Sanskrit Syntax

Sanskrit syntaxSelected papers presented at the seminar

on Sanskrit syntax and discourse structures,13-15 June 2013, Universit Paris Diderot,

with an updated and revised bibliographyby

HANS HENRICH HOCK

edited byPETER M. SCHARF

28 February 2015

ii

Scharf, Peter M., ed. Sanskrit syntax: Selected papers presentedat the seminar on Sanskrit syntax and discourse structures, 13-15June 2013, Universit Paris Diderot, with an updated and revisedbibliography by Hans Henrich Hock. Providence: The SanskritLibrary, 2015.

Copyright 2015 by The Sanskrit Library.

All rights reserved. Reproduction in any medium is restricted. Nopart of this publication may be reproduced or transmitted, exceptbrief quotations, in any form or by any means, electronic or me-chanical, including photocopying, recording, or any informationstorage or retrieval system, without prior written permission of thecopyright holder and publisher.

Published by:The Sanskrit Library89 Cole AvenueProvidence, RI 02906USAsanskritlibrary.org

Library of Congress Control Number: 2015934847

ISBN-10: 1943135002ISBN-13: 978-1-943135-00-4

1350047819439

ISBN 978-1-943135-00-490000

Preface

A strong tradition of linguistics developed in India in the firstmillenium BCE naturally associated with the heightened aware-ness of language engendered by the assiduous preservation of oralVedic texts. The curiousity as well as the need to understand thelanguage of these compositions already several hundred years oldinstigated the development of systematic linguistic analysis whichflourished throughout the two and a half millennia since Pan. inicomposed his comprehensive linguistic description of the Sanskritlanguage by the fourth century BCE. His unprecedented analysisof the language into basic units and reconstitution of utterancesin accordance with precise rules laid the foundation for the de-velopment of highly sophisticated discussions concerned with thestructure of verbal cognition and its relation to speech units rang-ing from roots and affixes to words, phrases, and sentences. Thestudy of syntax in India is intimately associated with semantics,and the analysis of the semantic content provides the foundationfor the generation of linguistic expressions. Relational structuresin the domain of consciousness are projected onto speech formswhose arrangement in the string of speech is subordinate.

Modern linguistics developed in Europe beginning in the lateeighteenth century as a direct result of the fascination Europeanscholars had with the resemblances of Sanskrit to classical Greekand Latin. Their excitement to discover the relationship amongthese and a gradually expanding number of languages in whatcame to be known as the Indo-European family led them to readaptthe precise phonetic rules of Indian linguists to historical soundchange in the science of historical and comparative linguistics.The great figures of the history of modern linguistics Jones,Bopp, de Saussure, Bloomfield, and many others all studied

iii

iv PREFACE

Sanskrit, and some worked directly on Sanskrit grammatical textsor with learned Sanskrit scholars (pan. d. ita) intimately familiar withsuch texts. Even Chomskys generative analysis of English was in-spired by Pan. inian methods.

Formal and computational linguistics, engendered in the U.S.,was dominated by English at its inception and developed in sub-sequent decades primarily in the environment of European lan-guages. More recently there has been a concerted effort to under-take formal linguistic analysis of a wide variety of languages, withparticular interest in those with dramatically different features, andto enrich syntactic theory to account for linguistic variety. Over thepast couple of decades there has been a growing interest among adiverse group including Sanskrit scholars and computer scientistswho collaborated together to form the Sanskrit Computational Lin-guistics Consortium. The Consortium has sponsored several sym-posia in Europe, the U.S. and India since 2007.

Against this background, I was inspired to undertake researchto develop universally adequate linguistic theory by formalizingthe sophisticated linguistic theories, structures and procedures de-veloped in the Indian linguistic tradition. To do so, for the year20122013, I was awarded a Chaire Internationale de RechercheBlaise Pascal, financed by the State of France and the RgiondIle-de-France and managed by the Fondation de lcole Nor-male Suprieure. The project investigated ways in which Indianlinguistics could contribute useful insights to contemporary for-mal linguistics, and designed ways in which Indian linguistic the-ories could be formalized and implemented computationally. Theproject focused on Indian semantic and syntactic theory and thesemantics-syntax interface where computational linguistic work isflourishing. In the midst of the project I was invited as a Visit-ing Professor to the Indian Institute of Technology Bombay. Theinvitation facilitated collaboration sponsored by the Blaise PascalChair with several junior Indian scholars. The project culminated

SANSKRIT SYNTAX v

in the organization of the Seminar on Sanskrit syntax and discoursestructures held at the Universit Paris Diderot, 1315 June 2013the program of which is posted on the Sanskrit Library Website(www.sanskritlibrary.org) under Events. The theme ofthe seminar was the exposition of current theoretical knowledgeconcerning Sanskrit syntax and the application of state of the artmethods of computational linguistics to Sanskrit. This book is thepublication of selected revised papers presented at that seminar.

The book includes twelve papers by seventeen contributorsfrom divergent backgrounds in European and American linguis-tics, Pan. inian grammar, and computer science that converge indealing with contemporary issues in Sanskrit syntax. Hans Hen-rich Hock provides a general survey of research on Sanskrit syn-tax since the publication of the collection on Sanskrit syntax heedited in 1991 (Hock 1991). He brings to our attention challeng-ing formal issues concerning word order, oblique subject agree-ment, clause coordination, and ditransitivity as well as contrastingdiscourse structures in different genres. Along with his presenta-tion, he provided a bibliography of research done in the twenty-five years since the publication of his and Madhav Deshpandescombined bibliographies on Sanskrit syntax (Deshpande and Hock1991). With the gracious consent of the authors and Motilal Banar-sidass I have revised and integrated their earlier bibliography withHocks update in a seventy-two page comprehensive bibliographyof Sanskrit syntax containing nearly a thousand entries. Supple-mented by the author and title indices at the back of this volume,Hocks survey of research should provide useful and convenientaccess to research on Sanskrit syntax generally.

George Cardona contributes two papers to the volume. In thefirst, he deals with how expectancy, ellipsis and suppletion presentin ordinary language use of Sanskrit are formalized in interpretiveprinciples (paribhas. as) in the metalanguage of Pan. inis As. t.adhya-y. Here he also reiterates the semantic foundation of Pan. inian

vi PREFACE

derivational procedure. In the second, he deals with the syntaxof the extension and removal of properties, and comparison withrespect to shared actions or properties. In that connection, he dis-cusses the two rules that account for the use of the affix vat andprovides a categorization of which sutras of the As. t.adhyay con-form to the syntactic pattern described by which rule.

The next paper, by Scharf, Goyal, Ajotikar and Ajotikar, de-scribes a computational implementation of Pan. inis rules that de-termine the use of Atmanepada and Parasmaipada verbal termina-tions under specific semantic and cooccurrence conditions. Thepaper necessarily touches upon the different categorization of con-ditions for voice in the Pan. inian and European traditions. Aussantexamines the presuppositions of European and Indian approachesto word classification generally. She traces contemporary schemesof tagging parts of speech in computational linguistics back to cat-egories described by Dionysius Thrax, and compares them withIndian classification schemes traced back to Pan. ini and Yaska.Joshi compares the concepts of concord and government in Eu-ropean grammar with the concepts of coreferrentiality (samana-dhikaran. ya) and unexpressed karaka roles (anabhihitatva) in Pa-n. inian grammar.

Several papers deal with the topic of word order and howfree it is. Scharfs paper, "Interrogatives," shows that interrog-atives locate in positions of focus and topic rather than movingfrom a position designated by a predetermined principle of word-ordering. Gillon examines various types of ellipsis and their re-lation to constituent units and the argument structure of lexicalitems. Amba Kulkarni, Shukla, Satuluri, and Shukl examine prox-imity constraints on sentences analyzed in dependency structuresto precisely determine the degree of freedom in Sanskrit word or-der. They find that dislocations in prose are limited to cases ofunilateral expectancy while in verse dislocations involving mutualexpectancy also occur. Scharf, Ajotikar, Savardekar, and Goyal

SANSKRIT SYNTAX vii

further examine differences between prose and poetic syntax com-putationally with significant results. Melnad, Goyal, and Scharfdescribe software they developed to identify metrical patterns. Fi-nally, Katira and Malhar Kulkarni, examine syntactic violations insentences identified as erroneous by Charudeva Shastri and presentparse trees for them.

As the research in the area of Sanskrit syntax continues toflourish, we plan to furnish bibliographic updates regularly. Wetherefore invite scholars in the field to inform us of current publi-cations as well as items missing from the bibliography supplied inthis volume. Please send complete citations to Hans Henrich Hockand to me at our email addresses provided in the list of contribu-tors.

Let me close by mentioning two practical points for readers.(1) While the reference lists for most contributions cover cita-tions in that contribution completely, since most of the citationsin Hocks survey of research are to references in the Sanskrit syn-tax bibliography, only references not included there are containedin the reference list at the end of his contribution. (2) Generallyaccents in Vedic passages in Devanagar are shown using marksproper to the particular Vedic school while accents in Romaniza-tion mark the underlying udatta with an acute accent mark () andindependent svarita with a grave accent mark (). Cardona marksaccents in the Romanization of words in derivation likewise but infinished forms borrows the marks used in Indic scripts for the com-mon system of accentuation described in the R

kpratisakhya. Hock

similarly borrows marks used in Indic scripts to mark accents inthe Romanization of Satapathabrahman. a passages in accordancewith their description in the Bhas. ikasutras.

Peter M. Scharf 28 February 2015

viii PREFACE

Contributors

ANUJA AJOTIKARLaboratoire dHistoire des Thories Linguistiques, UniversitParis Diderot, and Indian Institute of Technology [email protected]

TANUJA AJOTIKARLaboratoire dHistoire des Thories Linguistiques, UniversitParis Diderot, and Indian Institute of Technology [email protected]

MILIE AUSSANTCNRS, Laboratoire dHistoire des Thories Linguistiques, Univer-sit Paris Diderot, Sorbonne Paris Cit, [email protected]

GEORGE CARDONADepartment of Linguistics, University of Pennsylvania, [email protected]

BRENDAN S. GILLONDepartment of Linguistics, McGill University, [email protected]

PAWAN GOYALDepartment of Computer Science, Indian Institute of Technology,[email protected]

ix

x CONTRIBUTORS

HANS HENRICH HOCKUniversity of Illinois, [email protected]

PRASAD P. JOSHIDeccan College, [email protected]

DIPESH KATIRAShree Somnath Sanskrit University, [email protected]

AMBA KULKARNIDepartment of Sanskrit Studies, University of Hyderabad, [email protected]

MALHAR KULKARNIIndian Institute of Technology [email protected]

KESHAV S. MELNADIndian Institute of Technology [email protected]

PAVANKUMAR SATULURIDepartment of Sanskrit Studies, University of Hyderabad, [email protected]

SAMPADA SAVARDEKARIndian Institute of Technology [email protected]

SANSKRIT SYNTAX xi

PETER M. SCHARFThe Sanskrit Library, Laboratoire dHistoire des Thories Linguis-tiques, Universit Paris Diderot, and the Indian Institute of Tech-nology [email protected]

DEVANAND SHUKLMaharshi Sandipani Rashtriya Vedavidya Pratishthan, Ujjain, [email protected]

PREETI SHUKLADepartment of Sanskrit Studies, University of Hyderabad, [email protected]

xii CONTRIBUTORS

Table of contents

Preface iii

Contributors ix

Some issues in Sanskrit syntaxHANS HENRICH HOCK 1

Derivation and interpretation in Pan. inis systemGEORGE CARDONA 53

Extension rules and the syntax of As. t.adhyay sutraswith -vatiGEORGE CARDONA 109

Voice, preverb, and transitivity restrictionsin Sanskrit verb usePETER SCHARF, PAWAN GOYAL,ANUJA AJOTIKAR, and TANUJA AJOTIKAR 157

Interrogatives and word-order in SanskritPETER M. SCHARF 203

To classify words: European and Indiangrammatical approachesMILIE AUSSANT 219

Constituency and cotextual dependencein Classical SanskritBRENDAN S. GILLON 237

xiii

xiv SANSKRIT SYNTAX

How free is free word order in Sanskrit?AMBA KULKARNI, PREETI SHUKLA, PAVANKUMARSATULURI and DEVANAND SHUKL 269

Distinctive features of poetic syntax: preliminary resultsPETER SCHARF, ANUJA AJOTIKAR,SAMPADA SAVARDEKAR, and PAWAN GOYAL 305

Meter identification of Sanskrit verseKESHAV MELNAD, PAWAN GOYAL and PETER SCHARF 325

On concord and government relations in SanskritPRASAD P. JOSHI 347

Parse trees for erroneous sentencesDIPESH KATIRA and MALHAR KULKARNI 361

A bibliography of Sanskrit syntax 399

As. t.adhyay sutra index 471

Author index 477

Title index 487

Detailed table of contents

Preface iii

Contributors ixTable of contents . . . . . . . . . . . . . . . . . . . . . xiiiDetailed table of contents . . . . . . . . . . . . . . . . . xvList of tables . . . . . . . . . . . . . . . . . . . . . . . . xxvList of figures . . . . . . . . . . . . . . . . . . . . . . . xxvii

Some issues in Sanskrit syntaxHANS HENRICH HOCK 11 Introduction . . . . . . . . . . . . . . . . . . . . . 22 A brief survey of recent publications . . . . . . . . 23 Formal issues . . . . . . . . . . . . . . . . . . . . 4

3.1 Free Word Order and related issues . . . . 43.2 Relative-correlatives . . . . . . . . . . . . 93.3 Some issues of agreement . . . . . . . . . 133.4 Converbs, reflexives, oblique subjects, and

syntactic bracketing . . . . . . . . . . . . 193.5 Double Direct Object constructions and

Causatives . . . . . . . . . . . . . . . . . 263.6 Asamartha compounding . . . . . . . . . . 32

4 Functional issues that should be of interest to com-putational approaches . . . . . . . . . . . . . . . . 354.1 Fronting . . . . . . . . . . . . . . . . . . . 36

4.1.1 Initial strings in Vedic . . . . . . 364.1.2 Predicate-Subject order . . . . . 374.1.3 Narrative linkage and related is-

sues . . . . . . . . . . . . . . . 40

xv

xvi SANSKRIT SYNTAX

4.2 Extraposition . . . . . . . . . . . . . . . . 434.2.1 Gondas amplified sentences . . 434.2.2 Purpose datives in Vedic prose . 444.2.3 Kartr

backing and extraposition,

and politeness . . . . . . . . . . 455 Conclusions and implications for further research . 46References . . . . . . . . . . . . . . . . . . . . . . . . . 49

Derivation and interpretation in Pan. inis systemGEORGE CARDONA 531 Speakers and listeners . . . . . . . . . . . . . . . . 532 Derivation and a speakers perspective . . . . . . . 56

2.1 Katyayana on speech being prompted bymeaning . . . . . . . . . . . . . . . . . . . 56

2.2 Patajali on speech being caused by meaning 572.2.1 Number distinction . . . . . . . 572.2.2 Residual relations . . . . . . . . 612.2.3 Gender distinction . . . . . . . . 67

2.3 Meaning conditions and derivation . . . . . 693 Affixation meaning conditions . . . . . . . . . . . 714 A different approach suggested . . . . . . . . . . . 735 Interpreting Pan. inian sutras . . . . . . . . . . . . . 88

5.1 Introduction . . . . . . . . . . . . . . . . . 885.2 Contexts requiring metarules . . . . . . . . 89

5.2.1 :Sa; +a .~Ta;a;nea;ya;ea;ga;a (A. 1.1.49) . . . . 915.2.2 ta;/////////a;sma;a;a;a;ta ;a;na;a;dR ;e :pUa;vRa;~ya ta;sma;a-

;a;d;tyua:a:=+~ya (A. 1.1.6667) . . . . 945.2.2.1 Purpose of the rules. . . 945.2.2.2 Scope of 1.1.66. . . . . 965.2.2.3 Scope of 1.1.67. . . . . 99

6 Summary . . . . . . . . . . . . . . . . . . . . . . 100References . . . . . . . . . . . . . . . . . . . . . . . . . 103

TABLE OF CONTENTS xvii

Extension rules and the syntax of As. t.adhyay sutraswith -vatiGEORGE CARDONA 1091 Introduction . . . . . . . . . . . . . . . . . . . . . 1092 Varieties of extension . . . . . . . . . . . . . . . . 111

2.1 Karyatidesa . . . . . . . . . . . . . . . . . 1112.2 Sastratidesa . . . . . . . . . . . . . . . . . 1132.3 Rupatidesa . . . . . . . . . . . . . . . . . 1172.4 Tadatmyatidesa . . . . . . . . . . . . . . . 1172.5 Nimittatidesa . . . . . . . . . . . . . . . . 1192.6 Vyapadesatidesa . . . . . . . . . . . . . . 1202.7 Arthatidesa . . . . . . . . . . . . . . . . . 122

3 Extensions of the type X is Y . . . . . . . . . . . 1234 Syntactic structures . . . . . . . . . . . . . . . . . 127

4.1 Predication without comparison . . . . . . 1274.2 Comparisons . . . . . . . . . . . . . . . . 127

4.2.1 Comparison involving a com-mon action . . . . . . . . . . . . 128

4.2.2 Other comparisons . . . . . . . . 1314.3 Summary of syntactic structures . . . . . . 133

5 Semantic and pragmatic considerations . . . . . . . 1366 Summary . . . . . . . . . . . . . . . . . . . . . . 1407 As. t.adhyay sutras in which terms with -vati are used 142

7.1 List indicating extension type . . . . . . . 1427.2 Syntax of sutras with terms in -vati . . . . 146

7.2.1 Schema I (4.3) . . . . . . . . . 1467.2.2 Schema II . . . . . . . . . . . . 148

References . . . . . . . . . . . . . . . . . . . . . . . . . 151

Voice, preverb, and transitivity restrictionsin Sanskrit verb use

xviii SANSKRIT SYNTAX

PETER SCHARF, PAWAN GOYAL,ANUJA AJOTIKAR, and TANUJA AJOTIKAR 1571 Introduction . . . . . . . . . . . . . . . . . . . . . 158

1.1 Root markers . . . . . . . . . . . . . . . . 1611.2 Prayoga . . . . . . . . . . . . . . . . . . . 1611.3 Other semantic conditions . . . . . . . . . 1621.4 Preverbs . . . . . . . . . . . . . . . . . . . 1621.5 Other co-occurrence conditions . . . . . . 1631.6 Tabulation of voice conditions . . . . . . . 163

2 Related Work . . . . . . . . . . . . . . . . . . . . 1653 Preparing inputs for the implementation . . . . . . 171

3.1 XML Dhatupat.ha database . . . . . . . . . 1713.2 Preverb cooccurences attested by Pan. ini . . 175

4 Implementation details . . . . . . . . . . . . . . . 1774.1 Data Structure . . . . . . . . . . . . . . . 1774.2 Rule formulation . . . . . . . . . . . . . . 178

4.2.1 Phonetic conditions . . . . . . . 1784.2.2 Attribute conditions . . . . . . . 1784.2.3 Semantic conditions . . . . . . . 1794.2.4 Optionality . . . . . . . . . . . . 1794.2.5 Action . . . . . . . . . . . . . . 179

4.3 Control structure . . . . . . . . . . . . . . 1795 Results and related discussions . . . . . . . . . . . 181

5.1 Notations in the derivation tree . . . . . . . 1815.2 Handling meaning conditions . . . . . . . 1845.3 Handling rule interaction . . . . . . . . . . 1855.4 Handling praptavibhas.a . . . . . . . . . . 188

6 Conclusions and Future Work . . . . . . . . . . . . 189References . . . . . . . . . . . . . . . . . . . . . . . . . 195

Interrogatives and word-order in SanskritPETER M. SCHARF 2031 The problem . . . . . . . . . . . . . . . . . . . . . 203

TABLE OF CONTENTS xix

2 The evidence . . . . . . . . . . . . . . . . . . . . 2083 Parallels with other pronouns . . . . . . . . . . . . 2124 The solution . . . . . . . . . . . . . . . . . . . . . 214References . . . . . . . . . . . . . . . . . . . . . . . . . 216

To classify words: European and Indiangrammatical approachesMILIE AUSSANT 2191 Introduction . . . . . . . . . . . . . . . . . . . . . 2192 The European approach . . . . . . . . . . . . . . . 220

2.1 Parts of speech in computational linguistics 2202.2 Parts of speech in Graeco-Latin grammat-

ical tradition . . . . . . . . . . . . . . . . 2213 The Indian approach (Sanskrit Vyakaran. a) . . . . . 224

3.1 The naman-akhyata-upasarga-nipata clas-sification . . . . . . . . . . . . . . . . . . 224

3.2 The Pan. inian classification of padas . . . . 2273.3 Some remarks on a few other classifications 230

4 Concluding remarks . . . . . . . . . . . . . . . . . 232References . . . . . . . . . . . . . . . . . . . . . . . . . 233

Constituency and cotextual dependencein Classical SanskritBRENDAN S. GILLON 2371 Introduction . . . . . . . . . . . . . . . . . . . . . 2372 Parts of speech . . . . . . . . . . . . . . . . . . . 2393 Clausal constituents . . . . . . . . . . . . . . . . . 2404 Supra-lexical and sub-clausal constituents . . . . . 240

4.1 Context Dependence . . . . . . . . . . . . 2414.2 Proforms . . . . . . . . . . . . . . . . . . 2464.3 Ellipsis . . . . . . . . . . . . . . . . . . . 247

4.3.1 Interrogative ellipsis (sluicing) . 248

xx SANSKRIT SYNTAX

4.3.2 Appended coordination (strip-ping) . . . . . . . . . . . . . . . 249

4.3.3 Gapping . . . . . . . . . . . . . 2504.3.4 Verb phrase ellipsis . . . . . . . 2524.3.5 Copular complement ellipsis . . 2524.3.6 Nominal ellipsis . . . . . . . . . 253

5 Further evidence for Phrases . . . . . . . . . . . . 2556 Conclusion . . . . . . . . . . . . . . . . . . . . . 263References . . . . . . . . . . . . . . . . . . . . . . . . . 264

How free is free word order in Sanskrit?AMBA KULKARNI, PREETI SHUKLA, PAVANKUMARSATULURI and DEVANAND SHUKL 2691 Introduction . . . . . . . . . . . . . . . . . . . . . 2692 Word order in Sanskrit . . . . . . . . . . . . . . . 2733 Indian theories of expectancy and proximity . . . . 274

3.1 Akanks.a (expectancy) . . . . . . . . . . . 2753.2 Sannidhi (proximity) . . . . . . . . . . . . 277

4 Dependency parsing and word order . . . . . . . . 2784.1 Tree traversal and possible word orders . . 279

4.1.1 Generalising tree traversal . . . . 2804.2 Projectivity principle . . . . . . . . . . . . 2824.3 Weak non-projectivity (planarity) . . . . . 285

5 Empirical evaluation . . . . . . . . . . . . . . . . 2885.1 Cases of sannidhi violation from Gillons

data . . . . . . . . . . . . . . . . . . . . . 2885.1.1 Dislocation of a genitive . . . . . 2895.1.2 Dislocation of a vises.an. a . . . . 2915.1.3 Other relations . . . . . . . . . . 291

5.2 Sannidhi violation in the Bhagavadgta . . 2925.2.1 Sannidhi violation involving kartr

and karman . . . . . . . . . . . 293

TABLE OF CONTENTS xxi

5.2.2 Sannidhi violation involving twokartr

relations . . . . . . . . . . 298

5.2.3 Sannidhi violation involving kar-man and kriyavises.an. a . . . . . . 299

5.2.4 Sannidhi violation involving kar-man and apadana . . . . . . . . 299

6 Conclusion . . . . . . . . . . . . . . . . . . . . . 300References . . . . . . . . . . . . . . . . . . . . . . . . . 302

Distinctive features of poetic syntax: preliminary resultsPETER SCHARF, ANUJA AJOTIKAR,SAMPADA SAVARDEKAR, and PAWAN GOYAL 3051 Motivation . . . . . . . . . . . . . . . . . . . . . . 3062 Expectations . . . . . . . . . . . . . . . . . . . . . 3073 Methodology . . . . . . . . . . . . . . . . . . . . 310

3.1 Corpus preparation . . . . . . . . . . . . . 3103.2 An Indian cognitive linguistics depen-

dency relations tagset . . . . . . . . . . . . 3113.3 Queries . . . . . . . . . . . . . . . . . . . 313

4 Results . . . . . . . . . . . . . . . . . . . . . . . . 3144.1 Agent after verb . . . . . . . . . . . . . . 3144.2 Object after verb . . . . . . . . . . . . . . 3154.3 Instrument after verb . . . . . . . . . . . . 3154.4 Adverb after verb . . . . . . . . . . . . . . 3154.5 Qualifier after qualified . . . . . . . . . . . 3154.6 Genitive after what it limits . . . . . . . . 3184.7 Object after absolutive . . . . . . . . . . . 3184.8 Agent after object . . . . . . . . . . . . . . 3194.9 Summary . . . . . . . . . . . . . . . . . . 319

References . . . . . . . . . . . . . . . . . . . . . . . . . 322

xxii SANSKRIT SYNTAX

Meter identification of Sanskrit verseKESHAV MELNAD, PAWAN GOYAL and PETER SCHARF 3251 Introduction . . . . . . . . . . . . . . . . . . . . . 3252 Sources describing Sanskrit meters . . . . . . . . . 3273 Sanskrit prosody . . . . . . . . . . . . . . . . . . 328

3.1 Syllable weight . . . . . . . . . . . . . . . 3293.2 The basic unit of varn. avr

tta meter: gan. a . . 330

3.3 The basic unit of matravr

tta meter: catu-rmatrika . . . . . . . . . . . . . . . . . . . 331

3.4 Types of meters . . . . . . . . . . . . . . . 3314 Meter Identifying Tool (MIT) . . . . . . . . . . . . 333

4.1 Meter database . . . . . . . . . . . . . . . 3334.2 Input and output . . . . . . . . . . . . . . 3344.3 Algorithm . . . . . . . . . . . . . . . . . . 335

5 Results . . . . . . . . . . . . . . . . . . . . . . . . 3385.1 A samavr

tta verse in Vasantatilaka meter . 338

5.2 An ardhasamavr

tta verse in Vegavat meter 3395.3 A vis.amavr

tta verse in Lalita meter . . . . 340

5.4 A matravr

tta verse in Arya meter . . . . . . 3415.5 Performance . . . . . . . . . . . . . . . . 342

6 Conclusions and future work . . . . . . . . . . . . 343References . . . . . . . . . . . . . . . . . . . . . . . . . 345

On concord and government relations in SanskritPRASAD P. JOSHI 3471 Introduction . . . . . . . . . . . . . . . . . . . . . 3482 The importance of concord and government . . . . 3483 Concord . . . . . . . . . . . . . . . . . . . . . . . 349

3.1 Concord of substantive and adjective . . . 3493.1.1 Adjectives with taddhita-deletion 3523.1.2 Nouns in apposition . . . . . . . 354

3.2 Concord of a predicate verb and noun (ofagent and object) . . . . . . . . . . . . . . 355

TABLE OF CONTENTS xxiii

4 Government . . . . . . . . . . . . . . . . . . . . . 3574.1 Government of a verb and a noun . . . . . 3574.2 Government of an indeclinable and a noun 358

5 Conclusions . . . . . . . . . . . . . . . . . . . . . 359References . . . . . . . . . . . . . . . . . . . . . . . . . 360

Parse trees for erroneous sentencesDIPESH KATIRA and MALHAR KULKARNI 3611 Introduction . . . . . . . . . . . . . . . . . . . . . 3612 Modern Sanskrit . . . . . . . . . . . . . . . . . . . 3633 Error analysis . . . . . . . . . . . . . . . . . . . . 3634 The standard of correct Sanskrit . . . . . . . . . . 3645 Discussion of erroneous sentences . . . . . . . . . 367

5.1 Sentence 1 . . . . . . . . . . . . . . . . . 3675.1.1 Shastris comments . . . . . . . 3675.1.2 Discussion . . . . . . . . . . . . 3685.1.3 Directive rules . . . . . . . . . . 373

5.2 Sentence 2 . . . . . . . . . . . . . . . . . 3745.2.1 Shastris comments . . . . . . . 3755.2.2 Discussion . . . . . . . . . . . . 3765.2.3 Directive rule . . . . . . . . . . 378




xxiv SANSKRIT SYNTAX





5.10 Sentence 10 . . . . . . . . . . . . . . . . . 3955.10.1 Shastris comments . . . . . . . 3955.10.2 Directive rule . . . . . . . . . . 396

6 Conclusion . . . . . . . . . . . . . . . . . . . . . 397References . . . . . . . . . . . . . . . . . . . . . . . . . 397

A bibliography of Sanskrit syntax 399

As. t.adhyay sutra index 471

Author index 477

Title index 487

List of tables

Some issues in Sanskrit syntax 11 Hocks (1989a) account for example (8) . . . . . . 132 Locative absolute with normal and upside-

down agreement . . . . . . . . . . . . . . . . . . 183 Non-head relational nominal governing an exter-

nal head . . . . . . . . . . . . . . . . . . . . . . . 344 Node erasure, movement, and compounding in (35c) 34

Derivation and interpretation in Pan. inis system 531 Abbreviations . . . . . . . . . . . . . . . . . . . . 102

Extension rules and the syntax of . . . -vati 1091 Derivation of non-feminine forms of the interrog-

ative pronoun kim and noun vr

ks. a . . . . . . . . . 1122 Derivation of (14a) .tea;na tua;yMa va;tRa;tea and (14b) ta;d ;d ;tRa;tea 1303 Sutras with terms in -vati . . . . . . . . . . . . . . 1424 Ia. Sutras in which a nominative form is explicit . . 1465 Ib. Sutras in which a nominative form is understood 1476 IIa. B-vat = B6 iva . . . . . . . . . . . . . . . . . . 1487 IIb. B-vat = B7 iva . . . . . . . . . . . . . . . . . . 1508 Abbreviations . . . . . . . . . . . . . . . . . . . . 150

Voice, preverb, and transitivity restrictions 1571 Correspondence of voice with prayoga and pada . . 160

To classify words: European and Indian . . . approaches 2191 Sub-categories of noun in the Brown Corpus . . . . 2222 Abbreviations . . . . . . . . . . . . . . . . . . . . 233

xxv

xxvi SANSKRIT SYNTAX

Constituency and cotextual dependence 2371 English optional complementation . . . . . . . . . 2582 Abbreviations . . . . . . . . . . . . . . . . . . . . 264

How free is free word order in Sanskrit? 2691 Abbreviations . . . . . . . . . . . . . . . . . . . . 301

Distinctive features of poetic syntax 3051 Probability of x occurring before y in a relation

pair (x,y): comparing poetic sentences with prose . 321

Meter identification of Sanskrit verse 3251 Gan. a patterns . . . . . . . . . . . . . . . . . . . . 3312 Matra patterns . . . . . . . . . . . . . . . . . . . . 3323 Prototypes of each of the four types of meters and

the number of meters of each type in our database . 334

Parse trees for erroneous sentences 3611 Word sense in Sanskrit versus in modern Indian

languages . . . . . . . . . . . . . . . . . . . . . . 393

List of figures

Voice, preverb, and transitivity restrictions 1571 The control structure for the implementation. This

process is applied recursively until no rules aretriggered. . . . . . . . . . . . . . . . . . . . . . . 180

2 Derivation tree for the root bhu . . . . . . . . . . . 1833 Derivation tree for the root da . . . . . . . . . . . 1904 Derivation tree for the root gam . . . . . . . . . . . 1915 Derivation tree for the root ks. ip . . . . . . . . . . . 1926 Derivation tree for the root kr . . . . . . . . . . . 1937 Derivation tree for the root ram . . . . . . . . . . . 194

To classify words: European and Indian . . . approaches 2191 The main classes of speech forms distinguished by

Pan. ini . . . . . . . . . . . . . . . . . . . . . . . . 229

How free is free word order in Sanskrit? 2691 Traversals for sentence (1) with and without trans-

position . . . . . . . . . . . . . . . . . . . . . . . 2812 Dependency graph for sentence 3 . . . . . . . . . . 2823 Dependency structure with projections for sen-

tences 1.1 to 1.6 . . . . . . . . . . . . . . . . . . . 2834 Projections for sentences 2.1 to 2.6 . . . . . . . . . 2845 Projection for sentence (4) . . . . . . . . . . . . . 2846 Projection with Rearrangement of nodes . . . . . . 2857 Planar dependency graph for sentences (2.2) and

(2.5) . . . . . . . . . . . . . . . . . . . . . . . . . 2868 Planar dependency graph for sentence (4) . . . . . 2879 Dislocation without sannidhi violation . . . . . . . 287

xxvii

xxviii SANSKRIT SYNTAX

10 Dislocation of a genitive . . . . . . . . . . . . . . 29011 Dislocation of a genitive . . . . . . . . . . . . . . 29012 Dislocation of a vises.an. a . . . . . . . . . . . . . . 29113 Dislocation of an argument . . . . . . . . . . . . . 29214 Analysis of BhG. 6.34 . . . . . . . . . . . . . . . . 29415 Analysis of BhG. 8.25 . . . . . . . . . . . . . . . . 29516 Analysis of BhG. 10.16 . . . . . . . . . . . . . . . 29617 Analysis of BhG. 9.3 . . . . . . . . . . . . . . . . 29618 Analysis of BhG. 1.27 . . . . . . . . . . . . . . . . 29719 Modified Analysis of BhG. 1.27 . . . . . . . . . . 29820 Analysis of BhG. 8.19 . . . . . . . . . . . . . . . . 29921 Analysis of BhG. 1.37 . . . . . . . . . . . . . . . . 30022 Analysis of BhG. 18.75 . . . . . . . . . . . . . . . 301

Distinctive features of poetic syntax 3051 Relative position of an agent with respect the verb

that governs it: comparison of prose and poetrysyntax . . . . . . . . . . . . . . . . . . . . . . . . 316

2 Relative position of a direct object with respect tothe verb that governs it: comparison of prose andpoetry syntax . . . . . . . . . . . . . . . . . . . . 316

3 Relative position of an instrument with respect tothe verb that governs it: comparison of prose andpoetry syntax . . . . . . . . . . . . . . . . . . . . 317

4 Relative position of an adverb with respect to theverb that governs it: comparison of prose and po-etry syntax . . . . . . . . . . . . . . . . . . . . . . 317

5 Relative position of a qualifier with respect to whatit qualifies in a sentence: comparison of prose andpoetry syntax . . . . . . . . . . . . . . . . . . . . 318

6 Relative position of a genitive with respect to whatit limits in a sentence: comparison of prose andpoetry syntax . . . . . . . . . . . . . . . . . . . . 320

LIST OF FIGURES xxix

7 Relative position of a direct object with respect toan absolutive that governs it: comparison of proseand poetry syntax . . . . . . . . . . . . . . . . . . 320

8 Relative position of a direct object with respect toan agent of the same action governing them: com-parison of prose and poetry syntax . . . . . . . . . 321

Parse trees for erroneous sentences 3611 Sentence 1 . . . . . . . . . . . . . . . . . . . . . . 3732 Sentence 2 . . . . . . . . . . . . . . . . . . . . . . 3773 Sentence 3 . . . . . . . . . . . . . . . . . . . . . . 3794 Sentence 4 . . . . . . . . . . . . . . . . . . . . . . 3825 Sentence 5 . . . . . . . . . . . . . . . . . . . . . . 3866 Sentence 6 . . . . . . . . . . . . . . . . . . . . . . 3887 Sentence 7 . . . . . . . . . . . . . . . . . . . . . . 3898 Sentence 8 . . . . . . . . . . . . . . . . . . . . . . 3919 Sentence 9 . . . . . . . . . . . . . . . . . . . . . . 39510 Sentence 10 . . . . . . . . . . . . . . . . . . . . . 396

xxx SANSKRIT SYNTAX

Some issues in Sanskrit syntax

HANS HENRICH HOCK

Abstract: The purpose of this presentation is to provide a briefgeneral survey of major recent developments in Sanskritsyntactic studies since Deshpande and Hock 1991, fol-lowed by more in-depth discussion of formal and func-tional issues that should be of interest to scholars work-ing on computational approaches to Sanskrit syntax. For-mal syntactic topics are free word order, challengingsyntactic aspects of relative-correlative constructions, is-sues of agreement, oblique subjects and related issues,the syntax of double-direct-object constructions (includ-ing causatives), and the issue of asamartha compound-ing. The functional part focuses on the use of differentsyntactic alternatives in discourse and in different genres a topic that, to my knowledge, has not received sys-tematic discussion. Two major phenomena are examinedin some detail fronting and extraposition to the right,including their different functions in a variety of genres.Functional analytical approaches may yield interesting in-sights and challenges to formal syntactic approaches, al-though they clearly cannot be a substitute for formal anal-ysis. I conclude that there is ample room for more researchon Sanskrit syntax, especially in the post-Vedic languagewith its richer variety of different genres.

Keywords: Sanskrit, syntax, formal approaches, functionalapproaches

1

2 H. H. HOCK

1 Introduction

My presentation has three goals. One is to provide a survey of pub-lications in Sanskrit syntax since Deshpande and Hocks (1991)Sanskrit syntax bibliography. A second one is to focus in greaterdetail on a number of formal issues that, I believe, would be of in-terest both to linguists pursuing computational approaches to San-skrit syntax and to those working in linguistic theory. The thirdgoal is to discuss a selection of functional factors that influencethe use of particular syntactic structures in particular text types, anissue which I believe would also be interesting to those engaged incomputational work.

2 A brief survey of recent publications

The Sanskrit syntax bibliography provided at the end of this bookconsolidates Deshpande and Hocks (1991) A bibliography of writ-ings on Sanskrit syntax published in Hock 1991b and the bibliog-raphy of works on Sanskrit syntax published since then compiledfor presentation at the Seminar on Sanskrit Syntax and DiscourseStructures held in Paris, 1315 June 2013. The combined bibliog-raphy will also be published at the Sanskrit Library website withregular upadates. We look forward to receiving additional refer-ences so as to make the bibliography more comprehensive. At thispoint, let me briefly survey the distribution of recent publicationsin terms of chronology and general categories of subject matter.

Recent research concerning Sanskrit syntax includes over 200works, including article collections and bibliographies quite im-pressive for a period of less than 25 years, especially if comparedto the 474 entries in Deshpande and Hock 1991, which covers pub-lications from Burnouf (1824) to Brereton and Jamison To Ap-pear (published 1991). It is impressive, too, because, as in thepast, syntax receives much less scholarly attention than other as-

SOME ISSUES IN SANSKRIT SYNTAX 3

pects of Sanskrit linguistics. The following statistics are based ona smaller set of 186 entries that I had collected by 24 May 2013.

As in the past (see Deshpande and Hock 1991, as well as Hock1989b), the bulk of publications (147 out of 186) deals with theVedic period, especially the R

gveda, and commonly from a his-

torical, Indo-Europeanist perspective. The post-Vedic or classicalperiod is covered only in 45 publications, some of which addressboth Vedic and post-Vedic issues.

A common topic for the Vedic period is the issue of tense, as-pect, voice, and modality, which is dealt with in 29 publications.Another 23 publications address word and phrase order, includingthe issue of clitics and Wackernagels Law. Case syntax is cov-ered in 18 publications, especially by Hettrich (six publications,including the important 2007: Materialien zu einer Kasussyntaxdes R

gveda, which includes a rich bibliography). Issues of subor-

dination, both finite and non-finite (infinitives, converbs, locativeabsolute constructions, etc.) are dealt with in some 18 publica-tions.

Most of the 27 publications that approach Sanskrit syntax froma modern theoretical or typological approach are focused on thepost-Vedic language, and so are all of the 23 publications that dealwith, or refer to, the insights of the Indian grammatical tradition,especially Pan. ini. As a personal note, let me add that this relativelylimited reference to the Indian tradition is regrettable. In principle,all research on Sanskrit syntax and Sanskrit linguistics in gen-eral should treat the work of Pan. ini and the entire early gram-matical and phonetic tradition as earlier scholarship, just as it doesthe work of linguists like Delbrck, Wackernagel, and Whitney;see Hock 2009b and Hock Forthcoming(b).

Finally, 65 publications treat of issues of function and dis-course, in both the Vedic and the post-Vedic period. This categoryis heavily dominated by Jared Kleins publications, most of which(16 out of 22) focus on the issue of stylistic repetition of differ-

4 H. H. HOCK

ent grammatical structures and categories in the language of theR

gveda.

3 Formal issues

In the following section of my paper I address formal issues thatmay be of interest to scholars pursuing computational approachesto Sanskrit syntax as well those working on typology and syntactictheory. I draw to a large extent on my own research, both pub-lished and unpublished, but also include references to other recentresearch.

A recurrent theme is that we need to consider both Pan. i-nis generative approach and modern approaches, whether gen-erative or traditional-philological, and that, likewise, we need tokeep in mind both the empirical information conveyed by Pan. i-nis speaker-knowledge based grammar and the empirical data un-earthed by western philological approaches. The latter issue isespecially relevant, since as Deshpande (1983) suggested, Pan. inislocation on the northwestern periphery of the Sanskrit-speakingworld of his time may account for certain differences between thesyntactic structures predicted by his grammar and those found inthe tradition of Madhyadesa; see also Hock 1981, 2012b.

3.1 Free Word Order and related issues

It is well known that Sanskrit (like other early Indo-European lan-guages) exhibits a remarkable degree of free word order not justfree phrase order. In this section I discuss two major formal ap-proaches to this phenomenon. Schufele (1990, 1991a,b) followsthe major tradition of modern western scholarship (e.g. Delbrck1878, 1888; Speijer 1886, 1896; Lahiri 1933) in assuming a basicword and phrase order of the SOV type, with various movementprocesses accounting for marked orderings. The work of Gillon


(1996) and Gillon and Shaer (2005) adopts and modifies Staals(1967) notion of wild trees, i.e. trees without phrase-internallinear ordering. Neither approach adopts the possible alternativeof assuming complete non-configurationality along the lines sug-gested for other languages by Farmer (1980), and Ken Hale (1975,1983).

Schufeles most important findings are the following (1990:6163, 84):

In the majority of cases, phrases are continuous and exhibitall the features normally associated with hierarchical struc-ture. This is something that children learning the languagewould have to account for in their grammar, and it woulddiscourage them from positing a completely flat structure.

Similarly, in the majority of cases, phrases are head-final,although for PPs head-finality is only a statistical tendencyin Vedic. While Schufele does not pursue this issue ex-plicitly, the dominant head-finality too can be argued to besomething that children learning the language would have toaccount for in their grammar.

In PPs the adposition normally remains next to at least partof its complement if there is movement. This, again, sup-ports the assumption of hierarchical, rather than flat phrasestructure. Schufele (1990: 85) cites the examples in (1).Further examples can be found in the classical language; (2).Interestingly, Bolkestein (2001) and Snijders (2012) note thesame phenomenon in Latin.

Movement of individual words or combinations of words, asin (3),1 is made possible through a process of liberation ornode erasure (see Pullum 1982, Ross 1967: 5054 (1986)).

1These examples come from my collections.

6 H. H. HOCK

(1) a. .~ta;vEa :pu!a:=+a :pa;a;ya;R!a;a;d;nd ! +ma; Hstavai pura paryad ndram ahnah. (R

V. 3.32.14b)

I shall praise Indra before the fateful day.

b. O;;ta;Ea va;a ! :(;Ma ma;a;h;ma;!a;na;a;va;a;B!a;taH .s!a;}ba;BUa;va;tua:=,etau va a

svam mahima

nav abhi

tah. sa

mbabhuvatur (SB.10.6.4.1)These two jars appeared around = on both sides of thehorse. (Ss translation; my transcription)

(2) ;a .sa;mua;d +a:ua ;vEa :pUa;va;Ra;d;a .sa;mua;d +a:ua :pa;a;(a;ma;a;t,a ta;ya;ea;=e +va;a;nta:=M ;a;ga;ya;eRa:=+a;ya;Ra;va;t a ;a;va;du ;bRua;Da;aH

a samudrat tu vai purvad a samudrat tu pascimat.tayor eva + antaram giryor aryavartam vidur budhah. .. (Ma-nu 2.22)Wise people know (that) Aryavarta (extends) from the east-ern sea to the western sea, (and) between these two moun-tains (the himalayas and the vindhyas).

(3) a. O;;ta;!a;}vea;vE!a;S!a O;;t!a;smEa ;a;v!a;Sua;yRa;a;ea ;a;v!a;k +:a;////a;ntMa ;a;v!a;k +:ma;teaeta

mi v eva

+ es. a jeta

smai vi

s. n. ur j yajo vi

krantimi vikra-

mate (SB. 1.1.2.13)This Vis.n. u, the sacrifice, steps this (world-conqueringthree-fold) stepping for him (the sacrificer).

b. .tea;Sa;Mai Ba;a;ma;ea ma;h;a;ba;a;hu H :pa;a;a;TRa;va;a;na;Mai ma;h;a;tma;na;a;m,a ya;Ta;a;h ;k+.=+ea;tpUa:ja;Ma . . .tes. ami bhmo mahabahuh. parthivanami mahatmanam.yatharham akarot pujam . . . (Nala 2.11; MBh. 3.51.10*215)Strong-armed Bhma honored these noble rulers appro-priately . . .


Schufeles approach contrasts with that of Staal and Gillon.Staal starts with the claim that the Indian grammatical tradition,being silent on the issue, assumed that there are no rules for wordorder (360) and he goes on to argue for a (modern) generative ac-count operating with wild trees, i.e. trees without phrase-internallinear ordering.

Gillon (1996) adopts and modifies Staals (1967) notion ofwild trees, providing empirical support from two corpora theprose examples in Apte 1885 and 1,500 sentences from Dharma-krtis autocommentary on the Praman. avarttika. See also Gillonand Shaer 2005. The following discussion is based on the latterpublication.

Like Staal, as well as Schufele, Gillon and Shaer accept theneed for phrases, rather than a completely flat structure. UnlikeSchufele, they assume that there is no linear order within phrases.Moreover, they argue against a VP, instead postulating the flatclause structure in (4), without linear order. Further, they (2005:468) claim that . . . the strategy of deriving less common word or-ders with specialized information packaging functions from morebasic syntactic structures . . . seems to us less plausible than onesconsistent with the wild tree claim. The paper (2005: 48085)concludes with a section on Some remaining puzzles.

(4) S V, NPS, NPOIn support for the assumption of flat phrase-internal structures,

such as (4), they claim that their corpora exhibit both left- andright-headed phrases and that therefore there is no evidence forphrasal headedness (2005: 470). In addition, they accept move-ment processes that extract elements out of phrases and place themin left- or right-peripheral position within the clause (2005: 47580).

Certain features are shared between Schufeles and Gillonsapproaches the acceptance of phrases, rather than completely

8 H. H. HOCK

flat structures, and the fact that movement processes can extractand move elements out of phrases. For the purposes of computa-tional text analysis, therefore, there may be no significant differ-ence.

From the perspective of linguistic theory, however, the two ap-proaches differ considerably, and it is Schufeles approach thatprovides the better insights. His account of Sanskrit is completelycompatible with the linguistic typology of SOV languages, withhead-finality at all phrasal levels. In fact, Sanskrit also conformsto SOV typology in its complex syntax, by making extensive use ofnon-finite subordination as well as of relative-correlative construc-tions; Hock 1989a, 2005, 2014. Under the Staal-Gillon approachthese typological characteristics would be epiphenomenal at best.

Now, it is true that Pan. ini has no rules comparable to westerngeneralizations about word or phrase order. But Pan. ini also has norules comparable to western generalizations about phrases, suchas NP, VP, PP. True, there are rules regarding karakas and theirrealizations, but these do not address issues such as complex NPswith genitive modifiers. There is also the notion samanadhikara-n. a, but this presumably holds not only for agreement within NPsbut also relates surface subject NPs to their verbs (A. 1.4.104107)and must be assumed to hold also for agreement between subjects(kartr

s) and predicate nouns or adjectives (see 3.3 below for dis-

cussion). Even the notion sentence is a murky issue in the Pan. i-nian tradition; see e.g. Cardona 1976: 22324, Deshpande 1991,as well as Hock Forthcoming(b): 6. In all of these respects, andnot only as regards word order, the Indian grammatical traditionand modern generative approaches are orthogonal.

It is also true that Sanskrit offers frequent examples with non-final heads. But there are considerable differences between differ-ent texts. Consider major constituent order. As noted in Hock1984, while in mantra Vedic and Kalidasas dramatic dialogueverb-final structures amount to only about 65%, in Vedic prose


they are about 97% (see also Hock 1997). A similarly high ratioof verb-final structures is found in Patajalis Mahabhas. ya; see thestatistics in (5). Claims about headedness and any other aspectsof syntax must therefore be based on a large variety of differentgenres, not just on the two corpora examined by Gillon. And asthe evidence just cited shows, genres that do not make any claimsto stylistic or literary elegance are characterized by heavy predom-inance of head-final constituent order; in fact, even in other genresverb-finality runs to about 65%. The wild tree assumption thatphrases, including the sentence (S), have no internal order fails tocapture these facts.

(5) Word order in Patajalis Mahabhas. ya

a. Paspasa (Kielhorn-Abhyankar 1.1.11.3.5)2

V-final: 35V-initial: Not found in the sampleV + O in the formula . . . adhyeyam vyakaran. am: 73

sak + (O) + infinitive: 2

b. Sivasutras (Kielhorn-Abhyankar 1.15.21.16.18)V-final: 40V-initial (including after linker, such as tena): 5V + O/Pred: 3V + Other: 3V + [ ] iti: 8

3.2 Relative-correlatives

As in the case of word order, Pan. ini has remarkably little to sayabout the syntax of Sanskrit relativization. The closest he comes

2Vedic and other traditional citations are ignored. Gerundives and ta-participles used as main verbs are included.

3Contrast the formulaic use of the gerundive with the ordinary one in laghva-rtham cadhyeyam vyakaran. am. brahman. enavasyam sabda jeya iti.

10 H. H. HOCK

is in three sutras (A. 3.3.139140 and 3.3.156) that address the is-sue of modality in conditional structures which, as is well known,involve an adverbial form of the relative pronoun (yadi) or the par-ticle ced.

At least from the time of Speijer (1886, 1896) and Delbrck(1888), western scholarship has recognized that Sanskrit relativestructures consist of a relative clause, containing a relative pro-noun, and a main clause, containing a correlative pronoun and thatthe relative clause is not inserted into the main clause.4 Speijer(1886: 349, 1896: 349) refers to the relationship as one betweena protasis and an apodosis. Minard (1936) introduces the termdiptych for the construction which in typological and theoreticalliterature is now commonly referred to as relative-correlative.

The syntactic account of Sanskrit and other, similar relative-correlative constructions is further refined in the 1970s and 1980sby arguments that the relative clause is base-generated as AD-JOINED to the main or correlative clause; see e.g. Andrews1975 (1985), Ken K. Hale 1975, Dasgupta 1980, Keenan 1985,Lehmann 1984, Srivastav 1988.

Based on a broad range of evidence, Hock (1989a) goes onestep further and argues that relative clauses are syntactically CON-JOINED to their correlative clauses. While some of that evidenceappears to be restricted to Vedic, other evidence is also found inpost-Vedic. The nature of that evidence is, I believe, such thatboth those working in formal syntax and those working on com-putational analyses will find it interesting and challenging.

First, in some cases there is no clear relationship between therelative pronoun (or phrase) of the relative clause and the correl-ative pronoun (or phrase) of the main clause; see (6), where an

4Speijer (1886: 349) hedges on this issue by stating that preposing of the rel-ative clause before the main clause is much more used than inserting the relativesentence in the main one.


example similar to (6a) was brought to my attention in 1989 byKiparsky 1995, and (6c) by James Fitzgerald (March 2006). Struc-tures of this sort are typically best rendered as conditionals.

(6) a. ya I +.nd ! ya;ta;ya;~tva;!a Bxa;ga;va;e!a yea ..ca tua;u ! ;vuaH ma;m!a I+.du ;g{a (rua;Da;!a h;va;m,aya indra yatayas tva bhr

gavo ye ca tus. t.uvuh. . mama d u-

gra srudh havam (R

V. 8.6.18)Which Yatis, O Indra, (have praised) you, and which Bhr

-

gus have praised you, powerful Indra, nevertheless hearMY call. Even if the Yatis and Bhr

gus have praised you, power-

ful Indra, nevertheless hear MY call.

b. ya;a;sa;Ma na;a;d;d;tea Zua;kM a;a;ta;ya;ea na .sa ;a;va;k +:yaHyasam nadadate sulkam jatayo na sa vikrayah. (Manu3.45)Of which (women) the relatives do not appropriate the(bride) price, that is not a sale. If the relatives do not appropriate . . .

c. ya;(a;na;ea;+:ea ;a;h ;a;na;deR ;ZaH ;///a;~:a;ya;a mEa;Tua;na;txa;a;yea ta;~ya;a;sma;a:=+ya;ta;ea v.ya;+:ma;Da;ma;eRa na;a:a .sMa;Za;yaH yas canokto hi nirdesah. striya maithunatr

ptaye;

tasyasmarayato vyaktam adharmo natra samsayah. -. (MBh. 12.258.38)Which instruction to gratify ones wife sexually is notheard, of him who does not remember (this) it is clearly,no doubt, a breach of duty.Though there is no requirement to satisfy ones wife sex-ually, if a man does not remember this it is clearly a seri-ous infraction.

Secondly, there are some examples in which the relative clauseexhibits properties normally only associated with independentmain clauses, namely interrogation and imperative modality (7).

12 H. H. HOCK

(7) a. Z!a;ya;Ra;ta;ea h va;!a IR +a;M!a;.ca;ke ya;//a;tk+:m!a;k+.=M t!a;sma;a;a;d;d;ma;!a;pa;d ;!a;a;tasa

ryato ha va

ks. amcakre [yat kim a

karam]RC [ta

smad i-

dam a

padi]CC + iti (SB. 4.1.5.4)

Saryata thought, Because I have done what?, therefore Ihave gotten into this. . . . What have I done to get into this?(Thus also SB. 1.7.3.19; a similar structure with kvawhere at SB. 5.1.3.13)

b. tya:jea :pra;a;a;a;Ea;va d;d;a;Ma k+:pa;ea;tMa .sa;Ea;}ya;ea h:a;yMa ;a;kM na .ja;a;na;a;a;sa . . . ya;Ta;a :*:e +:ZMa ma;a ku +.+:Svea;h . . . na;a;hM k+:pa;ea;ta;ma;pRa;a;ya;Syea k+:TMa;a;.ca;t,atyaje pran. an naiva dadyam kapotamsaumyo hy ayam kim na janasi . . . .[yatha klesam ma kurus. va + iha . . . ]RC[naham kapotam arpayis. ye kathamcit]CC (MBh. 3, App.21/5.82)I abandon my life, but I may not at all give the dove;for he is gentle, dont you know. . . ? So that dont youmake trouble here ! . . . , I will not hand over the dove inany way. . . . so that you dont make trouble here . . .

Most important, example (8) shows clearly that the relativeclause must be CONJOINED to the two main clauses. It is simplyimpossible for the same clause to be simultaneously ADJOINEDto two different clauses; and deriving the relative clause from anunderlying center-embedded postnominal position would be pre-posterous how can a single clause be simultaneously embeddedunder two different NPs, in two different clauses? In Hock 1989aI therefore propose to conceive of the relation between the relativeclause and the two correlative clauses as in Table 1. The formalismis, of course, antiquated, but the syntactic relation must be some-thing along these lines. (Davison 2009 proposes CP adjunction for


S S

S S

S

CP CC RP RC CP CC

Table 1Hocks (1989a) account for example (8)

structures in which the relative clause precedes, in contrast to IPadjunction, which stands for the traditional adjunct analysis.)

(8) .sa;a ;vEa ;dE! ;va;a va;a;gy!a;ya;a y!a;d;a;de ;va v!a;d;a;ta t!a:a;; +va;a;ta[sai vai dai

v vag]CC[ya

yai ya

d yad j eva vadati]RC

[ta

t tad j bhavati]CC (BAU. 1.3.27)Thati is divine Speech by whichi whatever j one speaks, that jcomes about. Whatever one speaks by means of divine Speech comesabout.

3.3 Some issues of agreement

Pan. ini addresses some issues of agreement, in two places. One isthe ekases.a sutras which address the issue of gender resolution un-der the specific circumstance of one word taking the place of twoconjoined ones (A. 1.2.6473); the other are the sutras governingperson agreement between surface subjects (kartr

s or karmans) and

the la-kara of the verb (1.4.104107). But many aspects of agree-ment are not covered, except perhaps implicitly under the notionof samanadhikaran. a coreference.

14 H. H. HOCK

In this section I map out some issues of Sanskrit agreementthat I believe should be of interest, especially to linguists workingon computational analyses of Sanskrit syntax.

A fairly straightforward issue is the question of gender agree-ment with mixed-gender conjunct antecedents, where two differ-ent strategies can be observed. One is agreement with the nearestconjunct, as in (9); the other is gender resolution as in (10).

(9) k+:a;////a;nta;ma;ta;a .=+a:$ya;a;ma;dM ma;ma ..ca .ja;a;a;va;ta;ma;pya;d;a;pra;Bxa;a;ta tva;d;Da;a;na;m,akantimat rajyam idam mama ca jvitam apy adyaprabhr

ti

tvadadhnam (Das. 135)Kantimat [f.sg.], and this kingdom [n.sg.], and also my life[n.sg.] [is] from today under your control [n.sg.].

(10) a. .sa;e!a Y;a:(;!a;na;Ea ..ca .s!a:=+~va;ta;Ma ..ca;e!a;pa;!a;Da;!a;va;.cCe +.pa;!a;na;e!a Y;/////////a;sma n!a;mua;.ca;yea. . . (I+.t,aI).. .tea Y;b.rua;va;n,a . . .so

svinau ca sa

rasvatm co

pa

dha

vac chepa

no

smi na

mu-caye . . . (iti).. te bruvan . . . (SB. 12.7.3.12)He (Indra) went to the Asvins [m.du.] and Sarasva-t [f.sg.], (saying) I have sworn to Namuci . . . They[m.pl.] said . . .

b. mxa;dM ga;Ma ;dE ;va;tMa ;a;va;prMa ;Gxa;tMa ma;Dua ..ca;tua;Spa;Ta;m,a :pra;d;a:a;a;a;a;na ku +:va;Ra;tamr

dam gam daivatam vipram ghr

tam madhu catus. patha-m. pradaks. in. ani kurvta (Manu 4.39)He should keep on his right a lump of earth, a cow, anidol, a brahmin, ghee, honey, and a crossroads [n.pl.].

As I show in Hock 2012a, Speijers analysis for post-VedicSanskrit gender resolution (1886: 1920), going back to Borooah(1879), best accounts for the Vedic evidence: In the case of mixed-gender antecedents that are entirely human (or animate), genderresolution is in favor of the masculine; in all other cases, includ-ing cases like (10b), where non-human/inanimate and human an-tecedents are mixed, the result is neuter, except that in Vedic texts


some inanimate, but sacred antecedents such as the sun, the earth,or the sky may be treated as animate/human.

In the case of nearest-conjunct agreement, there is the a pri-ori possibility that a modifier to the left may show agreement withthe left conjunct, and one to the right with the right conjunct; seeArnold, Sadler, and Villavicencio 2007 for Portuguese and John-son 2008 for Latin. As it turns out, an example of this mirror-image agreement can also be found in Sanskrit (11). It remainsto be seen whether this kind of agreement occurs more frequently,and whether it does so in post-Vedic.

(11) v.ya;a;m!a;ma;a:a;Ea :p!a:a;Ea ..ca :pua;CM ..ca Ba;va;a;tavyamamatrau paks. au ca pucham ca bhavati (TS. 5.2.5.1)the two wings [m.du.] and the tail [n.sg.] are (lit. is [sg.3])measuring-a-fathom [m.du.].

While with the exception of the mirror-image agreement, thephenomena discussed so far are rather mundane, another type ofagreement presents greater challenges. This is what may be calledupside-down agreement.

The best-known variety of this agreement is widespread inVedic prose, as in (12), but is also found in the later language. Thisis the fact that pronoun subjects normally adopt the agreement fea-tures of their predicates, rather than the other way around. As faras I can tell, this usage was first introduced into the discussionof Sanskrit syntax by Speijer (1886: 18). The feminine singularmarking on sa in example (12) shows that at least in Vedic prosethis pattern of agreement is clause-bound, and that structures ofthis kind do not exhibit cross-clausal anaphoric gender agreement(which would have required nominative masculine te).

(12) yea tua;Sa;aH .sa;a tva;g,aye tus. ah. sa tvag (AB. 1.22.14)

16 H. H. HOCK

What (masc.) are the shells (masc.) that (fem.) is the skin(fem.).

As it turns out, upside-down agreement must also be pos-tulated for locative (and genitive) absolute constructions, such as(13); see Hock 2009a.

(13) a. vxa;tea tua .nEa;Sa;Dea BEa;}ya;a l+.ea;k+:pa;a;l+.a . . . na;l+.a;ya;a;;Ea va:=+a;nd;du Hvr

te tu nais. adhe bhaimya lokapala . . . nalayas. t.au varandaduh. (Nala 5.33; MBh. 3.54.28)The Nis.adhan having been chosen by Bhaim, the worldrulers gave Nala eight boons.

b. .tea;Sua ga;.cC+.tsua va;yMa .~Ta;a;~ya;a;maHtes. u gacchatsu vayam sthasyamah.With them having gone, we will stay.

c. ga;nta;v.yea na ;a;.ca:=M .~Ta;a;tua;a;ma;h Za;k+.a;m,agantavye na ciram sthatum iha sakyam(MBh. 1.150.4, Speijer 1886: 286)5

As/since we have to go, it is not possible to stay here forlong.Lit. (It) having to be gone, it is not possible to stay herefor long.

d. ga;ntMua ;a;na;a;(a;ta;.cea;ta;a;sa ;a;pra;ya;ta;mea .sa;veRa .sa;mMa :pra;//////a;~Ta;ta;aga;nta;v.yea .sa;a;ta .ja;a;a;va;ta;a;pra;ya;sua;&+.tsa;a;TRaH ;a;k+:mua tya:$ya;tea gantum niscitacetasi priyatame sarve samam prasthitagantavye sati jvitapriyasuhr

t sarthah. kim u tyajyate.

(Subhas. itaratnakos. a 1151)Together all set out to go to the determined-minded dear-est one. (It) having to be gone, how is the dear friend ofones life, having the same goal, getting left behind?

5The Critical Edition instead has gantavyam na ciram sthatum iha sakyam(MBh. 1.142.21)


The nearest analogue for analyzing such constructions wouldbe that of nominalization, which embeds a subordinate propositioninto a matrix clause by means of a nominal form of the verb, whosecase marking signals the status of that proposition within the ma-trix clause. See Yoon 1996 for an analysis of such structures.

The major difference between ordinary nominalizations andstructures of the type (13) is that the latter involve an adjectivalform, rather than a purely nominal one, a form which thereforemust be supplied with gender and number features, in addition tothe locative that signals the function of the construction within thematrix clause. Note that in the synchronic grammar of Sanskritthe locative case has to be assigned to the participle, not to its un-derlying subject, because of the fact that locative participial casemarking is not restricted to structures in which the participle hasa subject to agree with such as (13ab), but is also found in imper-sonal, subject-less structures like (13cd).

Note further that under this analysis, the subject of the par-ticiple, if any, is not in a position governed by a verb that couldassign case to it; the only features that the syntax can assign to itare gender and number.

A possible way to account for the fact that the participle nev-ertheless gets gender and number features agreeing with its under-lying subject, and that the subject, in turn, receives case, lies inadopting the approach of the post-syntactic distributed morphol-ogy of Halle and Marantz 1993. As illustrated in Table 2, in thisanalysis the syntactic output only has the abstract features pluralmasculine for the underlying subject of the locative absolute, andlocative for the participle. The rest of the features need to be filledin by the Morphology. The gender and number features of theparticiple are filled in by normal agreement control, but the casefeature of the subject is supplied by upside-down agreement fromthe participle. (In impersonal structures like (13cd), the participlereceives the usual neuter singular default features.)

18 H. H. HOCK

Syntactic Output tad gacchat vayam sthasyamah.[pl.m.] [loc.]

Morphology: Input tad- gacchat- vayam sthasyamah.[pl.m.] [loc.]

Coreference tad- gacchat-su vayam sthasyamah.[pl.m.]


account of whose action (there is) qualification of (another)action.6

3.4 Converbs, reflexives, oblique subjects, and syntacticbracketing

In discussions of modern South Asian syntax, converbs (variouslyreferred to as absolutives, conjunctive participles, gerunds, and thelike), combined with reflexivization and word order, play a signif-icant role as criteria that determine whether non-nominative con-stituents can be considered to be subjects or not. See for instancethe various contributions to Verma and Mohanan 1990.

Of the three features converbs, reflexivization and word order,only the syntax of converbs is addressed in the Pan. inian tradition.The discussion in Speijers Syntax, however, suggests that reflex-ives exhibit a similar syntactic behavior to converbs (1886: 200and 297298). More comprehensive discussions, which includenot only converbs and reflexives, but also word order, are Hock1986, 1990, 1991a (with references). This section surveys the ma-jor issues and findings.

Pan. inis account for the syntax of the converb (ktva)7 is wellknown (15), and its provision that ktva requires identity of kartr

s,

i.e. underlying subjects, is well motivated. The dominant pat-tern, at least for post-mantra-Vedic,8 is that this provision holds

6A priori yasya could refer to the agent of the action bhava, or to the wordexpressing the action. The latter is the usual interpretation and is made explicitin the Kasika Vr

tti: yasya ca kriyaya kriyantaram laks. yate tato bhavavatah. sa-

ptam vibhaktir bhavati locative case is also (used) after a word characterizingan action (bhavavat) by whose action another action is characterized. Joshi andRoodbergen (1980: 8788) interpret bhava as state, distinguishing it from kriyaaction. However, Cardona (1976: 197) (w. ref.) notes that both terms are usedto refer to actions.

7Here as elsewhere ktva also stands for its replacement lyap.8For the mantras, Hock (1982b, 1986, 1987, 1990, 1991a) finds some (lim-

ited) evidence for surface subject, rather than underlying subject (kartr

) control

20 H. H. HOCK

not only for active structures, where underlying and surface sub-ject are identical, but also for passive or passive-like structures,where they are not. See Hock 1986 for discussion.

(15) .sa;ma;a;na;k+:tRxa;k+:ya;eaH :pUa;vRa;k+:a;lesamanakartr

kayoh. purvakale (A. 3.4.21)

(ktva) is introduced under the condition of identity of kartr

sin reference to prior time

The syntax of reflexives is not covered in the Pan. inian tradi-tion, and most western discussions focus on Vedic and/or its Indo-European origins; see Vine 1997; Hock 2006; Kulikov 2007 forrecent discussions. In his coverage of reflexives, Speijer (1886:200) notes similar conditions for the use of reflexives as for thatof converbs (1886: 29798), without however trying to link thetwo phenomena. In a series of papers (Hock 1982b, 1986, 1987,1990, 1991a), I have shown that, just like converbs, reflexives arecontrolled by kartr

s, i.e. underlying, rather than surface subjects.

Moreover, in the same publications I have shown that word order,too, is sensitive to the notion kartr

, rather than surface subject.

Examples for kartr

control of converbs and reflexivizationabound; see e.g. (16), (17), and (18) which focus on instrumental-marked kartr

s. Note especially (18), which has both converb and

reflexive control.

(16) ta;ta;~ta;ma;a;ya;a;ntMa dx ; ;a :pa;a:a;Za;a;va;kE +.=, . . .k+:ea;l+.a;h;lH kx +:taHtatas tam ayantam dr

s. t.va paks. isavakair . . . kolahalah. kr

tah.

(Hit. 1.4)Then, upon seeing him coming, the young birds made aracket.

of converbs in passives and passive-like structures, and somewhat more robustevidence as regards word order and reflexive control. Zakharyin (1998) questionsthis finding, but his discussion only focuses on converbs and does not address thebroader evidence of word order and reflexivization.


(17) .~vea;Sua .~Ta;a;nea;Sva;va;a;h;tEa;BRa;a;va;ta;v.yMa Ba;va;a;; Hsves. u sthanes. v avahitair bhavitavyam bhavadbhih. (Vikram.1, p. 2; Speijer 1886: 199)Your lordships must be attentive on your own seats.

(18) ;Ta .tea;na tMa Za:Mua ma;tva;a;tma;a;nMa ta;~ya;ea;pa;a:= :pra;a:a;pya :pra;a;a;aH :pa;a:=+tya;+:aHatha tena tam satrum matva + atmanam tasyopari praks. i-pya pran. ah. parityaktah.(Pac. 70;Speijer 1886: 297)Then he1, considering him2 an enemy, threw himself1 on topof him2 and gave up his1 ghost.

Although converb and reflexive control by the kartr

(whethernominative or instrumental) is the most common pattern in San-skrit, there are examples where other constituents or no con-stituents in the same clause seem to exert control. See theexamples in (19)(25) which focus on converb control, with theexception of (24c) which shows that genitive-marked NPs also cancontrol reflexives.

(19) a. ;lM ;a;va;Sa;a;de ;na ;a;ba;lM :pra;a;va;Zya va;sa;a;ma .sa;veRa ya;a;d .=+ea;.ca;tea vaHalam vis. adena bilam pravisya (.) vasama sarve yadi ro-cate vah. (Ram. 4.52.31)Enough of entering the cave in despondency. All of usare staying if it pleases you.

b. ;pra;a;pya na;d ;Ma :pa;vRa;taH ;//////a;~Ta;taHaprapya nadm parvatah. sthitah. (Kas. on A. 3.4.20)Not having reached the river (i.e. on this side of the river)stands the mountain.

(20) ;ea! ;Sa;Da;a:jRa;gDva;!a;p!aH :pa;a;tva;a t!a;ta O;;Sa .=! +saH .sM!a;Ba;va;a;tao

s. adhr jagdhva+ apa

h. ptva ta

ta es. a rasah. sa

mbhavati(SB. 1.3.1.24)

22 H. H. HOCK

(The animals/somebody) having eaten the plants, havingdrunk the waters, from that arises this essence.

(21) a. ;a;a;ta;Tye!a;na ;vE!a :de ;va;!a I+. ;!a ta;!a;ntsa;m!a;d;a;va;nd;t,aatithye

na vai

deva

is. t.va

ta

n t-sama

d avindat (SB. 3.4.2.1)The Gods having sacrificed with the guest-offering discord befell them.

b. tM!a ;hE ;nMa dx ; ;a Ba;!a;a;vRa;vea;dta

m hainam dr

s. t.va bhr viveda (SB. 11.6.1.7)

Having seen him (i.e. someone else), fear befell him.

(22) a. (rua;tva;a ;//a;tva;d;mua;pa;a;K.ya;a;na;m,a . . .;nya;a .=+ea;.ca;tea [ta;smEa]srutva tv idam upakhyanam . . . anyan na rocate [tasmai](MBh. 1.2.236)(He) having heard this story, another (story) does notplease him/he does not like another (story).

b. ;a;d :ja ;a;sa;pra;a;na;d ;Ma ga;tva;a tua;Bya;ma;hM ma:n:Ma d;a;~ya;a;a;madvija sipranadm gatva tubhyam aham mantram dasyami(Vetalapacavimsati, ed. Emeneau 92.2021)O brahmin, I will give a mantra to you, (you) having goneto the river Sipranad.

(23) :pa;(a;a;dE ! :pa:= +!a;tya vx!a;Sa;a ya;e!a;Sa;a;m!a;a;Da;d +va;a;ta:pa;(a;!a;de ;vE!a;na;a;mea;t!a;tpa:= +!a;tya vx!a;Sa;a . . . (! );a;Da;d +a;va;ya;a;tapascad vai

par

tya vr

s. a yo

s. am adhidravati

pasca

d evainam eta

t par

tya vr

s. n. a . . . (a

)dhidravayati (SB.1.9.2.24)The bull mates with the female approaching her from be-hind. He makes the bull mate with her, (the bull) havingapproached her from behind.

(24) a. h;tva;a vxa:Ma ;a;va;a:ja;tya yua;Sma;a;a;Ba;meRa Y;yMa .sa;h .sa;ea;ma;pa;a;TaH(h)atva vr

tram vijitya yus. mabhir me yam saha somap-


tha(h. ) (KB. 15.2)Having slain Vr

tra, having conquered, this soma-

drinking with you is mine.

b. .sua;=+:a;a;tMa :pua:+:SMa dx ; ;a .~:a;a;a;Ma ;a;*:+:d;a;////a;nta ya;ea;na;yaHsusnatam purus. am dr

s. t.va strn. am klidyanti yonayah.

(Vetalapacavimsatika, ed. Uhle 15.3738)Having seen a well-bathed/graduated man, womensvaginas get wet.

c. .sa;a ;a;h .~va;a .=+a:ja;Da;a;na;asa hi sva rajadhan (Kathas. 39.163)for this is my (= the speakers) royal city

(25) I+.tyea;va k+:a;le Zyea;nea;na;a;na;a;ya Ka;a;d;a;ma;a;na;~ya .sa;pRa;~ya ga:=+lM ta;" +v.yea ;a;na;pa-;a;ta;ta;m,aity eva kale syenena + anya khadyamanasya sarpasya ga-ralam taddravye nipatitam(Vetalapacavimsati, ed. Emeneau 76.1113)At that very time, the venom of a snake being eaten by ahawk, (the hawk) having carried it off, fell into his food.

Faced with such a variety of different structures, some scholarsmay opt for claiming that there is no purely syntactic criterion forcontrol and that any element that is somehow salient may serve ascontroller. This is close to what Zakharyin (1998) proposes.

A heuristically and theoretically more interesting position is totry to determine whether some or all of these exceptional struc-tures can be accounted for by additional generalizations.

This is, of course, what Pan. ini has done for structures like(19a) by means of A. 3.4.18 (see also A. 3.4.19) and for (19b) byA. 3.4.20 with A. 3.4.21 taking care of elsewhere. In bothcases we are probably dealing with some kind of grammaticaliza-tion. The one in (19b) could be compared to later grammaticaliza-tions such as adhikr

tya about, adaya with, arabhya (starting)

24 H. H. HOCK

from which likewise do not seem to be sensitive to control by anyparticular constituent.

As far as (20) and (21a) are concerned, these seem to be pe-culiarities of Vedic prose; see Delbrck 1888: 408. Hock (1987)accounts for them under the notion reduced-clause structures, aphenomenon not limited to converbs but also found with partici-ples.

Example (21a), however, could also be analyzed as compara-ble to (21b) which, together with (22a), could be and has been taken as equivalent to Modern Indo-Aryan oblique-experiencer-subject constructions; see Hook 1976, 1984 for (22a), and Za-kharyin 1998 more generally.

Structures of this sort, however, are extremely rare and, in theaggregate, no more frequent than structures like (22b) in which anon-experiencer indirect object controls the converb, or (23) wherethe converbs dependence on the causee of adhidravayati may bedittological from the preceding non-causative construction with a-dhidravati. At any rate, all of these structures are quite rare andcan be dismissed as occasional examples of loose (or sloppy)control.9

The examples in (24), by contrast, exemplify a much morecommon pattern: control by genitive-marked possessor NPs. Thefact that examples of control by non-kartr

s (in Pan. inis sense) are

especially common with genitive-marked NPs was already notedby Speijer (1886: 298) who considered these NPs to be exemplarsof his dative-like genitive category. Focusing on converb and re-flexive control, as well as word order, I have argued (Hock 1990,1991a) that Possessor NPs must be recognized as a highly produc-

9Interestingly, if structures like (21b) and (22a) were to be analyzed asoblique-experiencer-subject constructions or as forerunners of such construc-tions accusative-marked experiencers would seem to occur more frequentlythan dative-marked ones. In Modern Indo-Aryan, it is dative-marking which pre-vails. On the syntax of ruc see also Cardona 1990; Deshpande 1990.


tive alternative to kartr

control. (In his very different approach tothe syntax of R

gvedic reflexive sva, Vine (1997) similarly finds

that genitive-marked NPs are the most common alternative to sub-ject NP controllers.)

Finally, example (25) shows the need for being sensitive tosyntactic bracketing. The converb anya is controlled by the kar-tr

(syena) of the participial structure headed by khadyamana, notby the kartr

(garala) of the matrix-clause verb nipatitam. See the

bracketing in (26). For further discussion see Hock 1986, 1987.10

(26) I+.tyea;va k+:a;leZyea;nea;na;a;na;a;ya Ka;a;d;a;ma;a;na;~ya.sa;pRa;~ya ga:=+lM ta;" +v.yea ;a;na;pa;a;ta;ta;m,a[ity eva kale[syenena + anya khadyamanasya]sarpasya garalam taddravye nipatitam](Vetalapacavimsati, ed. Emeneau 76.1113)At that very time, the venom of a snake being eaten by ahawk, (the hawk) having carried it off, fell into his food.

While this phenomenon is not overtly addressed in the Pan. i-nian tradition, there is nothing in that tradition that would preventit. Given that their suffixes replace la-karas (A. 3.2.124, 3.4.7071), participles are allowed to have their own kartr

s, and these kar-

tr

s can control converbs (and reflexives) in their own domain.Complications do however arise because participial structures

are normally integrated into their matrix clauses without being setoff by clear boundaries. As a consequence, in very similar struc-tures, such as (27ab), both involving the participle form gacchan,

10Speijer (1886: 29798) comes close to realizing the need for some kindof bracketing by noting that [apparent] control by locative NPs is common inlocative absolute constructions. Vine (1997) similarly weighs the possibility thatsome instances of apparent non-subject control of reflexives may be accountedfor by something like bracketing.

26 H. H. HOCK

it may be either the kartr

of the entire sentence or that of the par-ticipial structure that controls the converb. In fact, as (27c) shows,it is possible for one converb to be controlled by one kartr

, another

by the other.

(27) a. .sa ya;a;a;TeRa . . .C+.a;ga;mua;pa;k +:a;ya . . . ga;.cC+.n,a;DUa;tRa:a;yea;a;a;va;l+.ea;a;k+:taH[sa yajarthe . . . [chagam upakrya . . . gacchan]dhurtatrayen. a + avalokitah. ](Hitopadesa 43.56)He, having bought a goat for the purpose of sacrifice . . . ,(as he was) going was noticed by a trio of rogues.

b. tMa :pua:Ma d;ZRa;a;ya;tva;a;nea;na ga;.cC+*+;f;a;Da:=H .sa;ma;a;na;a;taH[tam putram darsayitva + anena gaccha jat.adharah. sa-mantah. ](Vetalapacavimsati, ed. Emeneau 28. 56)Having showed that boy to him he brought (back) themendicant (as he was) going.

c. ta;a;///a;nva;a:ja;tya ya;Ta;a;l+.ea;k+:ma;a;sa;a;na;a;a;na;nd O;;tya;a;b.ra;va;a;t,a[[tan vijitya yathalokam asnan]indra etya + abravt] (JB 1.156)Indra, having come up, said to them, (who were) sittingaccording to their own worlds, having won.

3.5 Double Direct Object constructions and Causatives

In Pan. inis sutras defining karman (28), A. 1.4.51 has met withconsiderable problems of interpretation. The commentatorial tra-dition agrees that it is intended to cover double-direct-object con-structions such as (29), but how it does so does not seem to havereceived a satisfactory explanation; see Deshpande 1987. The ev-idence of the textual tradition makes it clear that either of the twocomplements in these structures behaves like a true direct object,


being promotable to surface subject if it occurs by itself (30).However, if both complements are present, only the more agen-tive one can be promoted; (31). See Hock 1985, 2012c for furtherdiscussion; see also Ostler 1979; Van de Walle 1992.

(28) a. k+:tRua:= +a;/////a;psa;ta;ta;mMa k+:mRa ta;Ta;a;yua;M ..ca;a;na;a;/////a;psa;ta;m,akartur psitatamam karma. tathayuktam canpsitam (A.1.4.4950) That which is most desired by the agent is ka-rman, and also that which is not desired (but) linked (tothe action) in the same way;

b. ;k+:a;Ta;tMa ..caakathitam ca (A. 1.4.51)also what (is linked in the same way and) has not been asyet specified; (?)

c. ga;a;ta;bua;a:;d ;pra;tya;va;sa;a;na;a;TRa;Za;b.d;k+:ma;Ra;k+:mRa;k+:a;a;a;ma;a;a;k+:ta;Ra .sa .a;Ea &+.k +:ea:=+nya;ta:=+~ya;m,agatibuddhipratyavasanarthasabdakarmakarmakan. am a-n. ikarta sa n. au. hr

kror anyatarasyam (A. 1.4.5253)

also the non-causative agent in the causative of rootsmeaning go, understand, consume, communicate,(and) intransitives, and (optionally) of hr

and kr

.

(29) a. k+:du b.ra;vaH . . . nXa;n,akad u bravah. . . . nr

n (R

V. 10.10.6) (SPEAK)

What will you say to the men?

b. ta:va;a ya;a;a;ma . . .tat tva yami . . . (R

V. 1.24.11a) (ASK/ENTREAT)

. . . that I request from you.

c. du ;du ;h e! :pa;yaH . . .;a;Sa;m,aduduhre payah. . . . r

s. im (R

V. 9.54.1) (MILK)

They milked the milk (from) the sage.

28 H. H. HOCK

d. :de! ;va;a;na;sua-.=+aH y!a;a;ma-.ja;ya>+.s,adevan asurah. yajam ajayams (MS. 1.9.8) (WIN)The asuras won the sacrifice (from) the gods.

e. ya;d;mua;Sa;a;ta . . . :p!a;a;Ma ga;aHyad amus. n. ta . . . pan. m gah. (R

V. 1.93.4) (ROB)

. . . when you robbed the cows (from) the miser.

f. ta;a;nsa;h;~:Ma d;q+.yea;t,atan sahasram dan. d. ayet (Manu 9.234) (PUNISH). . . he should fine/punish them (with) a thousand.

(30) a. ().~ya va;a;gua;a;d;ta;a Ba;va;a;ta(a)sya vag udita bhavati (AB. 1.6.12) (SPEAK)His speech is spoken..sa :he ! ;nd e +a;ea;! ;a;sasa ha + i

ndren. a + ukta

asa (SB. 14.1.1.19)He was addressed by Indra.

b. .=! +a;ya;a;vRa;BUa;a;ta:= +a;ya;tea . . .rayr vbhutir yate . . . (R

V. 6.21.1) (ASK/ENTREAT)

Great wealth is implored..=+a:ja;a me!a;Da;a;a;Ba:=, IR +.ya;tear aja medhabhir yate (R

V. 9.65.16)

The king is implored with insight.

c. . . . ;a;pa;ba;tua du ! ;gDa;mM!a;Zua;m,a. . . pibatu dugdham amsum (R

V. 5.36.1) (MILK)

May he drink the milked (= expressed) soma.11

du ! ;h:a;ntea . . . ;De!a;na;va;ea11Hettrich (1994) cites dugdho amsuh. (RV 3.36.6d), glossed as der ausge-

molkene Stengel, as an example of the source, rather than the substance NPbecoming the passive subject. However, the present example suggests that amsuhas become simply an epithet of soma, the ingestible substance produced in theritual.


duhyante . . . dhenavo (AV. 7.73.2)The cows are being milked.

d. .~va;mR!a:+:tva;ta;a ;a:j!a;ta;m,asva`r marutvata jitam (R

V. 8.76.4) (WIN)

The sun has been conquered by (Indra) accompanied bythe Maruts.;a;su!a:= +a yu!a;Da;a ;a:j!a;ta;aasur yudha jit a (AV. 1.24.1)The asura woman, defeated in battle.

e. ne!a;de ;v!a na;+;a! I+.va mua;a;Sa;t!a I+.va Z!a;ya;a;ta;a I! +.ty,a . . .ne

d v eva

nagna

iva mus. itaiva sa

yata i

ty . . . (ROB)

(SB. 1.2.2.16). . . lest he lie naked as it were, robbed as it were.

(31) a. ;a;va:ja;ya;mua;+:~tEaHvijayam uktas taih. (Kathas. 18.247) (SPEAK). . . (was) told (about) the victory by them.

b. v!a;Za;a;a;ma;nd e +a ya;a;a;.c!a;taHvas am ndren. a yacitah. (AV. 12.4.50) (ASK/ENTREAT). . . asked by Indra for (his) cow

c. . . . na;Ba;ea du ;h:a;tea ;Gx!a;tMa :pa;yaH. . . nabho duhyate ghr

tam paya(h. ) (R

V. 9.74.4) (MILK)

The cloud is milked for ghee, milk.12

d. . . . .sa;vRa:$ya;a;a;nMa va;a .ja;a;ya;tea. . . sarvajyanim va jyate (KS. 29.6) (WIN). . . or he is defeated a complete defeat.

12Hettrich notes that the example is formally ambiguous, since both nabhah.and ghr

tam payah. can be both nominative and accusative. The singular on the

verb and the initial placement of nabhah. favor an interpretation that nabhah. is thesubject; but agreement with the nearest antecedent of the conjoined elementsghr

tam and payah. is a possible alternative. Geldner takes ghr

tam and payah. to

be the subject.

30 H. H. HOCK

e. ;a;h! ;mea;va :p!a;a;Ra mua;a;S!a;ta;a va;na;a;a;nahimeva parn. a mus. it a vanani (R

V. 10.68.10) (ROB)

. . . like trees robbed of their leaves by winter13

While this much is known, questions remain. First, it is notclear why only certain verbs that are subcategorized for two com-plements have double-direct-object constructions. Even more puz-zling is why some verbs belonging to the semantic sets SPEAK,ASK/ENTREAT, MILD, WIN, ROB fail to enter into double-direct-object constructions. Consider kath which to my knowledge onlytakes the addressee in the dative, or hr

take away which takes the

ablative for the source person. Presumably, the verbs participat-ing in the double-direct-object construction must be specificallylisted in the lexicon (together with alternative case markings, ifany; see Hock 1985. But this does not explain why many of theverb classes exhibit similar behavior in other Indo-European lan-guages; see Hock 2012c.

Problems of a different sort arise regarding A. 1.4.5253 whichclassifies the causees of certain verb classes as karman (optionallyfor hr

and kr

) and leaves others as kartr

s which, being anabhi-

hita, surface in the instrumental. As Speijer (1886: 3637 withreference) notes, a very different situation obtains in the classicallanguage, irrespective of verb class:

If one wants to say he causes me to do something, it isby his impulse I act, there is room for the [accusativecausee], but if it be meant he gets something done by

13As noted by Hettrich, formally this passage is ambiguous. However, thecontext favors the interpretation given here: himeva parn. a mus. it a vanani br

-

haspatinakr

payad valo gah. Like the trees robbed of their leaves by winter, Valamourned for the cows (taken from him) by Br

haspati. (Geldner: Wie die Bume

ihre vom Frost geraubten Bltter so vermite Vala die von Br

haspati (geraubten)Khe.)


me, I am only the agent or instrument through whichhe acts, the instrumental is in its place . . .

As shown in Hock 1981, this pragmatically sensitive marking con-vention is already found in the early Vedic prose texts; see (32)and (33), where the verb in (32) belongs to the categories of verbsthat by A. 1.4.52 should take a karman, and the verb in (33) doesnot. The phenomenon can therefore not be attributed to post-Pa-n. inian innovation. Rather, the difference between Pan. ini and thetextual tradition of Madhyadesa most likely reflects a difference inregional dialect. See Hock 1981, 2012b,c, as well as Deshpande(1983)s pioneering paper, Pan. ini as a frontier grammarian.

(32) a. d;a;a;va;pxa;a;T!a;va;a Bua;va;nea;Su!a ;a;pRa;teadyavapr

thiv bhuvanes. u arpite (TS. 4.7.13.2)

Heaven and

Sanskrit Syntax

Documents

resemblances of sanskrit

seminaron sanskrit syntax

sanskrit syntaxselected

sanskrit grammatical

learned sanskrit scholars

speech units

comparative linguistics

thestudy of syntax