions, as in (42), and constructions in which a single absolutive NP is modified by a genitive NP, as in (43). In my main text data, there are only three instances of single-absolutive constructions in which the possessor of the S argument occurs in the genitive case. Moreover, in my other text data, there is no occurrence of this type of single-absolutive construction, though the pc-second verb łuh-ine ‘to happen, etc.’ occurs 228 times. My main text data have 122 instances of single-absolutive intransitive verb constructions with the second verb łuh-ine ‘to happen, etc.’. In 31 of the 122 instances, the heads of the single absolutive NPs are nouns that would be used in bi-absolutive constructions such as body parts. As noted above, in three of the thirty-one instances, the single absolutive NP is modified by a genitive NP. Thus, in the other twenty-eight examples, the single absolutive NPs occur without their possessors. In these examples, one interpretation is that the absolutive possessor NPs in bi-absolutive constructions are omitted because their references are obvious from the context. Of course, another interpretation is that the genitive NPs of the single absolutive NPs do not explicitly occur. There remain ninety-one occurrences of single-absolutive intransitive verb constructions with single NPs that would not act as the S arguments in bi-absolutive constructions. Only ten of the ninety-one examples have inanimate absolutive single NPs, and the other eighty-one instances have animate single absolutive NPs. The verbs listed in (44) are examples of verbs occurring with animate S arguments.¹⁶ (44)

AM-ixxa-xoč-ize ‘to be confused’ (7 [11%]), baʕa-bakarł-ize ‘to become red’ (6 [25%]), AM-ox-ize ‘to become happy’ (5), razi-rakił-ize ‘to become happy’ (5 [62%]), rełł-ize ‘to become similar to’ (5), ccidax-ine ‘to become angry’ (4), ħinq’-ize ‘to be afraid’ (3), č’uħ-ize ‘to become proud’ (3), bercinł-ize ‘to become beautiful’ (3), xisa-AM-as-ize ‘to change’ (3)

List (44) seems to show that this pc-second verb is predominantly used with first verbs expressing change in emotion. With the second verb łuh-ine ‘to happen, etc.’, the first verbs are mostly intransitive.

¹⁶ The figures and percentages in list (44) are not restricted to those for constructions with an animate S argument.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

492

 

18.5 Linear adjacency In the discussion of complexes of two words, the linear adjacency between the two words is an important topic. The impossibility of attaching any particle to the end of the first word or the beginning of the second word in a combination of two words is crosslinguistically regarded as the evidence that this combination is morphologically created (see Kageyama, Chapter 2, this volume, for Japanese V-V complexes). Avar syntactic V-V complexes allow particles to attach to end of first verbs. Example (45) involves the particle gi ‘too, and’ and the pcsecond verb kk-eze ‘to happen, etc.’. (45)

he-l k’ijalgo cadaq that-. both together ‘They both fell asleep together.’

ł’iž-un=gi fall.asleep-=and

kk-ana. - (GG1-G: 107)

Japanese V-te V complexes allow particles to be inserted between first and second verbs, but they do not permit other classes of words to be inserted between first and second verbs (Kageyama, Chapter 2, this volume). Avar second verbs tend to immediately follow their first verbs. There are, however, many examples of second verbs occurring in a position distant from their first verbs. Sentence (46) is an example of the pc-second verb ine ‘to happen, etc.’.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(46)

bak’ł-un box-dul=gi become.heavy- leg-.=and ‘The legs became heavy.’

ana. . (MM-G: 185)

In my text data, the gp-second verb č’-eze ‘to stay, to stop’ and the pc-second verb AM-iç-ize ‘to see’ are always adjacent to their first verbs. This seems to be because these two second verbs are used rarely. Bokarev (1949: 136) contains an example of the gp-second verb č’-eze ‘to stay, to stop’ that stands apart from the first verb.

18.6 Agreement in V-V complexes Agreement patterns in V-V complexes have hardly been studied. In Avar, the stems of some verbs such as AM-uk’- of AM-uk’-ine ‘to be’, ha-AM- of ha-AM-ize ‘to do’, and žu-AM-a- of žu-AM-a-ze ‘to join’ involve agreement markers (the singular human male w-, singular human female j-, singular nonhuman b-, and plural r-), but the stems of the other verbs, such as łuʕ- of łuʕ-ize ‘to finish’, do not. In simplex verb constructions, if the stem of a verb contains an agreement marker, the verbal stem always agrees with the absolutive argument (S, O, etc.).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

493

In V-V complexes, if the stems of first verbs have agreement markers, they always agree with the absolutive NPs that are absolutive without second verbs: single absolutive NPs or first absolutive NPs in bi-absolutive constructions. Second verbs whose stems have an agreement marker include the i-second verbs žu-AM-a-ze ‘to join’, AM-eh-ize ‘to make something reach’, the gp-second verb AM-uk’-ine ‘to be’, the pc-second verbs AM-uk’-ine ‘to be’, AM-aq-ine ‘to stand up, etc.’, AM-ač’-ine ‘to come’, AM-a-ze ‘to sow, etc.’, and AM-içi-ze ‘to see’. Among them, the i-second verb žu-AM-a-ze ‘to join’, the gp-second verb AM-uk’ine ‘to be’, and the pc-second verb AM-uk’-ine ‘to be’ may occur in bi-absolutive constructions. In bi-absolutive constructions, the agreement marker of the second verb always agrees with the second absolutive NP: A, experiencer, goal, or the possessor of the first absolutive NP. This is exemplified by (47) for the i-second verb žu-AM-a-ze ‘to join’. (47)

dun xabar b-ic-ine I. story. -tell- ‘I began to tell a story.’

žu-w-ana. -- (SM1-N: 88)

All of the second verbs with agreement markers in their stems always or sometimes occur in single-absolutive constructions. The stem of i-second verb AM-eh-ize ‘to make something reach’ predominantly does not agree with any NP.¹⁷ When no agreement happens, the agreement marker for singular nonhuman occurs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(48)

do-j=gi j-ač’-ine that-.=too -come- ‘She may come there, too.’

b-eh-ula -..-

doje. there (RG2-U: 260)

The stems of the other second verbs always agree with single absolutive arguments, as in (5), (26), (34), (37), and (39). As noted, it is possible to embed a second verb under another second verb. (49)

(50)

Ǧazi ł’iž-un Gazi. fall.asleep- ‘Gazi was asleep.’

kk-un -

w-uk’-ana, --

toxtir-zabaz mun sax ha-w-ize doctor-. you healthy do-- ‘Doctors have begun to treat you.’

(AA5: 111) bajbiç-un -

b-ugo. -. (RG2-U: 92)

¹⁷ The second verb use of AM-eh-ize ‘to make something reach’ may have to be regarded as beh-ize with no agreement marker rather than AM-eh-ize for most native speakers. The change from an agreement slot to b is also observed for other verbs. For example, balah-ize is also used instead of AM-alah-ize ‘to look’. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

494

 

In (49) and (50), single-absolutive V-V complexes involving the pc-second verb kk-eze ‘to happen, etc.’ and the i-second verb bajbiç-ize ‘to begin’ are embedded under the pc-second verb AM-uk’-ine ‘to be’. In (49) the stem of the embedding second verb agrees with the absolutive NP, while in (50) it does not agree. When the i-second verb bajbiç-ize ‘to begin’ is embedded under a second verb whose stem always agrees with an absolutive NP, the stems of the embedding second verb only optionally agree with an absolutive NP. In my text data, there are many examples where the i-second verb bajbiç-ize ‘to begin’ does not require the embedding second verb to agree with an absolutive NP. When other second verbs are embedded under second verbs whose stems obligatorily agree with an absolutive NP, the stems of the embedding second verbs agree with absolutive NPs in all or most examples from my text data. Polinsky and Potsdam (2002: 247) note that in the Daghestanian language Tsez, the stems of the second verbs -oqa ‘begin’ and -iča ‘continue’ may agree with the ergative A rather than the absolutive O. No Avar second verb, however, shows such an agreement pattern.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

18.7 Conclusions This chapter has addressed itself to Avar syntactic V-V complexes. They include periphrastic tense forms, V-V complexes whose tense is indicated by inflectional forms of first verbs, and V-V complexes whose tense is expressed by tense forms of second verbs. I have mainly discussed V-V complexes of the third type and focused on second verbs that do not change the number of arguments. I will summarize the characteristics of second verbs such as I have presented in this chapter. Most second verbs of this type are aspectual or modal in the broad sense. The aspectual second verbs encode the progressive aspect, resultative aspect, ‘to begin’, ‘to continue’, ‘to finish’, ‘to stop’, the continuance of the result of an action, the completeness of an action, and repetition. The second verbs with modal or similar meaning express ‘may’, ‘have to do’, ‘succeed in doing’, and ‘try doing’. All but one of these second verbs are also used as lexical verbs. Most of the lexical verbs always or sometimes express movement or staying. Among aspectual second verbs, Movement verbs are used to indicate the starting or ending point of a process, and Stay verbs are employed to represent the continuance of a process or result. First verbs appear in the infinitive, general-tense participle, or perfective converb form. Second verbs representing the starting point of a process require their first verbs to be in the infinitive form, whereas those expressing the endpoint of a process demand the perfective converb form for their first verbs.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-   

495

All second verbs occurring with first verbs in the infinitive or generaltense participle form would be productive, while some second verbs used with first verbs in the perfective converb form combine with only a limited number of first verbs. Some second verbs occur in bi-absolutive constructions. Most bi-absolutive constructions contain an absolutive A. Two second verbs use other types of biabsolutive constructions, which assign the absolutive case to the dative and superlocative experiencer arguments of dative and locative verbs, the goal arguments of some intransitive verbs, and the possessors of the absolutive S and O arguments. Avar second verbs and their first verbs are not always adjacent. This demonstrates that Avar V-V complexes are syntactic. In V-V complexes, the stems of first verbs always agree with single absolutive NPs or first absolutive NPs. The stems of some second verbs obligatorily agree with single absolutive NPs or second absolutive NPs. There is a single second verb whose stem usually does not agree with any absolutive NP.

Acknowledgment This work was supported by JSPS KAKENHI Grant Number 23520484. I am grateful to two anonymous reviewers for comments on the first draft.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Sources of examples [AA5] Medžidowa, Čakar (ed.) 1991. Awar Adabijatałul Čiraq: 5 Klass. Makhachkala: Daguchipedgiz. [BZ-T] Batirowa, Zalmu 2006. T’erçunareb C’wa. Makhachkala: Epoxa. [DG-G] Daganow, ʕabdula. 1997. ʕadamal — dir c’wabi. Makhachkala: [selfpublished]. [DJu-A] Dadaew, Jusup. 1998. Aħul goħ — Dir rek’el buħi. Makhachkala: Jupiter. [GG1-G] Daganow, ʕabdula 1994. Ganč’al. Makhachkala: Dagestanskoe knizhnoe izdatel’stvo. [GG2-A] Ǧalbac’ow, Ǧazimuħamad 1994. Awaragzabi. Makhachkala: Istina. [HH-I] Ħažiew, Ħusen. 1995. Imam Ħamzat. Makhachkala: [self-published]. [MM-G] Muħamadow, Musa 1991. Dagestanskoe knizhnoe izdatel’stvo.

Goro-c’er

balelde

cebe.

Makhachkala:

[MP1-S] Murtazaliewa, Pat’imat 1990. Surat. Makhachkala: Daguchipedgiz. [MP2-K] Murtazaliewa, Pat’imat 1995. Kulakasul jas. Makhachkala: Dagestanskoe knizhnoe izdatel’stvo.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

496

 

[RG1-G] Rasulow, ʕarip. 1996. ʕadamalgi raʕadalgi. Makhachkala: Dagestanskoe knizhnoe izdatel’stvo. [RG2-U] Rasulow, ʕarip. 2006. Uzdenal. Makhachkala: Lotos. [ShG-K] Šaxtamanow, ʕumar-Ħaži 1994. Q’aral ʕor. Makhachkala: Dagestanskoe knizhnoe izdatel’stvo. [ShM1-C] Šamxalow, Muħamad 1982. C’udul was. Makhachkala: Daguchipediz. [ShM2-Q] Šamxalow, Muħamad. 2002. Q’isabi wa xarbal. Makhachkala [selfpublished] [SM1-N] Surxaew, Musalaw. 1990. Nux bit’agi. Makhachkala: Daguchipedgiz.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

[SM2-T] Surxaew, Musalaw. 1994. Tusnaqazda GULAGalda. Makhachkala: Jupiter.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

PART IV

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

LANGUAGES OF EAST AND SOUTHEAST ASIA

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

19 Verbal complexes in Thai Kingkarn Thepkanjana and Satoshi Uehara

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.1 Introduction In contrast to the SOV languages discussed in the other chapters of this volume, where the two members of a verb-verb complex typically occur consecutively in the form of a morphological compound or a syntactic cluster, Thai, an isolating language with an SVO word order, employs the so-called “serial verb constructions” (henceforth SVCs), where two verbs may occur sequentially or be separated by an object if the main verb is transitive. As discussed by Thepkanjana (1986), Thai SVCs do not constitute a homogeneous category but express a variety of notions related to the representation of multiple events in one sentence, such as causative, resultative, passive, modality, direction, aspect, and manner of action. In this chapter we will focus on a type of two-component SVC in Thai in which one component is the head and the other modifies the head component. This type of SVC can be considered a type of verbal complex¹ which is the object of study in this volume. The present chapter addresses an outstanding question of how Thai fits into the areal typology proposed by Masica (1976: Ch. 5). Masica (1976) holds that “explicator” compound verbs are an areal feature characterizing the languages of Asia to the exclusion of Chinese and the languages of Southeast Asia, including Thai. As seen from the map in Figure 1.1 of this volume, Masica (1976)² excludes Thai from the group of explicator-compound languages in Northeast, South, and Central Asia and instead subsumes it under the same group as Chinese that is characterized by “resultative” compound verbs (see Thepkanjana and Uehara 2009 for details of resultative SVCs in Thai). Our investigation shows that Thai does have SVCs that are semantically analogous to the explicator-type compound verbs. After an introductory exposition of compound verbs and SVCs in Section 19.2, Section 19.3 shows that some SVCs in Thai are identified as analogs of explicator compound verbs, and Section 19.3 presents their syntactic and

¹ In this chapter we prefer to use the term “verbal complex” rather than “verb-verb complex,” for reasons discussed in Section 19.3.3. ² Masica (2001: 250–2), however, changes his position and includes Mandarin Chinese in the group of languages that have explicator compound verb constructions. Kingkarn Thepkanjana and Satoshi Uehara, Verbal complexes in Thai In: Verb-Verb Complexes in Asian Languages. Edited by: Taro Kageyama, Peter E. Hook, and Prashant Pardeshi, Oxford University Press (2021). © Kingkarn Thepkanjana

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

500

    

semantic characterizations. Section 19.4 investigates Thai verbal complexes in more detail, and Section 19.5 concludes the chapter. In a nutshell, it is suggested that explicator/vector-like subsidiary verbs are not limited to contiguous verbverb sequences but are also available in serial verb constructions where the main verb and the subsidiary verb are syntactically separate.

19.2 V-V complex, compound verb, and SVC The type of V-V complex in Thai that is pertinent to the exploration in this volume is the explicator compound verbs in the sense of Masica (1976: Ch. 5). Masica employs the term “compound verb,” contrasting the Indian-type explicator compound verbs with the Chinese-type “resultative” compound verbs. In this section, we will clarify the syntactic status of SVCs in relation to compound verbs.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.2.1 V-V complexes and compound verbs According to Masica (1976), explicator compound verbs are characterized as a sequence of two verbs, V1 and V2, in which V2 is semantically subsidiary to V1 and which is schematically represented as “Vv.” Although Masica (1976) as well as Hook (1974) employs the term “compound,” discussion in the other chapters of this volume shows that the relevant verbal complexes in the languages of Northeast, South, and Central Asia may or may not make up morphological compound words. In this chapter, the term “compound verb” is used to refer to “Vv” sequences regardless of their morphological status as compound words. From the typological point of view, the semantic characteristic of explicator compound verbs is more important than their morphological property. Specifically, the semantic center of gravity of Vv sequences resides in the main verb (V1), with the explicator or “vector” verb (V2) acting as a modifier on V1 due to lexical emptying or semantic bleaching. According to Masica (1976: 146–7), the chief explicator verbs recurring in Indo-Aryan and Dravidian languages include verbs denoting motion or position, such as GO, COME, RISE, FALL, SIT, MOVE, GIVE, TAKE, PUT/KEEP, THROW, and LEAVE/RELEASE, and those in Altaic and other languages include GO, COME, RISE, FALL, SIT, MOVE, COME OUT, LIE, STAND, PASS, REMAIN, GIVE, TAKE, PUT/KEEP, THROW, LEAVE, BE USED UP, SEE, SHOW, SEND, WEAR, WRITE, and FINISH (individual languages differ in how many or which of these verbs they actually employ). These verbs, Masica claims, serve to add specification of features of the action denoted by the main verb in V1. Although Russian researchers such as Rastorgueva and Kerimova (1964: 212) assume that V2s express aspect, their meaning is not limited to aspect in the usual sense of perfective and imperfective (Comrie 1976).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

501

Some other researchers claim that V2s express “intensification.” As shown by Masica and by other chapters of this volume, however, the function of V2 is varied and manifold, comprising a diversity of notions involved in the unfolding of events, such as completion (e.g. GO), completion with forethought (e.g. PUT, KEEP), completion with violence (e.g. THROW), suddenness/inception (e.g. RISE), accident (e.g. FALL), attempt (e.g. SEE), benefaction (e.g. GIVE), regret (e.g. SIT), and so forth (see Masica 1976: 146–7; Abbi and Gopalakrishnan 1991). In a series of papers including Chapter 11 of this volume, Peter Hook argues that omissibility is a defining feature of vector/explicator verbs. In his definition, the explicator/vector compound verb is “a sequence of two verbs AB (main verb A plus auxiliary or ‘vector’ verb B) that alternates with main verb A with very little difference in meaning translatable into languages which do not have compound verbs (such as Sanskrit or English)” (Liang and Hook 2007: 109). In the HindiUrdu sentences in (1), for example, the compound verb aa jaataa in (1a) alternates with the simple verb aataa in (1b).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(1)

a. kabhii..kabhii dopahar mẽ bhii aa jaataa hai sometimes afternoon in also come  is ‘Sometimes he comes in the afternoon too.’ b. raat ko yahãã roz paamii pii-ne aataa hai night  here daily water drink- come is ‘Every night he comes here to drink water.’ (Liang and Hook 2007: 110)

Masica (1976: 141) points out, however, that the semantic bleaching, or in his terms “lexical emptying,” of V2 is a matter of degree. If lexical emptying is complete, the grammaticalized meaning of V2 in the compound verb is opaque and could not be easily rendered in English. In contrast, if much of the literal meaning of V2 is retained, the semantic contribution of V2 is rather transparent. Most cases come somewhere in between. In the examples in (2)–(5), taken from Masica (1976: 141), the meanings of explicator/vector verbs in the V2 position can be conveyed by manner adverbs or particles in English. (2)

kho baithnaa

‘lose - sit’

= ‘lose (ruefully?)’

(3)

le jaanaa

‘take - go’

= ‘take away’

(4)

likh deenaa

‘write - give’

= ‘write down for somebody’

(5)

gir parnaa

‘fall - lie there’ = ‘fall down’

In comparing the compound verbs in Hindi-Urdu and in Chinese, Liang and Hook (2006) distinguish two types of compound verbs in Chinese—(i) a main verb plus a resultative auxiliary (which generally corresponds to adjectives in

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

502

    

South Asian languages) and (ii) a main verb plus a directional auxiliary—and regard the second construction as a functional equivalent of Hindi-Urudu vector compound verbs. They found both parallels and differences between the vector compound verbs in Hindi-Urdu and Chinese. Some of the important parallels are as follows. The vector verbs in Hindi-Urdu and in Chinese may be used to create pairs of compound verbs that are similar in every respect except for their orientation. The compound verbs in both languages express accomplishments, whereas the unmodified (noncompound) verbs are interpreted conatively. Consequently, the vector verbs in both languages tend to express some sense of aspectuality. In both languages, complements of expressions of fear and anxiety, for example, prefer the use of explicator compound verbs. As for differences, it is found that the vector verbs in Hindi-Urdu are more grammaticalized than those in Chinese. Consequently, the vector verbs in Chinese, which are less grammaticalized, are more numerous than those in Hindi-Urdu. In addition, the explicator compound verbs in Hindi-Urdu have a lower frequency of occurrence than those in Chinese. Syntactically, Chinese allows the direct object of the main verb to intervene between V1 and V2 in compounds (if V2 is qù ‘go’ or lái ‘come’), whereas Hindi-Urdu allows only some adverbial particles, but not an object of the main verb, to intervene. In Hindi-Urdu, the vector verb can be transitive or intransitive depending on the valency of the main verb. In contrast, the vector verb in Chinese is almost always intransitive. Semantically, the vector verbs in Hindi-Urdu act as a kind of “semantic diacritic” (Liang and Hook 2006: 123) on the situation described by the clause as a whole, whereas those in Chinese have a narrower scope. Chinese vectors denote a distinct component of a complex event which is a consequence of the action denoted by the main verb. Such differences lead Liang and Hook (2006) to conclude that the Chinese compound verbs, which are less productive than their Hindi-Urdu counterparts and have radically different properties in syntax and semantics, are not properly subsumed under the same category as Indo-Aryan explicator/vector compound verbs. They thus maintain Masica’s (1976) original conception of a linguistic area comprising the languages of South, Central, and Northeast Asia but excluding Chinese and Southeast Asian languages like Thai.

19.2.2 SVC An SVC is a type of multiverb construction. Most researchers who study SVCs define this type of construction as a series of verbs appearing together without any overt linker. Series of verbs in an SVC act together as a single predicate and are conceptualized as a single event. They have a single intonational contour, one tense, aspect, and polarity. Each verb in an SVC can occur as a lexical verb on its own. Other types of multiverb construction are coordinated, subordinated, and converb constructions.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

503

Aikhenvald (2006) classifies SVCs into two types, namely, asymmetrical and symmetrical SVCs. Asymmetrical SVCs consist of one verb from a semantically or grammatically restricted (or closed) class. Asymmetrical SVCs denote a single event described by the verb from a non-restricted class, whereas the verb from a closed class modifies the verb from the non-restricted (or open) class often in terms of direction, orientation, or aspect. The verb from a closed class is usually a motion or posture verb. Aikhenvald (2006: 22) calls the verb from the open class the “major” verb and the verb from the closed class the “minor” verb.³ According to Aikhenvald, minor verbs in asymmetrical SVCs tend to get grammaticalized. A grammaticalized minor verb can function as a full lexical verb outside of an SVC. Some examples of asymmetrical SVCs are provided by Aikhenvald below. Cantonese (6)

lei5 lo2 di1 saam1 you take  clothing ‘Bring some clothes.’

lai4 come (Aikhenvald 2006: 21)

Kristang (7)

kora yo ja chegá nalí eli ja kaba bai when 1.  arrive there 3.  finish go ‘When I arrived there, he had gone.’ (Baxter 1988: 213, cited in Aikhenvald 2006: 23)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Aikhenvald (2006: 22–8) postulates the following semantics of asymmetrical SVCs as follows: (a) expressing direction and orientation; (b) expressing aspect, extent, and change of state; (c) expressing secondary concepts, such as obligation, probability, pretendtype, beginning-type, negator, ‘want’ concept, ‘intend’ concept; (d) introducing a complement of complement-clause taking verbs such as verbs of speech; (e) increasing valency to mark causatives, benefactives, instrumentals, comitatives, and specifying arguments to introduce direct objects and various other arguments; (f ) reducing valency such as in expressing a passive-like function; (g) marking comparatives and superlatives; (h) marking event-argument SVCs, which provide a manner modification to the event as a whole.

³ The terms “major” and “minor” verbs are from Durie (1997).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

504

    

On the other hand, all verbal components of symmetrical SVCs come from unrestricted or open classes. The order of components of symmetrical SVCs is iconic in that it reflects the temporal sequence of subevents the components denote. Symmetrical SVCs are not headed because all components have equal status. No component determines the semantic or syntactic properties of the SVC as a whole. In studying grammaticalization within the domain of the verb in East and Southeast Asian languages, Bisang (1991) claims that grammaticalization is tightly linked to verb serialization. He classifies verb serialization into two types, namely, verb serialization in a broad sense and in a narrow sense. Verb serialization in a narrow sense is influenced by grammaticalization, whereas verb serialization in a broad sense is not. Bisang postulates six products of grammaticalization in the context of verb serialization in the narrow sense, namely, resultative verbs, directional verbs, TAM (tense/aspect/modality) markers, coverbs, causative verbs, and conjunctional verbs. Bisang argues that three of them show a particularly high degree of grammaticalization, i.e. directional verbs, TAM markers, and coverbs. Aikhenvald (2006: 37–43) postulates four formal properties of SVCs: (i) contiguity vs. noncontiguity of components of a construction, (ii) wordhood of components, (iii) interaction between contiguity and wordhood, and (iv) expression and marking of grammatical categories. As for the first formal component, contiguous SVCs do not allow any other constituents to intervene between their components. Noncontiguous SVCs allow other constituents, such as noun phrases, to occur between the components. The second formal property characterizes SVCs as one-word and multiword constructions. Each component of an SVC may function as a predicate on its own, which results in a multiword SVC. On the other hand, the components may together form one word. The interaction between the (non)contiguity and wordhood of SVCs, which is the third formal property of SVCs, gives rise to four types (Durie 1997: 302–30, Aikhenvald 2006: 39). (i) (ii) (iii) (iv)

noncontiguous, multiword, e.g. Baule, Ewe, Thai; contiguous, multiword, e.g. Kristang, Tariana; contiguous, one-word, e.g. Igbo, Dâw, Alamblak; noncontiguous, one-word. This type has not been attested.

The fourth formal property of SVCs is expression and marking of grammatical categories such as person of the subject and object(s), tense, aspect, modality, evidentiality, negation, etc. Each of these categories within an SVC can be marked on every component or may be marked once per construction. The fourth formal property, which is morphological, is not applicable to isolating languages such as Thai.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

505

It is apparent that Aikhenvald’s asymmetric SVCs are more or less comparable to Bisang’s products of grammaticalization resulting from verb serialization in a narrow sense. We argue that it is grammaticalization of serial verbs which gives rise to the construction which is formally and functionally equivalent to the explicator compound verb in Hindi-Urdu and in Chinese. Section 19.3 examines this type of construction in Thai in detail.

19.3 Thai analogs of the “compound verb” To recapitulate, the explicator compound verb in Masica’s (1976) sense has two important properties. Syntactically, it is characterized by the V1+V2 structure, where the two verbs are contiguous. Semantically, V2 (explicator or vector) is delexicalized and bleached to a certain degree as a result of grammaticalization and is, as a consequence, subsidiary to the first verb, which functions as a main verb. In the rest of this section, these syntactic and semantic properties are applied to SVCs in Thai.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.3.1 The syntactic criterion As mentioned, the explicator compound verb in Masica’s sense has the V1+V2 structure, which entails that the two verbs must be contiguous to each other. In this section, we will examine Masica’s syntactic criterion of the compound verb against Thai data. Given that Thai is an isolating language having the SubjectVerb-Object (SVO) sentential word order, SVCs in Thai are noncontiguous and multiword, as remarked by Aikhenvald (2006). Since an object always follows a verb in Thai, an object NP can appear between verbs in an SVC, which results in a noncontiguous SVC. In other words, if the V1 of an SVC is a transitive verb, the object of V1, if not omitted, follows it, which results in the SVC having the [V1 NP1 V2 (NP2)] structure. The multiword property of an SVC in Thai arises from the ability of each verb in an SVC to function as a main verb on its own outside of an SVC. Examples (8) and (9) illustrate the noncontiguous and multiword properties of SVCs in Thai. (8)

khǎw láaŋ rót hây he wash car  ‘He washed a car for me.’

chǎn I

(9)

khǎw sɯ́ ɯ aahǎan wáy he buy food ... ‘He bought food for future use.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

506

    

V1s in the SVCs in (8) and (9), namely láaŋ ‘wash’ and sɯ́ ɯ ‘buy’, are transitive verbs followed by the object NPs rót ‘car’ and aahǎan ‘food’, respectively. The structures of the predicates in (8) and (9) are [V1 NP1 V2 NP2] and [V1 NP1 V2], respectively. The Thai predicates in (8) and (9) do not satisfy the syntactic criterion of compound verbs in Masica’s sense because they do not have the [V1V2] structure. The type of SVC in Thai compatible with his syntactic criterion is a two-verb sequence in which V1 is an intransitive verb and V2 does not take an object.⁴ In light of the general syntactic structure of Thai, it is not fitting to identify only those two-verb constructions with intransitive V1s as compound verbs while rejecting the serial verb constructions with transitive V1s as noncompounds. It is true that compactness of the construction as manifested in the form of two contiguous verbs is a good characteristic of the compound verb, but there is no good reason why the transitivity of V1s should determine the compound status of Thai serial verb constructions. Note that the data discussed by Masica is primarily drawn from Hindi, an IndoAryan language. The predominant sentential word order in Hindi is SubjectObject-Verb (SOV). If the V1 of the Hindi compound verb has an object, it always appears in front of the V1+V2 sequence. The SOV sentential word order allows only the [(NP) V1V2] structure, but not the [V1 NP V2] structure, to occur in Hindi. In addition, morphological compound verbs in Japanese, according to Kageyama (1989), also have the [V1V2] structure like Hindi compound verbs. Given the fact that Japanese is an SOV language like Hindi, it is not surprising that Japanese V-V complexes exhibit only the [V1V2] structure. In order to capture the “verb compounding” phenomena in both OV and VO languages, a [VP1—V2] structure is to be posited as a common syntactic structure where the V2 (vector verb) takes scope over VP1. In this syntactic structure, verb compounding (technically a reanalysis or morphological merger) turns [VP1—V2] into [V1-V2]. The [V1-V2] compound verb can then be structurally typologized into the [(O) V1-V2] type and the [V1 (O) V2] type, depending on the language’s basic word order of the verb phrase. Masica’s formal criterion can be applied to some extent to Chinese, even though it is also an SVO language like Thai. This is because Modern Chinese has many “SOV features” (Li and Thompson 1974) which are not found in Thai. The most notable feature is the bǎ construction in Chinese, which functions to mark the transitive verb object and prepose it to the preverbal position, as shown in (10): (10) a. Woˇ dǎ Zhāng-sān I hit Zhang-san ‘I hit Zhang-san.’

le 

⁴ Most of the postverbal grammaticalized morphemes derived from serial verbs are erstwhile transitive verbs. Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

    b. Woˇ bǎ Zhāng-sān I bǎ Zhang-san ‘I hit Zhang-san.’

dǎ hit

507

le  (Li and Thompson 1974: 203)

Li and Thompson (1974: 203) note that “if the verb is morphologically complex or modified, the bǎ-construction is usually preferred and often the only acceptable form.” The use of such a construction arguably has proliferated the contexts for verb-compounding, in which the first verb is transitive (see also Post 2007). Thai does not have such a productive object noun-preposing construction and is an SVO language in a strict sense.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.3.2 The semantic criterion According to Masica, V2 must be semantically bleached to a certain degree as a result of grammaticalization in order to combine with V1 to form an explicator compound verb. Thai shows a large number of grammaticalization processes based on nouns and verbs, some of which are so widespread across languages that they can be considered “cognitive universals,” whereas others are common in Southeast Asian languages. Since this chapter examines V2 functioning as a vector verb (a minor verb), we will discuss grammaticalization based on verbs with a special emphasis on postverbal grammaticalized morphemes. Diller (2001) found a number of grammaticalization processes based on verbs in the Tai language family, of which Thai is a member. All grammaticalized morphemes based on verbs are erstwhile verbs occurring in SVCs. Some of these grammaticalized morphemes occur before and others occur after main verbs. Diller (2001) found that the grammaticalized morphemes occurring after main verbs are much more numerous than those that occur preverbally. Some of the postverbal grammaticalized morphemes which are often discussed in Thai linguistic literature are listed in Table 19.1. Another class of verbs which are likely to be grammaticalized across languages are motion and directional verbs. Some examples of this class of verbs in addition to pay ‘go’ and maa ‘come’ are khɯ ˆ n ‘ascend’, loŋ ‘descend’, khâw ‘enter’, ɔ̀ɔk ‘exit’, khâam ‘cross’, phàan ‘pass’, and thɯ̌ ŋ ‘arrive’. It should be noted that these motion and directional verbs are semantically general in the sense that they do not specify means and manner of moving. According to Heine, Claudi, and Hünnemeyer (1991), verbs that are semantically general, such as verbs meaning ‘go’, ‘come’, ‘ascend’, and ‘descend’, tend to be grammaticalized rather than those that are semantically specific, such as verbs meaning ‘walk’, ‘run’, ‘crawl’, and ‘climb’. All of the verbs listed above are often mentioned in research works on grammaticalization in Thai. These verbs are

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

508

    

Table 19.1 Grammaticalized verbs in Thai V2 verbs

original meaning

grammaticalized usage

dây



modality marker (‘able, possible’)

wâa



complementizer

yùu

, 

progressive/ continuative aspect

hây



dative, benefactive, purposive marker

lǽæw



perfective aspect/ perfect marker

pen

 , , 

modality marker (‘be able’ in the sense of knowing how to do)

pay



‘excess’ marker

maa



perfect marker

kwàa

 

comparative adverb, temporal conjunctive adverb

tɔ̀ɔ



dative marker, distributive quantifier (‘per’), temporal marker

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

grammaticalized to varying degrees, which means that some verbs are more semantically bleached than others. The verbs that are grammaticalized at a relatively advanced stage retain little lexical meaning, which renders the relationship between the source and the grammaticalized meanings rather obscure.

19.3.3 Summary In short, the two-verb SVCs in Thai which are analogous to the explicator compound verbs in South Asian languages are characterized as follows. Syntactically, they have the structure [V1 (NP) V2], in which the direct object NP of the V1 is omissible. Such a syntactic structure, with an optional intervening NP, arises from the fact that the basic sentential word order in Thai is SVO, in which the object always appears after a verb. V1 functions as the main verb or the head, whereas V2 is the vector or the minor verb, which is grammaticalized to a certain extent and which provides a semantic specification in some way to the main verb. V2 cannot take any object and is semantically bleached to a certain degree. We therefore opt for the term verbal complexes, not V-V complexes or “compound verbs,” for the Thai phenomenon under discussion for the following reasons. First, when V1 is a transitive verb, a nonomissible NP object of V1 can appear after it because Thai has an SVO sentential word order. The term ‘V-V complexes’ applied to this construction might create a misunderstanding that no object can intervene between the two verbs in Thai. Second, grammaticalized minor verbs could appear both before and after main verbs. If the term

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

509

“compound verb” is adopted to refer to the structure [V1 (NP) V2] in which V2 is a grammaticalized minor verb, one might wonder whether the structure [V1 V2 (NP)] in Thai in which V1 is a grammaticalized minor verb should be categorized as a compound verb or not. Furthermore, a main verb can appear with either preverbal or postverbal grammaticalized verbs. In our opinion, the term “compound verb,” if applied to the Thai verb sequences under discussion, invites questions that are unnecessary for our purposes. We therefore opt for the more neutral term “verbal complexes” for the Thai verb sequences with the structure [V1 (NP) V2]. Different types of Thai verbal complexes will be examined in detail in Section 19.4.

19.4 Verbal complexes in Thai As discussed in Section 19.3, verbal complexes in Thai, which are considered to be analogous to “compound verbs,” have the structure [V1 (NP) V2], in which V1 can be either transitive or intransitive. However, there are many types of verbal constructions which fall under the scope of this structure. We will limit our data of verbal complexes to those which have additional formal properties, as discussed in Section 19.4.1.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.4.1 Additional formal properties of Thai verbal complexes We propose the following formal properties of verbal complexes in Thai in addition to the properties postulated in Section 19.3. I. If V2 is a transitive verb, it must not take any object. This syntactic property follows from the fact that V2 is grammaticalized to a certain degree as exemplified here. (11)

rew! phim khâw be.quick type  ‘Hurry! Type quickly!’

(12)

rawaŋ ná khăw cà wâa be.careful  he will scold ‘Be careful! He may scold (at you).’

(13)

sŏmsàk mòt Somsak be.gone ‘Somsak fainted.’

satìɁ consciousness

hây 

pay 

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

510

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The V2s in (11)–(13) are transitive verbs when used outside of the verb sequences. When they occur in the verbal complexes in (11)–(13), they are semantically bleached to a certain degree and have lost their syntactic properties as transitive verbs. We will examine the semantic contribution of V2 of verbal complexes later in this section. Notice that the same verbs occurring as V2, if combined with different verbs occurring as V1, can take objects as shown here. (14)

sŏmsàk dәәn khâw bâan Somsak walk  house ‘Somsak walked into the house.’

(15)

sŏmsàk khàp rót hây chăn Somsak drive car  I ‘Somsak drove a car for me.’

(16)

sŏmsàk sòŋ lûuk pay yîipùn Somsak send child  Japan ‘Somsak sent his child to Japan.’

It is apparent that V2s in (14)–(16) retain much of their original meaning when they occur in the verb sequences. Given the fact that they can take an object in (14)–(16) and that their lexical meaning is still retained to a certain extent, we can conclude that V2s in the three sentences above are inclined to be lexical verbs in contrast with those in (11)–(13). In other words, V2s in (11)–(13) are grammaticalized morphemes, whereas those in (14)–(16) are less grammaticalized, as evidenced by the fact that they can take objects after them. Whether they are highly grammaticalized morphemes or not obviously depends on which verbs they co-occur with. If directional verbs functioning as V2 appear with physical motion verbs, they are likely to be perceived as less grammaticalized morphemes, as in (14) and (16). In contrast, if they appear with nonmotion verbs, they are likely to be perceived as highly grammaticalized morphemes, as in (11) and (13). II. V2s cannot be negated, which means that the negative particle mây cannot intervene between V1 and V2. For example, V2s in (11)–(13) cannot be negated as shown in (17)–(19). (17)

*rew! be.quick

phim type

mây not

khâw 

(18)

*rawaŋ be.careful

ná 

khăw he

cà will

wâa scold

mây not

hây 

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

    (19)

*sŏmsàk Somsak

mòt be.gone

satìɁ consciousness

mây not

511

pay 

Sentences (17)–(19), in which V2s are negated, are ill-formed and uninterpretable. The fact that V2s in the three sentences above cannot be negated suggests that V2s in these sentences are grammaticalized to such a high degree that they have lost their verbal status and much of their lexical meaning. It is thus apparent that V2s do not play a major syntactic and semantic role in the sentences, which explains why V2s cannot be negated.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

III. Some of the V2s which have diphthongs are phonetically eroded, namely, lǽæw ‘finish’, sˇ1a ‘lose’, and pay ‘go’, as shown in (20)–(22). (20)

chǎn pay lá (lǽæw) I go finish ‘It is time for me to go.’

(21)

khâaw phəˆ ŋ sùk kin sá (sˇ1a) rice just cooked eat  ‘Rice has just been cooked. Eat now so that it will be over with.’

(22)

yàa phûut mâak tham pa (pay) don’t speak much do  ‘Don’t speak much. Go on doing.’

However, phonetic erosion does not occur with other verbs with diphthongs occurring as V2, namely, hây ‘give’, wáy ‘keep’, ‘khâw ‘enter’, aw ‘get a hold of ’, and lәәy ‘pass’. It can be speculated that the verbs lǽæw ‘finish’, sˇ1a ‘lose’, and pay ‘go’ have a relatively high frequency of occurrence as V2 in verbal complexes; therefore, they are likely to be phonetically eroded.

19.4.2 Semantic properties of V2s in Thai verbal complexes It is found that V2s in Thai verbal complexes are drawn from a limited set of verbs, as shown in (23). (23)

yùu lǽæw pay maa khâw ɔ̀ɔk

‘live, be located at’ ‘finish’ ‘go’ ‘come’ ‘enter’ ‘exit’

hây wáy sˇ1a lәәy aw duu

‘give’ ‘keep’ ‘lose’ ‘pass’ ‘get a hold of ’ ‘look’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

512

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Notice that all of the Thai lexical verbs which can function as V2 in verbal complexes listed in (23) denote concrete actions or processes. This observation corresponds with Heine, Claudi, and Hünnemeyer’s (1991) claim that the source concepts that are used for grammaticalization in most cases are concrete objects, processes, or locations. They are characterized as fundamental elements which are used frequently and generally in a typical speech situation. In the case of verbs, Sweetser (1988) makes the observation that verbs which are semantically commonest, represent the “basic vocabulary,” and correspond to superordinate categories in the level of categorization, such as those glossed as ‘go’, ‘come, ‘give’, ‘be’, ‘take, hold’, ‘have’, ‘stand’, ‘sit’, ‘stay’, ‘say’, are likely to be sources of grammaticalization. These “basic” verbs are largely cultureindependent. These characteristics hold true for the verbs which can function as V2 in verbal complexes listed in (23). Note that many of these verbs describe physical motions, especially directional ones, which are known to be common source concepts of grammaticalization across languages. It is found that the verbs which function as V2 listed here can be classified into two types based on the kind of aspectual meaning they exhibit, as discussed next. The first type is termed in this chapter “grammatical aspect vector verbs” and the second type is termed “pragmatically oriented vector verbs.”

19.4.2.1 Grammatical aspect vector verbs This class of vector verbs, which function as V2, consists of three verbs from the list given, namely, yùu ‘live, be located at’, lǽæw ‘finish’, and maa ‘come’, all of which express grammatical aspects. Grammatical aspects are defined by Comrie (1976: 3) as “different ways of viewing the internal temporal constituency of a situation.” Yùu indicates the continuative aspect, which refers to the aspect that expresses an incomplete action or state that is happening or in progress at a given time, as exemplified in (24). What is special about the continuative aspect marker yùu in Thai is that it expresses stativity of the situation described by a main verb. On the other hand, lǽæw and maa indicate different types of perfect aspect. According to Comrie (1976), the perfect indicates the continuing present relevance of a past situation. It expresses a relation between two time points, namely, the time of the state resulting from a prior situation and that of the prior situation. The perfect is rather different from other aspects in that it tells us nothing directly about the situation in itself, but rather relates some state to a preceding situation. According to Comrie’s classification of the perfect, lǽæw, the source meaning of which is ‘finish’, is categorized as the perfect of result, which indicates that a present state results from a past situation, as shown in (25). The other marker, maa, the source meaning of which is ‘come’, indicates the perfect of persistent situation, which describes a situation that started in the past but continues into the present, as shown in (26). Each of the vector verbs in the examples below is glossed as V2.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

    (24)

sŏmsàk àan nǎŋsɯɯ yùu Somsak read book V2 ‘Somsak is reading a book.’

(25)

sŏmsàk tham kaanbâan lǽæw Somsak do homework V2 ‘Somsak has already done homework.’

(26)

sŏmsàk dәәn maa sǎam chûamooŋ Somsak walk V2 three hour ‘Somsak has been walking for three hours.’

513

Note that it is more natural for the aspect marker maa, which expresses the perfect of persistent situation, to be accompanied by a time specification phrase such as sǎam chûamooŋ ‘for three hours’, as in (26). We note that there is a relatively high degree of semantic transparency in the three V2s in (24)–(26). In other words, it is relatively easy to understand how the aspectual meanings of the three V2s are derived from their source meanings. The semantic transparency is the most obvious in the case of lǽæw, the source meaning of which is ‘finish’ and the grammaticalized meaning of which is the perfect of result. The action of finishing is naturally related to the concept of result. As for the continuative aspect marker yùu, it was mentioned above that yùu implies stativity of the situation described by a main verb. The source meaning of yùu, which is concerned with a locative state of an entity, is metonymically related to the sense of stativity, which is incorporated in the aspectual meaning of yùu as a continuative marker. As for maa, it lexically expresses a motion of an entity toward the speaker’s location or point of view. It is considered a deictic verb because its interpretation is partially dependent on the speech situation in which this verb is used. In this case, the relevant element in the speech situation is the speaker’s location or point of view. The aspectual meaning of maa functioning as V2 is the perfect of persistent situation, which indicates that a situation took place at some point prior to the present time and proceeded continuously until the present time. The proceeding of a situation from the past until the present can be analogized to a movement of an entity further from the speaker toward the speaker’s present location. Notice that if the three V2s are omitted from the sentences, the aspectual meanings will be lost and, consequently, the propositional meaning of the sentences with omitted V2s will be affected. In other words, the sentences with unspecified aspectual meanings will be semantically different from those with aspectual meanings specified. Since yùu, lǽæw, and maa exhibit aspectual meanings which are easily identifiable, these three verbs can be considered postverbal auxiliaries. These three postverbal auxiliaries cannot be negated, unlike two other grammaticalized

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

514

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

postverbal auxiliaries, namely dây and pen, which are usually called modal auxiliaries. The source meanings of dây and pen are ‘acquire’ and ‘be alive’, respectively. The modal auxiliary dây indicates ability, permission, and accomplishment, whereas pen indicates ability or knowledge of how to perform a certain action. The examples below show the ill-formedness of yùu, lǽæw, and maa, on the one hand, and the well-formedness of dây and pen, on the other hand, when all of them occur with the negative particle. Note that the ill-formed examples (27)–(29) are uninterpretable. (27)

*sŏmsàk Somsak

àan read

nǎŋsɯ̌ ɯ book

(28)

*sŏmsàk Somsak

tham do

kaanbâan homework

mây not

(29)

*sŏmsàk Somsak

dәәn walk

mây not

sǎam three

(30)

sŏmsàk kûu ŋәn thanakhaan Somsak get.a.loan money bank ‘Somsak cannot get a loan from the bank.’

(31)

sŏmsàk wâaynám mây pen Somsak swim not  ‘Somsak cannot (does not know how to) swim.’

mây not

maa V2

yùu V2 lǽæw V2 chûamooŋ hour mây not

dây 

It is apparent that although yùu, lǽæw, maa, dây, and pen are all postverbal auxiliaries, they form two groups of postverbal auxiliaries which are both semantically and syntactically distinct. Yùu, lǽæw, maa, which are called vector verbs in this chapter, indicate grammatical aspectual meanings, whereas dây and pen indicate modality meanings. The vector verbs cannot be negated, whereas the modal auxiliaries can.

19.4.2.2 Pragmatically oriented vector verbs The remaining vector verbs in (23) form the second kind of vector verbs. Sentences (32)–(41) exemplify the use of the second kind of vector verbs, which are shown in bold. The source meaning of each V2 or each vector verb is put underneath in the examples. (32)

kin pay! yàa phûut mâak eat  don’t talk much ‘Eat (immediately, without delay)! Don’t talk much.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

   

515

(33)

yàa pìtbaŋ bɔ̀ɔk maa d1ˇawníi don’t hide tell  now ‘Don’t cover it up. Tell (me) now.’

(34)

mæˆæ sɔ̌ɔn wáy wâa yàa fûmfɯay mother teach ...  don’t be.extravagant ‘(My) mother taught (for future use) (me) not to be extravagant.’

(35)

kraprooŋ tua níi suˇ ay ɔ̀ɔk skirt  this be.beautiful  ‘This skirt is evidently beautiful (contrary to what you think).’

(36)

yàa yûa khăw dˇ1aw khăw dàa hây don’t provoke him in.a.moment he scold  ‘Don’t provoke him. He might scold (at you, which would be unfavorable for you).’

(37)

kin duu lǽæw caɁ ruú wâa arɔ̀y eat  then will know  delicious ‘Eat (it) and you will know that it is very delicious.’

(38)

rew! phim khâw be.quick type  ‘Hurry! Type quickly!’

(39)

khǎw dәәn aw dәәn aw dooy mây mɔɔŋ khray lәәy he walk  walk  by not look anybody at.all ‘He walked and walked (seriously and attentively) without looking at anybody.’

(40)

ˆ aphâa s1ˇa! sák sɯ yàa kèp wáy wash clothes  don’t keep ... ‘(You had better) Wash (your) clothes now (so that it will be over with)! Don’t keep them.

(41)

kin ləəy! mây tɔˆ ŋ rɔɔ eat  not must wait ‘Eat now (without delay)! You don’t have to wait.’

mâak very

V2s of the second kind are characterized as follows. (i) The V2s predominantly express a varied array of pragmatic meanings. However, it is not easy to spell out what pragmatic meaning each V2 indicates. Furthermore, it is almost impossible to translate each V2 into

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

516

    

(ii)

(iii)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(iv)

(v)

another language. What we did in the sentence translation of each sentence containing V2 in (32)–(41) was mainly trying to convey the speaker’s intention in making each utterance or trying to identify the context of each sentence. The V2 in each sentence can be omitted without any part of the propositional meaning of the sentence being lost.⁵ This property of V2s arises from the fact that V2s indicate pragmatic meanings. However, if a V2 is omitted, the sentence without it will sound unnatural. In most cases, sentences with a V2 must be used in marked contexts, not in neutral contexts. Some examples of marked contexts, as seen in (32)–(41), are a command context, as in (32), (33), (38), and (41), a suggestion context, as in (37) and (40), a context in which the speaker does not agree with the addressee’s prior belief, as in (35), a context in which the action described by the main verb is undesirable for the addressee, as in (36), and a context in which the action described by the main verb must be carried out without delay, as in (32), (38), and (41). The aspectual meanings they indicate are not as easily identifiable as those of the first kind of vector verbs. In other words, it is not always clear what aspectual meaning each vector verb of the second type expresses. As mentioned in Section 19.2.1, some Russian researchers such as Rastorgueva and Kerimova (1964: 212) claim that V2s express aspect but not aspect in the usual sense of perfective and imperfective. We can see that it is quite hard to identify the aspectual meaning of V2 in some examples, such as in (35), (36), (37), (38), and (41), whereas it is easier in other examples, such as in (32), (34), (39), and (40). In the latter case, V2 in (32) indicates that the speaker ordered that the addressee set out to eat without delay, which seems to correspond to the inceptive aspect. V2 in (34) implies that the mother had finished teaching or giving advice and the addressee remembered the mother’s advice for future use. V2 in (34) apparently indicates the perfective aspect. V2 in (39) indicates that the subject kept on walking or walking “repeatedly,” which corresponds to the iterative aspect. V2 in (40) implies that the speaker suggested that the addressee wash clothes until completion, which corresponds to the perfective aspect. However, the aspectual meanings of these V2s are not conventionalized meanings as those of V2 in the first type but have to be inferred from the sentence meaning and from context. The second type of V2s are less semantically transparent than the first type of V2s. In other words, it is more difficult to see semantic connections between the source meanings and the grammaticalized meanings of the

⁵ Hook (Chapter 11, this volume) argues that “vectors” may be omitted without affecting the propositional meaning.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

517

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

second type of V2s. However, among the V2s of the second type, some are more semantically transparent than others such as those in (32)–(37) repeated here for convenience as (42)–(47). (42)

kin pay! yàa phûut mâak eat  don’t talk much ‘Eat (immediately, without delay)! Don’t talk much.’

(43)

yàa pìtbaŋ bɔ̀ɔk maa d1ˇawníi don’t hide tell  now ‘Don’t cover it up. Tell (me) now.’

(44)

mæˆ æ sɔ̌ɔn wáy wâa yàa fûmfɯay mother teach ...  don’t be.extravagant ‘(My) mother taught (for future use) (me) not to be extravagant.’

(45)

kraprooŋ tua níi suˇ ay ɔ̀ɔk skirt  this be.beautiful  ‘This skirt is evidently beautiful (contrary to what you think).’

(46)

yàa yûa khăw dˇ1aw khăw dàa hây don’t provoke him in.a.moment he scold  ‘Don’t provoke him. He might scold (at you, which would be unfavorable for you).’

(47)

kin duu lǽæw caɁ ruú wâa arɔ̀y eat  then will know  delicious ‘Eat (it) and you will know that it is very delicious.’

mâak very

In (42), the source verb pay expresses physical motion away from the speaker’s position or the speaker’s point of view. The predicate in (42) realized as a verbal complex with the V2 pay indicates that the speaker ordered that the addressee start eating immediately. The fact that the addressee, who was presumably with the speaker at the time of utterance, started eating and the eating action proceeded for some time can be analogized with a motion of entity away from the speaker. In (43), the speaker ordered that the addressee tell him/her something. The fact that the addressee’s message reached the speaker can be compared to physical motion of an entity from the addressee toward the speaker, which is the source meaning of maa. The verbal complex in (44) indicates that the mother gave a piece of advice to the addressee so that the addressee could keep it for future use. In this case, the semantic transparency between the source meaning of wáy, namely, ‘keep for future use’, and the semantic contribution of wáy as V2 is obvious. The verbal complex in (45) indicates that the subject (the skirt) is evidently beautiful,

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

518

    

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

contrary to what the addressee thinks. The use of the V2 ɔ̀ɔk, the source meaning of which is ‘exit’, is motivated by the fact that the beauty of the skirt is evident or “comes out” to the speaker’s eyes. In (46), the action of scolding the addressee is sarcastically compared to the action of giving something to him/her. The verbal complex in (47) expresses a suggestion that the addressee should eat something and he/she would find it to be very delicious. The V2 duu implies that the addressee could “look” and “see” for himself/herself after eating that the thing recommended by the speaker was very delicious. As for (38)–(41) repeated here for convenience as (48)–(51), the semantic relationships between the source meanings of V2s and their semantic contribution in the verbal complexes in these sentences are rather obscure. Therefore, it is hard to explain why the V2s are used in the verbal complexes of these sentences. (48)

rew! phim khâw be.quick type  ‘Hurry! Type quickly!’

(49)

khǎw dәәn aw dәәn aw dooy mây mɔɔŋ khray lәәy he walk  walk  by not look anybody at all ‘He walked and walked (seriously and attentively) without looking at anybody.’

(50)

ˆ aphâa sák sɯ s1ˇa! yàa kèp wáy wash clothes  don’t keep ... ‘(You had better) Wash (your) clothes now (so that it will be over with)! Don’t keep them.’

(51)

kin ləəy! mây tɔˆ ŋ rɔɔ eat  not must wait ‘Eat now (without delay)! You don’t have to wait.’

It can be concluded that more elements of the source meanings of V2s in (41)–(45) are retained when they appear in verbal complexes than those of V2s in (48)–(51). That explains why the semantic contribution of V2s in (42)–(47) is more transparent. However, semantic bleaching of V2s in verbal complexes is a matter of degree, which accounts for the fact that some V2s in (48)–(51) are felt to be more semantically transparent than some others.

19.4.3 Summary We propose the structure [V1 (NP) V2], in which V1 can be either transitive or intransitive verb, for verbal complexes in Thai. The vector verbs or V2s in Thai

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   

519

verbal complexes are grammaticalized morphemes and are drawn from a limited set of verbs which are used frequently and generally in typical speech situations. Thai verbal complexes can be characterized based on formal and semantic properties of V2 as follows. Formally, a V2 that is lexically a transitive verb cannot take an object in a verbal complex nor can it be negated. Moreover, some V2s are phonetically eroded. V2s can be semantically classified into two types, namely, grammatical aspect vector verbs and pragmatically oriented vector verbs. The former type of vector verb includes yùu, lǽæw, and maa, the glosses of which are ‘live, be located at’, ‘finish’, and ‘come’, respectively. These three vector verbs indicate different grammatical aspects which are relatively easy to identify. In addition, they have a high degree of semantic transparency. The latter type of vector verb serves to provide pragmatic specification to the situation described by a main verb. However, the pragmatic meanings of these vector verbs are varied and rather difficult to spell out. The aspectual meaning of the second type of vector verb is much harder to identify than that of the first type of vector verb. Some vector verbs of the second type are more semantically transparent than others. It can be concluded at this point that the vector verbs in Thai have overlapping semantic properties with those in “compound verbs” in HindiUrdu, as postulated by Masica (1976), discussed in Section 19.2.1.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

19.5 Conclusion In this chapter, we have argued that although Thai is a language whose morphosyntactic structures are different from those of the rest of the languages examined in this volume, it contains a group of grammaticalized verbs in the postverbal position in a verbal complex. These postverbal grammaticalized verbs are called in this chapter “vector verbs,” and are characteristically similar to vector verbs in other languages, notably in Hindi-Urdu. Structurally speaking, Thai vector verbs depart drastically from those in other languages which are defined as the second verbs of V-V complexes, i.e. predicates consisting of two (or more) contiguous verbs. Thai vector verbs do not have to be contiguous with the preceding main verb because there can be an object NP of the first verb intervening between the two verbs. Therefore, we have postulated the notion of verbal complexes in place of V-V complexes to accommodate the Thai data. We have also identified language-specific syntactic criteria to distinguish vector verbs from other types of second verbs in verbal complexes in Thai. Thai vector verbs, thus identified, are described and illustrated with typical examples in this chapter. They are subclassified into two subclasses: grammatical aspect and pragmatically oriented vector verbs, based on the type of salient semantic element, i.e. grammatical aspect or pragmatic meaning, their omissibility, and degrees of semantic transparency. It has been demonstrated in this

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

520

    

chapter that Thai vector verbs have features both semantically and grammatically overlapping with those examined in the other languages in this volume.

Acknowledgments

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The authors would like to thank Taro Kageyama and John Haig for their helpful comments and their suggestions on textual improvements. This research is supported by the Grant for Research and Creative Works in Humanities, Faculty of Arts, Chulalongkorn University, (No. FAR22F010905) awarded to the first author and the second author’s Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (No. 16K02618).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

0

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

20 Verb-verb sequences in Mandarin and Hindi-Urdu Hsin-hsin Liang and Peter E. Hook

20.1 What is a compound verb?

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The phrase “compound verb” is a traditional term of art in South Asian linguistics literature. On the basis of shared properties in the languages of Northeast Asia (Japanese, Korean), Central Asia (Altaic, Avar), and South Asia, the compound verb may be defined as a sequence of two verbs (main verb “V” plus vector verb “v”) alternating with main verb used alone with little difference in meaning easily translatable into languages which do not have compound verbs (such as Sanskrit, French, or Spanish). Throughout this chapter¹ we will refer to concatenations of “V-Vs” or “verb-verb sequences” meeting this minimal definition as “compound verbs” or “CVs.” In these examples from Hindi-Urdu the SV [“simple verb”] ātā hai ‘comes’ (1a) alternates with the CV ā jātā hai ‘comes’ (1b): (1)

a. rāt2 ko yahãã roz pānī pī-ne ā-tā hai night  here daily water drink- come-  ‘Every night he (a leopard) comes here to drink water.’ (Godan 86.5 (See Premchand 1936) b. kabhī.kabhī dopahar mẽ bhī ā sometimes afternoon in also come ‘Sometimes he comes in the afternoon, too.’

jā-tā -

hai  (Godan 86.6)

Though the meaning of the non-CV (or SV) ātā hai in (1a) is not quite the same as that of the corresponding CV ā jātā hai in (1b), the difference in their meanings is not one that is easy to articulate or to translate into English. The use of the

¹ An earlier version of this chapter is available in Masica (2006). ² The transcription system used here for displaying data from Hindi-Urdu is one commonly found in the linguistics literature on modern South Asian languages. In it macrons and doubling of symbols indicate contrastive length. The digraphs /ai/ and /au/ represent the front low and back mid vowels. Chinese data are given in Pinyin.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

522

-    . 

vector jā- GO³ in (1b) may have to do with the speaker’s wish to amend or counter the meaning expressed in (1a).⁴ In any event the meaning of the vector jā- GO is not the same as the meaning of the full lexical verb jā- ‘go’ in example (2), taken from a ghazal by Dāgh: (2)

thake sab payāmī vahãã jā.te-jā.te yahãã ā.te-ā.te tired all messengers there going-going here coming-coming ‘All the messengers were exhausted going there, coming here.’

In (1b) the noncontradictory presence of vector jā- (which as a full lexical verb is an antonym of the main verb ā- ‘come’) reflects the semantic bleaching characteristic of grammaticalization.⁵

20.2 The geographical distribution of compound verbs in Asia

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

With the exceptions of Pashto, the isolate Burushaski, Shina of Gilgit, Palula, and some other Indo-Aryan languages of the Hindu Kush,⁶ examples of compound verbs in alternation with corresponding noncompound verbs can be adduced from most members of every South Asian language family (Indo-Aryan, Dravidian, Austro-Asiatic [Mundan], Tibeto-Burman,⁷ Iranian [Balochi], Kusunda [isolate]),

³ Capitals are used to indicate meanings of basic lexical items from which homophonous vectors are derived. ⁴ Other examples of compound verbs being used to amend as in (i), even cancel, as in (ii), a prior statement: (i)

is rog-mẽ prāy: jvar nahĩĩ ho-tā, kabhī.kabhī ho bhī jā-tā hai this disease-in usually fever not occur- sometimes occu also -  ‘With this disease usually there is no fever; sometimes [however] there is [fever].’ (Pancharatna, A., Vyadhi – Vigyan, vol. 2) (ii) ye sač nahĩĩ ki tere išāre nahĩĩ samajh-tā hũũ samajh to le-tā hũũ this true not that your hints not understand- am understand  - am magar sāre nahĩĩ samajh-tā hũũ but all not understand- am ‘It’s not that I don’t get your hints. I do get them but I don’t get all of them.’ (http://smsstatusk.blogspot.com/) ⁵ Other prominent vectors in Hindi-Urdu include de- GIVE, le- TAKE, par: - FALL, dāl: POUR, and baiTh SIT. For a fuller list of them, see Hook (1974: 119–44) and Slade (Chapter 10, this volume). Nespital (2008: xvii) lists forty-seven. ⁶ See Liljegren (2016) for Palula. ⁷ Vv sequences of the South Asian type are found in West Tibetan Balti (Read 1934: 60, 70, 102), Ladakhi (Zeisler: Chapter 14, this volume), the Kiranti languages of East Nepal (Driem 1987: 118–32; Ebert, ms.), Lahu (Matisoff 1973), Dzongkha, the national language of Bhutan (Driem 1992: 236-46), and Nangchenpa (Amdo). Limbu is a special case: Its Vvs show full desinence on both the main verb and the vector: (i) ku-sikla-?o putt-u-ŋ bīr-u-ŋ his-throat- grab-3.-1. -3.-1. ‘I grabbed him by the throat.’ (Driem 1987: 129, ex. 118)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

523

as well as from languages spoken in Central Asia, as in (3a, b), Northeast Asia, as in (4a, b), and the Eastern Caucasus:⁸ (3)

a. men şındıq-tı аyt-ıм-ın I truth- tell-1- ‘I will tell the truth.’ b. men şındıq-tı ayt-ıp I truth- tell- ‘I will tell the truth.’

(4)

(akikat.kazgazeta.kz/?p=6437)

ber-em-in -1-

a. motu ič-eoss-ta everything forget-- ‘(I) forgot everything.’ b. motu ič-eo everything forget- ‘(I) forgot everything.’

(Kazakh)

(data found online)

(Korean) (gsis.snu.ac.kr/facultyarticle)

peoli-eoss-ta -- (http://m.blog.daum.net/)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

20.3 Verb-verb sequences in Mandarin Chinese As functional and formal equivalents of the compound verb in South Asian, Central Asian, and Northeast Asian languages there are three candidates in Chinese: main verb plus either a resultative, a directional auxiliary, or a second verb. Some of these may be comparable to vector verbs; while others (e.g. míngbai in nòng míngbái ‘make (something) clear’ ! ‘realize’) clearly do not meet the criterion of relative omissibility (see (5)). For us the category of resultative includes items like bái ‘white’, míngbai ‘clear’, baˇo ‘full’, and zuì ‘drunken’, which correspond to adjectives in most South Asian and other languages. In order to limit the set of putative compound verbs to those sequences in which the second element is not a stative verb or adjective like bái ‘white’, míngbai ‘clear’, and zuì ‘drunken’, we will exclude all items that can occur as complements of the adverb hěn ‘very’ as being too close to the adjectival pole.⁹ (5)

xiàn.zài wˇo nòng now I do ‘Now I understand.’

*(míng.bái) 

le .

⁸ For Central Asia, see Chapters 15 (Shluinsky), 16 (Ohsaki and Ebata), and 17 (Kuribayashi). For the Eastern Caucasus, see Chapter 18 (Yamada) on Avar. Compound verbs alternating with noncompound counterparts are also found in some languages of northern Australia. See Wilson (1999: 82–3) and Merlan (1994). ⁹ This test is somewhat “over-exclusive”: For instance, polysemous items with both a verbal and an adjectival meaning (such as haˇo ‘do; be done’ and ‘good’) on the basis of their adjectival meaning will not pass it.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

524

-    . 

20.4 Parallels between Hindi-Urdu Vvs and Chinese Vvs In this section we present and illustrate ways in which verb-verb sequences in Mandarin are similar to the compound verb in Hindi-Urdu. (In Section 20.5 we will look at ways in which they differ.)

20.4.1 The verb z oˇ u as a main verb means ‘move’, ‘walk’, or ‘depart’. As a vector (or factor¹⁰) it optionally combines with verbs that express manner of motion such as fēi ‘fly’, paˇo ‘run’, tiào ‘hop’, táo ‘flee’, pá ‘crawl; slither’, and chuī ‘blow’ to express motion away: (6)

a. laˇo.yīng eagle

fēi fly

zˇou 

le 

b. laˇo.yīng fēi le eagle fly  ‘The eagle flew away.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(7)

a. fàn.rén prisoner

táo flee

zˇou 

le 

b. fàn.rén táo le prisoner flee  ‘The prisoner escaped.’ Depending on semantic and syntactic environment vector, z oˇ u MOVE alternates with its absence. In the progressive aspect (V + zài), for example, it cannot occur: (6)

c. laˇo.yīng zài fēi *(zˇou) eagle is.at fly  ‘The eagle is flying.’

(7)

c. fàn.rén zài táo *(zˇou) prisoner is.at flee  ‘The prisoner is escaping.’

¹⁰ A “factor verb” is one that has not been grammaticalized (bleached of its lexical meaning) to the same extent as a vector verb, yet does not express a meaning temporally independent of that expressed by the verb that it follows.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

525

Compare Hindi-Urdu, in which the progressive also disprefers the occurrence of a vector:¹¹ (8)

(9)

čīl ur: (*jā) rahā hawk fly  -ing ‘The hawk is flying.’

hai is

qaidī bhāg (*jā) rahā prisoner flee  -ing ‘The prisoner is escaping.’

hai is

20.4.2 In Hindi-Urdu, opposing vectors may be used to create pairs of compound verbs alike in every respect except their orientation. For instance, in the following pair the use of vector de GIVE as opposed to vector le TAKE with the main verb rakh ‘put; keep’ accords with the intention to put something out for others, as in (10), versus the desire to keep something for oneself hidden away from others, as in (11):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(10) tab-tak thor: āsā bhūsā nikāl-kar rakh then-until a.little straw take.out- put ‘In the meantime, put out a little straw (for Bholā).’

do  (Godan 23.23)

(11) gār: -kar rakh lo to kaun dekh-tā hai bury- put  then who see- is ‘If you bury it (keep it underground), then who will see it?’ (Godan 219.28) In Chinese, a similar contrast is possible using the outwardly versus inwardly directed vectors diào ‘drop, fall’ versus xià ‘descend,¹² come down’:

¹¹ The Hindi-Urdu compound verb does very occasionally occur in the progressive aspect: (i) ek čīl naulakhā.hār le-kar ur: -ā one hawk necklace take- fly-.? ‘A hawk was flying with a necklace (in its beak).’

jā-rahā -ing

thā was (Godan, 135.46)

Note the difference in morphology of main verb ur: - from that seen in nonprogressive compound verbs. For a brief discussion of progressive CVs, see Hook (1974: 258–60). ¹² The antonymy of diào versus xià is not so easily seen as that of de GIVE versus le TAKE. While both diào and xià mean ‘descend’ or ‘fall’, the descent denoted by diào is typically away from the subject, the topic, or the speaker (cf. the hair in (i)), while that denoted by xià is typically toward the subject, topic, or speaker (cf. the rain in (ii)): (i)

tóufa dōu { diào / hair all fall.away ‘All his hair fell out.’

*xià } / fall.toward

le 

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

526

-    . 

(12) baˇ fáng-zi zū diào ba  house rent   ‘Rent (out) the house (for others to live in).’

(Teng 1977: 16)

(13) baˇ fáng-zi zū xià ba  house rent   ‘Rent the house (for yourself to live in).’

(Teng 1977: 16)

The feasibility of arranging vector (and factor; see Hook 2014) verbs in sets and subsets on the basis of the semantic properties of the full lexical verbs with which they are homophonous depends on the distance they have traveled along an imagined path to grammaticalization. With the exception of Hindi-Urdu’s vector jā GO,¹³ it would seem that the more easily vectors can be put into such sets and subsets, the more they retain the semantics of their full lexical counterparts and the less grammaticalized they are.

20.4.3

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Both in Hindi-Urdu and in Chinese an unmodified (or noncompound) Accomplishment¹⁴ verb may be susceptible to a noncompletive or conative interpretation, as in this song from the film Birādarī: (14) tasavīr banā-tā hũũ; tasavīr nahĩĩ ban-tī picture make- am picture not be.made- ‘I try to picture (her) but the picture won’t form.’

(ii) yuˇ yˇ1.jīng {xià / rain already fall.toward / ‘It’s already rained.’

*diào} fall.away

le 

If the situation itself is atypical, the characteristic directionality can sometimes be canceled or reversed by using a vector verb: (iii) huī.chén dōu diào xià le dust all fall.away .  ‘All the dust fell down (toward me or the duster).’ (iv) yuˇ xià xià qù le rain fall.toward .   ‘It rained down away (from me standing on a cliff above the clouds).’ ¹³ In descriptions and lists of Hindi-Urdu’s vector verbs, vector jā GO is often twinned with vector ā COME. But ā, much more rarely figuring in CVs, is less grammaticalized than jā. ¹⁴ We are assuming that the reader has some familiarity with the Aristotelian-Vendlerian classification of verbs by their inherent aspectual properties as States, Activities, Achievements, and Accomplishments. For discussions of these categories, see (among others) Dowty (1979) and Van Valin (1990).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

527

Use of banā le (the compound verb corresponding to banā) precludes conative interpretation.¹⁵ (15) tasavīr banā-ī lekin tasavīr nahĩĩ ban-ī picture make- but picture not be.made- ‘I tried to picture (her) but the picture wouldn’t form.’ (16) tasavīr banā lī (*lekin tasavīr nahĩĩ ban-ī) picture make  but picture not be.made- ‘I pictured (her) (*but the picture wouldn’t form).’ In Mandarin Chinese the tendency of an unmodified Accomplishment verb to be interpreted conatively is, if anything, even stronger¹⁶ than it is in Hindi-Urdu: (17) wˇo huà-le huà kěshì méi huà haˇo I paint- picture but not picture done ‘I tried to paint a17 picture but the picture wouldn’t paint.’ (18) wˇo huà haˇo-le huà (*kěshì méi huà I paint - picture but not picture ‘I painted a picture (*but the picture wouldn’t paint).’

haˇo) done

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Work by Pederson (2008) on Tamil suggests that the susceptibility of noncompound forms of Accomplishment verbs to conative interpretation follows from

¹⁵ The meaning of completeness inherent in the compound verb is responsible for SV/CV pairs that occasionally correspond to lexically distinct pairs of verbs in languages like English. Thus, dhũũr : h le : [search TAKE] ‘find’ expresses the successful outcome of the action dhũũr : h- ‘search for’: : (i)

ek.bār us-ne us šīšī-ko dhũũr thā par : hā bhī : once he- that bottle- searched also was but ‘Once he had even looked for that bottle but it wasn’t to be found.’

nahĩĩ milī not turned.up (Bhandari 1979: 127)

(ii) ek.bār us-ne us šīšī-ko dhũũr : h liyā bhī thā (*par nahĩĩ mili) : once he- that bottle- search  also was but not turned.up ‘Once he had even found that bottle but it was not to be found.’ (= a contradiction) ¹⁶ In fact, according to Ross (1990: 64), to avoid conative interpretation all Accomplishment verbs in Chinese must be compounded. For discussion of “entailment cancelation” in Japanese Accomplishment verbs (moyashita, keredo moenakatta. ‘I burnt it but it didn’t burn.’), see Kato (2018: 174). The phenomenon seems not to exist in English. ¹⁷ Susceptibility of particular instances of huà ‘paint’ to conative interpretation depends too on properties of object: (i)

wˇ o huà le liăng-fu huà (*kěshì méi I paint  two- pictures but not ‘I painted two pictures (but the pictures didn’t get done).’

huà pictures

haˇ o) be.done

However, the point we wish to make is that a conative reading of the noncompound form is, under at least some conditions, possible. For the effect of classifiers on the interpretability of an Accomplishment verb as conative, see Soh and Kuo (2005).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

528

-    . 

inferences made by the language-learning child, who interprets each part of the compound verb as referring to a discriminate phase of telic events. However, as we show in Section 20.5.4, Chinese and Hindi-Urdu differ markedly in the extent to which the vector in and of itself expresses the telos of an action.

20.4.4 In Hindi-Urdu the compound versus noncompound alternation is suppressed in the phasal verb construction. Infinitival complements of verbs of phase (V-ne lag‘begin to V’, V-tā rah- ‘keep V-ing; continue to V’, V-nā čhor: - ‘stop V-ing’) cannot appear in compound form: (19) mazdūrõ-ne divār gir-ā dī workers- wall fall-  ‘The workers knocked down the wall.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(20) mazdūr divār gir-ā (*de) -ne lage workers wall fall-   began ‘The workers began to knock down the wall.’ In Chinese, depending on the class of the main verb, a phasal meaning is rendered in various ways. With Activity and Ektelic¹⁸ verbs, the “change-of-status” particle le or vectors¹⁹ are used to indicate phase. However, with verbs of Accomplishment full phasal predicates (e.g. kāishıˇ ‘begin to V’) may be used. As (22) shows, use of kāishˇ1 suppresses CVs: (21) gōng.rén zá daˇo 1e qiáng workers knock   wall ‘The workers knocked down the wall.’ (22) gōng.rén kāi.shˇ1 zá (*daˇo) qiáng workers begin knock  wall ‘The workers began to knock down the wall.’

¹⁸ Ektelic verbs are those whose telos or limit precedes (rather than follows a prefatory) activity: flee, leave (behind), keep, etc. Karen Ebert (pers. comm.) suggests “initiotransformative.” ¹⁹ For instance, the two-part lexicalized sequence xià-qù DESCEND, GO DOWN may be used with Activities in the sense of ‘keep on V-ing, continue’: (i)

shuō xià nˇ1 you speak  ‘Go on talking!’

qù 

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

(Teng 1977: 21)

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

529

If one thinks of the compound form of an Accomplishment verb as expressing an inherently complex situation comprising two phases, a prefatory Activity followed by the Achievement of a new State, then it follows that phasal verbs like HindiUrdu lagnā and Mandarin kāishˇ1 ‘begin’ may select only that part of the compound verb that most clearly refers to the prefatory Activity. For this reason phasal verbs, refusing compound verbs as their infinitival complements, take their noncompound counterparts instead.

20.4.5 In Hindi-Urdu there is a marked (if not absolute) preference for CVs in clauses that are complements of the conjunction jab tak ‘until’.²⁰ (23) yahãã tab-tak ruko jab-tak vo divār na girā dẽ here then-till stop when-until they wall  fall. . ‘Wait here until they knock down the wall.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

In Chinese the facts are more nuanced: Either compound or noncompound is possible depending on the situation. In (24) the demolition team has yet to arrive. In (25) they are present but have not yet succeeded in knocking down the wall. (24) děng zài zhèr zhí.dào tā.men zá qiáng wait at here until they knock wall ‘Wait here until they (come and) knock down the wall.’ (25) děng zài zhèr zhí.dào tā.men zá wait athere until they knock  ‘Wait here until they knock down the wall.’

daˇo wall

qiáng

²⁰ The negative particle in (23) is an expletive or otiose element. See Gaeffke (1967: 33–6) and Hook (1974: 217–20). It occurs in parallel constructions in Russian and in French: (i)

zvonite

poka

ne

otvetjat

ring.

until



answer.

‘Ring until they answer.’ (ii) partez avant qu’il leave. before that’it ‘Leave before it’s too late!

(Forsyth. 1970: 133) ne soit trop  be. too

tard ! late. (propx.skyrock.com/)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

530

-    . 

20.4.6 In both Hindi-Urdu and Mandarin Chinese complements of expressions of fear and anxiety strongly favor use of Vvs: (26) ham dar-rahe the kahĩĩ qaidī bhāg : we fear-ing were lest prisoner flee ‘We were afraid the prisoner might escape.’ (27) wˇo.men pà nà.ge fàn.rén huì táo we fear that prisoner could flee ‘We were afraid the prisoner might escape.’

na 

jā.e 

zˇou 21

Expressions of fear seem to focus attention on (unwanted) outcomes of action. This may explain the preference for Vv or V le or both in negative imperatives (28) and in warnings (29):²² (28) bié baˇ qiáng zá (*daˇo) le / . . . zá daˇo (*le) don’t  wall knock   / knock   ‘(Make sure you) don’t knock down the wall!’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(29) xiaˇo-xīn fàn.rén táo (*zˇou) le / . . . táo zˇou (*le) small-heart prisoner flee   / flee   ‘Watch out; the prisoner may escape!’ The data in these subsections (20.4.5 and 20.4.6) indicate that the alternation of compound with noncompound verb in both Hindi-Urdu and Mandarin Chinese has acquired at least some ability to express aspectual contrast of the “viewpoint” sort (see Smith 1997). Compare the reactances of perfective versus nonperfective forms to the corresponding environments in Russian²³ (Forsyth 1970: 259; Rassudova 1984: 174–5, 190–2; Hook 2001b: 215).

²¹ Notice the dispreference for the compound verb in the negative counterpart of (29): (i)

woˇ.men pà nà.ge fàn.rén bú (huì) táo we fear that prisoner not could flee ‘We were afraid the prisoner might not escape’. [i.e. We wanted him to escape.]

²² This suggests that in expressing fear or giving a warning either a vector or the perfective marker le does the same job. *боялся, как бы не опаздывать ²³ боялся, как бы не опоздать bojalsja kak.by ne opozdat’ bojalsja kak.by ne opazdyvat’ was.afraid lest  be.late. was.afraid lest  be.late.~ ‘X was afraid he’d be late.’

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

531

20.5 Ways in which the Hindi-Urdu and Mandarin Chinese verb-verb sequences differ 20.5.1 The vector verbs in Hindi-Urdu are highly grammaticalized; those of Chinese, less so. At times the elements in a Chinese Vv sequence express sequential components of a complex action, as in (30); at other times they express synchronous components, as in (31): (30) gōngrén baˇ qiáng zá daˇo workers  wall knock  ‘The workers knocked down the wall!’ (31) tā cóng chē.zhàn zˇou kāi he from station walk  ‘He walked away from the station.’

le 

le  (Teng 1977: 12)

Clearly componential (“factorial”) compound verbs are also found in Hindi-Urdu:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(32) dhaniyā-ko to vo zabardastī khĩĩč Dhaniya-  he by.force pull ‘ . . . but Dhaniya he can drag (here) by force.’

lā 

saktā-hai can-3. (Godan 47.41)

However, their frequency of occurrence appears to be lower than it is in Chinese.

20.5.2 Phrasal constituents may interrupt the components of a compound verb in Chinese (with vectors lái COME and qù GO): (33) tā qí dān.chē lái he ride bicycle  ‘He rides a bike (in this direction).’ (34) zˇou jìn jiào.shì walk enter classroom ‘Enter the classroom.’

lái  (Lu 1984: 84)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

532

-    . 

In Hindi-Urdu only discourse particles like hī ‘only; indeed’, bhī ‘also, even’, to (concession), na or nahĩĩ ‘not’, and hī kyõ na ‘why not just’ may separate a main verb from its vector. (35) le bhī lũũ to ghar.vālī mujhe jītā čhor: -egī ? take even . then wife me alive leave- ‘And even if I do take it, would my wife leave me in peace?’ (Godan 206.26) Unlike in Chinese,²⁴ full nouns or other lexical phrases in Hindi-Urdu rarely, if ever intervene.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

20.5.3 Perhaps as a consequence of being less grammaticalized, the vector verbs of Chinese are much more numerous than those of Hindi-Urdu. Even setting aside the quasiadjectival vectors like haˇo GOOD (whose functions and reactances are similar to the more clearly verbial vectors like z oˇ u MOVE), the list is longer than Hindi-Urdu’s. (See Appendices A and B.) Less grammaticalized vectors harmonizing more closely with the meanings of their main verbs necessarily occur with more limited sets of them. For example, in Hindi-Urdu the vector le TAKE occurs with centripetal main verbs belonging to a number of different subsets: verbs of perception (dekh ‘look; see’, sun ‘listen; hear’, etc.), verbs of consumption (khā ‘eat’, pī ‘drink’, etc.), verbs of mental action (soč ‘think’, samajh ‘understand’, etc.), and prehension (pakar: ‘catch’, čhīn ‘snatch’, etc.). In Chinese for each one of these subsets there is a different vector, one that is more complexly and specifically linked with the main verbs in the set: (36) tā kàn jiàn (or dào) le yì-zhī dà.xiàng (kàn *diào) (kàn *zhù) he look    one- elephant look   25 ‘He saw an elephant.’ (kàn ‘look’ takes SEE or ARRIVE as vector.) (37) tā hē diào (or hē dào) le yì-píng jiuˇ (hē *zhù or hē *jiàn) she drink  drink   one- wine drink  drink  ‘She drank up a bottle of wine.’ (hē ‘drink’ takes DROP or ARRIVE as vector.)’ (38) tā rèn.shì dào nà-ge wèn.tí (rèn.shì *jiàn or *diào or *zhù) he recognize  that- problem recognize    ‘He came to recognize the problem.’ (rèn.shì ‘recognize’ takes ARRIVE as vector.)

²⁴ Different classes of V-v sequences in Chinese have different degrees of separability. ²⁵ Vector zhù STAY does occur with kān (note the different tone) in the sense of ‘guard’.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     - (39) baˇ qì.chē tíng  car stop ‘Stop that car!’

zhù 

533

(tíng *jiàn) (tíng *diào) (tíng dmdào)26 stop  stop  stop  (tíng ‘stop’ takes STAY as vector.)

Although dào  is the most general, unlike Hindi-Urdu’s le TAKE, there is no single vector which can be used with all the members of these four subsets.²⁷

20.5.4 Vector verbs in Mandarin can be negated independently of main verbs. Not so in Hindi: (40) wˇo.men shēng bù zháo huˇo we light   fire ‘We (try to) light the fire but it doesn’t light.’ (Entailment cancelation. See fn 16.)

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(41) ham āg nahĩĩ lag-ā we fire  light- ‘We don’t light the fire.’

(*de-) 

-te 

hãĩ 

To express lack of success despite effort, Hindi-Urdu uses a very different construction, in which a transitive expresses the attempt and a corresponding intransitive is negated to express failure: (42) išq hai vo ātiš ki lagā-e na lage love is that fire that light-  lights ‘Love is the fire that doesn’t light no matter how hard you (try to) light it . . . ’ (Ghalib) The reason that the construction in (42) is the one that compares²⁸ more closely with the use of Vvs to express incapacity in Chinese is that it clearly divides the action of lighting a fire into two parts: (1) the activity of the agent, and (2) its result. The construction in (40) and the one in (42) allow the negative particle to apply to just that second component of the action, in effect to single it out for

²⁶ With dào the sense of tíng shifts to ‘park’ and requires a location phrase as its complement. ²⁷ The subsets themselves are internally coherent vis-à-vis vector choice: 1. tīng jiàn ‘hear’, qiáo jiàn ‘stare at’; 2. chī diào ‘eat up’, yòng diào ‘use up’; 3. xiaˇng dào ‘think of ’, zhù.yì dào ‘notice’; 4. daˇng zhù ‘block (a road)’, zhē zhù ‘eclipse (the moon)’; ‘cover up (a body part)’ etc. ²⁸ See discussion of Thai and Mandarin “success verbs” in Thepkanjana and Uehara (2004a: sec. 3).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

534

-    . 

negation, whereas the vector in the Hindi-Urdu compound verb (see (41)), while it does express completeness in the representation of a complex action, in itself does not express a discrete element or component of that action. Hence, it cannot be separately negated.

20.5.5 Unlike in Hindi-Urdu, where there is normally harmony in the valency of main and vector in a compound verb, in Chinese the vector is almost always intransitive (see Liu 2003: sec. 5.2). While the intransitive main verbs in both languages take intransitive vectors, transitive main verbs in Hindi-Urdu (phẽk ‘throw’, ut:hā ‘pick up’, dekh ‘look’, beč ‘sell’, batā ‘tell’, etc.) most commonly take one or the other of two unmarked transitive vectors: le TAKE (for reflexive or centripetal action: see Section 20.5.3) or de GIVE (with other two-place main verbs). Transitive main verbs in Chinese take intransitive vectors (with the exception of those taking jiàn SEE; MEET). Contrast and compare (43) with (44):

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(43) jhuniyā-ne t:opī utār-kar phẽk dī Jhuniya- cap doff- throw  ‘Jhuniya took off her cap and threw it away.’ (44) baˇ mào.zi tuō.le rán.hòu rēng diào  hat doffed then throw  ‘(She) took off her hat and threw it away.’

(Godan 193) le 

While the notional subject of the vector verb in Hindi-Urdu is the same as the subject of the main verb, the subject of the vector verb in Chinese is usually the object of the main verb (assuming the main verb is transitive). In (43) the outward orientation of the subject’s action determines the choice of the vector de. In (44) the choice of the intransitive diào DROP as vector reflects the motion imparted to the object of main verb rēng. If intransitive vectors are used with transitive main verbs in Hindi-Urdu, they refer either to the situation as a whole or to the subject. They never refer to the object. Thus, in (45) if the vector gayā WENT indicates any actual physical displacement at all,²⁹ it is the motion of the subject vo ‘he’, not that of asar ‘impression’:

²⁹ The vector jā GO when used with transitive main verbs may indicate a psychological rather than a physical movement. But even so the psychological movement is presented as something pertaining to or having its source in the subject, not the object.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     - (45) horī par na-mit:ne.vālā asar čhor: Hori on -erasable impression leave ‘(He) left an indelible impression on Hori.’

535

gayā  (Godan 356.05)

This may be the most important difference between the Hindi-Urdu and the Chinese compound verb. In Hindi-Urdu the vector acts as a kind of semantic diacritic on the situation represented by the clause as a whole. In Chinese the scope of the vector is narrower: It denotes a conceptually distinct component of a complex situation or event, one which is synchronous with the action expressed by the main verb and is a response or reaction to the action denoted by it. Rather than the common (and in the case of transitive V-Vs the misleading) phrase “resultative constructions,” we may term V-Vs like zá daˇo ‘knock fall’ as “action-reaction constructions.” By this difference, not only South Asia but Northeast Asia and most of Central Asia are distinguished from China. Note the transitive vectors de and kure homophonous with full verbs meaning ‘give’ in Hindi-Urdu (46) and Japanese (47), versus the intransitive chū and chut EMERGE in the corresponding Mandarin (48) and Cantonese (49): (46) us-ne apnī qābiliyat she- self’s ability ‘She showed her ability.’

dikhā show

dī 

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(47) Kenta-wa ōki.na yūki-o mise-te kure-ta30 Kenta- great courage- show- - ‘Kenta . . . showed great courage.’ (full-count.jp/2016/06/04/post34291/) (48) tā xiaˇn.shì chū le he reveal   ‘He showed his potential.’

qiánlì potential

(49) ta him.shì chut liao chim.le he reveal   potential ‘He showed his potential.’ (Cantonese31 of Heng Xian village, Zhuang-zu) The CVs in (46) and (47) may be said to exhibit “componential valency harmony.” Those in (48) and (49) do not³² have componential valency harmony. ³⁰ Using kure implies the beneficiary is the first person or belongs to the “inner” group. ³¹ Thanks to Mr. Teng-pong Liang for the Cantonese data. ³² Is chū in these examples really intransitive? It is essential to our demonstration of the syntactic difference between Hindi and Chinese compound verbs to establish that vectors like chū correspond not to a transitive, but to an intransitive main verb. While chū as a main verb does have some transitive uses, they are limited to fixed expressions such as chū shū ‘publish’, chū chéng.guˇo ‘produce results’, chū lıˇ ‘contribute energy’, and chū qián ‘provide the money’, or to clauses in which the subject is a (promoted or “ascended”) location or experiencer of an emergent event but not its agent:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

536

-    . 

Text counts of Premchand’s novel Godan show that in Hindi-Urdu compound verbs, over 90% of transitive main verbs are accompanied by etymologically transitive vectors; whereas a count of Vv sequences in a text in Mandarin (Lü 1999) shows at most 25% of transitive main verbs are followed by transitive vectors. Even among that 25% most of the transitive vectors are labile items such as kāi OPEN, whose etymological sources are indeterminate with respect to their inherent valency. The abstract structure³³ of the Chinese Vv or compound verb is different from its South, Central, Northwest, and Northeast Asian counterparts. It is a reduction or compression of a two-clause structure: (50) X did J to Y and Y K’ed. => X, Y (J) > Y (K) (where J is the cause of K and K is a (plausible) effect of J) This structural compression allows Chinese to render certain situations more succinctly than languages in South, Central, Northwest, or Northeast Asia can. Contrast the brevity of yaˇo sˇ1 in (51)³⁴ and qí sˇ1 in (53) with their lengthier (and heavier) Avar and Hindi counterparts ‘xánc’un ch`’wan réxxana in (52) and daur: ā-ke mār dālā in (54): :

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(51) laˇo.huˇ yaˇo sˇ1 le tā tiger bite die  him ‘The tiger bit and killed him.’ (i)

haí.zi chū má.zhěn le child emerge measles  ‘The child came down with measles.’

chū in (48) does not belong to either of these two sets of noncanonically transitive uses: (ii) * ta chū qiánlì. he emerge potential ‘He has potential.’ There is no fixed expression chū qiánlì ‘have potential’, nor can the subject of (48) be interpreted as the nonagentive location or source of potential. We are grateful to Colin Masica for calling our attention to the problem of determining the valency of chū. ³³ If we adopt the concept and terminology of the “syntactic pivot” (Dixon 1994; Bickel 2004), we can characterize the difference succinctly: In the languages of South Asia the vector verb operates on a Subject-Agent (or “nominative”) pivot, while in Chinese the V-V pivot is Subject-Object (“absolutive”). The only exception in the languages of South Asia is the “vector adverbs” of southern Rajasthan and eastern Sind which (as far as gender and number agreement is concerned) operate on a Subject-Object pivot, that is, absolutive pivot. See fn 41. ³⁴ Dumi, a Kiranti language spoken in eastern Nepal, has a similar construction but one which, owing to Dumi’s complex systems of agreement, mirrored on both the main and vector verb, reveals a biclausal identity more clearly than its Mandarin counterpart. (i)

ımma-?a hammıl sid-ini-ka beer- they. intoxicate-2/3- ‘The beer got them roaring drunk.’

ham-mit-a 3-die-2/3 (from Driem 1993: 228; glossing normalized)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     - (52) `gázhdarahay-all´ `xánc`-un ch`wa-n dragon- bite- kill- ‘The dragon bit (him) and killed (him).’

537

réxx-ana - (Charachidze 1981: 192)

(53) tā qí sˇ1 le maˇ he ride   horse ‘He rode the horse to death.’35

(Li 1990)

(54) kuttõ-ne čītal-ko daur: -ā-ke mār dālā : dogs- spotted.deer run-- kill . ‘The dogs ran the deer to death.’ (facebook.com/StarNews9Hindi/) When X = Y (i.e. where J is a nonstative intransitive verb), the schema in (50) reduces to a structure that seems to be parallel to the compound verb of the nonChinese type: (55) X did J (to itself ) and K’ed => X (J > K) (where J is the cause of K and K is a (plausible) effect of J) In Mandarin the presence of an intransitive vector even has the power to increase valency:

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

(56) X (J) > K (Y) => X did J and Y K’ed => X, Y (J) > Y (K) (where J and K are lexically intransitive) Compare lexically intransitive lèi ‘get tired’ in (57) with its contextually transitive counterpart in (58). Notice the two sentences³⁶ are identical save for the presence in (58) of the vector sˇ1 DIE: (57) tā lèi (*wˇo) le he tire (me)  ‘He got tired.’ (Including w oˇ does not yield the meaning ‘He tired me.’) (58) tā lèi sˇ1 wˇo le he tire  me  ‘He tired me to death.’ (See (56) and compare (58) with (51) and (53).)

³⁵ In its idiomatic meaning of ‘excessively’ vector si DIE is one of the few in Chinese able to modify ˇ in (53) can also mean either the object or the subject. According to Li (1990) the sequence qí sˇ1 le ma that the horse’s rider rode himself ‘to death’. ³⁶ See Chang (1998: 83) for further discussion of this kind of alternation in valency.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

538

-    . 

Another more dramatic example: In (59) the action of the intransitive kū ‘cry’ is paired with its reaction verb daˇo FALL (also intransitive) to create a transitive predicate kū-daˇo “cry down”: (59) mèng-jiāng nuˇ¨ kū-daˇo le cháng-chéng37 Meng-jiang lady cry-  Great-Wall ‘Lady Meng-jiang “cried down” the Great Wall.’ (Kageyama and Shen 2018: 194) In no South Asian language does the presence of an intransitive vector verb have the ability (through a kind of “inverse operation”³⁸) to induce an increase in the valency of an intransitive main verb. Nor, given the absence in South Asia of models like those in (51) and (53), should we expect to encounter analogical extensions of them like the ones seen in (58) and (59).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

20.6 Macrolinguistic areas in Asia In his most recent study, Masica poses the question: “are the ‘resultative’ verbs of Chinese, Thai, Burmese, etc., essentially the same phenomenon as the explicator verbs of South Asia?” (Masica 2001: 250). If we consider only intransitive verb + intransitive verb combinations, as in exx (6a)–(9), there seems to be no reason not to regard such “action-reaction” complexes as representing a single phenomenon that has diffused throughout South, Central, Northwest, Northeast, East, and nonMalay Southeast Asia. However, when we examine verb-verb combinations that involve transitive main verbs, we find distinct differences in syntax and semantics that divide East and Southeast Asia from South, Central, Northwest, and Northeast Asia: The East and Southeast Asia group as a rule have transitiveintransitive sequences (Thepkanjana and Uehara 2004b); the South, Central, Northwest, Northeast group, as a rule do not. Furthermore, setting one or two exceptions aside, languages in East and Southeast Asia do not have transitivetransitive V-V sequences, whereas in South, Central, Northwest, and Northeast Asia, transitive main verbs are normally followed by transitive vectors. This line of division corresponds closely to the VO versus OV isogloss that separates the same areas from South, Central, Northwest, and Northeast Asia. In other parts of the world verb-verb sequences of the South Asian type are ³⁷ This example alludes to a folk tale about Lady Meng Jiang, whose husband was pressed into service by imperial officials and sent as corvée labor to build the Great Wall of China. Lady Meng Jiang heard nothing after his departure, so she set out to bring him winter clothes. Unfortunately, by the time she reached the Great Wall, her husband had already died. Hearing the bad news, she wept so bitterly that a part of the Great Wall collapsed, revealing his bones. (en.wikipedia.org/wiki/Lady_Meng_Jiang) ³⁸ See Harris (1968: 95–9) for the notion of the “inverse operation.”

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

539

associated with OV languages like Quichua (Hook 1987), while those of the Chinese (and Southeast Asian) type are found in Atlantic Creoles (Schiller 1989) and VO languages spoken in West Africa (Crowley 2002). Of special interest are Sino-Tibetan languages like Lahu and Karen, which find themselves on the western side of this typological divide and exhibit both kinds of verb-verb sequences simultaneously: (60) nga21.hI ga qo? 21 chi54 to? 54 we must again lift  ‘We had to lift (it) out again for (them).’

pi54 

ve (Lahu)  (Matisoff 1973: 200)39

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

The line separating VO languages to the east from OV languages to the west runs near the Myanmar-Thai border. Although Lahu belongs to the Sino-Tibetan language family, most of whose non-Tibetan members are (or were) VO, it is spoken in an area lying to the west of the VO/OV isogloss and is now itself an OV language (Matisoff 1973: 40). However, it retains remnant features of word order more typical of VO languages, such as preverbal placements of the modals ga ‘must’ and qo? ²¹ ‘(do) again’ that are visible in (60). Lahu also retains the Sinitic kind of compound verb in which a transitive primary (such as chi⁵⁴ ‘lift’) is followed by an intransitive secondary (to? ⁵⁴ ‘emerge’) which modifies the object rather than the subject of the clause. In fact, except for the absence of any component answering to Lahu’s vector pi,⁵⁴ the order of verbal elements in (60) is identical to the order that is found in the corresponding Mandarin sequence deˇ1 zài quˇ chū: quˇ chū (yí-ge zhˇong.liú) (61) wˇo.men deˇ1 (weì bìng.rén) zài we must for patient again extract  (one tumor) ‘We have to take (the tumor) out again (for the patient).’ Lahu in (60) differs from Mandarin in (61) in its addition of the “South Asian”style vector pi⁵⁴ GIVE, which has closer parallels in Hindi-Urdu and other languages of South Asia than it does in Mandarin:⁴⁰ (62) hamẽ tyūmar phir.se nikāl de-nā us. tumor again extract - ‘We’ll have to remove the tumor again.’

par: egā must.

³⁹ Matisoff’s transcription of Lahu data has been modified for typographical convenience. The double numbers represent tonal contours based on a five-part range. See Matisoff (1973: 22) for definition and discussion. ⁴⁰ Mandarin and other forms of Chinese do have a grammaticalized form of a main verb geˇ1 GIVE, but like ba ˇ HOLD, zài BE AT (a place), and other “coverbs,” it is used as a preposition, not an auxiliary verb: (i)

tā geıˇ woˇ daˇ le he give me strike  ‘He made phone call for me.’

yí.ge a

diàn.huà telephone

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

540

-    . 

Our conclusion, then, is that Masica (1976) with its categorization of East and Southeast Asian verb-verb sequences as distinct from those of South,⁴¹ Central, and Northeast Asian, adheres more closely to the typological and geographical facts, and is preferable to Masica (2001: 250–2.). Indeed, Masica’s earlier view accommodates (and is supported by) the conflation of the two kinds of verb-verb sequences that is observed in the border area between South and Southeast Asia in languages like Lahu and Karen (Solnit 1997).

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

20.7 Relativizing linguistic areas The Indo-Turanian Linguistic Area proposed in Masica (1976) is to our knowledge the largest one ever proposed (unless we take the discussion and charts in Masica (2001) as the basis for an even larger one incorporating China and Southeast Asia). Since bilinguals operating in a bilingual or multilingual environment are regarded as providing the mechanism for the creation of linguistic areas, we must assume that the Indo-Turanian Linguistic Area (if it is a linguistic area) is the result of chains of bilingual encounters affecting the structures of a broad mosaic of different languages over a period of many thousands of years. In the absence of direct contact between the Japanese and Marathas or between Koreans and Tamils, we must suppose the convergence between the languages of South Asia and Northeast Asia to have been mediated by the diffusion of linguistic features along the links of smaller chains of social, political, and economic relations lasting over enormous spans of space and time. This network of change would have to persist despite the presence of competing networks radiating from different directions. The more extended these supposed chains are, the more fragile they become; and the more susceptible to influence (or interruption) from new sets of social, political, and economic relationships with speakers of typologically different languages, relationships that over such long periods of time and enormous distances must inevitably emerge. In this particular instance, we are confronted with groups of languages having different holistic typologies: the OV languages of South, Central, Northwest, and Northeast Asia (as well as Myanmar) and the VO languages of East and Southeast Asia. By merging differing kinds of V-V sequences into a single type, Masica’s (2001) chart makes it more difficult to ⁴¹ In the pattern of their agreement the “absolutive adverbs” of southern Rajasthan and eastern Sind are an exception. Concord is with the intransitive subject or with a transitive clause’s direct object in all tenses and moods. See Hook and Chauhan (1988) and Hook’s 1989 field notes on Thari, aka Dhātki: (i)

uena tehsan pug-āw her. station arrive- ‘Deliver her to the station.’

ri .. (from Ashok Kumar, r/o Thar Parkar, Sind)

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     -

541

identify and understand the ways in which languages of divergent typologies have interacted and, by further inflating an already mammoth linguistic area, makes it less likely that it actually is a linguistic area in the full historical and sociolinguistic sense of the term.⁴²

Appendix A List of Chinese vectors with glosses (in caps) of their lexical sources and examples of them used concretely and then more abstractly Vector

Gloss

qù

shū nˇ1 kě.yˇ1 ná qù book you may take go ‘You may take (away) the book.’ abstract yù.suàn jiaˇn qù le yí bàn budget shrink   one half ‘The budget was cut in half.’ COME qˇ1ng kuài huí lái please fast return come ‘Please hurry back (here).’ abstract kàn lái wèn.tí bú dà look  problem not big ‘It seems not to be much of a problem.’

lái

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

daˇo

GO

FALL

shù beì chuī daˇo le tree Pass blow fall  ‘The tree was blown down.’

abstract nˇ1 nán daˇo you difficult  ‘You stumped me.’ diào

(Compare (59)

le 

DROPintr baˇ jiù bào.zhˇ1 rēng diào  old newspaper throw drop ‘Throw away the old newspapers.’ abstract

kāi

wˇo me

niú.naˇ1 yˇ1.jīng huài diào le milk already spoil   ‘The milk’s already gone bad.’

OPENintr tā qīng.qīng.de tuī kāi mén she quietly push open door ‘She quietly pushed the door open.’ abstract

qˇ1ng ràng kāi please yield  ‘Please move over.’

⁴² While the co-authors of this chapter have exchanged views with Masica on its contents and the conclusions it reaches, many areas of disagreement remain: “one would expect there to be differences in origin of converging constructions where the underlying syntax is different.” (pers. comm.). We are grateful to him, to Timur Maisak, and to Karen Ebert for their careful reading and comments. The usual disclaimers apply.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

542

-    . 

xià

DESCEND tā fàng xià bào.zhˇ1 he put descend newspaper ‘He put down the newspaper.’ abstract qˇ1ng liú xià nˇ1.de please leave  your ‘Please leave your address.’

z ou ˇ

MOVEintr tā baˇ hái.zi dài she Acc child take ‘She led the child away.’

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

abstract

dì.zhˇ1 address

zˇou move

le 

wˇo cāi zˇou le nĭ.de dá.àn I guess   your answer ‘Ha, ha! I guessed your answer!’

dào

ARRIVE qiú daˇ dào wˇo.de tóu ball hit arrive my head ‘The ball hit me on the head.’ abstract SARS yˇ1ng.xiaˇ ng dào luˇ¨ .yóu yè SARS affect  tourism trade ‘SARS affects tourism.’

zhù

STAY

zhàn zhù stand stay ‘Freeze!’ (‘Don’t move!’) abstract xiān wěn zhù mínxīn first stabilize  people heart ‘First appease public opinion.’

sˇ1

DIE

yān sˇ1 le hěn duō rén drown die  very many ‘Many people were drowned.’

abstract wˇo máng sˇ1 le I busy   ‘I am extremely busy.’ duàn

jiàn

BREAKintr tā baˇ shù kaˇn duàn he  tree chop break ‘He chopped down the tree.’

abstract bié daˇ duàn wˇo.de huà don’t beat  my word ‘Don’t interrupt me while I’m talking.’ SEE; wˇo kàn jiàn le dà.xiàng MEET I look   elephant ‘I saw an elephant.’ abstract wˇo tīng jiàn tā.de I hear  his ‘I hear(d) his voice.’

chū

le 

shēng.yin voice

EMERGE xuésheng zˇou chū jiàoshì student walk emerge classroom ‘Students walk out of the classroom.’ abstract

tā xiaˇnshì chū qiánlì he reveal  potential ‘He showed his potential.’

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

-     - shàng

ASCEND hóu.zi pá shàng le monkey climb ascend Pfv ‘Monkeys climbed up a tree.’ abstract

zháo

diū

543

shù tree

tā ài shàng xiaˇ o lˇ1 he love  little Li ‘He falls in love with Little Li.’

ATTACH huˇo shaō zháo le căo fire burn attach  straw; grass ‘The straw has caught on fire.’ abstract wˇo zhaˇ o zháo le biăo I search   watch ‘I found the watch.’ LOSE

zài rén.qún-lˇ1 wˇo baˇ tā  crowd-in I hold him ‘I lost him in the crowd.’

gēn follow

diū-le lose-

abstract hái.zi zˇou diū le child walk  Pfv ‘The child wandered off.’

Appendix B

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

List of the ten⁴³ most common Hindi-Urdu vectors with glosses of their lexical sources and examples of them used first concretely and then more abstractly Vector

Gloss

jā

GO

mirzā bhī bāhar Mirza too outside ‘Mirza also went out.’

nikal exit

abstract 12 bajte-bajte vo šurū 12 striking-2 it begin ‘It began on the stroke of 12.’ de

GIVE

gae went Godan 169.2 ho happen

gayā  Godan 143.43

tamākhū hai ki lā dũũ tobacco is that bring give. ‘Have I any tobacco to bring and give you?’

abstract kyõ na čher: .čhār: šurū why not teasing begin ‘Why not start teasing (him)?’

kar do

Godan 109

de  Godan 141.45

⁴³ There are an additional dozen or so less common vector verbs in Hindi-Urdu: mār STRIKE, mar DIE, nikāl TAKE OUT, nikal EMERGE, V-kar chor: de RELEASE, etc. These are so rarely used that even a 350-page novel like Premchand’s Godan may not afford instances of them. For discussion and examples, see Hook 1974 (120–44).

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

544 le

dāl :

rakh

-    .  TAKE

us-ne ããčal sir-par khĩĩč liyā she-Erg border head-on pull took ‘She pulled the sari’s edge over her head.’

Godan 191.17

POUR Concrete use of dāl : as a vector not available. abstract us-ne khet.bārī sab beč-bāč dālī : he- fields all sell-  ‘While he had sold off his farm and all.’

Godan 363.31

PUT

tere-lie mūs pakar: rakhā hai you-for mouse catch put is ‘For you we’ve caught (and kept) a mouse.’ use gāy hī samajh it cow  consider ‘(I) had taken it to be a cow.’

bait:h

Godan 303

abstract log hãs-ẽge hãs lẽ people laugh-3 laugh . ‘People will laugh; let them laugh.’

SIT

rakhā 

Godan 43.41

thā was Godan 317.22

us vaqt jhuniyā jāg-kar ut:h bait:hī that time Jhuniya wake.up- rise sat ‘At that moment Juniya awoke and sat up.’

Godan 342.21

abstract krodh-mẽ mãĩ kyā kar bait:hũũ na kah saktī anger-in I what do . not say can ‘I can’t say what I’d do if I got mad.’ Godan 306.29

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

par:

FALL

punnī dhakkā khā-kar gir Punni shove eat- topple ‘Punni was shoved and fell down.’

par: ī fell

abstract ye kah-te kah-te vo muskarā this say-ing say-ing she smile ‘While saying this, she smiled.’ ut:h

ā

Godan 189.17

horī-kā xūn khaul ut:hā Hori-’s blood boil rose ‘Hori’s blood boiled.’ abstract mālatī adhīr ho ut:hī Malati impatient become  ‘Malati became impatient.’ COME mũh-mẽ pānī bhar āyā mouth-in water fill came ‘His mouth watered.’ RISE

abstract ããkhẽ sa-jal ho āyī eyes with-water become  ‘His eyes became wet (with tears).’ čal

Godan 34.7 par: ī 

MOVE

Godan 48.41

Godan 82.45

Godan 217.9 thĩĩ were

āie laut: čalẽ come return move. ‘Come. Let’s go back (together).’

Godan 347.26

Godan 294.22

abstract use kučh-kučh rāj.nītik gyān ho čalā hai him some-some political knowledge be  is ‘He’s become a bit politically aware.’ Godan 203.18

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

References Abbi, Anvita (ed.) (1991). India as a Linguistic Area Revisited (A Special Issue of Language Sciences 13(2)). Oxford: Pergamon Press. Abdusalamova, Madina Magomedovna (2006). Modal’nost’ Predlozhenija v Avarskom Jazyke v Sopostavlenii s Germanskimi Jazykami [Modality of Sentences in Avar in Comparison with Germanic Languages]. Сandidate’s Dissertation. Dagestanskij Gosudarstvennyj Pedagogicheskij Universitet. Makhachkala. Abe, Hiroshi (2011). ‘Jōdai nihongo no dōshi rensetsu tori- ni tsuite: fukugōdōshi no sonpi o chūshin ni’ [On verb sequences with tori- in Old Japanese: Focusing on the existence of verb compounds], Nagoya Linguistics [Nagoya University] 5: 1–14. Agarkar, G. G. (1888). Vākyamīmāmsā āni : : Vākyānnce Pr: thakkaran: [An Inquiry into the Nature of Sentences with an Analysis of Them]. Third edition. Pune: Aryabhushan Press. Aikhenvald, Alexandra Y. (2006). ‘Serial Verb Constructions in Typological Perspective’, in A. Y. Aikhenvald and R. M. W. Dixon (eds), Serial Verb Constructions: A Cross-linguistic Typology. Oxford: Oxford University Press, 1–68. Aikhenvald, Alexandra Y. (2011). ‘Multi-Verb Constructions: Setting the Scene’, in J. Birchall, P. Muysken, and A. Y. Aikhenvald (eds), Multi-Verb Constructions: A View from the Americas. Leiden and Boston, MA: Brill, 1–26. Aikhenvald, Alexandra Y. and R. M. W. Dixon (eds) (2006). Serial Verb Constructions: A Cross-linguistic Typology. Oxford: Oxford University Press. Alpatov, Vladimir M. and Vera Podlesskaya (1995). ‘Converbs in Japanese’, in M. Haspelmath and E. König (eds), Converbs in Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms-Adverbial Participles. Berlin: Mouton de Gruyter, 465–85. Alsina, Alex, Joan Bresnan, and Peter Sells (eds) (1997). Complex Predicates. Stanford, CA: CSLI Publications. Amberber, Mengistu, Brett Baker, and Mark Harvey (eds) (2010). Complex Predicates: Cross-Linguistic Perspectives on Event Structure. Cambridge: Cambridge University Press. Anantamurti, U. R. (2002). Samskara. (Reprint of the 1965 edition.) Heggodu: Akshara Prakashana. Andersen, Henning (1987). ‘From Auxiliary to Desinence’, in M. Harris and P. Ramat (eds) The Historical Development of Auxiliaries. Berlin: Mouton de Gruyter, 21–51. Andersen, Henning (1989). ‘Understanding Linguistic Innovations’, in L. E. Breivik and E. H. Jahr (eds), Language Change: Contributions to the Study of Its Causes. Berlin: Mouton de Gruyter, 5–27. Anderson, Gregory D. S. (2004). Auxiliary Verb Constructions in Altai Sayan Turkic. Wiesbaden: Harrassowitz Verlag. Anderson, Gregory D. S. (2006). Auxiliary Verb Constructions. Oxford: Oxford University Press. Anderson, Gregory D. S. and K. David Harrison (1999). Tyvan. Munich: Lincom Europa.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

546



Anderson, Stephen (1992). A-Morphous Morphology. Cambridge: Cambridge University Press. Annamalai, E. (1970). ‘On Moving from a Coordinate Structure’, in R. Binnick (ed.), Papers from the Sixth Regional Meeting, Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 131–46. Annamalai, E. (1971). ‘Lexical Insertion of a Mixed Language’, in Papers from the Seventh Regional Meeting, Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 20–7. Annamalai, E. (1985). Dynamics of Verbal Extension in Tamil. Trivandrum: Dravidian Linguistic Association of India. Annamalai, E. (2016). ‘Expanded Verbs in Dravidian’, in H. H. Hock and E. Bashir (eds), The Languages and Linguistics of South Asia: A Comprehensive Guide (The World of Linguistics 7). Berlin: De Gruyter Mouton, 550À9. Aoki, Hirofumi (2013). ‘Nihongo fukugōdōshi no rekishiteki-henka’ [The Historical Change of Japanese Compound Verbs], in T. Kageyama (ed.), Fukugōdōshi Kenkyū no Saisentan [Frontiers of Research on Compound Verbs]. Tokyo: Hituzi Syobo, 215–41. Arora, Harbir K. (2004). Syntactic Convergence: The Case of Dakkhini Hindi-Urdu. Delhi: Publications Division, University of Delhi. Asher, Ronald E. and T. C. Kumari (1997). Malayalam. London: Routledge. Badema (2012). Nihongo to Mongorugo ni okeru Hojodōshi no Taishōkenkyū [A Contrastive Study of Japanese and Mongolian Auxiliary Verbs]. Ph.D. dissertation, Kobe University. Baker, Brett and Mark Harvey (2010). ‘Complex Predicate Formation’, in M. Amberber, B. Baker, and M. Harvey (eds), Complex Predicates: Cross-Linguistic Perspectives on Event Structure. Cambridge: Cambridge University Press, 13–47. Baker, Mark and Carlos Fasola (2009). ‘Araucanian: Mapundungun’, in R. Lieber and P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press, 594–608. Baker, Mark, Kyle Johnson, and Ian Roberts (1989). ‘Passive Arguments Raised’, Linguistic Inquiry 20: 219–51. Bandhu, Vishva (ed.) (1963–6). Rgveda with the Padapāt:ha and Commentaries of : Skandasvāmin, Udgītha, Veṅkatamādhava and Mudgala. 8 vols. Hoshiarpur, Punjab: : Vishveshvaranand Vedic Research Institute. Baranova, Vlada V. (2013). ‘Grammaticalization and Semantics of Complex Predicates in Kalmyk’, Journal de la Société Finno-Ougrienne 94: 9–35. Bashir, Elena (1993). ‘Causal Chains and Compound Verbs’, in M. K. Verma (ed.), Complex Predicates in South Asian Languages. New Delhi: Manohar, 91–130. Bashirova, Roza Saadulaevna (2008). Analiticheskie Konstruktsii Glagola Avarskogo Literaturnogo Jazyka [Analytical Constructions of Standard Avar Verbs]. Сandidate’s Dissertation. Institut Jazyka, Literatury i Iskusstva DNTs RAN. Makhachkala. Baskakov, N.A. (1966). Severnye dialekty altajskogo (ojrotskogo) jazyka. Dialekt černevyx tatar (tuba-kiži) [Northern Dialects of Altai (Oirot). Tuba-Kizhi Dialect.] Vol. 2. Moscow: Nauka. Baskakov, N. A. and A. I. Inkiǰekova-Grekul (1953). Xakassko-russkiy slovar’. Moscow: Gosudarstvennoe izdatel’stvo inostrannïx i nacional’nïx slovarey. Baxter, Alan N. (1988). A Grammar of Kristang (Malacca Creole Portuguese). Canberra: Pacific Linguistics. Benveniste, Émile (1965). ‘Structure des relations d’auxiliarité’, Acta Linguistica Hafniensia 9: 1–15.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



547

Bhairappa, Es. El. (1973). dāt:u. Bangalore: Sahitya Bhandara. Bhandari, Mannu (1979). āp kā baNTī. Fourth edition. New Delhi: Akshar Prakashan. Bhaskararao, Peri (1980). Konekor Gadaba: A Dravidian Language. Pune: Deccan College Postgraduate and Research Institute. Bhaskararao, Peri (1998) ‘Gadaba’, in S. B. Steever (ed.), The Dravidian Languages. London: Routledge, 328–56. Bhat, D. N. S. (1979). ‘Vectors in Kannada’, International Journal of Dravidian Linguistics 8(2): 300À9. Bhatt, Rajesh (2006). Covert Modality in Non-Finite Contexts. Berlin: De Gruyter Mouton. Bickel, Balthasar (2004). ‘Syntax of Experiencers in the Himalayas’, in P. Bhaskararao and K. V. Subbarao (eds), Non-Nominative Subjects. Amsterdam: John Benjamins, 77–111. Bisang, Walter (1991). ‘Verb Serialisation, Grammaticalisation, and Attractor Positions in Chinese, Hmong, Vietnamese, Thai and Khmer’, in H. Seiler and W. Premper (eds), Partizipation: Das sprachliche Erfassen von Sachverhalten. Tübingen: Narr, 509–562. Bloch, Bernard (1946). ‘Studies in Colloquial Japanese I: Inflection’, Journal of the American Oriental Society 66: 97–109. Reprinted in R. A. Miller (ed.) 1969. Bernard Bloch on Japanese. New Haven, CT: Yale University Press, 233–65. Boeschoten, Hendrik (1998). ‘Uzbek’, in L. Johanson and É. Á. Csató (eds), The Turkic Languages. London and New York: Routledge, 357–78. Bokarev, A. A. (1949). Sintaksis Avarskogo Jazyka [The Syntax of Avar]. Moscow, Leningrad: Izdatel’stvo AN SSSR. Booij, Geert (2008). ‘Lexical Integrity as a Formal Universal: A Constructionist View’, in S. Scalise, E. Magni, and A. Bisetto (eds), Universals of Language Today. Dordrecht: Springer, 83–100. Booij, Geert (2010). Construction Morphology. Oxford: Oxford University Press. Borer, Hagit (2005). Structuring Sense Volume 2: The Normal Course of Events. Oxford: Oxford University Press. Bresnan, Joan and Sam A. Mchombo (1995). ‘The Lexical Integrity Principle: Evidence from Bantu’, Natural Language & Linguistic Theory 13: 181–254. Bresnan, Joan, Ash Asudeh, Ida Toivonen, and Stephen Wechsler (2015). LexicalFunctional Syntax. Second edition. Malden, MA: Wiley Blackwell. Bril, Isabelle (2004). ‘Complex Nuclei in Oceanic Languages: Contribution to an Areal Typology’, in I. Bril and F. Ozanne-Rivierre (eds), Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness. Berlin: De Gruyter Mouton, 1–48. Brinton, Laurie and Elizabeth Closs Traugott (2005). Lexicalization and Language Change. Cambridge: Cambridge University Press. Bubenik, Vit (1998). A Historical Syntax of Late Middle Indo-Aryan. Amsterdam: John Benjamins. Bugaeva, Anna (2018). ‘Ainu Complex Predicates with Reference to Japanese’, in P. Pardeshi and T. Kageyama (eds), Handbook of Japanese Contrastive Linguistics. Berlin and Boston, MA: De Gruyter Mouton, 247–72. Burrow, Thomas and S. Bhattacharya (1953). The Parji Language. Hertford: Stephen Austin and Sons. Burrow, Thomas and S. Bhattacharya (1970). The Pengo Language. Oxford: Oxford University Press. Burton-Page, John (1957). ‘Compound and Conjunct Verbs in Hindi’, Bulletin of the School of Oriental and African Studies 19: 469–78.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

548



Burzio, Luigi (1982). Italian Syntax: A Government and Binding Approach. Dordrecht: Foris. Butt, Miriam (1995). The Structure of Complex Predicates in Urdu. Stanford, CA: CSLI Publications. Butt, Miriam (1997). ‘Complex Predicates in Urdu’, in A. Alsina, J. Bresnan, and Peter Sells (eds), Complex Predicates. Stanford, CA: CSLI Publications, 107–49. Butt, Miriam (2003). ‘The Light Verb Jungle’, in G. Aygen, C. Bowern, and C. Quinn (eds), Papers from the GSAS/Dudley House Workshop on Light Verbs (Harvard Working Papers in Linguistics 9). Cambridge, MA: Harvard University Department of Linguistics, 1–50. Butt, Miriam (2010). ‘The Light Verb Jungle: Still Hacking Away’, in M. Amberber, B. Baker, and M. Harvey (eds), Complex Predicates: Cross-Linguistic Perspectives on Event Structure. Cambridge: Cambridge University Press, 48–78. Butt, Miriam and Wilhelm Geuder (2001). ‘On the (Semi)lexical Status of Light Verbs’, in N. Corver and H. van Riemsdijk (eds), Semi-Lexical Categories: The Function of Content Words and the Content of Function Words. Berlin and New York: De Gruyter Mouton, 323–70. Butt, Miriam and Wilhelm Geuder (2003). ‘Light Verbs in Urdu and Grammaticalisation’, in R. Eckardt, K. von Heusinger, and C. Schwarze (eds), Words in Time. Diachronic Semantics from Different Points of View. Berlin: De Gruyter Mouton, 295–350. Butt, Miriam and Aditi Lahiri (2002). ‘Historical Stability vs. Historical Change’. MS., Universität Konstanz, https://ling.sprachwiss.uni-konstanz.de/pages/home/butt/main/ papers/stability.pdf, accessed July 30, 2020. Butt, Miriam and Aditi Lahiri (2013). ‘Diachronic Pertinacity of Light Verbs’, Lingua 135: 7–29. Butt, Miriam and Alexandros Tantos (2004). ‘Verbal Semantic via Petri Nets’, https://web. stanford.edu/group/cslipublications/cslipublications/LFG/9/pdfs/lfg04.pdf, accessed November 16, 2020. Bybee, Joan (2001). ‘Mechanisms of Change in Grammaticalization: The Role of Frequency’, in B. B. Joseph and R. D. Janda (eds), The Handbook of Historical Linguistics. Oxford: Blackwell, 602–23. Bybee, Joan L. and Östen Dahl (1989). ‘The Creation of Tense and Aspect Systems in the Languages of the World’, Studies in Language 13: 51–103. Bybee, Joan, Revere Perkins, and William Pagliuca (1994). The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago, IL, and London: University of Chicago Press. Carnie, Andrew (2006). Syntax: A Generative Introduction. Second edition. Oxford: Blackwell. Chang, Claire Hsun-huei (1998). ‘V-V Compounds in Mandarin Chinese’, in J. L. Packard (ed.), New Approaches to Chinese Word Formation. Berlin: De Gruyter Mouton, 77–102. Charachidzé, Georges (1981). Grammaire de la langue avar: langue du Caucase nord-est. Saint-Sulpice-de-Favières: Éditions Jean-Favard. Chatterjee, Suniti Kumar (1926). The Origin and Development of the Bengali Language. Calcutta: Calcutta University Press. Revised, 1971; reprinted, New Delhi: Rupa, 1985. Chen, Yi-Ting and Yo Matsumoto (2018). Nihongo Goiteki Fukugōdōshi no Imi to Taikei [The Semantics and System of Japanese Lexical Compound Verbs]. Tokyo: Hituzi Syobo. Cho, Sungdai and John Whitman (2020). Korean: A Linguistic Introduction. Cambridge: Cambridge University Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



549

Ch’oe, Hyŏn-bae (1929). Uri Malbon. Kyŏngsŏng: Chŏngŭmsa. Reprinted in 1961, Seoul: Chŏngŭm Munhwasa. Chomsky, Noam (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: MIT Press. Chung, Taegoo (1993). Argument Structure and Serial Verbs in Korean. Ph.D. dissertation, University of Texas at Austin. Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Cleary-Kemp, Jessica (2015). Serial Verb Constructions Revisited: A Case Study from Koro. Ph.D. dissertation. University of California at Berkeley. Coelho, Gail (2018). ‘Complex Predicates in Betta Kurumba’, Journal of South Asian Languages and Linguistics 5(1):23À77. Collins, Chris (2002). ‘Multiple Verb Movement in ǂHoan’, Linguistic Inquiry 33: 1–29. Comrie, Bernard (1976). Aspect: An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Comrie, Bernard (1980). ‘Morphology and Word Order Reconstruction: Problems and Prospects’, in J. Fisiak (ed.), Historical Morphology. The Hague: De Gruyter Mouton, 71–82. Comrie, Bernard (2003). ‘Recipient Person Suppletion in the Verb “Give” ’, in M. R. Wise, T. N. Headland, and R. M. P. Brend (eds), Language and Life: Essays in Memory of Kenneth L. Pike. Arlington, TX: University of Texas at Arlington, 265–81. Corver, Norbert and Henk van Riemsdijk (2001). ‘Semi-Lexical Categories’, in N. Corver and H. van Riemsdijk (eds), Semi-Lexical Categories: The Function of Content Words and the Content of Function Words. Berlin and New York: De Gruyter Mouton, 1–19. Creissels, Denis (2010). ‘Benefactive Applicative Periphrases’, in F. Zúñiga and S. Kittilä (eds). Benefactives and Malefactives: Typological Perspectives and Case Studies. Amsterdam: John Benjamins, 29–70. Crowley, Terry (1987). ‘Serial Verbs in Paamese’, Studies in Language 11: 35–87. Crowley, Terry (2002). Serial Verbs in Oceanic: A Descriptive Typology. Oxford: Oxford University Press. Csató, Éva. Á. (2004) ‘On Copying in Kashkay’, in É. Á. Csató, B. Isaksson, and C. Jahani (eds), Linguistic Convergence and Areal Diffusion: Case Studies from Iranian, Semitic and Turkic. London and New York: Routledge Curzon, 271–84. Csató, Éva. Á. (2018). ‘Turkic Postverbial Constructions’. Handout Distributed at the International Symposium: Current Topics in Turkic linguistics Held at ILCAA, Tokyo University of Foreign Studies, March 3–4, 2018. Damle, M. K. (1970[1911]). šāstriy marāt:hi vyākaran: [A Scientific Grammar of Marathi]. Reprint. Pune: Deshmukh and Company. Dasgupta, Malasree (1990). ‘Composite Verbs in Bangla’, Indian Journal of Linguistics 17(1): 1–31. Dasgupta, Probal (1977). ‘The Internal Grammar of Bangla Compound Verbs’, Indian Linguistics 38(2): 68–85. De Boer, Elisabeth M. (2010). The Historical Development of Japanese Tone: From ProtoJapanese to the Modern Dialects. The Introduction of the Middle Chinese Tones in Japan. Wiesbaden: Harrassowitz Verlag. Delbrück, Berthold (1888). Altindische Syntax. Halle: Verlag der Buchhandlung des Waisenhauses. Diller, Anthony (2001). ‘Grammaticalization and Tai Syntactic Change’, in M.R. K. Tingsabadh and A. Abramson (eds), Essays in Tai Linguistics. Bangkok: Chulalongkorn University Press, 139–75.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

550



Dixon, R. M. W. (1994). Ergativity. Cambridge: Cambridge University Press. Doerfer, Gerhard (1971). Khalaj Materials. (Indiana University Publications Uralic and Altaic Series, Vol. 115). Bloomington, IN: Indiana University Publications. Doerfer, Gerhard (1998). Grammatik des Chaladsch. Wiesbaden: Harrassowitz. Dowty, David R. (1977). ‘Towards a Semantic Analysis of Verb Aspect and the English “Imperfective” Progressive’, Linguistics and Philosophy 1: 45–77. Dowty, David R. (1979). Word Meaning and Montague Grammar: The Semantics of Verbs and Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Kluwer Academic. Dowty, David R. (1991). ‘Thematic Proto-Roles and Argument Selection’, Language 67: 547–619. Driem, George van (1987). A Grammar of Limbu. Berlin and New York: De Gruyter Mouton. Driem, George van (1992). The Grammar of Dzongkha. Thimphu: Dzongkha Development Commission. Driem, George Van (1993). A Grammar of Dumi. Berlin and New York: De Gruyter Mouton. Durie, Mark (1997). ‘Grammatical Structures in Verb Serialisation’, in A. Alsina, J. Bresnan, and P. Sells (eds), Complex Predicates. Stanford, CA: CSLI Publications, 289–354. Dwyer, Arienne, Gülnar Eziz, and Major Travis (2013). ‘The Development of Complex Predication in Turkic: Uyghur Light Verbs’. Paper Presented at the Annual Meeting of the Linguistic Society of America, University of Kansas, January 5, 2013. Ebert, Karen H. (n.d.) Functions of Compound Verbs in Kiranti languages. MS. Elfenbein, Josef. (1998) ‘Brahui’, in S. B. Steever (ed.), The Dravidian Languages. London: Routledge, 388–414. Emeneau, Murray B. (1938). ‘Echo-Words in Toda’, New Indian Antiquary 1: 98–112. Emeneau, Murray B. (1945). ‘The Dravidian Verbs “come” and “give” ’, Language 21: 184–213. Emeneau, Murray B. (1961). Kolami. A Dravidian Language. Annamalainagar: Annamalai University. Emeneau, Murray B. (1984). Toda Grammar and Texts. Philadelphia, PA: American Philosophical Society. Ergin, Muharrem (1988). Orhun Abideleri (12. Baskı). İstanbul: Boğaziçi Yayınları. Fedson, Vijayarani Jotimuttu (1981). The Tamil Serial or Compound Verb. Ph.D. dissertation, University of Chicago. Fedson, Vijayarani Jotimuttu (1985). ‘A Note on Periphrastic Serial or Compound Verb Causative Constructions in Tamil’, in W. H. Eilfort, P. D. Kroeber, and K. L. Peterson (eds), Papers from the 21st Regional Meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 13–20. Foley, William and Mike Olson (1985). ‘Clausehood and Verb Serialization’, in J. Nichols and A. C. Woodbury (eds), Grammar Inside and Outside the Clause. Cambridge: Cambridge University Press, 17–60. Foley, William and Robert D. Van Valin, Jr. (1984). Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Forker, Diana (2012). ‘The Bi-Absolutive Constructions in Nakh-Daghestanian’, Folia Linguistica 46: 75–108. Forsyth, J. (1970). A Grammar of Aspect: Usage and Meaning in the Russian Verb. Cambridge: Cambridge University Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



551

Francke, August Herrmann. (1905–41). Gšamyulna bšadpahi : Kesargyi sgruŋs bžugs. A Lower Ladakhi Version of the Kesar Saga. Calcutta 1905–9: Royal Asiatic Society of Bengal. Frellesvig, Bjarke (2010). A History of the Japanese Language. Cambridge: Cambridge University Press. Frellesvig, Bjarke, Stephen Horn, Kerri Russell, and Peter Sells (2010). ‘Verb Semantics and Argument Realization in Pre-Modern Japanese: A Preliminary Study of Compound Verbs in Old Japanese’, Gengo Kenkyu 138: 25–65. Fukuda, Shin (2012). ‘Aspectual Verbs as Functional Heads: Evidence from Japanese Aspectual Verbs’, Natural Language & Linguistic Theory 30(4): 965–1026. Fukushima, Takenobu (2011). ‘-Te iru no seiritsu to sono hattatsu’ [Establishment of -te iru and Its Development], in H. Aoki (ed.), Nihongo Bunpō no Rekishi to Henka [History and Change in Japanese Grammar]. Tokyo: Kurosio Publishers. Gaeffke, Peter (1967). Untersuchungen zur Syntax des Hindi. The Hague: De Gruyter Mouton. Gagliardi, Annie, Michael Goncalves, Maria Polinsky, and Nina Radkevich (2014). ‘The Biabsolutive Construction in Lak and Tsez’, Lingua 150: 137–70. Genetti, Carol (2006). ‘The Participle Construction of Dolakha Newar. Syntactic Implications of an Asian Converb’, Studies in Language 29(1): 35–87. Godbole, R. S., N. S. Sathe, and N. D. Joshi (1897). Marāt:hī Bhās: ece Madhyam Vyākaran. : Ratnagiri: Bakul Press. Göksel, Aslı and Celia Kerslake (2005). Turkish: A Comprehensive Grammar. Second edition. London and New York: Routledge. Gopalkrishnan, D. and A. Abbi (1992). ‘The Vector Compound Verb: Some Definitional Issues and Criteria for Identification’, Indian Linguistics 53(1-4): 27–46. Graßmann, Hermann (1873). Wörterbuch zum Rig-Veda. Wiesbaden: Harrassowitz. Guru, Kamta Prasad (1920/1962). [Hindi Grammar]. Kashi: Lakshmi Narayan Press. Gurudutt (Vasanth Kumar Shivashankar Padukone) (2015). ‘bãt:vārā [Division], in gurudutt kī lokpriya kahāniyãã (Popular Stories of Gurudutt). Delhi: Prabhat Prakashan, 101–15. Gwynn, J. P. L. (1991). A Telugu-English Dictionary. New Delhi and New York: Oxford University Press. Hacker, Paul (1958). Zur Funktion einiger Hilfsverben im modernen Hindi. Wiesbaden: Steiner in Komm. Hahn, Michael (1996). Lehrbuch der klassischen tibetischen Schriftsprache. Seventh edition. Bonn: Indica et Tibetica. Hahn, Reinhard F. (1991). Spoken Uyghur. Seattle, WA: University of Washington Press. Hahn, Reinhard F. (1998). ‘Uyghur’, in L. Johanson and É. Á. Csató (eds), The Turkic Languages. London and New York: Routledge, 379–96. Hanser, Oskar (1977). Turkmen Manual: Descriptive Grammar of Contemporary Literary Turkmen, Texts, Glossary. Vienna: Verlag des Verbandes der wissenschaftlichen Gesellschaften Österreichs. Harris, Zellig (1968). Mathematical Structures of Language. New York: Interscience Publishers. Hasegawa, Yoko (1996). A Study of Japanese Clause Linkage. Stanford, CA: CSLI Publications and Tokyo: Kurosio Publishers. Haspelmath, Martin (1995). ‘The Converb as a Cross Linguistically Valid Category’, in M. Haspelmath and E. König (eds), Converbs in Cross-Linguistic Perspective: Structure

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

552



and Meaning of Adverbial Verb Forms—Adverbial Participles. Berlin and New York: Mouton de Gruyter, 1–55. Haspelmath, Martin (1999). ‘Why Is Grammaticalization Irreversible?’, Linguistics 37: 1034–68. Haspelmath, Martin (2004). ‘On Directionality in Language Change with Particular Reference to Grammaticalization’, in O. Fischer, M. Norde, and H. Perridon (eds), Up and Down the Cline: The Nature of Grammaticalization. Amsterdam: John Benjamins, 17–44. Haspelmath, Martin (2016). ‘The Serial Verb Construction: Comparative Concept and Cross-Linguistic Generalizations’, Language and Linguistics 17(3): 291–319. Hein, Veronica (2007). ‘The Mirative and Its Interplay with Evidentiality in the Tibetan Dialect of Tabo (Spiti)’, Linguistics of the Tibeto-Burman Area 30(2): 195–214. Heine, Bernd (1993). Auxiliaries: Cognitive Forces and Grammaticalization. Oxford: Oxford University Press. Heine, Bernd (2001). ‘Grammaticalization’, in B. Joseph and R. D. Janda (eds), The Handbook of Historical Linguistics. Oxford: Blackwell, 575–601. Heine, Bernd and Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer (1991). Grammaticalization. A Conceptual Framework. Chicago, IL: University of Chicago Press. Hendriksen, Hans (1944). Syntax of the Infinite Verb Forms of Pāli. Copenhagen: Einar Munksgaard. Hendriksen, Hans (1979). Himachali Studies: II. Texts. Copenhagen: Munksgaard. Herring, Susan C. (1993). ‘Aspectogenesis in South Dravidian: On the Origin of the “Compound Continuative” KONTIRU’, in H. Aertsen and R. J. Jeffers (eds), Historical Linguistics 1989: Papers from the 9th International Conference on Historical Linguistics: Rutgers University, August 14À18, 1989. Amsterdam: John Benjamins, 167À85. Higashitsuji, Yasukazu (1975). ‘Iwayuru fukugōdōshi-kōkō no igironteki-kōsatsu: Genji Monogatari o shiryō toshite’ [A Semological Study on the Second Members of Compound Verbs: Based on The Tale of Genji]. Kokubungakukō [Hiroshima University] 69: 27–39. Higashitsuji, Yasukazu, Yukio Okano, Yumiko Doi, and Katsuaki Katsumura (2003). Heianjidai Fukugōdōshi Sakuin [An Index of Compound Verbs in the Heian Period]. Osaka: Seibundō. Himeno, Masako (1999). Fukugōdōshi no Kōzō to Imi-Yōhō [Structures and Meaning/ Usage of Compound Verbs]. Tokyo: Hituzi Syobo. Hinrichs, Erhard, Andreas Kathol, and Tsuneko Nakazawa (eds) (1998). Complex Predicates in Nonderivational Syntax. San Diego, CA: Academic Press. Hock, Hans Henrich (1985). ‘Transitivity as a Gradient Feature: Synchronic and Diachronic Evidence from Indo-Aryan, Especially Sanskrit’, in A. R. K. Zide, D. Magier, and E. Schiller (eds), Proceedings of the Conference on Participant Roles in South Asia and Adjacent Areas. Bloomington, IN: Indiana University Linguistics Club, 247À63. Hock, Hans Henrich (1992). ‘Review of Tikkanen (1987): The Sanskrit Gerund: A Synchronic, Diachronic, and Typological Analysis’, Kratylos 37: 62À8. Hock, Hans Henrich (2008). ‘Dravidian Syntactic Typology: A Reply to Steever’, in R. Singh (ed.), Annual Review of South Asian Languages and Linguistics. Berlin: De Gruyter Mouton, 164À98. Hock, Hans Henrich (2016). ‘Complex Verbs: Introduction’, in H. H. Hock and E. Bashir (eds), The Languages and Linguistics of South Asia: A Comprehensive Guide (The World of Linguistics 7). Berlin: De Gruyter Mouton, 549À50.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



553

Hoji, Hajime (1985). Logical Form Constraints and Configurational Structures in Japanese. Ph.D. dissertation, University of Washington. Hojiev, A. (1973). Fe’l. Uzbekstan SSR. «Fan» Tashkent: Našriyoti. Hook, Peter E. (1974). The Compound Verb in Hindi. Ann Arbor, MI: Center for South and Southeast Asian Studies, University of Michigan. Hook, Peter E. (1977). ‘The Distribution of the Compound Verb in the Languages of North India and the Question of Its Origin’, International Journal of Dravidian Languages 6: 335À51. Hook, Peter E. (1987). ‘Linguistic Areas: Getting at the Grain of History’, in G. Cardona and N. H. Zide (eds), Festschrift for Henry Hoenigswald, on the Occasion of His Seventieth Birthday. Tübingen: Gunter Narr Verlag, 155–68. Reprinted in International Journal of Dravidian Linguistics 16: 220–37. Hook, Peter E. (1988). ‘Paradigmaticization: A Case Study from South Asia’, in S. Axmaker, A. Jaisser, and H. Singmaster (eds), Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, 293À303. Hook, Peter E. (1991). ‘The Emergence of Perfective Aspect in Indo-Aryan’, in B. Heine and E. C. Traugott (eds), Approaches to Grammaticalization. Vol. 2. Amsterdam and Philadelphia, PA: John Benjamins, 59–89. Hook, Peter E. (1993). ‘Aspectogenesis and the Compound Verb in Indo-Aryan’, in M. K. Verma (ed.), Complex Predicates in South Asian Languages. New Delhi: Manohar, 97À114. Hook, Peter E. (2001a). ‘The Hindi Compound Verb and the Constitution of India’, in D. W. Loenne (ed.), Tohfā-e-Dil. Festschrift Helmut Nespital. Reinbeck: Dr. Inge Wezler Verlag, 227–36. Hook, Peter E. (2001b). ‘Where Do Compound Verbs Come from? And Where Are They Going?’, in P. Bhaskarrao and K. V. Subbarao (eds), The Yearbook of South Asian Languages and Linguistics 2001. New Delhi, Thousand Oaks, CA, and London: Sage Publications, 101–30. Hook, Peter E. (2013a). ‘Competition between Vectored Verbs and Factored Verbs’, NINJAL Project Review 3: 178–89. Tokyo: National Institute for Japanese Language and Linguistics. Hook, Peter E. (2013b). ‘ “Bifunctional Auxiliaries” in Avar and the South and Central Asian Linguistic Area’. MS. Hook, Peter E. and M. M. Chauhan (1988). ‘The Perfective Adverb in Bhitrauti’, WORD 39(3): 177–86. Hook, Peter E. and Omkar N. Koul (1987). “Subject vs. Agent: A Study of the Kashmiri Phasal Verb hye ‘begin to’”, in E. Bashir, M. Deshpande, and P. E. Hook (eds), Select Papers from SALA-7. Bloomington, IN: Indiana University Linguistics Club, 199–219. Reprinted in Journal of the Oriental Institute (Baroda) 36: 115–32. Hook, Peter E. and Omkar N. Koul (1992). ‘On the Compound Verb in Kashmiri’, International Journal of Dravidian Linguistics 21: 1–16. Hook, Peter E. and Prashant Pardeshi (2005). ‘Are Vector Verbs Eternal?’: Paper Given at the South Asian Linguistic Analysis (SALA)-25 Conference, University of Illinois at Urbana-Champaign, September 16–18, 2005. Hopper, Paul J. and Elizabeth Closs Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul J. and Elizabeth Closs Traugott (2003). Grammaticalization. Second edition. Cambridge: Cambridge University Press. Hwang, Hyun Kyung (2011). ‘Distinct Types of Focus and Wh-question Intonation’, Proceedings of the 17th International Congress of Phonetic Sciences (ICPhS XVII), 922–25. City University of Hong Kong.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

554



Hyakutome, Yasuharu (2001). ‘Dōshi rensetsu kara fukugōdōshi e: iru no hojodōshika o chūshin ni’ [From Verb Sequence to Compound Verb: Focusing on the Change of iru (‘enter’) to an Auxiliary Verb], Bungei Kenkyū 152: 79–92. Nihon Bungei Kenkyūkai. Hyakutome, Yasuharu (2002). ‘Fukugōdōshi kōkō: dasu ni okeru imi no rekishiteki hensen’ [Second Members of Compound Verbs: Historical Changes in Meaning in dasu (‘put out’)], Bunka [Tohoku University] 66(1/2): 17–33. Hyakutome, Yasuharu (2015). ‘Asupekuto-fukugōdōshi -morasu no rekishiteki-kentō’ [A Historical Consideration of the Aspectual Compound Verb -morasu]. Shimane-Daigaku Kyōikugakubu Kiyō [Shimane University] 49: 89–106. Ibrahim, Ablahat (1995). Meaning and Usage of Compound Verbs in Modern Uighur and Uzbek. Ph.D. dissertation, University of Washington. Iida, Masayo and Peter Sells (2008). ‘Mismatches between Morphology and Syntax in Japanese Complex Predicates’, Lingua 118: 947–68. Ikegami, Yoshihiko (1991). ‘Do-Language and Become-Language: Two Contrasting Types of Linguistic Representation’, in Y. Ikegami (ed.), The Empire of Signs: Semiotic Essays on Japanese Culture. Amsterdam: John Benjamins, 285–326. Imaizumi, Tadayoshi (1971). Nippo Jisho no Kenkyū [A Study of Nippojisho]. Tokyo: Ōfūsha. Inoue, Kazuko (1976). Henkeibunpō to Nihongo, Ge: Imi-kaishaku o Chūshin ni [Transformational Grammar and Japanese 2: Semantic Interpretation]. Tokyo: Taishukan. Irwin, Mark (2005). ‘Rendaku-Based Lexical Hierarchies in Japanese: The Behaviour of Sino-Japanese Mononoms in Hybrid Noun Compounds’, Journal of East Asian Linguistics 14(2): 121–53. Isaev, M. G. (1975). Osobennosti Tl’anadinskogo Govora Antsukhskogo Dialekta Avarskogo Jazyka [Characteristics of the Tl’anada Subdialect of the Antsukh Dialect of Avar]. Сandidate’s Dissertation. Dagestanskij Gosudarstvennyj Universitet. Makhachkala. Israel, M. (1979). A Grammar of the Kuvi Language. Trivandrum: Dravidian Linguistics Association. Isshiki, Maiko (2016). ‘Nihongo no -oku no rekishiteki hensen’ [Historical Changes to Japanese -oku (‘put’)], in H. Aoki, T. Koyanagi, and Y. Takayama (eds), Nihongo Bunpōshi Kenkyū 3. Tokyo: Hituzi Syobo, 213–40. Jackendoff, Ray (1997). The Architecture of the Language Faculty. Cambridge, MA: MIT Press. Jackendoff, Ray (2010). Meaning and the Lexicon. Oxford: Oxford University Press. Jacobsen, Wesley M. (2016). ‘The Semantic Basis of Japanese Transitive-Intransitive Derivational Patterns’, in T. Kageyama and W. M. Jacobsen (eds), Valency and Transitivity Alternations: Studies on Japanese and Beyond. Berlin and Boston, MA: De Gruyter Mouton, 21–50. Janhunen, Juha A. (2012). Mongolian. Amsterdam: John Benjamins. Jeon, Mingi (2013). ‘Kankokugo no goiteki fukugō dōshi ni okeru asupekuto fukugō dōshi ni tsuite: V-nata, V-nayta, V-tulta no saikō to imi’ [On Aspectual Compound Verbs among Lexical Compound Verbs in Korean, Focusing on a Reconsideration and the Meaning of V-nata, V-nayta, and V-tulta], in T. Kageyama (ed.), Fukugōdōshi Kenkyū no Saisentan [Frontiers of Research on Compound Verbs]. Tokyo: Hituzi Syobo, 331–73. Johanson, Lars (1998a). ‘The Structure of Turkic’, in L. Johanson and É. Á. Csató (eds), The Turkic Languages. London and New York: Routledge, 30–66. Johanson, Lars (1998b). ‘The History of Turkic’, in L. Johanson and É. Á. Csató (eds), The Turkic Languages. London and New York: Routledge, 81–137. Johanson, Lars (2002). Structural Factors in Turkic Language Contacts. Richmond, Surrey: Curzon Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



555

Joshi, R. B. (1925 [1889]). Praudhabodh Marāt:hī Vyākaran: [A Higher Marathi Grammar]. : Seventh edition. Pune: Madhav Joshi Publisher. Juldašev, Axnef A. (1977). Sootnošenie deepričastnyx i ličnyx form glagola v tjurkskix jazykax [Correlation of Converbial and Finite Verbal Forms in Turkic languages]. Moscow: Nauka. Kachru, Yamuna. (1979). ‘Pragmatics and Verb Serialization in Hindi-Urdu’, Studies in the Linguistic Sciences 9(2): 157–69. Kachru, Yamuna and Rajeshwari Pandharipande (1980). ‘Towards a Typology of Compound Verbs in South Asian Languages’, Studies in the Linguistic Sciences 10(1): 113–24. Kageyama, Taro (1989). ‘The Place of Morphology in the Grammar: Verb-Verb Compounds in Japanese’, in G. Booij and J. van Marle (eds), Yearbook of Morphology 2. Dordrecht: Foris: 73–94. Kageyama, Taro (1993). Bunpō to Gokeisei [Grammar and Word Formation]. Tokyo: Hituzi Syobo. Kageyama, Taro (1996). Dōshi Imiron: Gengo to Ninchi no Setten [Verb Semantics: The Interface of Language and Cognition]. Tokyo: Kurosio Publishers. Kageyama, Taro (2009a). ‘Isolate: Japanese’, in R. Lieber and P. Štekauer (eds), The Oxford Handbook of Compounding. Oxford: Oxford University Press, 512–26. Kageyama, Taro (2009b). ‘Dekigoto o arawasu ukemi’ [Passives Denoting Events], in T. Kageyama (ed.), Nichiei-Taishō Keiyōshi/Fukushi no Imi to Kōbun [Syntax and Semantics of Adjectives and Adverbs: A Contrastive Study of Japanese and English]. Tokyo: Taishukan, 78–119. Kageyama, Taro (2010). ‘Variation between Endocentric and Exocentric Word Structures’, Lingua 120: 2405–23. Kageyama, Taro (2013). ‘Goiteki-fukugōdōshi no shin-taikei’ [A New System of Lexical compound verbs], in T. Kageyama (ed.), Fukugōdōshi Kenkyū no Saisentan [Frontiers of Research on Compound Verbs]. Tokyo: Hituzi Syobo, 3–46. Kageyama, Taro (2016a). ‘Verb-Compounding and Verb-Incorporation’, in T. Kageyama and H. Kishimoto (eds), Handbook of Japanese Lexicon and Word Formation. Berlin and Boston, MA: De Gruyter Mouton, 273–310. Kageyama, Taro (2016b). ‘Agents in Anticausative and Decausative Compound Verbs’, in T. Kageyama and W. M. Jacobsen (eds), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin and Boston, MA: De Gruyter Mouton, 89–124. Kageyama, Taro (2018). ‘Compound and Complex Predicates in Japanese’, The Oxford Research Encyclopedia of Linguistics. Oxford University Press, https://oxfordre.com/ linguistics, accessed August 4, 2020. Kageyama, Taro (2020). ‘Morphology in Japonic Languages’, The Oxford Research Encyclopedia of Linguistics. Oxford University Press, https://oxfordre.com/linguistics, accessed, August 4, 2020. Kageyama, Taro and Kyoko Kanzaki (2014). Compound Verb Lexicon (online database). Tokyo: National Institute for Japanese Language and Linguistics, https://db4.ninjal.ac.jp/ vvlexicon/db/, accessed August 23, 2020. Kageyama, Taro and Li Shen (2018). ‘Resultative Constructions in Japanese from a Typological Perspective’, in P. Pardeshi and T. Kageyama (eds), Handbook of Japanese Contrastive Linguistics. Berlin and Boston, MA: De Gruyter Mouton, 193–226. Kang, Seung-Man (1993). ‘Serial Verb Constructions in Korean and Their Implications’, Studies in Generative Grammar 3(1): 79–109. Kapp, Dieter B. (2006). Ālu-Kurmba-Texte. Wiesbaden: Harrassowitz Verlag.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

556



Karimata, Shigehisa (2017). ‘The Linguistic Difference between Northern and Southern Ryukyuan from the Perspective of Human Movement’. Paper read at the Workshop ‘Human Evolution in Eurasia Elucidated Through Genetics, Archeology, and Linguistics’. Mishima: National Institute of Genetics. Karpov, V. G. (1975). ‘Glagol’, in N. A. Baskakov (ed.), Grammatika xakasskogo yazïka. Moscow: Izdatel’stvo Nauka, 163–245. Kato, Atsuhiko (2018). ‘Entailed and Intended Results in Japanese and Burmese’, in P. Paradeshi and T. Kagayema (eds), Handbook of Japanese Contrastive Linguistics. Berlin and Boston, MA: De Gruyter Mouton, 175–92. Kaul, Vijay Kumar (2006). Compound Verbs in Kashmiri. Delhi: Indian Institute of Language Studies. Kellogg, S. H. (1893). A Grammar of the Hindi Language, in Which Are Treated the High Hindi, Braj, and the Eastern Hindi of the Ramayana of Tulsidas, also the Colloquial, Dialects of Rajputana, Kumaon, Avadh, Riwa, Bhojpur, Magadha, Maithila, etc. London: Paul, Trench, Trubner & Co. Reprinted 1990, Delhi: Munshiram Manoharlal. Kher, A. K. (1899 [1895]). A Higher Marathi Grammar. Second edition. Bombay: R. A. Sagoon. Kimmig, Rainer (2018). ‘Hindi-Urdu Compound Verbs and Russian Aspect: The Evidence of Translations’. Paper Presented at 34th South Asian Languages Analysis Roundtable (SALA), University of Konstanz, Germany, June 2018. Kindaichi, Haruhiko (1953). ‘Kokugo-akusentoshi no kenkyū ga nani ni yakudatu ka’ [What Is the Research on the History of Japanese Accent Good for?], in Kindaichihakushi Koki-kinen Gengo-minzoku Ronsō [Papers on Language and Folklore in Honor of Dr. Kyosuke Kindaichi on His Seventieth Birthday]. Tokyo: Sanseidō, 1–26. Kinsui, Satoshi (2004). ‘Bunmyaku teki kekka jōtai ni motozuku nihongo jodōshi no imi kijutsu’ [A Semantic Description of Japanese Auxiliaries Based on Resultative States in Context], in T. Kageyama and H. Kishimoto (eds), Nihongo no Bunseki to Gengo Ruikei [Analyses of Japanese and Linguistic Typology]. Tokyo: Kurosio Publishers, 47–56. Kinsui, Satoshi (2006). Nihongo Sonzai Hyōgen no Rekishi. [A History of Existential Expressions in Japanese]. Tokyo: Hituzi Syobo. Kishimoto, Hideki (2005). Tōgo Kōzō to Bunpō Kankei [Syntactic Structures and Grammatical Relations]. Tokyo: Kurosio Publishers. Kishimoto, Hideki (2006). ‘Japanese Syntactic Nominalization and VP-Internal Syntax’, Lingua 116: 771–810. Kishimoto, Hideki (2009). ‘Hobun o toru dōshi to keiyōshi: Kontorōru to jōshō’ [Complement Taking Verbs and Adjectives: Control and Raising], in T. Kageyama (ed.), Nichiei-Taishō Keiyōshi/Fukushi no Imi to Kōbun [Syntax and Semantics of Adjectives and Adverbs: A Contrastive Study of Japanese and English]. Tokyo: Taishukan, 152–90. Kishimoto, Hideki (2010). ‘Subjects and Constituent Structure in Japanese’, Linguistics 48: 629–70. Kishimoto, Hideki (2013). ‘Tōgoteki fukugōdōshi no kaku to tōgotokusei’ [Case and Syntactic Properties of Syntactic Compound Verbs], in T. Kageyama (ed.), Fukugōdōshi Kenkyū no Saisentan [Frontiers of Research on Compound Verbs]. Tokyo: Hituzi Syobo, 143–83. Kishimoto, Hideki (2016). ‘Bun-no kōzō to kaku-kankei’ [Sentence Structures and Case Relations], in K. Murasugi, M. Saito, Y. Miyamoto, and K. Takita (eds), Nihongo Bunpō Handobukku [A Handbook of Japanese Grammar]. Tokyo: Kaitakusha, 102–45.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



557

Kitagawa, Kazuhide (1982). Shoku Nihongi Senmyō: Kōhon Sōsakuin [The Senmyō (Imperial Edicts) in the Shoku Nihongi: Critical Edition and Indices]. Tokyo: Yoshikawa Kōbunkan. Klamer, Marian (1998). A Grammar of Kambera. Berlin and New York: De Gruyter Mouton. Ko, Heejeong and Daeyoung Sohn (2015). ‘Decomposing Complex Serialization: The Role of v’, Korean Linguistics 17(1): 78–125. Kobayashi, Masato (2012). Texts and Grammar of Malto. Vizianagaram: Kotoba Books. Koizumi Masatoshi (1999). Phrase Structure in Syntax. Tokyo: Hituzi Syobo. Kojima, Satoko (1999). ‘Fukugōdōsi kōkō yuku no hensen [History of yuku as a Second Member of Compound Verbs]. Kokugo-kokubun 76: 57–69. Komatsu, Hideo (1971). Nihongo Seichōshi Ronkō [A Study of the History of Japanese Prosody]. Tokyo: Kasama Shoin. Kononov, Andrej N. (1960). Grammatika sovremennogo uzbekskogo literaturnogo jazyka [A Grammar of Contemporary Literary Uzbek]. Moscow and Leningrad: Izdatel’stvo AN SSSR. Krifka, Manfred (1992). ‘Thematic Relations as Links between Nominal Reference and Temporal Construction’, in I. Sag and A. Szabolsci (eds), Lexical Matters. Chicago: University of Chicago Press, 30–52. Krippes, Karl A. (1996). Uzbek-English Dictionary. Kensington, MD: Dunwoody Press. Krippes, Karl A. (1998). Kyrgyz-English Dictionary. Kensington, MD: Dumwoody Press. Krishnamurti, Bh. (1969). Konda or Kubi: A Dravidian Language. Hyderabad: Tribal Cultural Research and Training Institute. Krishnamurti, Bh. (2003). The Dravidian Languages. Cambridge: Cambridge University Press. Krishnamurti, Bh. and J. P. L. Gwynn (1985). A Grammar of Modern Telugu. Delhi: Oxford University Press. Kubota, Yusuke (2007). ‘The Scope Interpretation of Complex Predicates in Japanese: A Unified Lexicalist Analysis’, Journal of Linguistics 43: 489–530. Kubozono, Haruo (1999). Nihongo no Onsei [Sounds of Japanese]. Tokyo: Iwanami. Kudaybergenov, S. (1987). ‘Glagol’, in O. V. Zaxarova (ed.), Grammatika kirgizskogo literaturnogo yazïka 1: Fonetika i morfologiya. Frunze: Ilim, 207–94. Kudo, Mayumi (1995). Asupekuto-tensu-taikei to Tekusuto: Gendai-nihongo no Jikan no Hyōgen [Tense-Aspect System and Text: Temporal Expressions in Modern Japanese]. Tokyo: Hituzi. Kuno, Susumu (1973). The Structure of the Japanese Language. Cambridge, MA: MIT Press. Kuno, Susumu (1978a). Danwa no Bunpō [Discourse Grammar]. Tokyo: Taishukan. Kuno, Susumu (1978b). Shin Nihon Bunpō Kenkyū [A New Study of Japanese Grammar]. Tokyo: Taishukan. Kuno, Susumu (1986). Functional Syntax: Anaphora, Discourse and Empathy. Chicago, IL: University of Chicago Press. Kuribayashi, Yuu (2009). ‘Contact Induced Changes in Southwestern Turkic –Emergence of Analytic Strategy for Modals –’, in S. Ay, Ö. Aydın, İ. Ergenç, S. Gökmen, S. İşsever, and D. Peçenek (eds), Essays on Turkish Linguistics. Wiesbaden: Harrassowitz Verlag, 413–20. Kuribayashi, Yuu (2018). ‘Verb-Verb Compounding in Japanese and Turkish’, in P. Pardeshi and T. Kageyama (eds), Handbook of Japanese Contrastive Linguistics. Berlin and Boston, MA: De Gruyter Mouton, 227–46. Kuteva, Tania (2001). Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

558



Lapointe, Steven G. (1980). A Theory of Grammatical Agreement. Ph.D. dissertation, University of Massachusetts, Amherst. Lee, Sook-hee (1992). The Syntax and Semantics of Serial Verb Constructions. Ph.D. dissertation, University of Washington. Lee, Sun Myoung (2006). Kukŏ ŏgan pokhapŏ yŏngu [Research on Root Compounds in Korean]. Seoul: Thaehaksa. Lee, Youngjoo (2003). ‘Two Kinds of Structural Relationships in SVCs’, in G. Iverson and S.-C. Ahn (eds), Explorations in Korean Language and Linguistics. Seoul: Hankook Publishing Company, 443–58. Levinson, Stephen C. (1983). Pragmatics. Cambridge: Cambridge University Press. Levinson, Stephen C. (2000). Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press. Li, Chao (2007). ‘Split Ergativity and Split Intransitivity in Nepali’, Lingua 117: 1462–82. Li, Charles N. and Sandra A. Thompson (1974). ‘An Explanation of Word Order Change SVO ! SOV’, Foundations of Language 12: 201–14. Li, Charles N. and Sandra A. Thompson (1981). Mandarin Chinese: A Functional Reference Grammar. Berkeley and Los Angeles, CA: University of California Press. Li, Yafei (1990). ‘On V-V Compounds in Chinese’, Natural Language & Linguistic Theory 8: 177–207. Liang, Hsin-hsin and Peter Hook (2007). ‘The Compound Verb in Chinese and HindiUrdu and the Plausibility of Macro Linguistic Areas’, in C. P. Masica (ed.), Old and New Perspectives on South Asian Languages: Grammar and Semantics. Delhi: Motilal Banarsidass, 43. Lieber, Rochelle and Pavol Štekauer (2009). ‘Introduction: Status and Definition of Compounding’, in R. Lieber and P. Štekauer (eds), The Oxford Handbook of Compounding. New York: Oxford University Press, 3–18. Liljegren, Henrik (2016). A Grammar of Palula. Berlin: Language Science Press Lindholm, James (1969). The Role of iru ‘be’ in a Grammar of Tamil. MA thesis, University of Chicago. Liu, Haiyong (2003). Complex Predicates in Mandarin Chinese: Three Types of bu-yu Structures. Ph.D. dissertation, University of California, Los Angeles. Los, Bettelou, Corrien Blom, Geert Booij, Marion Elenbaas, and Ans van Kemenade (2012). Morphosyntactic Change: A Comparative Study of Particles and Prefixes. Cambridge: Cambridge University Press. Lu, Jianji (1984). ‘Directional Complement: A Pedagogical View’, Journal of Chinese Language Teachers’ Association 19: 79–94. Lü, Xiao-xin (1999). mào nián yì wǎng (Recalling the Past While in my Eighties). Taiwan: published by the author. Lyons, John (1977). Semantics, 2 vols. Cambridge: Cambridge University Press. Lyutikova, Ekaterina A., Sergei G. Tatevosov, Anna G. Pazelskaya, Mikhail Ju. Ivanov, and Andrey B. Shluinsky (2006). Struktura sobytija i semantika glagola v karačaevobalkarskom jazyke [Event Structure and Verb Semantics in Karachay-Balkar]. Moscow: Imli RAN. McGregor, R. S. (1968). The Language of Indrajit of Orchā: A Study of Early Braj Bhās: ā Prose. Cambridge: Cambridge University Press. Magomedova, Dzhavgarat Abdurakhmanovna (2010). Tipologija Sredstv Vyrazhenija Funktsional’no-semanticheskoj Kategorii Aspektual’nosti v Avarskom i Anglijskom Jazykakh [Typology of Ways of Expression of Functional-Semantic Categories of Aspectuality in Avar and English]. Сandidate’s Dissertation. Institut Jazyka, Literatury i Iskusstva DNTs RAN. Makhachkala.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



559

Magomedova, Patimat Aripovna (2006). Semantika i Sintaksis Avarskogo Glagola [Semantics and Syntax of Avar Verbs]. Ph.D. Dissertation. Institut Jazykoznanija RAN. Moscow. Mahapatra, B. P. (1979). Malto: An Ethnosemantic Study. Mysore: Central Institute of Indian Languages. Mallaeva, Z. M. (2000). Vido-vremennaja Sistema Glagola Avarskogo Literaturnogo Jazyka [Aspect and Time System of Standard Avar Verbs]. Ph.D. dissertation, Institut Jazyka, Literatury i Iskusstva DNTs RAN. Makhachkala. Mallaeva, Z. M. (2007). Glagol Avarskogo Jazyka: Struktura, Semantika, Funktsii [Avar Verbs: Structure, Semantics, Functions]. Makhachkala: Institut Jazyka, Literatury i Iskusstva DNTs RAN. Margetts, Anna (1999). Valence and Transitivity in Saliba, an Oceanic Language of Papua New Guinea. (MIP Series in Psycholinguistics 12.) Nijmegen: Max Planck Institute for Psycholinguistics. Martin, Samuel E. (1975). A Reference Grammar of Japanese. New Haven, CT: Yale University Press. Martin, Samuel E. (1987). The Japanese Language through Time. New Haven, CT, London: Yale University Press. Martin, Samuel (1993). A Reference Grammar of Korean. Rutland, VT: Tuttle. Martin, Samuel (1997). ‘Un-Altaic Features of the Korean Verb’, in H.-M. Sohn and J. Haig (eds), Japanese/Korean Linguistics 6. Stanford, CA: CSLI Publications, 3–40. Masica, Colin P. (1976). Defining a Linguistic Area: South Asia. Chicago, IL, and London: University of Chicago Press. Masica, Colin P. (1991). The Indo-Aryan Languages. Cambridge: Cambridge University Press. Masica, Colin P. (2001). ‘The Definition and Significance of Linguistic Areas: Methods, Pitfalls, and Possibilities (with Special Reference to the Validity of South Asia as a Linguistic Area)’, in P. Bhaskararao and K. V. Subbarao (eds), South Asia Yearbook 2001: Papers from the Symposium on South Asian Languages: Contact, Convergence and Typology. Delhi: SAGE, 205–67. Masica, Colin P. (ed.) (2006). Old and New Perspectives on South Asian Languages: Grammar and Semantics. Delhi: Motilal Banarsidass. Masuoka, Takashi (2012). ‘Nihongo dōshi no katsuyō saihō’ [Verbal Inflection in Japanese Revisited], in K. Mihara and Y. Nitta (eds), Katsuyō-ron no Zensen [Advances in the Studies of Verb Conjugation]. Tokyo: Kurosio Publishers. Masuoka, Takashi (2014). ‘Nihongo no chūritsu-kei setsuzoku to te-kei setsuzoku no kyōgō to kyōzon’ [Competition and Coexistence of Linkage with the Neutral Form and the –te Form in Japanese], in T. Masuoka et al. (eds), Nihongo Fukubun Kōbun no Kenkyū [Studies in Japanese Complex Sentences]. Tokyo: Hituzi, 521–42. Matasović, Ranko (1996). A Theory of Textual Reconstruction in Indo-European Linguistics. Frankfurt am Main: Peter Lang Verlag. Matisoff, James. A. (1973). The Grammar of Lahu. University of California Publications in Linguistics, 75. Berkeley, CA: University of California Press. Matisoff, James A. (1982). The Grammar of Lahu. Second edition. Berkeley, CA: University of California Press. Matsumoto, Noriko (2009). ‘The Asymmetry between the iki (go) + V and the ki (come) + V Constructions’, in Y. Takubo et al. (eds), Japanese/Korean Linguistics 16. Stanford, CA: CSLI Publications, 289–303. Matsumoto, Takanobu (ed.) (1970). Hyakunijukkubon Heike Monogatari [The Hyakkunijukku Manuscript of the Heike Monogatari]. Tokyo: Kyūko Shoin.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

560



Matsumoto, Yo (1990). ‘On the Syntax of the Japanese Intransitivizing -te aru Construction: Non Lexical Function Changing’, in M. Ziolkowski, M. Noske, and K. Deaton (eds), Papers from the 26th Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 277–92. Matsumoto, Yo (1991). ‘On the Lexical Nature of Purposive and Participial Complex Motion Predicates in Japanese’, in L. A. Sutton (ed.), Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, 180–91. Matsumoto, Yo (1996). Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion ‘Word’. Stanford, CA: CSLI Publications and Tokyo: Kurosio Publishers. Matsumoto, Yo (1997). ‘Kūkan-idō no gengo-hyōgen to sono kakuchō’ [Linguistic Expressions of Motion in Space and Their Extensions], in S. Tanaka and Y. Matsumoto (eds), Kūkan to Idō no Hyōgen [Expressions of Space and Motion]. Tokyo: Kenkyusha, 125–230. Matsumoto, Yo (1998). ‘Nihongo no goiteki-fukugōdōshi ni okeru dōshi no kumiawase’ [The Combinatory Possibilities in Japanese V-V Lexical Compounds], Gengo Kenkyu 114: 37–83. Matsumoto, Yo (2003). ‘Typologies of Lexicalization Patterns and Event Integration: Clarifications and Reformulations’, in S. Chiba et al. (eds), Empirical and Theoretical Investigations into Language: A Festschrift for Masaru Kajita. Tokyo: Kaitakusha, 403–17. Matsumoto, Yo (2010). ‘Aspects of the Semantics of Logophoricity: Comparison of Malayalam with Yoruba and Japanese’, in L. Uyechi and L.-H. Wee (eds), Reality Exploration and Discovery: Pattern Interaction in Language and Life. Stanford, CA: CSLI Publications, 269–84. Matsumoto, Yo (2017a). ‘Nihongo ni okeru idō-zishō-hyōgen no taipu to keiro no hyōgen’ [Types of Mmotion-Event Expressions and Expressions of Path in Japanese], in Y. Matsumoto (ed.), Idō-hyōgen no Ruikeiron [The Typology of Motion Expressions]. Tokyo: Kurosio Publishers, 247–73. Matsumoto, Yo (2017b). ‘Idō-hyōgen no ruikeisei’ [Typological Nature of Motion Event Descriptions], in Y. Matsumoto (ed.), Idō-hyōgen no Ruikeiron [The Typology of Motion Expressions]. Tokyo: Kurosio Publishers, 337–53. Matsumoto, Yo (2018). ‘Motion Event Descriptions in Japanese: Typological Perspectives’, in T. Kageyama and P. Pardeshi (eds), Handbook of Japanese Contrastive Linguistics. Berlin and Boston, MA: De Gruyter Mouton, 273–89. Matsumoto, Yo, Kimi Akita, and Kiyoko Takahashi (2017). ‘The Functional Nature of Deictic Verbs and the Coding Patterns of Deixis: An Experimental Study in English, Japanese, and Thai’, in I. Ibarretxe-Antuñano (ed.), Motion and Space across Languages and Applications. Amsterdam: John Benjamins, 95–122. Matsuo, Sutejiro (1936). Kokugohō Ronkō [Treatise on Japanese Grammar]. Tokyo: Hakuteisha. Matsuse, Ikuko (2017). ‘Newāru-go no idō-hyōgen’ [Motion Event Descriptions in Newar], in Y. Matsumoto (ed.), Idō-hyōgen no Ruikeiron [The Typology of Motion Expressions]. Tokyo: Kurosio Publishers, 65–94. Matthews, David (1998). A Course in Nepali. London: Curzon. Meltem, Filiz and Erdem Uçar (2016). ‘Çağatay Türkçesinde tasvir fiilleri’ [Descriptive Verbs in Chaghatay Turkish], Türkbilig 32: 21–56. Merlan, F. C. (1994). A Grammar of Wardaman. Berlin: De Gruyter Mouton.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



561

Mihara, Kenichi (2011). ‘Te-kei-dōshi no imi-ruikei’ [A Semantic Typology of –te Verbs], Nihongo Nihon-bunka Kenkyū [Osaka University] 21: 1–12. Minami Fujio (1993). Gendai Nihongo Bunpō no Rinkaku [Outline of the Grammar of Modern Japanese]. Tokyo: Taishūkan. Minorsky, Vladimir (1940). ‘The Turkish Dialect of the Khalaj’, Bulletin of the School of Oriental Studies 10(2): 417. (Turkish version: F. Güley, Halac Türk Diyalekti, TDED, C. 4.1–2, 1950: 83–106). Miyagawa, Shigeru (1987). ‘Restructuring in Japanese’, in K. Imai and M. Saito (eds), Issues in Japanese Linguistics. Dordrecht: Foris, 273–300. Miyagawa, Shigeru (1989). Structure and Case Marking in Japanese. San Diego, CA: Academic Press. Miyaji, Hitoshi (1981). ‘Keigoshi-ron’ [A Study of the History of Honorifics], in H. Miyaji (ed.), Keigoshi [History of Honorifics]. Tokyo: Meiji Shoin, 1–25. Mori, Yuta (2016). Hatsuwa-kōi kara Mita Nihongo Juju-hyōgen no Rekishiteki-kenkyū [A Historical Study of Japanese Expressions of Giving and Taking from the Perspectives of Speech Acts]. Tokyo: Hituzi. Muhamedowa, Raihan (2016). Kazakh: A Comprehensive Grammar. London and New York: Routledge. Muraki, Masatake (1978). ‘The shika nai Construction and Predicate Raising’, in J. Hinds and I. Howard (eds), Problems in Japanese Syntax and Semantics. Tokyo: Kaitakusha,155–77. Musaev, Kenesbay Musaevič (2008). Kazaxskiy yazïk. Moscow: Izdatel’skaya firma Vostočnaya literatura. Myhill, John (1992). Typological Discourse Analysis. Oxford: Basil Blackwell. Myhill, John and Junko Hibiya (1988). ‘The Discourse Function of Clause-chaining’, in J. Haiman and S. A. Thompson (eds), Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins, 361–98. Nakatani, Kentaro (2003). ‘Analyzing -te’, in W. McClure (ed.), Japanese/Korean Linguistics 12. Stanford, CA: CSLI Publications, 377–87. Nakatani, Kentaro (2013). Predicate Concatenation: A Study of the V-te V Predicate in Japanese. Tokyo: Kurosio Publishers. Nakatani, Kentaro (2016). ‘Complex Predicates with -te Gerundive Verbs’, in T. Kageyama and H. Kishimoto (eds), Handbook of Japanese Lexicon and Word Formation. Berlin and Boston, MA: Mouton de Gruyter, 387–423. Narrog, Heiko (2010). ‘The Order of Meaningful Elements in the Japanese Verbal Complex’, Morphology 20: 205–37. Narrog, Heiko and Toshio Ohori (2011). ‘Grammaticalization in Japanese’, in Heiko Narrog and Bernd Heine (eds), The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, 775–85. Narrog, Heiko, Seongha Rhee, and John Whitman. (2018). ‘Grammaticalization in Japanese and Korean’, in Heiko Narrog and Bernd Heine (eds), Grammaticalization from a Typological Perspective. Oxford: Oxford University Press, 166–88. Nash, Léa and Pollet Samvelian (eds) (2016). Approaches to Complex Predicates. Leiden and Boston, MA: Brill. Navalkar, G. R. (1880). The Student’s Marathi Grammar. Bombay: Education Society’s Press. Nayar, Prabodachandran V. R. (1979). ‘Aspectual System in Malayalam’, International Journal of Dravidian Linguistics 8(2): 289À99. Nedjalkov, Igor (1997). Evenki. London: Routledge.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

562



Nedjalkov, Igor’ V. and Vladimir P. Nedjalkov (1987). ‘Karačaevo-balkarskaja forma na =b/ =p tura so značenijami nastojaščego i prošedšego vremeni’ [Karachay-Balkar Form on =b/=p tura with Present and Past Tense Meaning], in A. M. Muxin et al. (eds), Lingvističeskie issledovanija 1987. Funkcional’no-semantičeskie aspekty grammatiki [Linguistic Research 1987. Functional-Semantic Aspects of Grammar]. Moscow: Institut jazykoznanija AN SSSR, 113–21. Nedjalkov, Vladimir P. (1995). ‘Some Typological Parameters of Converbs’, in M. Haspelmath and E. König (eds), Converbs in Cross-Linguistic Perspective. Berlin and New York: De Gruyter Mouton, 97–136. Nedjalkov, Vladimir P. and Sergej E. Jaxontov (1983). ‘Tipologia rezul’tativnyx konstrukcij’ [A Typology of Resultative Constructions], in V. P. Nedjalkov (ed.), Tipologia rezul’tativnyx konstrukcij [A Typology of Resultative Constructions]. Leningrad: Nauka, 5-41. English translation: Nedjalkov, Vladimir. P. and Sergej E. Jaxontov (1988). ‘The Typology of Resultative Constructions’, in V. P. Nedjalkov (ed.), Typology of Resultative Constructions. Amsterdam and Philadelphia, PA: John Benjamins, 3–62. Nespital, Helmuth (1989). ‘Verbal Aspect and Lexical Semantics in Indo-Aryan languages: The Typology of Verbal Expressions (Compound Verbs) and Their Relation to Simple Verbs’, Studien zur Indologie und Iranistik 15: 159–96. Nespital, Helmut (1997a). Hindī kriyā-koś / Dictionary of Hindi Verbs. Allahabad: Lokbharati. Nespital, Helmut (1997b). ‘Les Verbes composés dans les langues d’Asie méridionale [Philippe Bourdin, traducteur]’. Faits de langues 5(10): 37–50. Nespital, Helmuth (1997c). ‘The Use of Non-Perfective Simple Verbs and of Perfective Verbal Expressions [“Compound Verbs”] in New Indo-Aryan Languages: Grammatical, Lexical and Pragmatic Criteria for Their Selection’, International Journal of Dravidian Linguistics 26(1): 39–56. Nespital, Helmut (2008). Dictionary of Hindi Verbs ( ). Allahabad: Lok Bharti. Newmeyer, Frederick J. (2001). ‘Deconstructing Grammaticalization’. Language Sciences 23: 187–229. NINJAL (National Institute for Japanese Language and Linguistics) (2014). The Compound Verb Lexicon [Original Data], https://db4.ninjal.ac.jp/vvlexicon/, accessed July 1, 2020. Nishigauchi, Taisuke (1993). ‘Long Distance Passive’, in N. Hasegawa (ed.), Japanese Syntax in Comparative Grammar. Tokyo: Kurosio Publishers, 79–114. Nishiyama, Kunio (1998). ‘V-V Compounds as Serialization’, Journal of East Asian Linguistics 7: 175–217. Nishiyama, Kunio (2008). ‘V-V compounds’, in S. Miyagawa and M. Saito (eds), The Oxford Handbook of Japanese Linguistics. Oxford: Oxford University Press, 320–47. Nishiyama, Kunio and Yoshiki Ogawa (2014). ‘Auxiliation, Atransitivity and Transitivity Harmony in Japanese V-V compounds’, Interdisciplinary Information Sciences 20: 71–101. Nitta, Tetsuo (2010). ‘Ishikawa-ken Shiramine hōgen no fukugōdōshi no akusento’ [The Accent of Verb Compounds in the Dialect of Shiramine in Ishikawa Prefecture], in Z. Uwano (ed.), Nihongo Kenkyū no 12 shō [12 Chapters on the Japanese Language]. Tokyo: Meiji Shoin, 413–28. Nitta, Yoshio (1991). Nihongo no Modaritī to Ninshō [Japanese Modality and Person]. Tokyo: Hituzi. Nitta, Yoshio (1995). ‘Shite-kei setsuzoku o megutte’ [Concerning the Linkage with the shite-form], in Y. Nitta (ed.), Fukubun no Kenkyū (Jo) [Studies on Complex Sentences I]. Tokyo: Kurosio Publishers, 87–126.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



563

Nolan, Brian and Elke Diedrichsen (eds) (2017). Argument Realisation in Complex Predicates and Complex Events: Verb-Verb Constructions at the Syntax-Semantic Interface. Amsterdam: John Benjamins. Norde, Muriel (2009). Degrammaticalization. Oxford: Oxford University Press. Okimori, Takuya (1990). ‘Kotengo no fukugōdōshi’ [Compound Verbs in Classical Japanese], Kokubungaku Extra Number 38: Koten Bunpō Hikkei [Gakutōsha]: 55–63. Ono, Tsuyoshi (1992). ‘The Grammaticization of the Japanese Verbs oku and shimau’, Cognitive Linguistics 3(4): 367–90. Ono, Tsuyoshi and Ryoko Suzuki (1992). ‘The Development of a Marker of Speaker’s Attitude: The Pragmatic Use of the Japanese Grammaticalized Verb shimau in Conversation’, in L. A. Buszard-Welcher, J. Evans, D. Peterson, L. Wee, and W. Weigel (eds), Proceedings of the Eighteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, 204–13. Oprina, Florin (2014). ‘V-V Predicates and Restructuring’, in Y. Yumoto and H. Kishimoto (eds), Fukuzatsu-Jutsugo Kenkyū no Genzai [Current Issues in Complex Predicate Research]. Tokyo: Hituzi Syobo, 151–77. Oyakawa, Takatsugu (1975). ‘On the Japanese shika-nai Construction’, Gengo Kenkyu 67: 1–20. Öztürk, Balıkız (2004). Case, Referentiality and Phrase Structure. Ph.D. dissertation, Harvard University. Paia, Mirose and Jacques Vernaudon (2004). ‘Complex Predicates in Tahitian’, in I. Bril and F. Ozanne-Rivierre (eds), Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness. Berlin: De Gruyter Mouton, 251–62. Pajzulaeva, Patimat Gamzatovna (2009). Deeprichastie v Avarskom Jazyke [Converbs in Avar]. Сandidate’s dissertation, Dagestanskij Gosudarstvennyj Pedagogicheskij Universitet. Makhachkala. Pandharipande, Rajeshwari. (1990). ‘Serial Verb Construction in Marathi’, in B. D. Joseph and A. M. Zwicky (eds), When Verbs Collide: Papers from the 1990 Ohio State MiniConference on Serial Verbs. Columbus, OH: Deptartment of Linguistics, Ohio State University, 178–99. Pandharipande, Rajeshwari (1993). ‘Serial Verb Construction in Marathi’, in M. K. Verma (ed.), Complex Predicates in South Asian Languages. New Delhi: Manohar, 177–95. Paolillo, John C. (1989). ‘Deictic and Dynamic Interactions in Sinhala Verb-Verb Compounds’. MS., Stanford University. Paranavitana, Senarat (1956). Sigiri Graffiti, Sinhalese Verses of the Eighth, Ninth and Tenth Centuries. London: Oxford University Press. Pardeshi, Prashant (2001). ‘The Compound Verb in Marathi: Definitional Issues and Criteria for Identification’, Kobe Papers in Linguistics [Kobe University] 38: 94À111. Pardeshi, Prashant (2003). ‘The Compound Verb in Marathi: Definitional Issues and Criteria for Identification’, Indian Linguistics 64: 19–36. Pardeshi, Prashant, Peter E. Hook, and Sung-Yeo Chung (2006). ‘In Search of the Origins of Compound Verbs in Marathi’. Paper Presented at SALA 26, CIIL, Mysore. Payne, Thomas E. (1997). Describing Morphosyntax. Cambridge: Cambridge University Press. Pazelskaya, Anna G. and Natalija V. Rybinceva (2009). ‘Akcional’nye klassy glagolov i smežnye voprosy’ [Actional Classes of Verbs and Related Issues], in S. G. Tatevosov (ed.), Tubalarskie ètjudy [Essays on Tubalar]. Moscow: IMLI RAN, 54–77. Pazelskaya, Anna G. and Andrey B. Shluinsky (2007). ‘Obstojatel’stvennye predloženija’ [Adverbial Clauses], in E. A. Lyutikova et al. (eds), Mišarskij dialekt tatarzkogo jazyka:

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

564



očerki po sintaksisu i semantike. [Mishar Dialect of Tatar: Essays on Syntax And Semantics]. Kazan: Magarif, 38–82. Pederson, Eric (2008). ‘Event Realization in Tamil’, in M. Bowerman and P. Brown (eds), Crosslinguistic Perspectives on Argument Structure. New York: Lawrence Erlbaum, 331–57. Perlmutter, David (1970). ‘The Two Verbs begin’, in R. Jacobs and P. Rosenbaum (eds), Readings in English Transformational Grammar. Waltham, MA: Ginn, 107–19. Peterson, John (2002). ‘The Nepali Converbs: A Holistic Approach’, in R. Singh (ed.), The Yearbook of South Asian Languages and Linguistics. New Delhi: Sage and London: Thousand Oaks, 93À133. Pokharel, Balkrishna ( 2020 [=1963 ]). pañc śaya vars: a. Kathmandu: Sajha Prakashan. Pokharel, Madhav P. (1991). ‘Compound Verbs in Nepali’, Contributions to Nepalese Studies 18(2): 149À73. Reprinted in Y. P. Yadava and W. W. Glovers (eds) (1999), Topics in Nepalese Linguistics. Kathmandu: Royal Nepal Academy, 185–208. Polinsky, Maria and Eric Potsdam (2002). ‘Backward Control’, Linguistic Inquiry 33: 245–82. Pořízka, Vincenc (1967–69). ‘On the Perfective Verbal Aspect in Hindi’, Archiv orientální. 35: 64–88, 208–31; 36: 233–51; 37: 19–47, 345–64. Post, Mark W. (2007). ‘Grammaticalization and Compounding in Thai and Chinese: A Text-Frequency Approach’, Studies in Language 31(1): 117–75. Poudel, Tikaram (2006). Tense, Aspect and Modality in Nepali and Manipuri. Ph.D. dissertation, Tribhuvan University, Kathmandu. Pray, Bruce (1970). Topics in Hindi-Urdu Grammar. Berkeley, CA: Center for South and Southeast Asia Studies, University of California. Premchand (1936). godan (Gift of a Cow). Allahabad: Saraswati Press, http:// sahityachintan.com/books-detail/317/godan.html, accessed August 18, 2020. Pustejovsky, James (1995). The Generative Lexicon. Cambridge, MA: MIT Press. Radford, Andrew (2009). Analyzing English Sentences: A Minimalist Approach. Cambridge: Cambridge University Press. Raina, Achla M. (2011). ‘The Co-Eventual Verb in Hindi’, in Omkar N. Koul (ed.), IndoAryan Linguistics. Mysore: Central Institute of Indian Languages, 135À52. Ramchand, Gillian (2008). Verb Meaning and the Lexicon: A First-Phase Syntax. Cambridge and New York: Cambridge University Press. Ramstedt, John Gustav (1939). A Korean Grammar. Helsinki: Suomalais-ugrilainen Seura. Rappaport Hovav, Malka and Beth Levin (1998). ‘Building Verb Meanings’, in M. Butt and W. Geuder (eds), The Projection of Arguments: Lexical and Compositional Factors. Stanford, CA: CSLI Publications, 97–134. Rassudova, O. P. (1984). Aspectual Usage in Modern Russian. Moscow: Russkij Jazyk. Rastorgueva, V. S. and Aza Alimovna Kerimova (1964). Sistema Tadzhikskogo Glagola. Moscow: Nauka. Read, A. F. C. (1934). Balti Grammar. London: Royal Asiatic Society. Rentzsch, Julian (2008). ‘Notes on Uyghur Verb Morphology’, Turkic Languages 12: 161–9. Rizzi, Luigi (1990). Relativized Minimality. Cambridge, MA: MIT Press. Roberts, Ian and Anna Roussou (2003). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Ross, Claudia (1990). ‘Resultative Verb Compounds’, Journal of the Chinese Language Teachers Association 25: 61–83. Ross, Malcolm (2004). ‘The Grammaticization of Directional Verbs in Oceanic Languages’, in I. Bril and F. Ozanne-Rivierre (eds), Complex Predicates in Oceanic Languages: Studies in the Dynamics of Binding and Boundness. Berlin: De Gruyter Mouton, 297–330.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



565

Rüstəmov, Rüfət (2007). Türk Dilinin Morfologiyası [Morphology of the Turkish Language]. Baku: Nurlan nəşriyyatı. Saidov, Magomedsajid (1967). Avarsko-russkij Slovar’ [Avar-Russian Dictionary]. Moscow: Sovetskaja Éntsiklopedija. Saito, Mamoru (1982). ‘Case Marking in Japanese: A Preliminary Study’, MS., MIT. Sakuma, Kanae (1936). Gendai-nihongo no Hyōgen to Gohō [The Expressions and Usage of Modern Japanese]. Tokyo: Kōseikaku. Samedov, Dzhalil Samedovich (1996). Slozhnoe Predlozhenie v Avarskom Jazyke v Sopostavlenii s Russkim [Complex Sentences in Avar in Comparison with Russian]. Doctoral dissertation. Dagestanskij Gosudarstvennyj Universitet. Makhachkala. Sasaguri, Junko (1999). ‘Meishiku no modariti toshite no koto’ [‘The Formal Noun koto as a Modal Element of Noun Phrases’], in Y. Alam Sasaki (ed.), Gengogaku to Nihongo Kyōiku [Linguistics and Japanese Language Education]. Tokyo: Kurosio Publishers, 161–76. Sawada, Jun (2014). ‘Nihongo no juyo-dōshi-kōbun no kōbun-patān no ruikeika: Ta-gengo to no hikaku taishō to awasete’ [Constructional Patterns of Japanese Benefactive Constructions: A Cross-Linguistic Perspective], Gengo Kenkyu 145: 27–60. Ščerbak, Aleksandr Mixaylovič (1981). Očerki po sravnitel’noy morfologii tyurkskix yazïkov (Glagol). Leningrad: Nauka. Schiffman, Harold (1969). A Transformational Grammar of Tamil Aspectual System. Ph.D. dissertation, University of Chicago. (Published as Volume 8 of Studies in Linguistics and Language Teaching. Seattle, WA: University of Washington.) Schiller, Eric (1989). ‘On the Phrase Structure of Serial Verb Constructions’, in C. Wiltshire, R. Graczyk, and B. Music (eds), Papers from the 25th Annual Regional Meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 405–19. Schmidt, Ruth Laila (2004). ‘Compound Verbs in the Shina of Kohistan’, Acta Orientalia 65: 19À31. Schmidt, Ruth Laila and Razwal Kohistani (2008). A Grammar of the Shina Language of Indus Kohistan. Wiesbaden: Harrassowitz. Sebba, Mark (1987). The Syntax of Serial Verbs: An Investigation into Serialization in Sranan and Other Languages. Amsterdam: John Benjamins. Seki, Kazuo (1977). Kokugo Fukugōdōshi no Kenkyū [A Study of Japanese Compound Verbs]. Tokyo: Kasama Shoin. Sells, Peter (1990). ‘VP in Japanese: Evidence from -te Complements’, in H. Hoji (ed.), Japanese/Korean Linguistics. Stanford: CSLI Publications, 319–33. Sells, Peter (1995). ‘Korean and Japanese Morphology from a Lexical Perspective’, Linguistic Inquiry 26: 277–325. Shakhova, Mar’jam Khilichevna (1998). Kategorija Modal’nosti i Ee Vyrazhenie v Avarskom Literaturnom Jazyke [The Category of Modality and Its Expression in Standard Avar]. Сandidate’s dissertation. Dagestanskij gosudarstvennyj universitet. Makhachkala. Shankara Bhat, D. N. (1967). Descriptive Analysis of Tulu. Pune: Deccan College Postgraduate and Research Institute. Sharma, Tara Nath (1980). The Auxiliary in Nepali. Ph.D. dissertation, University of Wisconsin at Madison. Shauq, Shafi (1996). dag (Pains). In Sheeraza 30. Srinagar, https://www.rekhta.org/ emagazines/sheeraza, accessed July 30, 2020. Shibatani, Masayoshi (1973). ‘Where Morphology and Syntax Clash: A Case in Japanese Aspectual Verbs’, Gengo Kenkyu 64: 65–96. Shibatani, Masayoshi (1978). Nihongo no Bunseki [An Analysis of Japanese]. Tokyo: Taishukan.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

566



Shibatani, Masayoshi (2003). ‘Directional Verbs in Japanese’, in E. Shay and U. Seibert (eds), Motion, Direction and Location in Languages: In Honor of Zygmunt Frajzyngier. Amsterdam: John Benjamins, 259–86. Shibatani, Masayoshi (2007a). ‘Grammaticalization of Motion Verbs’, in B. Frellesvig, M. Shibatani, and J. C. Smith (eds), Current Issues in the History and Structure of Japanese. Tokyo: Kurosio Publishers, 107–33. Shibatani, Masayoshi (2007b). ‘Grammaticalization of Converb Constructions: The Case of Japanese -te Conjunctive Constructions’, in J. Rehbein and C. Hohenstein (eds), Connectivity in Grammar and Discourse. Amsterdam: John Benjamins, 21–49. Shibatani, Masayoshi (2009). ‘On the Form of Complex Predicates: Toward Demystifying Serial Verbs’, in J. Helmbrecht, et al. (eds), Form and Function in Language Research. Berlin: De Gruyter Mouton, 309–36. Shibatani, Masayoshi and Sung Yeo Chung (2007). ‘On the Grammaticalization of Motion Verbs: A Japanese-Korean Comparative Perspective’, in N. McGloin and J. Mori (eds), Japanese/Korean Linguistics 15. Stanford, CA: CSLI Publications, 21–40. Shibatani, Masayoshi and Taro Kageyama (1988). ‘Word Formation in a Modular Theory of Grammar: Post-Syntactic Compounds in Japanese’, Language 64: 451–84. Shimoji, Michinori (2009). ‘Foot and Rhythmic Structure of Irabu Ryukyuan’, Gengo Kenkyu 135: 83–115. Shimoji, Michinori (2010). ‘Ryukyuan Languages: An Introduction’, in M. Shimoji and T. Pellard (eds), An Introduction to Ryukyuan Languages. Tokyo: Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies, 1–13. Shimoji, Michinori (2017). A Grammar of Irabu, a Southern Ryukyuan Language. Fukuoka: Kyushu University Press. Shluinsky, Andrey B. (2006a). ‘Akcional’nost’ i aspektual’nyj pokazatel’: konstrukcija so vspomogatel’nym glagolom il- v čuvašskom jazyke’ [Actionality and Aspectual Marker: Construction with the il- Auxiliary in Chuvash]. Vestnik MGU, serija 9 Filologija 1: 32–47. Shluinsky, Andrey B. (2006b). ‘Akcional’nye klassy glagolov v xakasskom jazyke’ [Actional Classes of Khakas Verbs], in Arsenij P. Vydrin et al. (eds), Tret’ja Konferencija po tipologii grammatike dlja molodyx issledovatelej [The Third Conference on Typology and Grammar for Young Researchers]. St. Petersburg: Nestor-Istorija, 158–64. Shluinsky, Andrey B. (2009a). ‘Biverbal’nye konstrukcii i ix leksičeskie ograničenija’ [TwoVerb Constructions and Their Lexical Restrictions], in S. G. Tatevosov (ed.), Tubalarskie ètjudy [Essays on Tubalar]. Moscow: Imli RAN, 6–53. Shluinsky, Andrey B. (2009b). ‘Individual-Level Meanings in the Semantic Domain of Pluractionality’, in P. Epps and A. Arkhipov (eds), New Challenges in Typology: Transcending the Borders and Refining the Distinctions. Berlin and New York: De Gruyter Mouton, 175–97. Singh, Bhagat (2005). . New Delhi: Hemkunt Press. Singh, Udaya Narayana (1998). ‘Compound Verbs in Selected South Asian Languages’, in L. V. Khokhlova and A. Daswani (eds), Vaagbhaaratii: Proceedings of the International Conference on South Asian Languages (July 1-4. 1997). Moscow: Institute of Asian and African Studies, Moscow State University, 152–82. Singh, Udaya Narayana, Karumuri Venkata Subbarao, and S. K. Bandyopadhyay (1986). ‘Classification of Polar Verbs in Selected South Asian Languages’, in B. Krishnamurti (ed.), South Asian Languages: Structure, Convergence and Diglossia. Delhi: Motilal Banarsidass, 244À69. Sjoberg, Andrée F. (1963). Uzbek Structural Grammar. Bloomington, IN: Indiana University and The Hague: Mouton.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



567

Skalbzaŋ Hgyurmed (1981). [The Very Clear Mirror: A Á Guide to the Understanding of Tibetan Grammar]. [Chengdu: Sichuan Nationalities Publishing House]. Skalbzaŋ Hgyurmed (1992). Le Clair Miroir. Enseignement Á de la grammaire tibétaine. Translated, adapted, and with a commentary by H. Stoddard and N. Tournadre. Arvillard: Éditions Prajñā, Collection Tibétaine. Slade, Benjamin (2008 [2010]). ‘How (Exactly) to Slay a Dragon in Indo-European? PIE *bheid- {h3égwhim, kwr: m ́ i-}’, Historische Sprachforschung 121: 3À53. Slade, Benjamin (2013). ‘The Diachrony of Light and Auxiliary Verbs in Indo-Aryan’, Diachronica 30(4): 531À578. Corrected draft online at: https://ling.auf.net/lingbuzz/ 002198/. Slade, Benjamin (2016). ‘Compound Verbs in Indo-Aryan’, in H. H. Hock and E. Bashir (eds), The Languages and Linguistics of South Asia: A Comprehensive Guide (The World of Linguistics 7). Berlin: De Gruyter Mouton, 559À67. Slepcov, P. A. (ed.) (1972). Slovar’ yakutskogo yazïka. Moscow: Izdatel’stvo sovetskaya Enciklopediya. Slepcov P. A. et al. (eds) (2010). Tolkovïj slovar’ yakutskogo yazïka. Vol. 9. Novosibirsk: Nauka. Slobin, Dan I. and Ayhan Aksu-koҫ (1982). ‘Tense, Aspect, and Modality in the Use of the Turkish Evidential’, in P. J. Hopper (ed.), Tense-Aspect: Between Semantics and Pragmatics, Vol. 1. Amsterdam: John Benjamins, 185–200. Smith, Carlota (1997). The Parameter of Aspect. Dordrecht: Kluwer Academic Publishers. Soga, Matsuo (1983). Tense and Aspect in Modern Colloquial Japanese. Vancouver: University of British Columbia Press. Soh, Hooi Ling and Jenny Yi-Chun Kuo (2005). ‘Perfective Aspect and Accomplishment Situations in Mandarin Chinese’, in H. J. Verkuyl, H. de Sward, and A. van Hout (eds), Perspectives on Aspect. Dordrecht: Springer, 199–216. Sohn, Ho-min (1976). ‘Semantics of Compound Verbs in Korean’, Enŏ (Language) 1(1): 142–50. Solnit, David (1997). Eastern Kayah Li: Grammar, Texts, Glossary. Honolulu: University of Hawai‘i Press. Soper, John (1996). Loan Syntax in Turkic and Iranian. Bloomington, IN: Eurolingua. Sridhar, S. N. (1990). Kannada. London: Routledge. Stankiewicz, Edward (1962). ‘The Interdependence of Paradigmatic and Derivational Patterns’, Word 18: 1–23. Steever, Sanford B. (1981). Selected Papers on Tamil and Dravidian Linguistics. Madurai: Muttu Patippakam. Steever, Sanford B. (1988). The Serial Verb Formation in the Dravidian languages. Delhi: Motilal Banarsidass. Steever, Sanford B. (1993). Analysis to Synthesis; The Development of Complex Verb Morphology in the Dravidian Languages. New York and London: Oxford University Press. Steever, Sanford B. (1998). The Dravidian Languages. London: Routledge. Steever, Sanford B. (2005). The Tamil Auxiliary Verb System. London: Routledge. Stewart, John M. (1963). ‘Some Restrictions on the Object in Twi’, Journal of African Languages 2: 145–9. Stewart, Osamuyimen T. (2001). The Serial Verb Construction Parameter. New York: Garland. Stilo, L. Donald (2004). ‘Iranian as Buffer Zone between Turkic and Semitic’, in É. Á. Csató, B. Isakssonand, and E. Jahani (eds), Linguistic Convergence and Areal Diffusion: Case

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

568



Studies from Iranian, Semitic and Turkic. London and New York: Routledge Curzon, 35–63. Stoynova, Natalya M. (2007). ‘Konstruktsija vzjat’ i sdelat’ v russkom jazyke i jejë tipologicheskie paralleli’ [The “TAKE and V” Construction in Russian and Its Typological Parallels]. Moscow State University (term paper), https://studiorum-ruscorpora.ru/wpcontent/themes/bootstrap-basic4-child/texts/stoynova_construction_paper.pdf, accessed July 1, 2020. Strauss, Susan (2003). ‘Completive Aspect, Emotion, and the Dynamic Eventive: The Case of Korean V-a/e pelita, Japanese V-te shimau, and Spanish se’, Linguistics 41(1): 653–79. Subbarao, Karumuri V. (1979). ‘Secondary Verbs in Telugu’, International Journal of Dravidian Linguistics 8(2): 268À76. Sugar, Alexander (2017). ‘Uygur -ip as a Verb Linker in Multiple Constructions’, Proceedings of the Linguistic Society of America 2, 16: 1–14. Sugar, Alexander (2018). ‘Monoclausal Double Negation in Uyghur’. Oral Presentation at the 19th International Conference on Turkish Linguistics held at Nazarbayev University, Kazakhstan (August 17–19 2018). Sugioka, Yoko (1986). Interaction of Derivational Morphology and Syntax in Japanese and English. New York: Garland. Sweetser, Eve E. (1988). ‘Grammaticalization and Semantic Bleaching’, in S. Axmaker, A. Jaisser, and H. Singmaster (eds), Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, 389–405. Takano, Yuji (2010). ‘Scrambling and Control’, Linguistic Inquiry 41: 83–110. Takeuçi, Kazuo (1990). Uigurugo Yonshūkan [New Uyghur in 4 Weeks]. Tokyo: Daigakushorin. Takubo, Yukinori (2007). ‘An Overt Marker for Individual Sublimation in Japanese’, in B. Frellesvig, M. Shibatani, and J. C. Smith (eds), Current Issues in the History and Structure of Japanese. Tokyo: Kurosio Publishers, 135–51. Talmy, Leonard (1991). ‘Path to Realization: A Typology of Event Conflation’, in L. A. Sutton (ed.), Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on The Grammar of Event Structure. Berkeley, CA: Berkeley Linguistics Society, 480–519. Tarkhadkar, D. P. (1899 [1836]). mahārāštrbhā seʦe vyākran: [A Grammar of the : : Maharashtra Language]. Eleventh edition. Bombay: Education Society. Tatevosov, Sergei (2002). ‘The Parameter of Actionality’, Linguistic Typology 6(3): 317–401. Tekin, Talat (1968). A Grammar of Orkhon Turkic. Indiana University Uralic and Altaic Series, C. 69. Bloomington, IN: Indiana University and The Hague: De Gruyter. Tekin, Talat (2010 [2008]). Orhon Yazıtları [Orkhon Texts]. Ankara: Türk Dil Kurumu. Teng, Shou-hsin (1977). ‘A Grammar of Verb Particles in Chinese’, Journal of Chinese Linguistics 5: 1–25. Teramura, Hideo (1984). Nihongo no Shintakkusu to Imi, Dai-ni-kan [The Syntax and Semantics of Japanese 2]. Tokyo: Kurosio Publishers. Terrill, Angela (2003). A Grammar of Lavukaleve. Berlin: De Gruyter Mouton. Thepkanjana, Kingkarn (1986). Serial Verb Constructions in Thai. Ph.D. dissertation, University of Michigan. Thepkanjana, Kingkarn and Satoshi Uehara (2004a). ‘Grammaticalization of Directional Verbs into Success Markers in Thai’, in L. A. Sutton (ed.), Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on The Grammar of Event Structure. Berkeley, CA: Berkeley Linguistics Society, 649–64.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.



569

Thepkanjana, Kingkarn and Satoshi Uehara (2004b). ‘Semantic Types of Resultative Predicate in Transitive-Based Resultative Constructions in Thai’, in S. Burusphat (ed.), SEALS 11: 731–47. Tempe: Program for SE Asian Studies, Arizona State University. Thepkanjana, Kingkarn and Satoshi Uehara (2009). ‘Resultative Constructions with “Implied-Result” and “Entailed-Result” Verbs in Thai and English: A Contrastive Study’, Linguistics 47(3): 589–618. Tikkanen, Bertil (1987). The Sanskrit Gerund: A Synchronic, Diachronic and Typological Analysis. Helsinki: Finnish Oriental Society. Tournadre, Nicolas and Konchok Jiatso (2001). ‘Final Auxiliary Verbs in Literary Tibetan and in the Dialects’, Linguistics of the Tibeto-Burman Area 24(1): 49–110. Traugott, Elizabeth Closs (2001). ‘Constructions in Grammaticalization’, in B. B. Joseph and R. D. Janda (eds), The Handbook of Historical Linguistics. Oxford: Blackwell, 624–47. Traugott, Elizabeth Closs and Graeme Trousdale (2013). Constructionalization and Constructional Changes. Oxford: Oxford University Press. Trousdale, Graeme and Muriel Norde (2013). ‘Degrammaticalization and Constructionalization: Two Case Studies’. Language Sciences 36: 32–46. Tsujimura, Natsuko (2016). ‘The Role of Lexical Semantics in the Reorganization of the Resultative Construction’, in T. Kageyama and W. M. Jacobsen (eds), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: De Gruyter Mouton, 155–80. Tulpule S. G. and Anne Feldhaus (1999). A Dictionary of Old Marathi. Mumbai: Popular Prakashan. Turner, Ralph Lilley (1931). A Comparative and Etymological Dictionary of the Nepali Language. London: Routledge and Kegan Paul. Reprinted 1980, New Delhi: Allied Publishers. Ubrjatova, E. I. (1972). ‘Kratkiy grammatičeskiy očerk yakutskogo yazïka’, in P. A. Slepcov (ed.), Yakutsko-russkiy slovar’. Moscow: Izdatel’stvo sovetskaya enciklopediya, 569–605. Ueno, Kazuaki (2011). Heikyoku-Fuhon ni yoru Kinsei Kyōto Akusento no Shiteki Kenkyū. [A Historical Study of Pitch Accent in Early-Modern Kyoto Japanese Based on Scores to Heikyoku]. Tokyo: Waseda Daigaku Shuppanbu. Vale, Ramchandra Narayan (1948). Verbal Composition in Indo-Aryan. Pune: Deccan College Postgraduate and Research Institute. Van Valin, Robert (1990). ‘Semantic Parameters of Split Intransitivity’, Language 66: 221–60. Vendler, Zeno (1957). ‘Verbs and Times’, The Philosophical Review 66(2): 143–60. Vendler, Zeno (1967). Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Verma, Manindra K. (ed.) (1993). Complex Predicates in South Asian Languages. New Delhi: Manohar. Verma, Manindra K. and T. N. Sharma (1979). Intermediate Nepali Reader. New Delhi: Manohar. Vovin, Alexander (2009). A Descriptive and Comparative Grammar of Western Old Japanese. Part 2: Adjectives, Verbs, Adverbs, Conjunctions, Particles, Postpositions. Folkestone: Global Oriental. Wallace, William D. (1982). ‘The Evolution of Ergative Syntax in Nepali’, Studies in the Linguistic Sciences 12(2): 147À211. Watkins, Calvert (1995). How to Kill a Dragon: Aspects of Indo-European Poetics. New York: Oxford University Press. Watters, James K. (2017). ‘Verb-Verb Compounds and Argument Structure in Tepehua’, in B. Nolan and E. Diedrichsen (eds), Argument Realisation in Complex Predicates and

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

570



Complex Events: Verb-Verb Constructions at the Syntax-Semantic Interface. Amsterdam: John Benjamins, 277–303. Weiss, Daniel (2012). ‘Verb Serialization in North East Europe: The Case of Russian and Its Finno-Ugric Neighbours’, in B. Wiemer, B. Wälchli, and B. Hansen (eds), Grammatical Replication and Borrowability in Language Contact. Berlin and Boston, MA: De Gruyter Mouton, 611–46. Whitman, John (2012). ‘The Relationship between Japanese and Korean’, in N. Tranter (ed.), The Languages of Japan and Korea. London: Routledge, 24–38. Whitney, William Dwight (1879/1889). Sanskrit Grammar, Including Both the Classical Language and the Older Dialects of Veda and Brahmana. First/second edition. Leipzig: Breitkopf and Härtel. Reprinted 1995, Delhi: DK Publications. Wiklund, Anna-Lena (2009). ‘The Syntax of Surprise: Unexpected Event Readings in Complex Predication’, Working Papers in Scandinavian Syntax 84: 181–224. Wilson, Stephen (1999). Coverbs and Complex Predicates in Wagiman. Stanford, CA: CSLI Publications. Xaritonov, L. N. (1960). Formï glagol’nogo vida v yakutskom yazïke. Moscow and Leningrad: Nauka. Yamada, Hisanari (2010). ‘Abārugo niokeru dōshi no ketsugōka’ [Verbal Valence in Avar], Ajia Afurika no Gengo to Gengogaku 4: 85–109. Yamada, Hisanari (2013). ‘Reciprocal Constructions in Avar’, Lingua 126: 150–71. Yamada, Toshihiro (2004). Nihongo no Benefakutibu: -te yaru, -te kureru, -te morawu no Bunpō [Benefactives in Japanese: A Grammar of -te yaru, -te kureru, and -te morawu]. Tokyo: Meiji Shoin. Yanaike, Makoto (2004). ‘Heian-jidai Kyōto hōgen no akusento katsuyō’ [Accent on Predicates in the Kyoto Dialect in the Heian Era], Onsei kenkyū 8(2): 46–57. Yi, Eun-Young (1997). De- and Re-Constructing Coordination in Korean. Ph.D. dissertation, Cornell University. Yoon, James (1997). ‘Coordination (A)symmetries’, in S. Kuno et al. (eds), Harvard Studies in Korean Linguistics 7. Seoul: Hanshin, 3–30. Yoshida, Taeko (2012). Nihongo Dōshi te-kei no Asupekuto [Aspect of –te Verbs in Japanese]. Kyoto: Koyo Shobo. Yoshizawa, Norio (1952). ‘Fukugōdōshi ni tsuite’ [On Compound Verbs], Nihon Bungaku Kenkyū 10: 32–42. Yudaxin, Konstantin Kuz’mič (1965). Kirgizsko-Russkiy slovar’. Moscow: Izdatel’stvo Sovetskaya enciklopediya. Yuldašev, Axnef A. (1965). Analitičeskie formï glagola v tyurkskix yazïkax. Moscow: Nauka. Yumoto, Yoko (1996). ‘Gokeisei to goigainen-kōzō’ [Word Formation and Lexical Conceptual Structure], in Okuda Hiroyuki-kyōju Taikankinen-ronbunshū-kankōkai (ed.), Gengo to Bunka no Shosō [Aspects of Language and Culture]. Tokyo: Eihosha, 105–18. Yumoto, Yoko (2005). Fukugōdōshi Haseidōshi no Imi to Tōgo [Semantics and Syntax of Compound Verbs and Derived Verbs]. Tokyo: Hituzi Syobo. Yumoto, Yoko (2009). ‘Modularity of Word Formation: Differences between Two Types of Japanese Compound Verbs’, in H. Hoshi (ed.), The Dynamics of the Language Faculty. Tokyo: Kurosio Publishers, 203–30. Zbavitel, Dušan (1970). Non-Finite Verbal Forms in Bengali. Prague: Czechoslovak Academy of Science. Zeisler, Bettina (2004). Relative Tense and Aspectual Values in Tibetan Languages. A Comparative Study. Berlin and New York: De Gruyter Mouton.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi



571

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Zeisler, Bettina (2011). ‘Kenhat, the Dialects of Upper Ladakh and Zanskar’, in M. Turin and B. Zeisler (eds), Himalayan Languages and Linguistics. Studies in Phonology, Semantics, Morphology and Syntax. Leiden: Brill, 235–301. Zeisler, Bettina (2014). ‘Modal Verbs and Modal Constructions in Ladakhi’, Journal of South Asian Languages and Linguistics 1(1): 31–57. Zubizaretta, Maria Luisa (1987). Levels of Representation in the Lexicon and in the Syntax. Dordrecht: Foris. Zubizarreta, Maria Luisa and Eunjeong Oh (2007). On the Syntactic Composition of Manner and Motion. Cambridge, MA: MIT Press.

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Index of languages Ainu 1 Alu Kurumba 337; see also Kurumba Amdo 359, 522 Anatri Chuvash 1, 397, 401, 410–13 Austro-Asiatic 1, 522 Avar 1, 471–95 Azerbaijanian 462, 466, 468 Badaga 327 Balti 357–9, 365, 368, 370, 374, 522 Bangla 242–4 Bashkir 1, 398–401 Bengali 8, 270, 272 Betta Kurumba 271; see also Kurumba Bhitrauti 242 Brahui 327, 329, 339 Burmese 354, 538 Burushaski 328, 522

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Chinese 1, 10, 33, 164, 354, 499–502, 506, 521–44 Chuvash 1, 397–404, 410–14 Dhangar 327; see also Kurux Domkhar (dialect) 359, 369, 371, 374, 379, 387 Dravidian languages 1, 250, 259, 270–4, 300, 327–31, 334, 338, 340–1, 349–53, 522 Dutch 99–100 English 1, 4, 10, 24, 31, 101, 275, 302, 349, 354, 365, 501, 527 Gondi 327, 339 Gujarati 1, 242, 245, 281–2 Hindi 1, 4, 6, 19, 26, 41, 242, 244, 249–55, 261–70, 272–3, 275–98, 329, 357, 376, 501–2, 506, 521, 524–36, 543, 522 Hindi-Urdu, see Hindi Indo-Aryan languages 1, 9, 223, 230, 237, 241–2, 249–51, 271–3, 279, 282, 319, 354, 357, 502 Early Modern Indo-Aryan 257 Middle Indo-Aryan 254 Irabu, see Ryukyuan

Iranian 522 Irula 327 Japanese 1, 3, 15–43, 70–102, 103–38, 139–63, 165, 171–3, 179, 182–4, 189–92, 197, 200, 204–8, 211, 213–16, 302, 304, 306, 308, 319, 324, 330–1, 351, 353, 398, 430, 446, 477, 506 Early Middle Japanese 44, 72, 83, 91 Late Middle Japanese 44–5, 78, 86, 88, 95, 97 Middle Japanese 44–5, 72, 77 Modern Japanese 44, 54, 57, 96–7 Old Japanese 42, 44–5, 53–4, 57, 68, 71, 77, 81–2, 91, 220 Present-day Japanese 3, 44, 51, 62–3, 72, 82 Japonic languages 1, 165 Kannada 327, 328, 336, 339, 343, 346, 350 Karachay-Balkar 379, 402–9, 413 Karen 539 Kashkay 455, 469–70 Kashmiri 1, 19, 242, 279, 282, 287 Kazakh 1, 430, 439, 441, 443, 446–7, 455, 466, 523 Kenhat 355, 361, 372, 374 Khakas 1, 398, 401–2, 430, 447 Khalaj 455, 469–70 Kham 359 Kodagu 327 Kolami 327, 339, 349 Konda 327, 334, 335 Konekor Gadaba 327, 336, 346, 347 Korean 1, 3, 4, 9, 26, 57, 193–220, 523, 540 Kota 327 Kui 327, 328, 336, 344 Kurumba 327, 335 Kurux 327, 328 Kuvi 327, 337, 343, 344 Kyrgyz 430–1, 439–44, 446–8 Ladakhi 1, 354–9, 361–2, 365–6, 368–72, 376, 381, 389 Lahu 522, 539 Leh 361, 374, 376 Limbu 1, 522 Malayalam 1, 26, 158, 271, 327, 337 Malto 1, 327, 328, 329, 336, 342, 343, 348, 350

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

574

  

Manda 327, 344 Mandarin, see Chinese Marathi 1, 3–5, 20, 223–5, 228–30, 234–46, 250, 267–8, 272, 248, 279–82, 285–7 Marwari 242–4 Mishar Tatar 397, 399, 401 Miyako, see Ryukyuan Munda 1, 250, 522 Naiki 327 Nepali 1, 250, 252–3, 257–68, 270, 272–3, 278–82, 287 Oghuz Turkic 469 Ollari 327; see also Konekor Gadaba Oriya 270, 272, 329 Orkhon Turkic 466–8 Ottoman Turkish 1 Pali 254, 256, 270 Panjabi 244, 282 Parji 327, 329, 334, 338 Pengo 1, 327, 328, 333, 334, 335, 342, 344, 346, 351

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Russian 12, 290, 293, 383, 386, 470, 529–30 Ryukyuan 1, 165, 173, 179 Irabu (dialect) 165–92 Miyako 165, 173 Sagay Khakas 398, 401–2 Sakha 1, 430–9, 444, 455 Sanskrit 4, 224, 237, 241, 249, 251–4, 270, 277, 521 Shamskat 361, 374, 376

Shina 270, 272, 357, 369–71, 374, 376, 522 Sinhala 1, 251, 254–7, 266, 272–4 Spanish 26 Spiti 372 Swedish 12 Tamil 1, 26, 271, 300–6, 309, 317–18, 322, 324–7, 330–3, 335–7, 341–2, 344–5, 347, 349, 527 Old Tamil 274, 340, 343, 346 Tatar 1, 397, 399, 401 Telugu 1, 271, 327, 336, 346, 348–9 Thai 1, 10, 354, 499–519, 538 Tibetan 1, 354–5, 357–61, 366, 372 Central Tibetan 359 Middle/Classical Tibetan 355, 357, 359 Old Tibetan 355, 366 Standard Spoken Tibetan 357–8 Written Tibetan 358 Tibetic languages 1, 10, 355–9, 365, 372, 388 Tibeto-Burman 1, 250, 382, 522 Toda 327, 328, 339 Tofa 438 Tsez 1, 494 Tubalar Altai 1, 397, 400–1, 404, 413–29 Turkic languages 1, 397–400, 413, 423, 428 Turkish 1, 10, 430, 455–6, 458, 460–8 Turkmen 430, 455, 466–7 Tuvan 1, 430–9, 441, 443, 455 Tyvan, see Tuvan Uyghur 1, 430–1, 439, 443–6, 455–68 Uzbek 1, 164, 401, 430, 439, 441, 443–6 Yakut, see Sakha

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

Index of subjects abductive innovation 55 absolutive 10, 249, 256, 261, 263, 361, 376, 380, 387, 478–81, 490–4, 540 accentuation 61–3, 196, 198, 202, 378 accomplishment (verb) 101, 181, 208, 213, 217–18, 309, 502, 526–9 achievement (verb) 155, 181, 200, 213–14, 235, 458, 529 actional class 403, 406, 409, 413–14, 417, 428 actionality 403, 405, 409, 412 activity (verb) 100, 155–6, 181, 208, 213–14, 302, 332, 338, 343–9, 462, 487, 528–9 adjacency 113, 145, 147, 208, 475, 492 adjectival 30, 57, 340, 354, 523 adjunct 74, 110, 151, 288, 378, 402 adverbialization 54 agentivity 428 agglutination 40, 42, 71 agreement 209, 225, 236, 268, 333, 369, 481, 492–4, 536, 540 aktionsart 36, 40–2, 45, 49–58, 62, 66–8, 74, 77, 83, 90, 100, 205, 213, 216, 219, 400 aktionsart verb 40, 42, 49–59, 67–8, 74, 77 alternation 233, 276, 278, 350, 355, 390, 522, 530 A-movement 115–18 anticausative 102 anticipation 8, 21, 297, 307 anti-perfect tense 338 areal feature 41, 250, 499 areal typology 2, 499 argument structure 30, 34, 49, 53–4, 71–3, 90–1, 100, 276, 318, 320, 431, 475 atelic 25, 413–15, 424, 428 attitudinal 8, 19–22, 38, 40–2, 181–3, 306–7, 333, 334, 338, 478 autogeny 291 auxiliary verb 19–21, 25, 39–42, 45–6, 66–8, 77, 81–4, 91, 132–4, 170, 181, 186, 224, 230, 233–5, 255, 257, 303, 332, 335, 437, 439, 445, 455–7, 459–63, 466–8 auxiliary verb construction 45–6, 66–8, 165, 170, 181, 430, 457

benefactive 19–22, 38, 40–2, 142–3, 145–7, 183–4, 199, 200, 231, 332, 338, 400, 417–19, 431, 436, 457–60, 503 self-benefactive 410, 414–16, 436 bi-absolutive construction 478–81 cancellation of entailment/results 101, 290 capacitive 414, 423–4 case absolutive case 380, 471, 478–81, 490 accusative case 104, 114, 116, 123–6, 200, 209, 465 dative case 122–3, 128–9, 319, 481 ergative case 361, 479–80 genitive case 44, 117, 122, 479, 490 nominative case 44, 119, 122 case-marking 118–23, 263, 268, 359, 386 absolutive-ergative 471, 478 nominative-accusative 118–21, 134, 479 categorial reinterpretation 55–6 causative 28, 62, 79, 92, 131, 170, 175–7, 187–8, 205–6, 212, 219, 226, 231, 234, 305, 320, 322, 344, 355, 390, 475, 503 clausal idiom 107, 135 clausality 7, 46 biclausal 7, 16–17, 21, 29, 45–6, 53–5, 64–7, 80, 143, 148, 150–1, 179, 190, 238, 304, 318, 362, 367, 369–71, 381, 383, 389–90, 460 monoclausal 16–17, 46–7, 49, 54, 64, 70, 146, 150–1, 162, 190, 209, 238, 302, 305–7, 322, 363, 370, 390 clause chaining 179, 191, 356, 359–61, 388–9 collocability 7, 238, 276 complementation 32, 35, 49, 73–4, 94, 103, 114, 208, 330 completive 25, 31, 199, 208, 267, 273, 340, 358, 433–4 complex motion predicate 139, 154, 324 complex predicate 16–22, 38, 45–55, 62–8, 78, 80, 99, 140, 143–8, 180, 253–4, 300, 318, 320, 361–2, 463, 468 complex verb 44–5, 66, 90, 107, 223–5, 228–31, 384, 430, 439, 446, 468, 471, 481, 483, 485, 487

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

576

  

compositionality 198, 356, 389, 392 compound verb 3, 17, 28, 57, 104, 106, 143, 147, 171, 198, 224, 230, 275, 302, 327, 330, 332, 390, 398, 500, 521 auxiliary compound verb 327, 331, 332 balance compound verb 346 echo compound verb 349 explicator compound verb 2, 33, 354, 468, 499, 502, 505 lexical aspectual compound verb 17, 37, 66, 73–6, 96 lexical compound verb 327, 331, 341 reduplicating compound verb 348 resultative compound verb 33, 354, 500 syntactic compound verb 16, 28–31, 37–8, 45–6, 79, 80, 90, 105, 108, 136, 148, 151, 155–61, 170, 173–8, 191, 220, 302 thematic compound verb 16, 17, 34–5, 48, 54, 61, 66, 71–3, 78, 80, 90–6, 193 vector compound verb 501–2 conative 8, 101, 183, 199, 201, 205, 414, 419, 428, 526–7 condemnation 8, 39, 43 conjunction 167, 295, 297, 304, 359, 529, conjunctive 194, 207, 213, 330, 333–4, 337–40, 347, 455, 466 conjunctive participle 266, 279, 354, 361 conjunct verb 229 constative 293, 395 constructionalization 72, 90, 99 contraction 22–4, 335, 344, 351, 443 control verb 21, 108–10, 115–16, 125, 131–2, 156 converb 19, 22, 35, 73, 140, 164, 167–8, 199, 251, 257, 267, 279, 281–2, 297, 323, 358–61, 370–2, 387, 389–90, 398–400, 402, 410, 432–3, 437–40, 444–69 472–4, 477 creativity 101 deductive innovation 55 degrammaticalization 27, 89, 98 deictic motion verb 140, 156–8 delexicalization 42, 130, 133, 274 delexicalized verb 6, 10, 19, 25, 41–2, 132, 397–8, 400–5, 409–10, 413–14, 417, 429 derogatory 6, 43 desemanticization 25, 41, 72, 90, 98–9 desemanticized 3–5, 70, 229, 231 directionality 36, 141, 233, 365, 370, 457 ditransitive verb 115, 128–9, 200, 328 durative 25, 75, 110, 136, 253, 315, 384, 392 embedding 79, 131, 136, 162, 174, 177, 188, 361, 473 epistemic perfect 337

evidential 43, 308, 311–13, 338, 355, 360, 372, 376–7, 472, 504 expectation 26–7, 182, 309–10, 371 explicator verb 2, 9, 27, 33, 41, 223, 246, 322, 457, 467–8, 499–502, 538; see also vector verb factor verb 238, 240, 524, 535 flux (text frequency) 241, 246, 279, 281, 289, 293 focus marking 170, 186–7 functional meaning 8, 15–17, 20, 37, 41, 88, 94 fusion 378–80 general-tense participle 472–4 geographical distribution 9, 430, 522 gerund 251, 282, 477 gerundive 15, 18–19, 25, 38, 132, 140, 179, 189–90, 194, 208, 217, 340, 455–7, 459–64, 466 grammatical aspect 37–8, 75, 85, 90, 512, 514, 519 grammatical change 91, 98, 468 grammaticalization 7–9, 17, 23–5, 38, 41, 43. 45, 55, 67–8, 71–2, 88, 98, 130, 186, 234, 246, 250, 259–61, 275, 328, 363, 390, 402, 414, 457, 504, 507, 512, 526 grammaticalization cline 9, 23, 38, 43, 67, 71, 98, 235, 462 Grice’s principle 310 habitual 31, 105, 199, 225, 228, 409, 462 honorific 16, 28, 62, 77–80, 141, 150, 158–60, 200, 209, 259, 366 imperfective 19, 28, 43, 100–1, 403–4, 406–8, 422, 435 inanimate subject 21, 108, 120, 317 inchoative 155, 211, 357, 401, 417 indicative 310, 333–5, 337 infinitive 15–18, 32, 39, 45–6, 54, 58, 63–6, 91, 143, 169, 172–5, 193, 199, 205, 207, 220, 283, 301–2, 331, 333, 336, 339, 472–5 inflection 2, 16, 18, 19, 23, 60, 70–2, 91, 166–9, 220, 302, 318, 323, 328, 332–5, 341, 345, 350, 445, 462, 473 integration of two verbs 22, 73–4, 163, 438, 441–3, 459 interrogative 170, 178, 305, 459 intransitive verb 30, 102, 104, 186, 303, 317, 362, 374, 387, 479–81, 487–91, 506, 537–8 language contact 10, 357, 465, 468–9 lexical aspect 8, 16–17, 20, 49, 205, 213, 216, 219; see also aktionsart lexical integrity 7, 17, 29, 60, 78, 80, 106, 144, 170, 305, 321, 459

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

   lexicalization 42, 72, 90, 91, 174, 186, 323 lexical verb 6, 8, 9, 16–17, 24–6, 39–43, 49, 54, 86, 88–90, 104, 130–4, 181, 184, 186, 231, 234–6, 276, 302, 306, 315, 317, 322, 385, 398, 402, 405–28, 436, 475–7, 503, 510, 522, 526 light verb 9, 24, 41, 42, 67, 230, 245–6, 251–3, 256–7, 259–66, 269–74, 301–3, 305–25, 364 light verb construction 29, 177, 466 limitative 410–14 linear adjacency 145, 208, 475, 492 linguistic area 2, 9, 224, 328, 502, 538, 541 Indo-Altaic linguistic area 2 Indo-Turanian linguistic area 9, 224, 540

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

minimality condition 115–16 modal auxiliary 2, 333, 336, 514 modification 34–5, 55, 73, 99, 112, 148–9, 369, 381, 390, 468, 503 modifier 100, 136, 232, 317, 322, 329, 354, 500 mood 167, 229, 261, 281, 328, 333, 337, 355, 360, 382–3, 463 morphological integrity of words, see lexical integrity motion verb 22, 132, 140, 156–7, 357, 362, 365–7, 385, 387, 510 negation 40, 78, 146, 201, 203, 210, 283–5, 297, 306, 340, 378, 381–4, 389–90, 444–5, 483, 504 noncompound 198, 203, 208, 277–8, 282, 296, 506, 522, 527–30 nonfinite verb form 6, 16, 140, 164, 167–8, 184, 220, 223, 225, 237, 325, 347, 359, 471, 474 omissibility 501, 523 ontogeny 224, 245 participle 251, 253, 263, 266, 278–9, 297, 301–3, 354, 361, 432, 437–9, 472–4 passive 21, 28–30, 79, 92, 115–18, 123, 175, 211, 227, 234, 273, 302, 333, 441–2, 460, 464–5 double passive 460, 465 long-distance passive 104, 114–18, 121–9 pejorative 9, 39, 42, 339, 349 perfective 19, 23, 25–7, 100, 113, 140, 153–5, 199, 205, 242, 245, 253, 267–8, 283, 287, 330, 340, 403, 407, 472, 477, 508, 516, 530 perfect tense 332, 337 perspectival 144, 161–3 phasal verb 39, 40, 91, 238, 282, 290, 528–9 phonetic erosion/reduction 24, 39, 41–2, 72, 356, 378, 459, 511 polar opposite 319 polite 8, 20, 39–42, 131

577

pragmatic 4, 6, 19, 24–8, 309, 419, 425, 457, 514–16 present perfect 153, 335, 338, 377 PRO 110, 115, 146, 150, 479 productivity 9, 32, 54, 66, 70, 79, 81, 101–4, 173–4, 194, 241, 398, 403–4, 413, 432–3, 457, 463, 478 progressive 19, 25, 153–5, 181, 186, 188, 200, 456–9, 472, 482, 508, 524–5 prosodic phrase 180, 190 prospective tense 336 punctual 25, 339, 405, 409, 413, 417, 425–6 quasi-auxiliary verb 2–5, 467 raising verb 21, 108–9, 113, 125, 131, 134 reanalysis 28–9, 72, 90–1, 241, 251, 256, 506 recipient 20, 319, 320, 368 reciprocal voice 339 reduplication 62, 92–4, 144, 349 reinterpretation 53, 55–7, 66 resultative aspect 20, 25, 101, 472, 482 scope of negation 136, 201, 306, 381, 461, scrambling 117, 331 semantic bleaching 6, 25, 28, 43, 72, 186, 234–5, 277, 314, 393, 500–1, 522 semantic structure 38, 72, 74–5 semantic transparency 315, 319, 513, 517, 519 semelfactive 342, 410, 413 semi-lexical category 17 serial verb 28, 46, 77–80, 90, 170, 179, 184, 209, 229–31, 240–1, 277, 304, 326, 335, 340, 343, 345, 347, 350, 358–60, 369–71, 379, 389–90, 499–500, 506 social deixis 17, 36, 51, 71, 159 speech style 8, 39 stative (verb) 154, 200, 309, 405–9, 411–13, 420–3, 523 stem nucleus 175–7, 187 subject sharing 151–2 subordination 45, 105, 140, 360–2 syntactic concatenation 7, 457 TAM (tense-aspect-modality) 17, 38, 504 telic 25, 101, 201, 208, 405–13, 415–20, 423–4, 528 tone 166, 378–9 transitive verb 32–3, 95, 102, 123, 205, 329, 362, 374, 376, 387, 441, 476, 480, 485, 487–9, 505, 508–10 transitivity 7, 30, 100, 186, 263, 268–70, 476–7, 506 transitivity harmony 32, 35, 66, 151, 186, 211 unaccusative (verb) 32, 33, 86, 200, 211 univerbation 23, 378–9

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.

OUP CORRECTED PROOF – FINAL, 22/1/2021, SPi

578

   volitional 26, 33, 150, 317, 358, 385, 387, 457 volitivity 419 verbs of giving 140, 158–9 wordhood 45, 57, 77, 170, 202, 351, 504 word order 35, 40, 54, 216, 440, 469, 505–6 OV word order 2, 539 VO word order 499, 539

Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.

vector verb 3–10, 19, 37, 143, 224, 233–6, 241–2, 249, 276–7, 377, 381, 390, 467, 500–2, 507, 512, 514, 519, 521, 523, 531–4 verbhood 39, 305, 313, 315, 319 V-te V complex 16–17, 21, 23, 90, 139–58, 160–3, 492 verb stem 58, 140, 167, 173, 263, 355, 370, 432, 447, 456 verb-verb complex 2, 6, 230–2

Verb-Verb Complexes in Asian Languages, edited by Taro Kageyama, et al., Oxford University Press USA - OSO, 2021.