Sanskrit Syntax

Sanskrit syntax
Selected papers presented at the seminar

on Sanskrit syntax and discourse structures,
13-15 June 2013, Universit Paris Diderot,
with an updated and revised bibliography
by
H ANS H ENRICH H OCK
edited by
P ETER M. S CHARF
28 February 2015
ii
Scharf, Peter M., ed. Sanskrit syntax: Selected papers presented
at the seminar on Sanskrit syntax and discourse structures, 13-15
June 2013, Universit Paris Diderot, with an updated and revised
bibliography by Hans Henrich Hock. Providence: The Sanskrit
Library, 2015.
Copyright 2015 by The Sanskrit Library.
All rights reserved. Reproduction in any medium is restricted. No
part of this publication may be reproduced or transmitted, except
brief quotations, in any form or by any means, electronic or mechanical, including photocopying, recording, or any information
storage or retrieval system, without prior written permission of the
copyright holder and publisher.
Published by:
The Sanskrit Library
89 Cole Avenue
Providence, RI 02906
USA
sanskritlibrary.org
Library of Congress Control Number: 2015934847
ISBN-10: 1943135002
ISBN-13: 978-1-943135-00-4
ISBN 978-1-943135-00-4
90000
9 781943 135004
Preface
A strong tradition of linguistics developed in India in the first
millenium BCE naturally associated with the heightened awareness of language engendered by the assiduous preservation of oral
Vedic texts. The curiousity as well as the need to understand the
language of these compositions already several hundred years old
instigated the development of systematic linguistic analysis which
flourished throughout the two and a half millennia since Pan.ini
composed his comprehensive linguistic description of the Sanskrit
language by the fourth century BCE. His unprecedented analysis
of the language into basic units and reconstitution of utterances
in accordance with precise rules laid the foundation for the development of highly sophisticated discussions concerned with the
structure of verbal cognition and its relation to speech units ranging from roots and affixes to words, phrases, and sentences. The
study of syntax in India is intimately associated with semantics,
and the analysis of the semantic content provides the foundation
for the generation of linguistic expressions. Relational structures
in the domain of consciousness are projected onto speech forms
whose arrangement in the string of speech is subordinate.
Modern linguistics developed in Europe beginning in the late
eighteenth century as a direct result of the fascination European
scholars had with the resemblances of Sanskrit to classical Greek
and Latin. Their excitement to discover the relationship among
these and a gradually expanding number of languages in what
came to be known as the Indo-European family led them to readapt
the precise phonetic rules of Indian linguists to historical sound
change 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
P REFACE
Sanskrit, and some worked directly on Sanskrit grammatical texts

or with learned Sanskrit scholars (pan.d.ita) intimately familiar with
such texts. Even Chomskys generative analysis of English was inspired by Pan.inian methods.
Formal and computational linguistics, engendered in the U.S.,
was dominated by English at its inception and developed in subsequent decades primarily in the environment of European languages. More recently there has been a concerted effort to undertake formal linguistic analysis of a wide variety of languages, with
particular interest in those with dramatically different features, and
to enrich syntactic theory to account for linguistic variety. Over the
past couple of decades there has been a growing interest among a
diverse group including Sanskrit scholars and computer scientists
who collaborated together to form the Sanskrit Computational Linguistics Consortium. The Consortium has sponsored several symposia in Europe, the U.S. and India since 2007.
Against this background, I was inspired to undertake research
to develop universally adequate linguistic theory by formalizing
the sophisticated linguistic theories, structures and procedures developed in the Indian linguistic tradition. To do so, for the year
20122013, I was awarded a Chaire Internationale de Recherche
Blaise Pascal, financed by the State of France and the Rgion
dIle-de-France and managed by the Fondation de lcole Normale Suprieure. The project investigated ways in which Indian
linguistics could contribute useful insights to contemporary formal linguistics, and designed ways in which Indian linguistic theories could be formalized and implemented computationally. The
project focused on Indian semantic and syntactic theory and the
semantics-syntax interface where computational linguistic work is
flourishing. In the midst of the project I was invited as a Visiting Professor to the Indian Institute of Technology Bombay. The
invitation facilitated collaboration sponsored by the Blaise Pascal
Chair with several junior Indian scholars. The project culminated
S ANSKRIT SYNTAX
in the organization of the Seminar on Sanskrit syntax and discourse

structures held at the Universit Paris Diderot, 1315 June 2013
the program of which is posted on the Sanskrit Library Website
(www.sanskritlibrary.org) under Events. The theme of
the seminar was the exposition of current theoretical knowledge
concerning Sanskrit syntax and the application of state of the art
methods of computational linguistics to Sanskrit. This book is the
publication of selected revised papers presented at that seminar.
The book includes twelve papers by seventeen contributors
from divergent backgrounds in European and American linguistics, Pan.inian grammar, and computer science that converge in
dealing with contemporary issues in Sanskrit syntax. Hans Henrich Hock provides a general survey of research on Sanskrit syntax since the publication of the collection on Sanskrit syntax he
edited in 1991 (Hock 1991). He brings to our attention challenging formal issues concerning word order, oblique subject agreement, clause coordination, and ditransitivity as well as contrasting
discourse structures in different genres. Along with his presentation, he provided a bibliography of research done in the twentyfive years since the publication of his and Madhav Deshpandes
combined bibliographies on Sanskrit syntax (Deshpande and Hock
1991). With the gracious consent of the authors and Motilal Banarsidass I have revised and integrated their earlier bibliography with
Hocks update in a seventy-two page comprehensive bibliography
of Sanskrit syntax containing nearly a thousand entries. Supplemented by the author and title indices at the back of this volume,
Hocks survey of research should provide useful and convenient
access to research on Sanskrit syntax generally.
George Cardona contributes two papers to the volume. In the
first, he deals with how expectancy, ellipsis and suppletion present
in ordinary language use of Sanskrit are formalized in interpretive
principles (paribhas.a s) in the metalanguage of Pan.inis As..ta dhyay. Here he also reiterates the semantic foundation of Pan.inian
vi
P REFACE
derivational procedure. In the second, he deals with the syntax

of the extension and removal of properties, and comparison with
respect to shared actions or properties. In that connection, he discusses the two rules that account for the use of the affix vat and
provides a categorization of which sutras of the As..ta dhyay conform to the syntactic pattern described by which rule.
The next paper, by Scharf, Goyal, Ajotikar and Ajotikar, describes a computational implementation of Pan.inis rules that de
termine the use of Atmanepada
and Parasmaipada verbal terminations under specific semantic and cooccurrence conditions. The
paper necessarily touches upon the different categorization of conditions for voice in the Pan.inian and European traditions. Aussant
examines the presuppositions of European and Indian approaches
to word classification generally. She traces contemporary schemes
of tagging parts of speech in computational linguistics back to categories described by Dionysius Thrax, and compares them with
Indian classification schemes traced back to Pan.ini and Yaska.
Joshi compares the concepts of concord and government in European grammar with the concepts of coreferrentiality (samanadhikaran.ya) and unexpressed karaka roles (anabhihitatva) in Pan.inian grammar.
Several papers deal with the topic of word order and how
free it is. Scharfs paper, "Interrogatives," shows that interrogatives locate in positions of focus and topic rather than moving
from a position designated by a predetermined principle of wordordering. Gillon examines various types of ellipsis and their relation to constituent units and the argument structure of lexical
items. Amba Kulkarni, Shukla, Satuluri, and Shukl examine proximity constraints on sentences analyzed in dependency structures
to precisely determine the degree of freedom in Sanskrit word order. They find that dislocations in prose are limited to cases of
unilateral expectancy while in verse dislocations involving mutual
expectancy also occur. Scharf, Ajotikar, Savardekar, and Goyal
S ANSKRIT SYNTAX
vii
further examine differences between prose and poetic syntax computationally with significant results. Melnad, Goyal, and Scharf
describe software they developed to identify metrical patterns. Finally, Katira and Malhar Kulkarni, examine syntactic violations in
sentences identified as erroneous by Charudeva Shastri and present
parse trees for them.
As the research in the area of Sanskrit syntax continues to
flourish, we plan to furnish bibliographic updates regularly. We
therefore invite scholars in the field to inform us of current publications as well as items missing from the bibliography supplied in
this volume. Please send complete citations to Hans Henrich Hock
and to me at our email addresses provided in the list of contributors.
Let me close by mentioning two practical points for readers.
(1) While the reference lists for most contributions cover citations in that contribution completely, since most of the citations
in Hocks survey of research are to references in the Sanskrit syntax bibliography, only references not included there are contained
in the reference list at the end of his contribution. (2) Generally
accents in Vedic passages in Devanagar are shown using marks
proper to the particular Vedic school while accents in Romanization mark the underlying udatta with an acute accent mark () and
independent svarita with a grave accent mark (). Cardona marks
accents in the Romanization of words in derivation likewise but in
finished forms borrows the marks used in Indic scripts for the common system of accentuation described in the Rkpratisa khya. Hock
to mark accents in
similarly borrows marks used in Indic scripts
the Romanization of Satapathabr

ahman.a passages in accordance
with their description in the Bhas.ikasutras.
Peter M. Scharf
28 February 2015
viii
P REFACE
Contributors
A NUJA A JOTIKAR
Laboratoire dHistoire des Thories Linguistiques, Universit
Paris Diderot, and Indian Institute of Technology Bombay
[email protected]
TANUJA A JOTIKAR
Laboratoire dHistoire des Thories Linguistiques, Universit
Paris Diderot, and Indian Institute of Technology Bombay
[email protected]
MILIE AUSSANT
CNRS, Laboratoire dHistoire des Thories Linguistiques, Universit Paris Diderot, Sorbonne Paris Cit, France
[email protected]
G EORGE C ARDONA
Department of Linguistics, University of Pennsylvania, Philadelphia
[email protected]
B RENDAN S. G ILLON
Department of Linguistics, McGill University, Montreal
[email protected]
PAWAN G OYAL
Department of Computer Science, Indian Institute of Technology,
Kharagpur
[email protected]
ix
C ONTRIBUTORS

University of Illinois, Urbana-Champaign
[email protected]
P RASAD P. J OSHI
Deccan College, Pune
[email protected]
D IPESH K ATIRA
Shree Somnath Sanskrit University, Veraval
[email protected]
A MBA K ULKARNI
Department of Sanskrit Studies, University of Hyderabad, India
[email protected]
M ALHAR K ULKARNI
Indian Institute of Technology Bombay
[email protected]
K ESHAV S. M ELNAD
[email protected]
PAVANKUMAR S ATULURI
[email protected]
S AMPADA S AVARDEKAR
[email protected]
S ANSKRIT SYNTAX
xi
P ETER M. S CHARF
The Sanskrit Library, Laboratoire dHistoire des Thories Linguistiques, Universit Paris Diderot, and the Indian Institute of Technology Bombay
[email protected]
D EVANAND S HUKL
Maharshi Sandipani Rashtriya Vedavidya Pratishthan, Ujjain, India
[email protected]
P REETI S HUKLA
[email protected]
xii
C ONTRIBUTORS
Table of contents
Preface
iii
Contributors
ix
Some issues in Sanskrit syntax

Derivation and interpretation in Pan.inis system

G EORGE C ARDONA
53
Extension rules and the syntax of As..ta dhyay sutras

with -vati
G EORGE C ARDONA
109
Voice, preverb, and transitivity restrictions

in Sanskrit verb use
P ETER S CHARF, PAWAN G OYAL,
A NUJA A JOTIKAR, and TANUJA A JOTIKAR
157
Interrogatives and word-order in Sanskrit

P ETER M. S CHARF
203
To classify words: European and Indian

grammatical approaches
MILIE AUSSANT
219
Constituency and cotextual dependence

in Classical Sanskrit
B RENDAN S. G ILLON
237
xiii
xiv
S ANSKRIT SYNTAX
How free is free word order in Sanskrit?

A MBA K ULKARNI, P REETI S HUKLA, PAVANKUMAR
S ATULURI and D EVANAND S HUKL
269
Distinctive features of poetic syntax: preliminary results
P ETER S CHARF, A NUJA A JOTIKAR,
S AMPADA S AVARDEKAR, and PAWAN G OYAL
305
Meter identification of Sanskrit verse
K ESHAV M ELNAD, PAWAN G OYAL and P ETER S CHARF 325
On concord and government relations in Sanskrit
P RASAD P. J OSHI
347
Parse trees for erroneous sentences

D IPESH K ATIRA and M ALHAR K ULKARNI
361
A bibliography of Sanskrit syntax
399
As..ta dhyay sutra

index
471
Author index
477
Title index
487
Detailed table of contents

Preface
Contributors
Table of contents . . . .
Detailed table of contents
List of tables . . . . . . .
List of figures . . . . . .
iii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
Introduction . . . . . . . . . . . . . . . . . . . .
2
A brief survey of recent publications . . . . . . .
3
Formal issues . . . . . . . . . . . . . . . . . . .
3.1
Free Word Order and related issues . . .
3.2
Relative-correlatives . . . . . . . . . . .
3.3
Some issues of agreement . . . . . . . .
3.4
Converbs, reflexives, oblique subjects, and
syntactic bracketing . . . . . . . . . . .
3.5
Double Direct Object constructions and
Causatives . . . . . . . . . . . . . . . .
3.6
Asamartha compounding . . . . . . . . .
4
Functional issues that should be of interest to computational approaches . . . . . . . . . . . . . . .
4.1
Fronting . . . . . . . . . . . . . . . . . .
4.1.1
Initial strings in Vedic . . . . .
4.1.2
Predicate-Subject order . . . .
4.1.3
Narrative linkage and related issues . . . . . . . . . . . . . .
xv
.
.
.
.
ix
xiii
xv
xxv
xxvii
1
. 2
. 2
. 4
. 4
. 9
. 13
. 19
. 26
. 32
.
.
.
.
35
36
36
37
. 40
xvi
S ANSKRIT SYNTAX
4.2
Extraposition . . . . . . . . . . . . . . .
4.2.1
Gondas amplified sentences .
4.2.2
Purpose datives in Vedic prose
4.2.3
Kartr backing and extraposition,
and
politeness . . . . . . . . .
5
Conclusions and implications for further research
References . . . . . . . . . . . . . . . . . . . . . . . .
. 43
. 43
. 44
. 45
. 46
. 49

G EORGE C ARDONA
53
1
Speakers and listeners . . . . . . . . . . . . . . . . 53
2
Derivation and a speakers perspective . . . . . . . 56
2.1
Katyayana on speech being prompted by
meaning . . . . . . . . . . . . . . . . . . . 56
2.2
Patajali on speech being caused by meaning 57
2.2.1
Number distinction . . . . . . . 57
2.2.2
Residual relations . . . . . . . . 61
2.2.3
Gender distinction . . . . . . . . 67
2.3
Meaning conditions and derivation . . . . . 69
3
Affixation meaning conditions . . . . . . . . . . . 71
4
A different approach suggested . . . . . . . . . . . 73
5
Interpreting Pan.inian sutras . . . . . . . . . . . . . 88
5.1
Introduction . . . . . . . . . . . . . . . . . 88
5.2
Contexts requiring metarules . . . . . . . . 89
5.2.1
:Sa; +a .~Ta;a;nea;ya;ea;ga;a (A. 1.1.49) . . . . 91
5.2.2
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) . . . . 94
5.2.2.1
Purpose of the rules. . . 94
5.2.2.2
Scope of 1.1.66. . . . . 96
5.2.2.3
Scope of 1.1.67. . . . . 99
6
Summary . . . . . . . . . . . . . . . . . . . . . . 100
References . . . . . . . . . . . . . . . . . . . . . . . . . 103
TABLE OF CONTENTS
xvii
Extension rules and the syntax of As..ta dhyay sutras

with -vati
G EORGE C ARDONA
109
1
Introduction . . . . . . . . . . . . . . . . . . . . . 109
2
Varieties of extension . . . . . . . . . . . . . . . . 111
2.1
Karyatidesa . . . . . . . . . . . . . . . . . 111
astratidesa . . . . . . . . . . . . . . . . . 113
2.2
S
2.3
Rupatidesa . . . . . . . . . . . . . . . . . 117
2.4
Tadatmyatidesa . . . . . . . . . . . . . . . 117
2.5
Nimittatidesa . . . . . . . . . . . . . . . . 119
2.6
Vyapadesa tidesa . . . . . . . . . . . . . . 120
2.7
Arthatidesa . . . . . . . . . . . . . . . . . 122
3
Extensions of the type X is Y . . . . . . . . . . . 123
4
Syntactic structures . . . . . . . . . . . . . . . . . 127
4.1
Predication without comparison . . . . . . 127
4.2
Comparisons . . . . . . . . . . . . . . . . 127
4.2.1
Comparison involving a common action . . . . . . . . . . . . 128
4.2.2
Other comparisons . . . . . . . . 131
4.3
Summary of syntactic structures . . . . . . 133
5
Semantic and pragmatic considerations . . . . . . . 136
6
Summary . . . . . . . . . . . . . . . . . . . . . . 140
7
As..ta dhyay sutras in which terms with -vati are used 142
7.1
List indicating extension type . . . . . . . 142
7.2
Syntax of sutras with terms in -vati . . . . 146
7.2.1
Schema I (4.3) . . . . . . . . . 146
7.2.2
Schema II . . . . . . . . . . . . 148
References . . . . . . . . . . . . . . . . . . . . . . . . . 151
in Sanskrit verb use
xviii
S ANSKRIT SYNTAX

1
Introduction . . . . . . . . . . . . . . . . . . .
1.1
Root markers . . . . . . . . . . . . . .
1.2
Prayoga . . . . . . . . . . . . . . . . .
1.3
Other semantic conditions . . . . . . .
1.4
Preverbs . . . . . . . . . . . . . . . . .
1.5
Other co-occurrence conditions . . . .
1.6
Tabulation of voice conditions . . . . .
2
Related Work . . . . . . . . . . . . . . . . . .
3
Preparing inputs for the implementation . . . .
3.1
XML Dhatupa.tha database . . . . . . .
3.2
Preverb cooccurences attested by Pan.ini
4
Implementation details . . . . . . . . . . . . .
4.1
Data Structure . . . . . . . . . . . . .
4.2
Rule formulation . . . . . . . . . . . .
4.2.1
Phonetic conditions . . . . .
4.2.2
Attribute conditions . . . . .
4.2.3
Semantic conditions . . . . .
4.2.4
Optionality . . . . . . . . . .
4.2.5
Action . . . . . . . . . . . .
4.3
Control structure . . . . . . . . . . . .
5
Results and related discussions . . . . . . . . .
5.1
Notations in the derivation tree . . . . .
5.2
Handling meaning conditions . . . . .
5.3
Handling rule interaction . . . . . . . .
5.4
Handling praptavibhas.a . . . . . . . .
6
Conclusions and Future Work . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
157
158
161
161
162
162
163
163
165
171
171
175
177
177
178
178
178
179
179
179
179
181
181
184
185
188
189
195
Interrogatives and word-order in Sanskrit

P ETER M. S CHARF
203
1
The problem . . . . . . . . . . . . . . . . . . . . . 203
TABLE OF CONTENTS
2
The evidence . . . . . . . .
3
Parallels with other pronouns
4
The solution . . . . . . . . .
References . . . . . . . . . . . . .
xix
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

MILIE AUSSANT
1
Introduction . . . . . . . . . . . . . . . . . . . . .
2
The European approach . . . . . . . . . . . . . . .
2.1
Parts of speech in computational linguistics
2.2
Parts of speech in Graeco-Latin grammatical tradition . . . . . . . . . . . . . . . .
3
The Indian approach (Sanskrit Vyakaran.a) . . . . .
3.1
The naman-akhyata-upasarga-nipata classification . . . . . . . . . . . . . . . . . .
3.2
The Pan.inian classification of padas . . . .
3.3
Some remarks on a few other classifications
4
Concluding remarks . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . .
B RENDAN S. G ILLON
1
Introduction . . . . . . . . . . . . . . . . . . . .
2
Parts of speech . . . . . . . . . . . . . . . . . .
3
Clausal constituents . . . . . . . . . . . . . . . .
4
Supra-lexical and sub-clausal constituents . . . .
4.1
Context Dependence . . . . . . . . . . .
4.2
Proforms . . . . . . . . . . . . . . . . .
4.3
Ellipsis . . . . . . . . . . . . . . . . . .
4.3.1
Interrogative ellipsis (sluicing)
.
.
.
.
.
.
.
.
208
212
214
216
219
219
220
220
221
224
224
227
230
232
233
237
237
239
240
240
241
246
247
248
xx
S ANSKRIT SYNTAX
4.3.2
Appended coordination (stripping) . . . . . . . . . . . . . .

4.3.3
Gapping . . . . . . . . . . . .
4.3.4
Verb phrase ellipsis . . . . . .
4.3.5
Copular complement ellipsis .
4.3.6
Nominal ellipsis . . . . . . . .
5
Further evidence for Phrases . . . . . . . . . . .
6
Conclusion . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
249
250
252
252
253
255
263
264

269
1
Introduction . . . . . . . . . . . . . . . . . . . . . 269
2
Word order in Sanskrit . . . . . . . . . . . . . . . 273
3
Indian theories of expectancy and proximity . . . . 274
an ks.a (expectancy) . . . . . . . . . . . 275
3.1
Ak
3.2
Sannidhi (proximity) . . . . . . . . . . . . 277
4
Dependency parsing and word order . . . . . . . . 278
4.1
Tree traversal and possible word orders . . 279
4.1.1
Generalising tree traversal . . . . 280
4.2
Projectivity principle . . . . . . . . . . . . 282
4.3
Weak non-projectivity (planarity) . . . . . 285
5
Empirical evaluation . . . . . . . . . . . . . . . . 288
5.1
Cases of sannidhi violation from Gillons
data . . . . . . . . . . . . . . . . . . . . . 288
5.1.1
Dislocation of a genitive . . . . . 289
5.1.2
Dislocation of a vises.an.a . . . . 291
5.1.3
Other relations . . . . . . . . . . 291
5.2
Sannidhi violation in the Bhagavadgta . . 292
5.2.1
Sannidhi violation involving kartr
and karman . . . . . . . . . . . 293
TABLE OF CONTENTS
xxi
5.2.2
Sannidhi violation involving two

kartr relations . . . . . . . . . .
5.2.3
Sannidhi
violation involving karman and kriyavises.an.a . . . . . .
5.2.4
Sannidhi violation involving karman and apadana . . . . . . . .
6
Conclusion . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . .
Distinctive features of poetic syntax: preliminary results
P ETER S CHARF, A NUJA A JOTIKAR,
S AMPADA S AVARDEKAR, and PAWAN G OYAL
1
Motivation . . . . . . . . . . . . . . . . . . . . .
2
Expectations . . . . . . . . . . . . . . . . . . . .
3
Methodology . . . . . . . . . . . . . . . . . . .
3.1
Corpus preparation . . . . . . . . . . . .
3.2
An Indian cognitive linguistics dependency relations tagset . . . . . . . . . . .
3.3
Queries . . . . . . . . . . . . . . . . . .
4
Results . . . . . . . . . . . . . . . . . . . . . . .
4.1
Agent after verb . . . . . . . . . . . . .
4.2
Object after verb . . . . . . . . . . . . .
4.3
Instrument after verb . . . . . . . . . . .
4.4
Adverb after verb . . . . . . . . . . . . .
4.5
Qualifier after qualified . . . . . . . . . .
4.6
Genitive after what it limits . . . . . . .
4.7
Object after absolutive . . . . . . . . . .
4.8
Agent after object . . . . . . . . . . . . .
4.9
Summary . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . .
298
299
299
300
302
.
.
.
.
305
306
307
310
310
.
.
.
.
.
.
.
.
.
.
.
.
.
311
313
314
314
315
315
315
315
318
318
319
319
322
xxii
S ANSKRIT SYNTAX

K ESHAV M ELNAD, PAWAN G OYAL and P ETER S CHARF
1
Introduction . . . . . . . . . . . . . . . . . . . . .
2
Sources describing Sanskrit meters . . . . . . . . .
3
Sanskrit prosody . . . . . . . . . . . . . . . . . .
3.1
Syllable weight . . . . . . . . . . . . . . .
3.2
The basic unit of varn.avrtta meter: gan.a . .
3.3
The basic unit of matra
vrtta meter: caturmatrika . . . . . . . . .. . . . . . . . . .
3.4
Types of meters . . . . . . . . . . . . . . .
4
Meter Identifying Tool (MIT) . . . . . . . . . . . .
4.1
Meter database . . . . . . . . . . . . . . .
4.2
Input and output . . . . . . . . . . . . . .
4.3
Algorithm . . . . . . . . . . . . . . . . . .
5
Results . . . . . . . . . . . . . . . . . . . . . . . .
5.1
A samavrtta verse in Vasantatilaka meter .
5.2
An a rdhasamavr
tta verse in Vegavat meter
5.3
A vis.amavrtta verse in Lalita meter . . . .
verse in Ary
a meter . . . . . .
5.4
A matravrtta
5.5
Performance . . . . . . . . . . . . . . . .
6
Conclusions and future work . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . .
On concord and government relations in Sanskrit
P RASAD P. J OSHI
1
Introduction . . . . . . . . . . . . . . . . . . . . .
2
The importance of concord and government . . . .
3
Concord . . . . . . . . . . . . . . . . . . . . . . .
3.1
Concord of substantive and adjective . . .
3.1.1
Adjectives with taddhita-deletion
3.1.2
Nouns in apposition . . . . . . .
3.2
Concord of a predicate verb and noun (of
agent and object) . . . . . . . . . . . . . .
325
325
327
328
329
330
331
331
333
333
334
335
338
338
339
340
341
342
343
345
347
348
348
349
349
352
354
355
TABLE OF CONTENTS
xxiii
Government . . . . . . . . . . . . . . . . . . . . .
4.1
Government of a verb and a noun . . . . .
4.2
Government of an indeclinable and a noun
5
Conclusions . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . .
1
Introduction . . . . . . . . . . . . . .
2
Modern Sanskrit . . . . . . . . . . . .
3
Error analysis . . . . . . . . . . . . .
4
The standard of correct Sanskrit . . .
5
Discussion of erroneous sentences . .
5.1
Sentence 1 . . . . . . . . . .
5.1.1
Shastris comments
5.1.2
Discussion . . . . .
5.1.3
Directive rules . . .
5.2
Sentence 2 . . . . . . . . . .
5.2.1
Shastris comments
5.2.2
5.2.3
Directive rule . . .
5.3
Sentence 3 . . . . . . . . . .
5.3.1
Shastris comments
5.3.2
5.3.3
5.4
Sentence 4 . . . . . . . . . .
5.4.1
Shastris comments
5.4.2
5.4.3
5.5
Sentence 5 . . . . . . . . . .
5.5.1
Shastris comments
5.5.2
5.5.3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
357
357
358
359
360
361
361
363
363
364
367
367
367
368
373
374
375
376
378
378
378
379
380
380
380
381
381
382
382
383
385
xxiv
S ANSKRIT SYNTAX
5.6
Sentence 6 . . . . . . . . . .
5.6.1
Shastris comments
5.6.2
5.6.3
5.7
Sentence 7 . . . . . . . . . .
5.7.1
Shastris comments
5.7.2
5.7.3
5.8
Sentence 8 . . . . . . . . . .
5.8.1
Shastris comments
5.8.2
5.8.3
5.9
Sentence 9 . . . . . . . . . .
5.9.1
Shastris comments
5.9.2
5.9.3
5.10 Sentence 10 . . . . . . . . . .
5.10.1
Shastris comments
5.10.2
6
Conclusion . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
386
386
387
387
387
388
388
389
390
390
390
392
392
392
392
395
395
395
396
397
397
399
As..ta dhyay sutra

index
471
Author index
477
Title index
487
List of tables
1
Hocks (1989a) account for example (8) . . . . . .
2
Locative absolute with normal and upsidedown agreement . . . . . . . . . . . . . . . . . .
3
Non-head relational nominal governing an external head . . . . . . . . . . . . . . . . . . . . . . .
4
Node erasure, movement, and compounding in (35c)
1
13
18
34
34

53
1
Abbreviations . . . . . . . . . . . . . . . . . . . . 102
Extension rules and the syntax of . . . -vati
1
Derivation of non-feminine forms of the interrogative pronoun kim and noun vrks.a . . . . . . . . .
2
Derivation of (14a) .tea;na tua;yMa
va;tRa;tea and (14b) ta;d;d;tRa;tea
3
Sutras with terms in -vati . . . . . . . . . . . . . .
4
Ia. Sutras in which a nominative form is explicit . .
5
Ib. Sutras in which a nominative form is understood
6
IIa. B-vat = B6 iva . . . . . . . . . . . . . . . . . .
7
IIb. B-vat = B7 iva . . . . . . . . . . . . . . . . . .
8
Abbreviations . . . . . . . . . . . . . . . . . . . .
109
112
130
142
146
147
148
150
150

157
1
Correspondence of voice with prayoga and pada . . 160
To classify words: European and Indian . . . approaches 219
1
Sub-categories of noun in the Brown Corpus . . . . 222
2
Abbreviations . . . . . . . . . . . . . . . . . . . . 233
xxv
xxvi
S ANSKRIT SYNTAX

237
1
English optional complementation . . . . . . . . . 258
2
Abbreviations . . . . . . . . . . . . . . . . . . . . 264
269
1
Abbreviations . . . . . . . . . . . . . . . . . . . . 301
Distinctive features of poetic syntax
305
1
Probability of x occurring before y in a relation
pair (x,y): comparing poetic sentences with prose . 321
325
1
Gan.a patterns . . . . . . . . . . . . . . . . . . . . 331
2
Matra patterns . . . . . . . . . . . . . . . . . . . . 332
3
Prototypes of each of the four types of meters and
the number of meters of each type in our database . 334
361
1
Word sense in Sanskrit versus in modern Indian
languages . . . . . . . . . . . . . . . . . . . . . . 393
List of figures
1
The control structure for the implementation. This
process is applied recursively until no rules are
triggered. . . . . . . . . . . . . . . . . . . . . .
2
Derivation tree for the root bhu . . . . . . . . . .
3
Derivation tree for the root da . . . . . . . . . .
4
Derivation tree for the root gam . . . . . . . . . .
5
Derivation tree for the root ks.ip . . . . . . . . . .
6
Derivation tree for the root kr . . . . . . . . . .
7
Derivation tree for the root ram . . . . . . . . . .
157
.
.
.
.
.
.
.
180
183
190
191
192
193
194
To classify words: European and Indian . . . approaches 219

1
The main classes of speech forms distinguished by
Pan.ini . . . . . . . . . . . . . . . . . . . . . . . . 229
1
Traversals for sentence (1) with and without transposition . . . . . . . . . . . . . . . . . . . . . .
2
Dependency graph for sentence 3 . . . . . . . . .
3
Dependency structure with projections for sentences 1.1 to 1.6 . . . . . . . . . . . . . . . . . .
4
Projections for sentences 2.1 to 2.6 . . . . . . . .
5
Projection for sentence (4) . . . . . . . . . . . .
6
Projection with Rearrangement of nodes . . . . .
7
Planar dependency graph for sentences (2.2) and
(2.5) . . . . . . . . . . . . . . . . . . . . . . . .
8
Planar dependency graph for sentence (4) . . . .
9
Dislocation without sannidhi violation . . . . . .
xxvii
269
. 281
. 282
.
.
.
.
283
284
284
285
. 286
. 287
. 287
xxviii
10
11
12
13
14
15
16
17
18
19
20
21
22
S ANSKRIT SYNTAX
Dislocation of a genitive . . . .
Dislocation of a genitive . . . .
Dislocation of a vises.an.a . . . .
Dislocation of an argument . . .
Analysis of BhG. 6.34 . . . . . .
Analysis of BhG. 8.25 . . . . . .
Analysis of BhG. 10.16 . . . . .
Analysis of BhG. 9.3 . . . . . .
Analysis of BhG. 1.27 . . . . . .
Modified Analysis of BhG. 1.27
Analysis of BhG. 8.19 . . . . . .
Analysis of BhG. 1.37 . . . . . .
Analysis of BhG. 18.75 . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
Relative position of an agent with respect the verb
that governs it: comparison of prose and poetry
syntax . . . . . . . . . . . . . . . . . . . . . . .
2
Relative position of a direct object with respect to
the verb that governs it: comparison of prose and
poetry syntax . . . . . . . . . . . . . . . . . . .
3
Relative position of an instrument with respect to
the verb that governs it: comparison of prose and
poetry syntax . . . . . . . . . . . . . . . . . . .
4
Relative position of an adverb with respect to the
verb that governs it: comparison of prose and poetry syntax . . . . . . . . . . . . . . . . . . . . .
5
Relative position of a qualifier with respect to what
it qualifies in a sentence: comparison of prose and
poetry syntax . . . . . . . . . . . . . . . . . . .
6
Relative position of a genitive with respect to what
it limits in a sentence: comparison of prose and
poetry syntax . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
290
290
291
292
294
295
296
296
297
298
299
300
301
305
. 316
. 316
. 317
. 317
. 318
. 320
L IST OF FIGURES
7
xxix

an absolutive that governs it: comparison of prose
and poetry syntax . . . . . . . . . . . . . . . . . . 320
an agent of the same action governing them: comparison of prose and poetry syntax . . . . . . . . . 321

1
Sentence 1 . . . . . . . . .
2
Sentence 2 . . . . . . . . .
3
Sentence 3 . . . . . . . . .
4
Sentence 4 . . . . . . . . .
5
Sentence 5 . . . . . . . . .
6
Sentence 6 . . . . . . . . .
7
Sentence 7 . . . . . . . . .
8
Sentence 8 . . . . . . . . .
9
Sentence 9 . . . . . . . . .
10 Sentence 10 . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
361
373
377
379
382
386
388
389
391
395
396
xxx
S ANSKRIT SYNTAX

Abstract: The purpose of this presentation is to provide a brief
general survey of major recent developments in Sanskrit
syntactic studies since Deshpande and Hock 1991, followed by more in-depth discussion of formal and functional issues that should be of interest to scholars working on computational approaches to Sanskrit syntax. Formal syntactic topics are free word order, challenging
syntactic aspects of relative-correlative constructions, issues of agreement, oblique subjects and related issues,
the syntax of double-direct-object constructions (including causatives), and the issue of asamartha compounding. The functional part focuses on the use of different
syntactic alternatives in discourse and in different genres
a topic that, to my knowledge, has not received systematic discussion. Two major phenomena are examined
in some detail fronting and extraposition to the right,
including their different functions in a variety of genres.
Functional analytical approaches may yield interesting insights and challenges to formal syntactic approaches, although they clearly cannot be a substitute for formal analysis. I conclude that there is ample room for more research
on Sanskrit syntax, especially in the post-Vedic language
with its richer variety of different genres.
Keywords: Sanskrit, syntax, formal approaches, functional
approaches
H. H. H OCK
Introduction
My presentation has three goals. One is to provide a survey of publications in Sanskrit syntax since Deshpande and Hocks (1991)
Sanskrit syntax bibliography. A second one is to focus in greater
detail on a number of formal issues that, I believe, would be of interest both to linguists pursuing computational approaches to Sanskrit syntax and to those working in linguistic theory. The third
goal is to discuss a selection of functional factors that influence
the use of particular syntactic structures in particular text types, an
issue which I believe would also be interesting to those engaged in
computational work.
A brief survey of recent publications
The Sanskrit syntax bibliography provided at the end of this book

consolidates Deshpande and Hocks (1991) A bibliography of writings on Sanskrit syntax published in Hock 1991b and the bibliography of works on Sanskrit syntax published since then compiled
for presentation at the Seminar on Sanskrit Syntax and Discourse
Structures held in Paris, 1315 June 2013. The combined bibliography will also be published at the Sanskrit Library website with
regular upadates. We look forward to receiving additional references so as to make the bibliography more comprehensive. At this
point, let me briefly survey the distribution of recent publications
in terms of chronology and general categories of subject matter.
Recent research concerning Sanskrit syntax includes over 200
works, including article collections and bibliographies quite impressive for a period of less than 25 years, especially if compared
to the 474 entries in Deshpande and Hock 1991, which covers publications from Burnouf (1824) to Brereton and Jamison To Appear (published 1991). It is impressive, too, because, as in the
past, syntax receives much less scholarly attention than other as-
S OME ISSUES IN S ANSKRIT SYNTAX
pects of Sanskrit linguistics. The following statistics are based on

a 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 Hock
1989b), the bulk of publications (147 out of 186) deals with the
Vedic period, especially the Rgveda, and commonly from a his
torical, Indo-Europeanist perspective.
The post-Vedic or classical
period is covered only in 45 publications, some of which address
both Vedic and post-Vedic issues.
A common topic for the Vedic period is the issue of tense, aspect, voice, and modality, which is dealt with in 29 publications.
Another 23 publications address word and phrase order, including
the issue of clitics and Wackernagels Law. Case syntax is covered in 18 publications, especially by Hettrich (six publications,
including the important 2007: Materialien zu einer Kasussyntax
des Rgveda, which includes a rich bibliography). Issues of subor
dination,
both finite and non-finite (infinitives, converbs, locative
absolute constructions, etc.) are dealt with in some 18 publications.
Most of the 27 publications that approach Sanskrit syntax from
a modern theoretical or typological approach are focused on the
post-Vedic language, and so are all of the 23 publications that deal
with, or refer to, the insights of the Indian grammatical tradition,
especially Pan.ini. As a personal note, let me add that this relatively
limited reference to the Indian tradition is regrettable. In principle,
all research on Sanskrit syntax and Sanskrit linguistics in general should treat the work of Pan.ini and the entire early grammatical and phonetic tradition as earlier scholarship, just as it does
the work of linguists like Delbrck, Wackernagel, and Whitney;
see Hock 2009b and Hock Forthcoming(b).
Finally, 65 publications treat of issues of function and discourse, in both the Vedic and the post-Vedic period. This category
is heavily dominated by Jared Kleins publications, most of which
(16 out of 22) focus on the issue of stylistic repetition of differ-
H. H. H OCK
ent grammatical structures and categories in the language of the

Rgveda.
Formal issues
In the following section of my paper I address formal issues that

may be of interest to scholars pursuing computational approaches
to Sanskrit syntax as well those working on typology and syntactic
theory. I draw to a large extent on my own research, both published and unpublished, but also include references to other recent
research.
A recurrent theme is that we need to consider both Pan.inis generative approach and modern approaches, whether generative or traditional-philological, and that, likewise, we need to
keep in mind both the empirical information conveyed by Pan.inis speaker-knowledge based grammar and the empirical data unearthed by western philological approaches. The latter issue is
especially relevant, since as Deshpande (1983) suggested, Pan.inis
location on the northwestern periphery of the Sanskrit-speaking
world of his time may account for certain differences between the
syntactic structures predicted by his grammar and those found in
the 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 languages) exhibits a remarkable degree of free word order not just
free phrase order. In this section I discuss two major formal approaches to this phenomenon. Schufele (1990, 1991a,b) follows
the major tradition of modern western scholarship (e.g. Delbrck
1878, 1888; Speijer 1886, 1896; Lahiri 1933) in assuming a basic
word and phrase order of the SOV type, with various movement
processes 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-internal
linear ordering. Neither approach adopts the possible alternative
of assuming complete non-configurationality along the lines suggested 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 exhibit
all the features normally associated with hierarchical structure. This is something that children learning the language
would have to account for in their grammar, and it would
discourage 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 tendency
in Vedic. While Schufele does not pursue this issue explicitly, the dominant head-finality too can be argued to be
something that children learning the language would have to
account for in their grammar.
In PPs the adposition normally remains next to at least part
of its complement if there is movement. This, again, supports the assumption of hierarchical, rather than flat phrase
structure. 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 the
same phenomenon in Latin.
Movement of individual words or combinations of words, as
in (3),1 is made possible through a process of liberation or
node erasure (see Pullum 1982, Ross 1967: 5054 (1986)).
1 These
examples come from my collections.
H. H. H OCK
(1) a. .~ta;vEa :p!au :=+a :pa;a;ya;R!a;
a;d;nd!+ma; H
stavai pura pa ryad ndram a hnah. (RV. 3.32.14b)
day.
I shall praise Indra before the fateful
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 asvam
mahimanav abhitah. sambabhuvatur (SB.
10.6.4.1)
These two jars appeared around = on both sides of the
horse. (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;ta ;
a;va;du;bRua;Da;aH
a samudrat tu vai purvad a samudrat tu pascimat.
tayor eva + antaram
vidur budhah... (Ma giryor a ryavartam
nu 2.22)
avarta (extends) from the eastWise people know (that) Ary
ern sea to the western sea, (and) between these two mountains (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;ntM
/ /a ;
a;v!a;k+:ma;tea
etami v eva + es.a j etasmai vis.n.ur j yajo vikrantim
i vikra
mate (SB. 1.1.2.13)

This Vis.n.u, the sacrifice, steps this (world-conquering
three-fold) stepping for him (the sacrificer).
b. .tea;Sa;Mai Ba;a;ma;ea ma;h;a;ba;a;huH :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.a m
i mahatmanam.
i bhmo mahabahuh. parthivanam
yatharham
akarot pujam
. . . (Nala 2.11; MBh. 3.51.10
*215)
Strong-armed Bhma honored these noble rulers appropriately . . .
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 word
order (360) and he goes on to argue for a (modern) generative account operating with wild trees, i.e. trees without phrase-internal
linear ordering.
Gillon (1996) adopts and modifies Staals (1967) notion of
wild trees, providing empirical support from two corpora the
prose examples in Apte 1885 and 1,500 sentences from Dharmakrtis autocommentary on the Praman.avarttika. See also Gillon
and Shaer 2005. The following discussion is based on the latter
publication.
Like Staal, as well as Schufele, Gillon and Shaer accept the
need for phrases, rather than a completely flat structure. Unlike
Schufele, they assume that there is no linear order within phrases.
Moreover, they argue against a VP, instead postulating the flat
clause structure in (4), without linear order. Further, they (2005:
468) claim that . . . the strategy of deriving less common word orders with specialized information packaging functions from more
basic syntactic structures . . . seems to us less plausible than ones
consistent with the wild tree claim. The paper (2005: 48085)
concludes with a section on Some remaining puzzles.
(4) S V, NPS, NPO
In support for the assumption of flat phrase-internal structures,
such as (4), they claim that their corpora exhibit both left- and
right-headed phrases and that therefore there is no evidence for
phrasal headedness (2005: 470). In addition, they accept movement processes that extract elements out of phrases and place them
in left- or right-peripheral position within the clause (2005: 475
80).
Certain features are shared between Schufeles and Gillons
approaches the acceptance of phrases, rather than completely
H. H. H OCK
flat structures, and the fact that movement processes can extract
and move elements out of phrases. For the purposes of computational text analysis, therefore, there may be no significant difference.
From the perspective of linguistic theory, however, the two approaches differ considerably, and it is Schufeles approach that
provides the better insights. His account of Sanskrit is completely
compatible with the linguistic typology of SOV languages, with
head-finality at all phrasal levels. In fact, Sanskrit also conforms
to SOV typology in its complex syntax, by making extensive use of
non-finite subordination as well as of relative-correlative constructions; Hock 1989a, 2005, 2014. Under the Staal-Gillon approach
these typological characteristics would be epiphenomenal at best.
Now, it is true that Pan.ini has no rules comparable to western
generalizations about word or phrase order. But Pan.ini also has no
rules comparable to western generalizations about phrases, such
as NP, VP, PP. True, there are rules regarding karakas and their
realizations, but these do not address issues such as complex NPs
with genitive modifiers. There is also the notion samanadhikaran.a, but this presumably holds not only for agreement within NPs
but also relates surface subject NPs to their verbs (A. 1.4.104107)
and must be assumed to hold also for agreement between subjects
(kartrs) and predicate nouns or adjectives (see 3.3 below for dis
cussion).
Even the notion sentence is a murky issue in the Pan.inian tradition; see e.g. Cardona 1976: 22324, Deshpande 1991,
as well as Hock Forthcoming(b): 6. In all of these respects, and
not only as regards word order, the Indian grammatical tradition
and modern generative approaches are orthogonal.
It is also true that Sanskrit offers frequent examples with nonfinal heads. But there are considerable differences between different texts. Consider major constituent order. As noted in Hock
1984, while in mantra Vedic and Kalidasas dramatic dialogue
verb-final structures amount to only about 65%, in Vedic prose
they are about 97% (see also Hock 1997). A similarly high ratio
of verb-final structures is found in Patajalis Mahabhas.ya; see the
statistics in (5). Claims about headedness and any other aspects
of syntax must therefore be based on a large variety of different
genres, not just on the two corpora examined by Gillon. And as
the evidence just cited shows, genres that do not make any claims
to stylistic or literary elegance are characterized by heavy predominance of head-final constituent order; in fact, even in other genres
verb-finality runs to about 65%. The wild tree assumption that
phrases, including the sentence (S), have no internal order fails to
capture these facts.
(5) Word order in Patajalis Mahabhas.ya
a. Paspasa (Kielhorn-Abhyankar 1.1.11.3.5)2
V-final: 35
V-initial: Not found in the sample
V + O in the formula . . . adhyeyam
vyakaran.am: 73
sak + (O) + infinitive: 2
b. Sivas
utras (Kielhorn-Abhyankar 1.15.21.16.18)
V-final: 40
V-initial (including after linker, such as tena): 5
V + O/Pred: 3
V + Other: 3
V + [ ] iti: 8
3.2
Relative-correlatives
As in the case of word order, Pan.ini has remarkably little to say

about the syntax of Sanskrit relativization. The closest he comes
2 Vedic and other traditional citations are ignored. Gerundives and taparticiples used as main verbs are included.
3 Contrast the formulaic use of the gerundive with the ordinary one in laghvartham
cadhyeyam
sabda jeya iti.
vyakaran.am. brahman.enavasyam
10
H. H. H OCK
is in three sutras (A. 3.3.139140 and 3.3.156) that address the issue of modality in conditional structures which, as is well known,
involve an adverbial form of the relative pronoun (yadi) or the particle ced.
At least from the time of Speijer (1886, 1896) and Delbrck
(1888), western scholarship has recognized that Sanskrit relative
structures consist of a relative clause, containing a relative pronoun, and a main clause, containing a correlative pronoun and that
the relative clause is not inserted into the main clause.4 Speijer
(1886: 349, 1896: 349) refers to the relationship as one between
a protasis and an apodosis. Minard (1936) introduces the term
diptych for the construction which in typological and theoretical
literature is now commonly referred to as relative-correlative.
The syntactic account of Sanskrit and other, similar relativecorrelative constructions is further refined in the 1970s and 1980s
by arguments that the relative clause is base-generated as AD JOINED to the main or correlative clause; see e.g. Andrews
1975 (1985), Ken K. Hale 1975, Dasgupta 1980, Keenan 1985,
Lehmann 1984, Srivastav 1988.
Based on a broad range of evidence, Hock (1989a) goes one
step further and argues that relative clauses are syntactically CON JOINED to their correlative clauses. While some of that evidence
appears to be restricted to Vedic, other evidence is also found in
post-Vedic. The nature of that evidence is, I believe, such that
both those working in formal syntax and those working on computational analyses will find it interesting and challenging.
First, in some cases there is no clear relationship between the
relative pronoun (or phrase) of the relative clause and the correlative pronoun (or phrase) of the main clause; see (6), where an
4 Speijer
(1886: 349) hedges on this issue by stating that preposing of the relative clause before the main clause is much more used than inserting the relative
sentence in the main one.
11
example similar to (6a) was brought to my attention in 1989 by

Kiparsky 1995, and (6c) by James Fitzgerald (March 2006). Structures of this sort are typically best rendered as conditionals.
(6) a. ya I +.nd! ya;ta;ya;~tva;!a Bax ;ga;va;e!a yea . ca tua;u!;vuaH ma;m!a I+.du ;g{a (rua;Da;!a h;va;m,a
ya indra yatayas tva bhrgavo ye ca tus..tuvuh.. mama d ugra srudh havam (RV.
8.6.18)
(have praised) you, and which BhrWhich Yatis, O Indra,
gus have praised you, powerful Indra, nevertheless hear

MY call.
Even if the Yatis and Bhrgus have praised you, powerful Indra, nevertheless hearMY call.
b. ya;a;sa;Ma na;a;d;d;tea Zua;kM a;a;ta;ya;ea na .sa ;
a;va;k+:yaH
yasam
nadadate sulkam
jatayo na sa vikrayah. (Manu
3.45)
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 maithunatrptaye;
tasyasmarayato vyaktam adharmo n

atra sam
sayah.. (MBh. 12.258.38)
Which instruction to gratify ones wife sexually is not
heard, 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 sexually, if a man does not remember this it is clearly a serious infraction.
Secondly, there are some examples in which the relative clause
exhibits properties normally only associated with independent
main clauses, namely interrogation and imperative modality (7).
12
H. H. H OCK
(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;ta
[yat kim akaram]
saryato ha va ks.a mcakre
RC [tasmad i
dam a padi]CC + iti (SB.

4.1.5.4)
ata thought, Because I have done what?, therefore I

Sary
have gotten into this.
. . . What have I done to get into this?
1.7.3.19; a similar structure with kva
(Thus also SB.
where 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,a
tyaje pran.a n naiva dadyam
kapotam
saumyo hy ayam
kim
na janasi . . . .
[yatha klesam
ma kurus.va + iha . . . ]RC
[naham
CC (MBh. 3, App.
kapotam arpayis.ye kathamcit]
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 you
make trouble here ! . . . , I will not hand over the dove in
any way.
. . . so that you dont make trouble here . . .
Most important, example (8) shows clearly that the relative
clause must be CONJOINED to the two main clauses. It is simply
impossible for the same clause to be simultaneously ADJOINED
to two different clauses; and deriving the relative clause from an
underlying center-embedded postnominal position would be preposterous how can a single clause be simultaneously embedded
under two different NPs, in two different clauses? In Hock 1989a
I therefore propose to conceive of the relation between the relative
clause and the two correlative clauses as in Table 1. The formalism
is, of course, antiquated, but the syntactic relation must be something along these lines. (Davison 2009 proposes CP adjunction for

S
S
0
CP
13
CC
RP
RC
CP
CC
Table 1
Hocks (1989a) account for example (8)
structures in which the relative clause precedes, in contrast to IP
adjunction, 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 daiv vag]CC
[yayai yad yad j eva vadati]RC
[tat tad j bhavati]CC (BAU. 1.3.27)
That
i is divine Speech by whichi whatever j one speaks, that j
comes about.
Whatever one speaks by means of divine Speech comes
about.
3.3
Some issues of agreement
Pan.ini addresses some issues of agreement, in two places. One is

the ekases.a sutras which address the issue of gender resolution under the specific circumstance of one word taking the place of two
conjoined ones (A. 1.2.6473); the other are the sutras governing
person agreement between surface subjects (kartrs or karmans) and
the la-kara of the verb (1.4.104107). But many aspects of agreement are not covered, except perhaps implicitly under the notion
of samanadhikaran.a coreference.
14
H. H. H OCK
In this section I map out some issues of Sanskrit agreement

that I believe should be of interest, especially to linguists working
on computational analyses of Sanskrit syntax.
A fairly straightforward issue is the question of gender agreement with mixed-gender conjunct antecedents, where two different strategies can be observed. One is agreement with the nearest
conjunct, as in (9); the other is gender resolution as in (10).
(9) k+:a;a;nta;ma;t
/ / a;a .=+a:$ya;a;ma;dM ma;ma . ca .ja;a;
a;va;ta;ma;pya;d;a;pra;Bax ;a;ta tva;d;Da;a;na;m,a
kantimat rajyam idam
mama ca jvitam apy adyaprabhrti
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 sarasvatm
copadhavac chepano smi namu
caye . . . (iti).. te bruvan . . . (SB.

12.7.3.12)
He (Indra) went to the Asvins [m.du.] and Sarasvat [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;ta
mrdam
gam
daivatam
vipram
ghrtam
madhu catus.patha pradaksina ni kurvta (Manu
m.
4.39)
. .
He should keep on his right a lump of earth, a cow, an
idol, a brahmin, ghee, honey, and a crossroads [n.pl.].
As I show in Hock 2012a, Speijers analysis for post-Vedic
Sanskrit gender resolution (1886: 1920), going back to Borooah
(1879), best accounts for the Vedic evidence: In the case of mixedgender antecedents that are entirely human (or animate), gender
resolution is in favor of the masculine; in all other cases, including cases like (10b), where non-human/inanimate and human antecedents are mixed, the result is neuter, except that in Vedic texts
15
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 priori possibility that a modifier to the left may show agreement with
the left conjunct, and one to the right with the right conjunct; see
Arnold, Sadler, and Villavicencio 2007 for Portuguese and Johnson 2008 for Latin. As it turns out, an example of this mirrorimage agreement can also be found in Sanskrit (11). It remains
to 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;ta
vyamama trau 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, the
phenomena discussed so far are rather mundane, another type of
agreement presents greater challenges. This is what may be called
upside-down agreement.
The best-known variety of this agreement is widespread in
Vedic prose, as in (12), but is also found in the later language. This
is the fact that pronoun subjects normally adopt the agreement features of their predicates, rather than the other way around. As far
as I can tell, this usage was first introduced into the discussion
of Sanskrit syntax by Speijer (1886: 18). The feminine singular
marking on sa in example (12) shows that at least in Vedic prose
this pattern of agreement is clause-bound, and that structures of
this 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,a
ye tus.a h. sa tvag (AB. 1.22.14)
16
H. H. H OCK
What (masc.) are the shells (masc.) that (fem.) is the skin
(fem.).
As it turns out, upside-down agreement must also be postulated 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;duH
vrte tu nais.adhe bhaimya lokapala . . . nalayas..tau varan
daduh
. (Nala 5.33; MBh. 3.54.28)
The Nis.adhan having been chosen by Bhaim, the world
rulers gave Nala eight boons.
b. .tea;Sua ga;.cC+.tsua va;yMa .~Ta;a;~ya;a;maH
tes.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,a
gantavye 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 for
long.
Lit. (It) having to be gone, it is not possible to stay here
for 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;a
//
ga;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
prasthita
gantavye sati jvitapriyasuhrt sarthah. kim u tyajyate.
(Subhas.itaratnakos.a 1151)
Together all set out to go to the determined-minded dearest one. (It) having to be gone, how is the dear friend of
ones life, having the same goal, getting left behind?
5 The Critical Edition instead has gantavyam
na ciram
sthatum iha sakyam
(MBh. 1.142.21)
17
The nearest analogue for analyzing such constructions would

be that of nominalization, which embeds a subordinate proposition
into a matrix clause by means of a nominal form of the verb, whose
case marking signals the status of that proposition within the matrix clause. See Yoon 1996 for an analysis of such structures.
The major difference between ordinary nominalizations and
structures of the type (13) is that the latter involve an adjectival
form, rather than a purely nominal one, a form which therefore
must be supplied with gender and number features, in addition to
the locative that signals the function of the construction within the
matrix clause. Note that in the synchronic grammar of Sanskrit
the locative case has to be assigned to the participle, not to its underlying subject, because of the fact that locative participial case
marking is not restricted to structures in which the participle has
a subject to agree with such as (13ab), but is also found in impersonal, subject-less structures like (13cd).
Note further that under this analysis, the subject of the participle, if any, is not in a position governed by a verb that could
assign case to it; the only features that the syntax can assign to it
are gender and number.
A possible way to account for the fact that the participle nevertheless gets gender and number features agreeing with its underlying subject, and that the subject, in turn, receives case, lies in
adopting the approach of the post-syntactic distributed morphology of Halle and Marantz 1993. As illustrated in Table 2, in this
analysis the syntactic output only has the abstract features plural
masculine for the underlying subject of the locative absolute, and
locative for the participle. The rest of the features need to be filled
in by the Morphology. The gender and number features of the
participle are filled in by normal agreement control, but the case
feature of the subject is supplied by upside-down agreement from
the participle. (In impersonal structures like (13cd), the participle
receives the usual neuter singular default features.)
18
H. H. H OCK
Syntactic Output
tad
[pl.m.]
gacchat
[loc.]
Morphology: Input
tad-
[pl.m.]
sthasyamah.
gacchat- vayam
[loc.]
Coreference
tad-
[pl.m.]
sthasyamah.
gacchat-su vayam
[loc.pl.m.]
<
Upside-down
vayam
sthasyamah.
te-s.u
sthasyamah.
gacchat-su vayam
[loc.pl.m.]
[loc.pl.m.]
]
Table 2
Locative absolute
with normal and upside-down agreement
This analysis is similar in spirit to that of Pan.inis account for
the locative absolute, which assigns locative case to the form expressing the subordinate verbal action (bhava), rather than to a
nominal constituent (14). However, Pan.inis focus, if I understand
it correctly, is on the (implicit) subject of the participial structure.
The fact that structures such as (13cd), without subject, also have
locative, expressed only on the verb and with the usual default
neuter singular agreement, shows that case assignment has to be
on the participle first and then percolates from the participle to
the subject, if any. (Although Pan.ini does not provide an explicit
account for the locative case marking on the agent of a locative absolute construction, we can infer that he would do so in the same
manner as for any other cases of agreement, namely under coreference (samanadhikaran.atva).)
(14) ya;~ya . ca Ba;a;vea;na Ba;a;va;l+a;a;m,a
yasya ca bhavena bhavalaks.an.am (A. 3.2.37)
the locative ending (is) also introduced (after an element) on
19
account of whose action (there is) qualification of (another)

action.6
3.4
Converbs, reflexives, oblique subjects, and syntactic

bracketing
In discussions of modern South Asian syntax, converbs (variously

referred to as absolutives, conjunctive participles, gerunds, and the
like), combined with reflexivization and word order, play a significant role as criteria that determine whether non-nominative constituents can be considered to be subjects or not. See for instance
the 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 reflexives exhibit a similar syntactic behavior to converbs (1886: 200
and 297298). More comprehensive discussions, which include
not only converbs and reflexives, but also word order, are Hock
1986, 1990, 1991a (with references). This section surveys the major issues and findings.
Pan.inis account for the syntax of the converb (ktva)7 is well
known (15), and its provision that ktva requires identity of kartrs,
i.e. underlying subjects, is well motivated. The dominant pat8

tern, at least for post-mantra-Vedic, is that this provision holds
6A
priori yasya could refer to the agent of the action bhava, or to the word
expressing the action. The latter is the usual interpretation and is made explicit
in the Kasika Vrtti: yasya ca kriyaya kriyantaram
laks.yate tato bhavavatah. saptam vibhaktirbhavati locative case is also (used) after a word characterizing
an action (bhavavat) by whose action another action is characterized. Joshi and
Roodbergen (1980: 8788) interpret bhava as state, distinguishing it from kriya
action. However, Cardona (1976: 197) (w. ref.) notes that both terms are used
to refer to actions.
7 Here as elsewhere ktv
a also stands for its replacement lyap.
8 For the mantras, Hock (1982b, 1986, 1987, 1990, 1991a) finds some (limited) evidence for surface subject, rather than underlying subject (kartr) control
20
H. H. H OCK
not only for active structures, where underlying and surface subject 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;le
samanakartrkayoh. purvakale (A. 3.4.21)
(ktva) is introduced
under the condition of identity of kartrs
in reference to prior time

The syntax of reflexives is not covered in the Pan.inian tradition, and most western discussions focus on Vedic and/or its IndoEuropean origins; see Vine 1997; Hock 2006; Kulikov 2007 for
recent discussions. In his coverage of reflexives, Speijer (1886:
200) notes similar conditions for the use of reflexives as for that
of converbs (1886: 29798), without however trying to link the
two phenomena. In a series of papers (Hock 1982b, 1986, 1987,
1990, 1991a), I have shown that, just like converbs, reflexives are
controlled by kartrs, 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 reflexivization
and (18) which focus on instrumentalabound; see e.g. (16), (17),
marked kartrs. 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+:taH
tatas tam a yantam
drs..tva paks.isa vakair . . . kolahalah. krtah.
(Hit. 1.4)
Then, upon seeing him coming, the young birds made a
racket.
of converbs in passives and passive-like structures, and somewhat more robust
evidence as regards word order and reflexive control. Zakharyin (1998) questions
this finding, but his discussion only focuses on converbs and does not address the
broader evidence of word order and reflexivization.
21
(17) .~vea;Sua .~Ta;a;nea;Sva;va;

a;h;tEa;BRa;
a;va;ta;v.yMa Ba;va; a;;
H
sves.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;+:aH
atha tena tam
satrum
matva + a tmanam
tasyopari praks.ipya pran.a h. parityaktah.
(Pac. 70;Speijer 1886: 297)
Then he1 , considering him2 an enemy, threw himself1 on top
of him2 and gave up his1 ghost.
Although converb and reflexive control by the kartr (whether
in Sannominative or instrumental) is the most common pattern
skrit, there are examples where other constituents or no constituents in the same clause seem to exert control. See the
examples in (19)(25) which focus on converb control, with the
exception of (24c) which shows that genitive-marked NPs also can
control 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 vaH
alam
vis.a dena bilam
pravisya (.) vasama sarve yadi rocate vah. (Ram. 4.52.31)
Enough of entering the cave in despondency. All of us
are staying if it pleases you.
b. ;pra;a;pya na;d ;Ma :pa;vRa;taH ;a;~Ta;taH
//
aprapya 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;ta
osadhr jagdhva + apah. ptva tata es.a rasah. sambhavati
(SB. 1.3.1.24)
22
H. H. H OCK
(The animals/somebody) having eaten the plants, having
drunk 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,a
3.4.2.1)
a tithyena vai deva is..tva tan t-samad avindat (SB.
The Gods having sacrificed with the guest-offering

discord befell them.
b. tM!a ;hE ;nMa dx;;a Ba;!a;
a;vRa;vea;d
11.6.1.7)
tam
hainam
drs..tva bhr viveda (SB.
Having seen him (i.e. someone else), fear befell him.

(22) a. (rua;tva;a ;a;tva;d;mu
/ a;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 not
please 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;ma
dvija sipranadm
gatva tubhyam aham
mantram
dasyami
(Vetalapacavim
sati, ed. Emeneau 92.2021)
O brahmin, I will give a mantra to you, (you) having gone
to the river Sipranad.
x! ya;e!a;Sa;a;m!a;a;Da;d+va;a;ta
(23) :pa;(a;a;dE! :pa:=+!a;tya va;Sa;a
x!
:pa;(a ;!a;de ;vE!a;na;a;mea;t!a;tpa:=+!a;tya va;Sa;a
. . . (! );a;Da;d+a;va;ya;a;ta
pascad vai partya vrs.a yos.a m adhidravati
pascad evainam etat partya vrs.n.a . . . (a)dhidravayati (SB.
1.9.2.24)
The bull mates with the female approaching her from behind. He makes the bull mate with her, (the bull) having
approached 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 vrtram
vijitya yus.mabhir me yam
saha somap
23
tha(h.) (KB. 15.2)

Having slain Vrtra, having conquered, this somadrinking with youis mine.
b. .sua;=+:a;a;tMa :pua:+:SMa dx;;a .~:a;a;a;Ma ;
a;* :+:d;a;a;nta
/ / ya;ea;na;yaH
susnatam
purus.am
drs..tva strn.a m
klidyanti yonayah.
a, ed. Uhle 15.3738)
(Vetalapacavim
satik
Having seen a well-bathed/graduated man, womens
vaginas get wet.
c. .sa;a ;
a;h .~va;a .=+a:ja;Da;a;na;a
sa 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,a
ity eva kale syenena + a nya khadyamanasya sarpasya garalam
taddravye nipatitam
(Vetalapacavim
At that very time, the venom of a snake being eaten by a
hawk, (the hawk) having carried it off, fell into his food.
Faced with such a variety of different structures, some scholars
may opt for claiming that there is no purely syntactic criterion for
control and that any element that is somehow salient may serve as
controller. This is close to what Zakharyin (1998) proposes.
A heuristically and theoretically more interesting position is to
try to determine whether some or all of these exceptional structures 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) by
A. 3.4.20 with A. 3.4.21 taking care of elsewhere. In both
cases we are probably dealing with some kind of grammaticalization. The one in (19b) could be compared to later grammaticalizations such as adhikrtya about, a daya with, a rabhya (starting)
24
H. H. H OCK
from which likewise do not seem to be sensitive to control by any

particular constituent.
As far as (20) and (21a) are concerned, these seem to be peculiarities of Vedic prose; see Delbrck 1888: 408. Hock (1987)
accounts for them under the notion reduced-clause structures, a
phenomenon not limited to converbs but also found with participles.
Example (21a), however, could also be analyzed as comparable to (21b) which, together with (22a), could be and has been
taken as equivalent to Modern Indo-Aryan oblique-experiencersubject constructions; see Hook 1976, 1984 for (22a), and Zakharyin 1998 more generally.
Structures of this sort, however, are extremely rare and, in the
aggregate, no more frequent than structures like (22b) in which a
non-experiencer indirect object controls the converb, or (23) where
the converbs dependence on the causee of adhidravayati may be
dittological from the preceding non-causative construction with adhidravati. At any rate, all of these structures are quite rare and
can be dismissed as occasional examples of loose (or sloppy)
control.9
The examples in (24), by contrast, exemplify a much more
common pattern: control by genitive-marked possessor NPs. The
fact that examples of control by non-kartrs (in Pan.inis sense) are
especially common with genitive-marked NPs was already noted
by Speijer (1886: 298) who considered these NPs to be exemplars
of his dative-like genitive category. Focusing on converb and reflexive control, as well as word order, I have argued (Hock 1990,
1991a) that Possessor NPs must be recognized as a highly produc9 Interestingly, if structures like (21b) and (22a) were to be analyzed as
oblique-experiencer-subject constructions or as forerunners of such constructions accusative-marked experiencers would seem to occur more frequently
than dative-marked ones. In Modern Indo-Aryan, it is dative-marking which prevails. On the syntax of ruc see also Cardona 1990; Deshpande 1990.
25
tive alternative to kartr control. (In his very different approach to

the syntax of Rgvedic reflexive sva, Vine (1997) similarly finds
that genitive-marked
NPs are the most common alternative to subject NP controllers.)
Finally, example (25) shows the need for being sensitive to
syntactic bracketing. The converb a nya is controlled by the kartr (syena) of the participial structure headed by khadyamana, not
the kartr (garala) of the matrix-clause verb nipatitam. See the
by
bracketingin (26). For further discussion see Hock 1986, 1987.10
(26) 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,a
[ity eva kale
[syenena + a nya khadyamanasya]
sarpasya garalam taddravye nipatitam]
(Vetalapacavim
At that very time, the venom of a snake being eaten by a
hawk, (the hawk) having carried it off, fell into his food.
While this phenomenon is not overtly addressed in the Pan.inian tradition, there is nothing in that tradition that would prevent
it. Given that their suffixes replace la-karas (A. 3.2.124, 3.4.70
71), participles are allowed to have their own kartrs, and these kar domain.
trs can control converbs (and reflexives) in their own
Complications do however arise because participial structures
are normally integrated into their matrix clauses without being set
off by clear boundaries. As a consequence, in very similar structures, such as (27ab), both involving the participle form gacchan,
10 Speijer (1886: 29798) comes close to realizing the need for some kind
of bracketing by noting that [apparent] control by locative NPs is common in
locative absolute constructions. Vine (1997) similarly weighs the possibility that
some instances of apparent non-subject control of reflexives may be accounted
for by something like bracketing.
26
H. H. H OCK
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.a dharah. samantah.]
(Vetalapacavim
sati, ed. Emeneau 28. 56)
Having showed that boy to him he brought (back) the
mendicant (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 a snan]
indra etya + abravt] (JB 1.156)
Indra, having come up, said to them, (who were) sitting
according 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 with

considerable problems of interpretation. The commentatorial tradition agrees that it is intended to cover double-direct-object constructions such as (29), but how it does so does not seem to have
received a satisfactory explanation; see Deshpande 1987. The evidence of the textual tradition makes it clear that either of the two
complements in these structures behaves like a true direct object,
27
being promotable to surface subject if it occurs by itself (30).

However, if both complements are present, only the more agentive one can be promoted; (31). See Hock 1985, 2012c for further
discussion; see also Ostler 1979; Van de Walle 1992.
(28) a. k+:tRua:=+a;a;psa;ta;ta;mM
/ /
a k+:mRa ta;Ta;a;yua;M . ca;a;na;a;a;psa;ta;m,
/ / a
kartur psitatamam karma. tathayuktam
canpsitam (A.
1.4.4950) That which is most desired by the agent is karman, and also that which is not desired (but) linked (to
the action) in the same way;
b. ;k+:a;Ta;tMa . ca
akathitam
ca (A. 1.4.51)
also what (is linked in the same way and) has not been as
yet 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,a
gatibuddhipratyavasanarthasabdakarmakarmakan.a m an.ikarta sa n.au. hrkror anyatarasyam (A. 1.4.5253)
also the non-causative

agent in the causative of roots
meaning go, understand, consume, communicate,
(and) intransitives, and (optionally) of hr and kr.
(29) a. k+:du b.ra;vaH . . . naX ;n,a

kad u bravah. . . . nrn (RV. 10.10.6)
(SPEAK)
What will you say to the men?

b. ta:va;a ya;a;a;ma . . .
tat tva yami . . . (RV. 1.24.11a)
. . . that I requestfrom you.
c. du;du;he! :pa;yaH . . . ;
a;Sa;m,a
duduhre payah. . . . rs.im (RV. 9.54.1)
(from)
They milked the milk

the sage.
(ASK / ENTREAT)
(MILK)
28
H. H. H OCK
d. :de!;va;a;na;su-.a =+aH y!a;a;ma-.ja;ya>+.s,a
deva n a surah. yajam ajayams
(MS. 1.9.8)
The asuras won the sacrifice (from) the gods.
e. ya;d;mua;Sa;a;ta . . . :p!a; a;Ma ga;aH
yad a mus.n.ta . . . pan.m
ga h. (RV. 1.93.4)
(from) the miser.
. . . when you robbed the cows
f. ta;a;nsa;h;~:Ma d;q+.yea;t,a
(WIN)
(ROB)
tan sahasram
(PUNISH)
dan.d.ayet (Manu 9.234)
. . . 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)
His speech is spoken.
.sa :he!;nde+a;ea;! ;a;sa
14.1.1.19)
sa ha + indren.a + ukta a sa (SB.
He was addressed by Indra.

b. .=!+a;ya;
a;vRa;BUa;a;ta:=+a;ya;tea . . .
rayr vbhutir yate . . . (RV. 6.21.1)
Great wealth is implored.

.=+a:ja;a me!a;Da;a;a;Ba:=, IR +.ya;tea
ra ja medha bhir yate (RV. 9.65.16)
The king is implored
with insight.
(SPEAK)
(ASK / ENTREAT)
c. . . . ;
a;pa;ba;tua du!;gDa;mM!a;Zua;m,a
. . . pibatu dugdham am
(MILK)
su m (RV. 5.36.1)
11
May he drink the milked (= expressed) soma.
du!;h:a;ntea . . . ;De!a;na;va;ea
11 Hettrich (1994) cites dugdh
o am
su h. (RV 3.36.6d), glossed as der ausgemolkene Stengel, as an example of the source, rather than the substance NP
becoming the passive subject. However, the present example suggests that am
su
has become simply an epithet of soma, the ingestible substance produced in the
ritual.
29
duhyante . . . dhenavo (AV. 7.73.2)

The cows are being milked.
d. .~va;mR!a:+:tva;ta;a ;a:j!a;ta;m,a
svàr marutvata jitam (RV. 8.76.4)
(WIN)
The sun has been conquered

by (Indra) accompanied by
the Maruts.
;a;su!a:=+a yu!a;Da;a ;a:j!a;ta;a
a sur yudha jita (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 . . .
ned v eva nagna iva mus.ita iva sayata ity . . .
(ROB)
(SB.
1.2.2.16)
. . . lest he lie naked as it were, robbed as it were.
(31) a. ;
a;va:ja;ya;mua;+:~tEaH
vijayam uktas taih. (Kathas. 18.247)
. . . (was) told (about) the victory by them.
b. v!a;Za;a;a;ma;nde+a ya;a;a;.c!a;taH
vasa m ndren.a yacitah. (AV. 12.4.50)
. . . asked by Indra for (his) cow
(ASK / ENTREAT)
x! a :pa;yaH
c. . . . na;Ba;ea du;h:a;tea ;Ga;tM
. . . nabho duhyate ghrtam
paya(h.) (RV. 9.74.4)
12
The cloud is milkedfor ghee, milk.
d. . . . .sa;vRa:$ya;a;a;nMa va;a .ja;a;ya;tea
. . . sarvajyanim
va jyate (KS. 29.6)
. . . or he is defeated a complete defeat.
12 Hettrich
(SPEAK)
(MILK)
(WIN)
notes that the example is formally ambiguous, since both nabhah.

and ghrtam
payah. can be both nominative and accusative. The singular on the
the initial placement of nabhah favor an interpretation that nabhah is the
verb and
.
.
subject; but agreement with the nearest antecedent of the conjoined elements
ghrtam and payah. is a possible alternative. Geldner takes ghrtam and payah. to
bethe subject.
30
H. H. H OCK
e. ;
a;h!;mea;va :p!a;a;Ra mua;
a;S!a;ta;a va;na;a;a;na
himeva parn.a mus.ita vanani (RV. 10.68.10)
. . . like trees robbed of their leaves

by winter13
(ROB)
While this much is known, questions remain. First, it is not

clear why only certain verbs that are subcategorized for two complements have double-direct-object constructions. Even more puzzling is why some verbs belonging to the semantic sets SPEAK,
ASK / ENTREAT , MILD , WIN , ROB fail to enter into double-directobject constructions. Consider kath which to my knowledge only
takes the addressee in the dative, or hr take away which takes the
ablative for the source person. Presumably,

the verbs participating in the double-direct-object construction must be specifically
listed in the lexicon (together with alternative case markings, if
any; see Hock 1985. But this does not explain why many of the
verb classes exhibit similar behavior in other Indo-European languages; see Hock 2012c.
Problems of a different sort arise regarding A. 1.4.5253 which
classifies the causees of certain verb classes as karman (optionally
for hr and kr) and leaves others as kartrs which, being anabhi
hita,surfacein the instrumental. As Speijer
(1886: 3637 with
reference) notes, a very different situation obtains in the classical
language, irrespective of verb class:
If one wants to say he causes me to do something, it is
by his impulse I act, there is room for the [accusative
causee], but if it be meant he gets something done by
13 As
noted by Hettrich, formally this passage is ambiguous. However, the

context favors the interpretation given here: himeva parn.a mus.ita vanani br
haspatinakrpayad valo ga h. Like the trees robbed of their leaves by winter, Vala
the cows (taken from him) by Brhaspati. (Geldner: Wie die Bume
mourned for
Vala die von Brhaspati (geraubten)
ihre vom Frost geraubten Bltter so vermite
Khe.)
31
me, I am only the agent or instrument through which

he acts, the instrumental is in its place . . .
As shown in Hock 1981, this pragmatically sensitive marking convention is already found in the early Vedic prose texts; see (32)
and (33), where the verb in (32) belongs to the categories of verbs
that by A. 1.4.52 should take a karman, and the verb in (33) does
not. The phenomenon can therefore not be attributed to post-Pan.inian innovation. Rather, the difference between Pan.ini and the
textual tradition of Madhyadesa most likely reflects a difference in
regional 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;tea
dya vaprthiv bhuvanes.u a rpite (TS. 4.7.13.2)
and earth have been made to reach the worlds.
Heaven
b. ! ;mMua .te!a Zua;k, ;C+.tua . . . ya;me!a;va :de;
a;! ta;ma;~ya ua! ;Da;a . ca Zua! ;.ca;a . ca;pRa;ya;a;ta
amum
te su k. rchatu . . . yam eva dves..ti tam asya ks.udha
ca suca ca + arpayati
(TS. 5.4.4.12; similarly passim)
Your pain should go to him; whom indeed he hates, to
him he makes his hunger and pain go.
Or: . . . him he afflicts with his hunger and pain.
(33) a. ;ea;Sa;Da;a;=e +va :P+.lM g{a;a;h;ya;a;ta
os.adhr eva phalam
grahayati (KS. 26.5)
He causes the plants to take fruit.
b. va:+:ea;nEa;v!a Bra;a;txa;v.yMa g{a;a;h;a;y!a;tva;a b.ra; ;a;a .~txa;ua;tea
varun.enaiva bhra trvyam
grahayitva brahman.a strn.ute
(TS. 2.1.8.2, similarly KS. 13.4)

Having caused Varun.a to seize the enemy, he lays him
low with the sacrificial formula.
32
3.6
H. H. H OCK
Asamartha compounding
The syntax of constructions such as (34), called asamartha

compounds by Gillon (1993), was first addressed in Patajalis
commentary on A. 2.1.1 (Kielhorn-Abhyankar edition 1.359.21
361.24). The fact that in the reading of (34), devadattasya does
not modify the head (kula) of gurukulam, but the non-head (guru), leads to a lengthy discussion, with one side arguing that the
interpretation is acceptable and another one that it is not, since
only heads can have external modification. The issue is in effect
left unresolved. By contrast, Bhartrhari (Vakyapadya 3.14.47) accepts the grammaticality of such structures if the non-head is a
relational noun, such as father, son, teacher, student.
(34) :de;va;d:a;~ya gua: - ku+:l+.m,a
devadattasya guru-kulam Devadattas teachers family
That this issue is not just an idle invention of the grammarians
is shown by the fact that structures of this sort do in fact occur,
such as (35a); and as shown by (35bc) other structures that would
not be amenable to Bhartrharis account are also found. As far as I
can tell, Whitney (1889: 515) was the first western scholar to note
the existence of such structures.
(35) a. d;nta;a;Ga;a;ta;~ya . . . du;
a;h;tuaH :pa;+a;a;va;tya;a ;Da;a:a;a;~va;sa;a;h;m,a
dantaghatasya . . . duhituh. padmavatya dhatrsvasaham
(Vetalapacavim
sati, Emeneau edition, 16.1617)
I am the sister of the nurse of Padmavat, the daughter of
Dantaghata . . .
b. ta;~ya;Ma :a;=+:
/a;gDa;dx;a;a .sUa;a;.ca;ta;a;a;Ba;l+.a;SaH
tasyam
a 3.9.16)
snigdhadrs..tya sucitabhilas.ah. (Sakuntal
. . . whose affection was indicated by his gaze fixed on

her.
33
c. ;a;.ca:a-:pra;ma;a;a;Ta;na;a ba;a;l+.a :de;va;a;na;a;ma;

a;pa .sua;nd:=+a
citta-pramathin bala devanam api sundar (Nala 1.18;
MBh. 3.50.13)
A beautiful girl disturbing the minds even of the Gods.
In a recent paper, Kiparsky (2009: 48) argues that structures
like (34) can be interpreted as apparent syntax/morphology mismatches [that] should be treated in terms of a semantic inheritance
mechanism whereby properties of individuals become properties
of groups to which individuals belong, as in a laughing group of
children which really means a group of laughing children rather
than merely a laughing group consisting of children.
A more comprehensive analysis is that of Gillon (1993, 1995).
Adopting Bhartrharis notion of relational noun, he concludes
are associated with a karaka or whose meaning
that non-heads that
presupposes some kind of relation are permitted to take external
heads. This allows him to account not only for the type (34)/(35a)
but also for (35b), under the assumption that unlike languages such
as English, Sanskrit allows transmission of unsaturated argument
positions not only for heads but also for non-heads; see Table 3.14
Molina Muoz (2014) notes that while Gillons account works
for (35a) and (35b), it does not for the type (35c), where citta is not
a relational noun and where devanam is not an argument of citta,
but syntactically merely an adjunct. She therefore argues that a
different explanation is required which, in principle, works for all
subtypes under (35). Starting out with Schufeles notion of liberation or node erasure and Pan.inis account for compounds as
combining full words (saha supa, A. 2.1.4), she proposes to derive
(productive) compounds in a post-syntactic component along the
lines of Halle and Marantzs (1993) distributed morphology. Compounding, under this account, can take place between two neighboring semantically compatible words in the output of the syntax
14 This
is Molina Muozs (2014) rendition of Gillons diagram.
34
H. H. H OCK
NPINSTR < >
NPLOC
NINSTR <LC>
NLOC
A <LC, PT>
NINSTR
tasyam on her
snigdha fixed <LC, PT>
drs..tya gaze
Table 3
Non-head relational nominal governing an external head
and after liberation, which erases syntactic nodes and potentially,
but not necessarily moves words or phrases to other positions in
the clause. Example (35c) serves as an excellent example, since
in this case the movement of devanam and bala out of their original complex NP provides positive evidence for node erasure; see
Table 4. (Note that api moves along with devanam because of
its quasi-clitic nature.) Structures such as (35a) and (35b), then,
would also be analyzed as involving node erasure, but without any
Node erasure
Movement
Compounding
Suffix erasure
Final output
[NP [AP [NP [NP [N devanam] [PCL api]] [N cittam]]

[A pramathin]] [A sundar] [N bala]]
devanam=api cittam pramathin sundar bala
.
+
t1 cittam pramathin t2 bala devanam=api1 sundar2
[cittam-pramathin]
cittapramathin
cittapramathin bala devanam api sundar
Table 4
Node erasure, movement, and compounding in (35c)
35
overt movement.
Functional issues that should be of interest

to computational approaches
In this section I address several functional or usage aspects of

particular grammatical phenomena, including genre-based usages,
that should be of interest to those engaged in computational analysis of Sanskrit texts. In fact, attention to genre differences and
their influence on the choice of syntactic processes can also pay
off for formal analyses. Consider the interaction between genre
and word order discussed in 3.1 and the more comprehensive discussion in Hock 1997, 2000. Conversely, as we will see, discourse
phenomena can raise interesting questions for formal analyses.
One difficulty with functional investigations is that they tend
to be highly specific. For instance, Gonda (1942) observes that
although, using recent terminology, Sanskrit is a pro-drop language, personal pronouns are common in dialogues. Jamison
(1991b,a) similarly focuses on dialogues in Vedic prose which, as
she notes, exhibit interesting differences from the technical discourse that surrounds them, including a much greater use of deictics such as idam, adas, rather than demonstratives such as tad,
etad. As regards the Vedic-prose difference between tad and etad,
Hock (1982a) finds that etad is preferred in cataphoric contexts,
while tad is anaphoric or unmarked. My impression is that this
difference holds also for classical Sanskrit. M. R. Hale (1991)
comes to different conclusions for the Taittirya Samhit
a. This is
an issue that deserves fuller study. Still in the area of pronoun usage, Van de Walle (1991, 1993: 11920) notes that, while plural
may be used for politeness (as in bhavantah. for bhavan), second
plural pronouns are rarely used with singular reference. Normally
either the second singular pronoun or a form of bhavat is used
36
H. H. H OCK
sometimes even within the same (complex) sentence. (Van de

Walles work, however, is not limited to pronoun use but addresses
the broader issue of linguistic politeness and distancing in
classical Sanskrit.)
Tsiang-Starcevics (1997) study is a broad investigation of the
use of finite (i.e. relative clause) vs. non-finite (participle, converb) subordination in Sanskrit narratives. See also Tsiang and
Watanabe 1987, which focuses on the rhetoric of fable narratives.
An important finding is that non-finite structures dominate in narrative portions and that finite relative clauses tend to be restricted
to dialogues embedded in the narratives.
In his monumental study of relative-clause syntax, Hettrich
(1988: 74557) claims that appositive or non-restrictive relative
clauses were a feature of Proto-Indo-European poetic language,
surviving in mantra Vedic, but becoming rare in Vedic prose and
disappearing in post-Vedic. Hock (1993) argues that the difference
between mantra and prose can be explained in terms of genre and
that non-restrictive relative clauses continue to be used in the classical language. It would be interesting to investigate whether different classical genres exhibit differences similar to those between
mantra and prose Vedic.
In the following I take a more detailed look at fronting and extraposition, two general movement processes, both of which have
interesting discourse, genre, and grammatical characteristics.
4.1
Fronting
Fronting processes play a significant role in a number of different

genres, both in Vedic and in the classical language.
4.1.1
Initial strings in Vedic
Vedic-prose texts are characterized by complex initial strings, such

as (36a), consisting of topicalized elements (commonly nominal or
37
pronominal), demonstrative and other pronouns, as well as particles. Similar, but generally less complex and shorter strings are
also found in the mantra language. M. R. Hale (1987, 1996) proposes syntactic movement accounts, with some prosodic readjustment, for strings of this sort; see also Keydana 2011. By contrast,
Hock (1982a, 1996, 1997) proposes a fully prosodic account, both
in terms of a template defining their linear order and in terms of
their domain of occurrence. Especially relevant in this regard is
the evidence of the mantra language, where initial strings may occur line-initially (36b) or even post-caesura (36c) in run-on lines,
i.e. in prosodically defined domains, rather than the syntactically
defined domain of clause-initiality.
(36) 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;ntM
/ /a ;
a;v!a;k+:ma;tea
etami v eva + es.a j etasmai vis.n.ur j yajo vikrantim
i vikra
(SB.

1.1.2.13)
mate
This Vis.n.u, the sacrifice, steps this (world-conquering
three-fold) stepping for him (the sacrificer).
b. h;ea;ta;a:=M ;
a;v!a:(;a;vea;d;sM!a ( ) .sMa ;
a;h tva;!a ;
a;va;Za I! +.nDa;t,aO;
hotaram
visvavedasam
(.) sam
h tva vsa indhate (RV.
1.44.7ab)
For the clans light you as the all-knowing hotr.
c. ;
a;va;pr!aH :prea;! H: .sa h:ea;Sa;Ma ba;BUa;va
vprah. pres..thah.: sa hy e` s.a m
babhuva (RV. 10.61.23c)
For he was of/for them the dearest singer.

To my knowledge, strings of this sort do not survive in the
post-Vedic language; but see 4.1.3 for linkage strings.
4.1.2
Predicate-Subject order
The fact that predicates frequently precede their subjects, as in

(37), has attracted western scholars attention from an early period.
38
H. H. H OCK
Speijer (1886: 10) considers this the normal order in Sanskrit, but
notes that Pronouns, it seems, may be put indiscriminately before or behind their noun-predicate (1886: 10); and in his 1896
monograph, he adds the further restriction that subject pronouns
normally precede their predicates. Delbrck (1878: 27), by contrast, while acknowledging the pervasive presence of predicatesubject order in Vedic prose, considers it marked: Der Grund fr
diese Stellung liegt auf der Hand. Das Subject is nmlich bekannt,
das Praedikatsnomen aber bringt etwas Neues hinzu, und tritt also
nach dem allgemeinen Gesetz der occasionellen Wortstellung vor.
[The reason for this position is obvious. The subject is known,
but the predicate nominal adds something new and hence moves to
the front according to the general law of occasional word order.]
(37) a. .tea:ja;ea ;vEa b.ra; ga;a;ya:a;a
TEJO vai BRAHMA g
ayatr (KS. 25.5)15
The gayatr is brilliance, brahman.
b. ma;na;ea ;vEa bax ;h;d, va;a;g,a .=+Ta;nta:=+m,a . . .
rathantaram . . . (JB. 1.128)
MANO vai brhad V AG
The brhat is mind, the rathantaram is speech . . .
For the post-Vedic language, Lahiri (1933) finds that contrary

to Speijer, the normal order is subject-predicate; see also Hock
2013.
Even for Vedic prose it is possible to show that predicate-initial
order is marked. First, there are many examples like (38) in which
only part of the predicate appears in initial, pre-subject position,
while the rest remains stranded downstairs. Second, in longer
series of equational structures, the order predicate-subject tends
to break down, reverting to subject-predicate order; (39). Finally,
as noted in Hock 2013, predicate-initial order is regular only in a
15 Here as elsewhere in this discussion, subjects are marked by bold, predicates
by SMALL CAPS.
39
sub-genre of Vedic prose, namely passages that equate instruments

of the sacrifice (such as the meters, or body parts of the sacrificial
animal) with more abstract or supramundane phenomena or qualities such as the Year (writ large), the Mind (again, writ large), or
the World and its components.
(38) a. mua! ;K.ya;Ea va;!a ;a;va;M!a ya;!a;~ya .~va;ea
4.1.5.16)
MU KHYAU v
a a vam
YAJA SYA svo (SB.

We two are the chiefs of the sacrifice.
b. BUa;ya;a;n,a ;vEa b.ra;a;;aH a;
a:a;ya;a;d,
vai br
(AB. 7.15.8)
AN
BH UY
ahman.ah. KS. ATRIY AD
A brahmin is better than a ks.atriya.
(39) o+.Sa;a va;a :(;a;~ya mea;Dya;~ya ; a;Za:=H
.sUa;yRa;(a:ua:=,
va;a;taH :pra;a;a;ea
Ra:(;a;a;na:=H
v.ya;a:a;ma;a;+;a;vE
.sMa;va;tsa:= ;a;tma;a
+:va;DyMa ;a;sa;k+:ta;aH
;a;sa;nDa;va;ea gua;d;aH
ya;kx+:a * :+:ea;ma;a;na;(a :pa;vRa;ta;aH
v
US. A
a asvasya medhyasya sirah.
S URYA S caks.ur
V ATAH
. pran.o
ANARAH
vyattam AGNIR VAI SV

.
SA MVATSARA a
tma
.
uvadhya
m
SIKAT AH
SINDHAVO gud
ah.
. (BAUM.
yakrc ca klomanas ca PARVAT AH
1.1.1)
The head of the sacrificial horse is the dawn; the eye, the
sun; the breath, the wind; the open mouth, Agni Vaisvanara;
the body, the year . . . ; the food in the stomach, the sand; the
blood vessels, the rivers; the liver and lungs, the mountains.
40
H. H. H OCK
Predicate-initiality, thus, is not basic, but results from (partial) fronting and, moreover, is a feature sensitive to genre and
discourse. Moreover, as in the case of major constituent order,
sensitivity to genre differences makes it possible to argue in favor of one formal analysis in preference to another. (See also 5
below.)
4.1.3
Narrative linkage and related issues
The use of converbs as narrative linkers at or near the beginning

of the clause, as in (40a,d) is often considered a feature reflecting
Dravidian contact; see e.g. Bloch 1930; Emeneau 1971. Under
the name tail-head linkage, the phenomenon of nonfinite recapitulation has been shown to be more widespread in (folk) narratives, irrespective of syntactic typology (Thompson and Longacre
1985: 20913); and under the term catena it has been shown to
occur also in Ancient Greek (Migron 1993). In Sanskrit, the use
of converbs alternates with that of ta-participles (40b) and locative absolutes (40c) in a system of switch reference (Haiman and
Munro 1983)), where converbs indicate kartr continuity, transitive
ta-participles a switch to the karman of thepreceding action, and
locative absolutes a switch to some other actant. (See Hock Forthcoming(a).)
(40) a. na;a:=+d;~ya tua ta;d;a;k+.aM (rua;tva;a va;a;k+.a;
a;va;Za;a:=+dH
:pUa:ja;ya;a;ma;a;sa ;Da;ma;Ra;tma;a .sa;h; a;Za;Sya;ea ma;h;a;mua;a;naH
naradasya tu tad vakyam
srutva vakyavisa radah..
pujayam a sa dharmatma sahasis.yo mahamunih...
b. ya;Ta;a;va;tpUa;a:ja;ta;~tea;na :de;va;
a;SRa;na;Ra:=+d;~ta;d;a
;a;pxa;E;va;a;Bya;nua;a;a;taH .sa .ja;ga;a;ma ;
a;va;h;a;ya;sa;m,a
yathavat pujitas
tena devars.ir naradas tada.
a prs..tvaivabhyanujatah. sa jagama vihayasam..
41
c. .sa mua;hU;ta ga;tea ta;a;sma;nde

/ / / ;va;l+.ea;kM mua;a;na;~ta;d;a
.ja;ga;a;ma ta;ma;sa;a;ta;a:=M .ja;a;+:v.ya;a;~tva;
a;va;dU:=+taH
sa muhurtam
gate tasmin devalokam
munis tada.
jagama tamasatram
jahnavyas tv aviduratah...
d. .sa tua ta;a:=M .sa;ma;a;sa;a;d;a ta;ma;sa;a;ya;a ma;h ;a;pa;a;taH
; a;Za;Sya;ma;a;h ;a;~Ta;tM
/ / a :pa;a:(;eRa dx;;a ta;a;TRa;ma;k+:dR;ma;m,a
sa tu tram
samasadya tamasaya mahpatih..
sis.yam a ha sthitam
parsve drs..tva trtham akardamam..
(Ramayan.a 1.2.14)
When the eloquent one (Valmiki) had heard this speech of
Narada, the righteous great sage (Valmiki) and his disciples
honored him. When the divine seer Narada had been duly
honored by him at that time, he went to heaven, having asked
for permission to leave and received it. When he (= Narada)
had gone to the heavenly world, the sage (Valmiki) at that
time went after a while to the bank of the Tamasa, not far
from the Jahnav (= the Ganga). But when he reached the
bank of the Tamasa, the ruler of the earth, seeing a bathing
spot free from mud, spoke to his pupil who was standing next
to him.
As shown in Hock 1994a,b, narrative discourse linkage is
accomplished by a variety of other fronting processes, including the fronting of demonstratives, finite verbs, conjunctions, or
conjunction-like adverbs; see for instance (41). Moreover, verbs,
whether finite or non-finite, may be accompanied by complements
and other satellites; and all the fronting processes may apply together, yielding linkage strings that can become quite complex,
as in (42).
(41) a. ta;TEa;va;a;sa;a;
a;d;d;BeRa;Sua Ba;a;ma;ea Ba;a;ma;pa:=+a;k+:maH . . .
.sa :pra:ja;a;TeRa :pa:=M ya;a;ma;k+.=+ea;t,a . . .
42
H. H. H OCK
ta;m,a ;Bya;ga;.cC+.d " ;;
a;SRa;dR ;ma;na;ea na;a;ma . . .
tMa .sa Ba;a;maH . . . ta;ea;Sa;ya;a;ma;a;sa ;Da;mRa;
a;va;t,a
tathaivasd vidarbhes.u bhmo bhmaparakramah.. . . . ..
sa prajarthe param
yatnam akarot . . . .
tam abhyagacchad brahmars.ir damano nama . . . ..
tam
sa bhmah. . . . tos.ayam a sa dharmavit.
(MBh. 3.50.57)
Likewise there was among the Vidarbhans Bhma of terrible prowess . . . He made the utmost effort for the sake of
progeny . . . To him came a brahmin sage, named Damana
. . . Him that Bhma gladdened, knowing dharma.
b. ;b.ra;va;a;d, ;tua;pa;Ra;~ta;m,a . . .
abravd rtuparn.as tam . . . (MBh. 3.70.16)
(Then)Rtuparn.a said to him . . .
c. ta;ta;ea Y;nta;a:=;a;ga;ea va;a;.cMa v.ya;a:ja;h;a:= . . .

tato ntariks.ago vacam
vyajahara . . . (MBh. 3.50.19)
Then the bird said a speech . . .
(42) a. ta;ta;~ta;a .nEa;Sa;DMa dx;;a . . .

tatas ta nais.adham drs..tva . . . (MBh. 3.52.14)
16
adv. + tad + SAT + conv.
Then they, having see the Nis.adhan . . .
b. .tea tua hM ;sa;aH .sa;mua;tpa;tya . . .
te tu hams
ah. samutpatya . . . (MBh. 3.50.21)
tad + conj. + SAT + conv.
But those swans, having flown up . . .
c. :pra;
a;va;Za;nta;Ma tua ta;Ma dx;;a . . .
pravisantm
tu tam
drs..tva . . . (MBh. 3.62.20)
16 Abbreviations:
adv. = conjunction-like adverb, conj. = conjunction, conv. =

converb, pres. pple. = present participle, SAT = satellite, tad = demonstrative
pronoun
43
pres.pple + conj. + tad + conv.

But having seen her entering . . .
As noted in Hock Forthcoming(a), linkage strings are the
strongest indicators of narrativity, especially in the epics. But
even in fable literature, which adopts a more concise narrative
style (Tsiang and Watanabe 1987), linkage strings are more clearly
linked with narrative than the use of individual linkers by themselves.
4.2
Extraposition
The functions of extraposition to post-verbal position are an issue

that deserves further study. I am aware of three studies that treat
certain aspects of the phenomenon.
4.2.1
Gondas amplified sentences
The earliest is Gonda 1959, with its principle of amplification

a common phenomenon both in mantra Vedic and in the epics, see
e.g. (43). As Gonda puts it, the structure before and including the
verb (in this case agnm .le) forms a complete sentence or proposition in itself; what follows is additional information that elaborates
on what precedes (in this case on the object agnm). In both mantra
Vedic and the epics, extraposition serves to expound on the good
and desirable qualities of the deity or the hero or heroine.
a;a;Le [:pau :=+ea;
a;h ;tMa y!a;a;~ya :de!;va;ma;x a;tva:
(43) a. ! ;a;+;a;m
!
! / ja;m,a h;ea;ta;a:=M .=+!a;Da;a;ta;ma;m,a]
agnm .le [purohitam
yajasya devam rtvjam. hotaram
ratnadha tamam] (RV. 1.1.1)

foremost God of the sacrifice, the
I invoke Agni, the
priest, the hotr, most bestowing treasure.
b. ta:a .sma Bra;a:ja;tea BEa;ma;a [.sa;va;Ra;Ba:=+a;BUa;

a;Sa;ta;a
.sa;Ka;a;ma;Dyea Y;na;va;d;a;a;*: +.a ;
a;va;d;au;tsa;Ea;d;a;ma;na;a ya;Ta;a
44
H. H. H OCK
...
;a;.ca:a;pra;ma;a;a;Ta;na;a ba;a;l+.a :de;va;a;na;a;ma;
a;pa .sua;nd:=+a ]
tatra sma bhrajate bhaim [sarvabharan.abhus.ita.
sakhmadhye navadyan g vidyut saudaman yatha.
...
cittapramathin bala devanam api sundar..]
(Nala 1.1214; MBh. 3.50.1213)
There Bhaim radiated, adorned with all ornaments, surrounded by her friends, having entirely praiseworthy
limbs, like monsoon lightning . . . the girl robbing the
minds even of the Gods, beautiful.
4.2.2
Purpose datives in Vedic prose
A peculiarity of Vedic prose, which generally is heavily and

some might say, unimaginatively verb-final, is the frequent
extraposition of dative purpose phrases, as in (44). Delbrck
(1888: 25) considers such structures a Satzanhang (clause appendix). More specifically, Hock (2014) argues for a genre-based
use of extraposition, indicating a benefit that extends beyond the
simple sacrificial action. The relation between the preceding structure and the extraposed dative phrase can thus be interpreted as
an iconic reflex of the contrast between ritual-internal action and
ritual-external benefit.
(44) a. ta;a;nya;a;hu;na;Ra;na;ea;pea;tya;a;a;na na;a;nea;va va;a I+.mea l+.ea;k+:aH O;;Sa;Ma l+.ea;k+:a;na;Ma
;
a;va;Dxa;tya;a I+.a;ta
tany a hur nanopetyani. naneva va ime lokah.. es.a m
lokanam
vidhrtya iti (JB. 2.218)
They saythese (samans?) are to be undertaken variously
these worlds are various, as it were for keeping apart
these worlds.
45
b. t!a;a;ma;nd+a;+;a ;!a ;nua;s!a;ma;ta;nua;ta;a;m,a :pra:ja;!a;na;a;}.pr!a:ja;a;tyEa
tam indragn anusamatanutam. prajanam prajatyai (SB.
4.3.1.2)
Indra and Agni preserved him (Soma); for the procreation of creatures.
c. ;
a;d;gva;;
+va;a;ta Bra;a;txa;v.ya;~ya;a;pa;nua:yEa
digvad bhavati bhratrvyasyapanut[t]yai (PB. 12.4.10)
It contains the worddirection, for repelling the enemy.
4.2.3
Kartr backing and extraposition, and politeness
Wallace (1984) finds that backing of kartrs to post-karman position

in passives, gerundives, and ta-participleconstructions, as in (45a),
is a common feature of dialogues in the Vetalapacavim
sati, while
narrative portions typically have the kartr in the unmarked initial
he argues that this reposition. Based on the contextual evidence,
ordering serves purposes of politeness, indicating either modesty
on the part of the speaker or deference to an addressee. As it turns
out, the phenomenon is not limited to passive-like structures but
also occurs in actives; see e.g. (45b). Moreover, similar considerations probably account for the frequent extraposition of addressees
or speakers, as in (46),17 and both reordering and extraposition are
a widespread feature of dialogues. What seems to be shared by
both processes is that they downgrade the addressee or speaker,
thereby avoiding the threatening of face (see Van de Walle 1991,
1993 on this latter issue).
(45) a. ta;va :pa;ea;Sa;a;ma;a;va;a;Bya;Ma k+:tRa;v.ya;m,a
tava pos.an.am a vabhyam
sati,
kartavyam (Vetalapacavim
Emeneau edition) We will take care of you.
17 Some
of these, such as (46a), are included in Wallaces data.
46
H. H. H OCK
b. ta;sma;a;n}.a;Ma Ba;va;Ma;~tya:ja;tua
tasman mam
bhavams
sati, Eme tyajatu (Vetalapacavim
neau edition, 70.34) Therefore may your lordship release me.
(46) a. ku+.a ga;tva;a ;a;~Ta;tM

/ / a Ba;va;ta;a
kutra gatva sthitam
bhavata (Vetalapacavim
Where did your lordship go and stay (so long)?
b. ya;d;a;
a;d;Za;a;ta :de;vaH
yad a disati deva(h.) (Vetalapacavim
As the lord commands.
c. yua;va;ya;ea:=, . . . ;a;Ka;l+.
mea;va k+:a;Ta;tMa ma;ya;a
yuvayor . . . akhilam eva kathitam
maya
(Vetalapacavim
sati, Emeneau edition 16.2021)
I have told everything to the two of you.
d. va;.ca;sa;a ma;na;sa;a . cEa;va ya;Ta;a na;a;a;Ba;.ca:=+a;}ya;h;m,a
.tea;na .sa;tyea;na ;
a;va;bua;Da;a;~ta;mea;va :pra;
a;d;Za;ntua mea
vacasa manasa caiva yatha nabhicaramy aham .
tena satyena vibudhas tam eva pradisantu me.. (Nala
5.18)
As I do not transgress by speech or mind, by that truth let
the very wise ones (the Gods) point him (Nala) out to me.
Conclusions and implications for further research
What is remarkable is that extraposition seems to have at least two

very different purposes. In the context of politeness it serves to
down-grade the addressee or the speaker, in Gondas amplification as well as the more restricted phenomenon of Vedic-prose
47
purpose-dative extraposition it serves to provide additional and important information. Presumably there would be a prosodic difference, with politeness extraposition being realized with low or
reduced pitch, while there would be no such prosodic reduction
in amplifications and, perhaps, even a raised pitch for Vedic-prose
purpose-datives. Nevertheless, there seems to be no clear syntactic
difference between the different types of extraposition.
Following the reasoning in Hock 1993 (for different pragmatic
uses of non-restrictive relative clauses), it is possible to account for
this situation by assuming that extraposition is one syntactic process, which leads to syntactically marked structures. Such marked
structures, in turn, make it possible for the speaker to invite the
hearer to assume that there is a special reason for using them, along
the lines of Grices (1975) notion of implicature or invited inference, with the precise pragmatic inference being determined by
discourse and genre.
If this line of reasoning is correct, we must conclude that functional accounts, however interesting and important for textual interpretation they may be, cannot substitute for formal syntactic accounts, and that the latter must be formulated irrespective of the
uses to which different possibilities permitted by the grammar can
be put.
At the same time, as already noted, functional accounts can be
helpful in assessing conflicting formal accounts such as the issue
of major constituent order (3.1), the question of subject-predicate
vs. predicate-subject ordering (4.1.2), or the phenomenon of
marked kartr backing (4.2.3). In fact, the idea of kartr backing
makes it possible
to account for a large number of structures
with
predicate-subject order, beyond the Vedic-prose equational structures discussed in 4.1.2. See e.g. the examples in (47), where
kartr or subject backing can be attributed to the same politeness
concerns
as in 4.2.3: modesty on the part of the speaker or deference to an addressee.
48
H. H. H OCK
(47) a. .sa;a;Dva;a Ba;va;ta;a

sadhv bhavat (Vetalapacavim
sati, Emeneau edition,
68.13)
Your ladyship is good. You (polite) are a good
woman.
b. . . . :pra;ma;a;Ma tua Ba;va;nta;a;~:
/ a;d;Zea:(;a:=+aH
. . . praman.am
tu
bhavantas
tridasesvarah. (Nala 4.31;
MBh. 3.53.21)
. . . but your lordships, rulers of the thirty(-three Gods),
are the authority.
c. a;a;a;nta;Z
/ / a;a;l+.ea na;a;ma k+:a;pa;a;a;l+.k+:ea Y;hM ma;h;a;ya;ea;ga;a
ks.a ntislo nama kapaliko ham
mahayog
(Vetalapacavim
I (am) a mendicant, Ks.a ntisla by name, a great yogi.
d. .=+a:ja;pua:a;a;va;a;va;Ma :pa;yRa;f;na;Za;a;l+.a;va:a;a;ya;a;ta;Ea
rajaputrav a vam
paryat.anaslav atrayatau
(Vetalapacavim
We are princes, in the habit of wandering come here.
There may even be cases where functional evidence may create
interesting challenges for formal syntactic accounts. Let us take
another look at (43b), specifically its passage cittapramathin bala
devanam api sundar, which in Table 4 was analyzed as involving
movement of denanam api and sundar out of their matrix NP. That
the structure involves movement is suggested by the particle api
which indicates special emphasis on the preceding devanam even
of the Gods; and sundar can be considered an elaboration
la Gonda of cittapramathin bala, which would be a complete
structure in its own right.
From the functional, pragmatic perspective such an account is,
I believe, entirely reasonable. From the formal perspective, however, the account presents a challenge, since it operates with the
49
idea of extraposition within a phrase, rather than a sentence. More

than that, the phrase itself has been extraposed within the larger
matrix sentence. Put differently, we would have to assume an extraposition within an extraposed structure.
Clearly, such an account is highly unorthodox, and its lack of
orthodoxy might be taken to favor the Staal/Gillon wild-tree account which does not stipulate linear order within phrases. However, under a wild-tree account the functional difference between
(48a) and (48b) would merely be epiphenomal. The advantage
of operating with base-generated linear order and accounting for
all deviations through movement, whether within the phrase or beyond, is that it invites attempts to account for such differences; and
that, I believe, is a methodological strong point.
(48) a. ;a;.ca:a;pra;ma;a;a;Ta;na;a ba;a;l+.a :de;va;a;na;a;ma;
a;pa .sua;nd:=+a
cittapramathin bala devanam api sundar
b. :de;va;a;na;a;ma;
a;pa ;a;.ca:a;pra;ma;a;a;Ta;na;a .sua;nd:=+a ba;a;l+.a
devanam api cittapramathin sundar bala
Still, I would hope that advocates of the wild-tree approach
will continue pushing their account, testing its predictive power
and comparing it to theories that operate with base-generated linear order. Whatever the outcome, we are bound to gain additional
insights into Sanskrit syntax.
There is, moreover, ample room for further research, especially
on the syntax of classical Sanskrit which offers a much broader
range of texts and genres than the Vedic tradition and which, in
part because of that, has received much less coverage.
Areas of further research that I would personally find interesting are relative-correlatives as well as non-finite subordination,
agreement (especially in complex-numeral constructions), and the
extent to which post-syntactic distributed-morphology accounts
may provide insights.
50
H. H. H OCK
References
The following reference list includes only references that are not
included in the Sanskrit syntax bibliography; for references to publications dealing with Sanskrit syntax see the Sanskrit syntax bibliography (p. 399).
Andrews Avery Delano, III. 1975. Studies in the syntax of relative
and comparative clauses. Ph.D. diss. Cambridge, Mass.: MIT.
[Revised as Andrews 1985.]
. 1985. Studies in the syntax of relative and comparative
clauses. Outstanding dissertations in linguistics. New York;
London: Garland Pub. [Lightly retouched version of Andrews
1975.]
Arnold, Doug, Louisa Sadler, and Aline Villavicencio. 2007. Portuguese: Corpora, coordination and agreement. Roots: Linguistics in search of its evidential base, ed. by Sam Featherston and Wolfgang Sternefeld, pp. 928. Studies in Generative Grammar 96. Berlin; New York: Mouton de Gruyter. URL:
http : / / privatewww . essex . ac . uk / ~louisa /
newpapers/.
Bloch, Jules. 1930. Some problems of Indo-Aryan philology.
Bulletin of the School of Oriental Studies 5: 71956.
Bolkestein, A. Machteld. 2001. Random scrambling? Constraints
on discontinuity in Latin noun phrases. De lingua Latina
novae quaestiones, ed. by C. Moussy, pp. 24558. Louvain:
Peeters.
Cardona, George. 1976. Pan.ini: A survey of research. The Hague;
Paris: Mouton.
Dasgupta, Probal. 1980. Questions and relative and complement
clauses in a Bangla grammar. Ph.D. diss. New York: New
York University.
R EFERENCES
51
Farmer, Anne. 1980. On the interaction of morphology and syntax. Ph.D. diss. Cambridge, Mass.: MIT.
Grice, H. P. 1975. Logic and conversation. Syntax and semantics, 3: Speech acts, ed. by P. Cole and J. Morgan, pp. 4158.
New York: Academic Press.
Haiman, John and Pamela Munro, eds. 1983. Switch reference and
universal grammar. Amsterdam: John Benjamins.
Hale, Kenneth. 1975. Gaps in grammar and culture. Linguistics and anthropology: In honor of C. F. Voegelin, ed. by
M. D. Kinkade et al., pp. 295315. Lisse: de Ridder.
. 1983. Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory 3: 547.
Halle, Morris and Alec Marantz. 1993. Distributed Morphology
and the pieces of inflection. The view from Building 20, ed. by
Kenneth Hale and S. Jay Keyser, pp. 11176. Cambridge: MIT
Press.
Johnson, Cynthia. 2008. Mixed-gender antecedent agreement in
Latin. B.A. thesis. Urbana: University of Illinois.
Joshi, S. D. and J. A. F. Roodbergen. 1980. Patajalis Vyakaran.aMahabhas.ya: Vibhaktyahnika (P. 2.3.182.3.45). Pune: University of Poona, Centre for Advanced Study in Sanskrit.
Keenan, Edward L. 1985. Relative clauses. Language typology
and syntactic description, ed. by T. Shopen, vol. 2, pp. 14170.
Cambridge: Cambridge University Press.
Pullum, Geoffrey K. 1982. Free word order and phrase structure
rules. Proceedings of the Twelfth Annual Meeting of the North
Eastern Linguistic Society, ed. by James Pustejovsky and Peter Sells, pp. 20920. Amherst: University of Massachusetts
Graduate Linguistics Student Association.
Ross, John Robert. 1967. Constraints on variables in syntax.
Ph.D. diss. Cambridge, Mass.: MIT. [Revised as Ross 1986.]
. 1986. Infinite syntax! Norwood, New Jersey: Ablex. [Slightly
updated reprint of Ross 1967.]
52
H. H. H OCK
Snijders, Liselotte. 2012. Issues concerning constraints on discontinuous NPs in Latin. Proceedings of the LFG12 Conference, ed. by Miriam Butt and Tracy Holloway King, pp. 565
81. Stanford: Center for the Study of Language and Information, Stanford University.
Srivastav, Veneeta. 1988. Relative clauses in Hindi and learnability. Cornell Working Papers in Linguistics 8: 13360.
Thompson, Sandra A. and Robert E. Longacre. 1985. Adverbial
clauses. Language typology and syntactic description, ed. by
Timothy Shopen, vol. 1, pp. 171234. Cambridge: Cambridge
University Press.
Yoon, James Hye Suk. 1996. A syntactic approach to categorychanging phrasal morphology: Nominalizations in Korean and
English. Morphosyntax in generative grammar, ed. by Heedon Ahn, M. Y. Kang, Y. S. Kim, and S. Lee, pp. 6386. Seoul:
Hankuk Publishing Company.
Derivation and interpretation

in Pan.inis system
G EORGE C ARDONA
Abstract: Pan.inyas are aware from the outset that there are
two aspects to communication: speakers (prayoktr) form
they
utterances, which they utter to communicate what
wish to convey, and listeners (srotr) interpret what they
hear. In his derivation system accounting

for Sanskrit
usage, Pan.ini operates from the standpoint of a speaker:
meanings to be conveyed are made the causes for the formation of utterances (vakya) and their constituent words
(pada). The As..ta dhyay, moreover, is a set of rules for
such derivations which presuppose native speakers as their
interpreters, so that Pan.ini formulates certain metarules in
accordance with conventions known to such speakers. I
shall discuss evidence relating to these two aspects, emphasizing that Pan.inis derivations do indeed start from
meaning and taking up illustrative points concerning how
rules are interpreted.
Keywords: Pan.ini, grammar, syntax, semantics, metalanguage
Speakers and listeners
Pan.inyas recognize the interaction between a speaker (prayoktr)
and a listener (srotr) in communication. In the vrtti on his Vakya
53
54
G. C ARDONA
padya, Bhartrhari describes such an interaction.1 In his commen he puts it as follows. Speech units (sabdah) inextary on VP. 2.19,
.
tricably bound with the capacity to convey meanings (samsr
s..tasa (praktayah.) occur in the internal organ called buddhi of speakers
yoktrn.a m) who are possessed of speech (samavis..tavacam). They
there first in a state such that there is no sequence of sounds
occur
(kramasamh
aren.a). Through a speakers effort (prayatnena) in
setting the flow of air and articulators in motion, however, these
speech units reach a state of occurring in sequence (kramavrttitam) in the airflow (pran.e) and articulators (karan.es.u). Once
they
have gone through (anubhuya after experiencing) this state
that is, have been articulated as sounds they reach a state where
they enter into hearers also (pratipattrs.v api) with their sequence
2 Commenting on
done away with (kramapratyastamayenaiva).
VP. 1.44, Bhartrhari cites a verse saying that a unitary (avibha speech unit which signifies a meaning (arthaktah. undivided)
sya vacakah.) arises from divided entities (vibhaktebhyah.), that is,
from perceived sounds.3 Preceding this verse, Bhartrhari presents
two viewpoints, that of a speaker and that of a listener.

Due to
the activity (vyaparat) of speech organs, a speech unit acquires
modification (pratilabdhavikriyavises.ah.) that is, is manifested
as articulated sounds and becomes an object of the auditory
sense (srotranupat); it thus is employed (prayujyate) for mean1 Repeated
attempts have been made to discredit the traditionally held view

that the vrtti is an autocommentary, but, in my opinion, these have not succeeded.
See mostrecently Cardona 2012.
2 .sMa;sxa;;Za;+:ya;(
a k+:ma;sMa;h;a:=e+a .sa;ma;a;
a;va;;va;a;.ca;Ma :pra;ya;ea;X+:a;Ma Za;b.d;a bua:;d ;Ea :pra;ya;ea;na :pra;a;ea
k+.=+ea;Sua . ca k+:ma;vax ; a:a;ta;a;ma;nua;BUa;ya :pra;a;ta;pa:xa;Sva;
a;pa k+:ma;pra;tya;~ta;ma;yea;nEa;va .sa;ma;a;vea;ZMa :pra;a;ta;pa;d;a;ntea
(VPVr. 2.19 [201.1012]).
a h:a;a;h ;
a;va;Ba;+:ea ;
a;va;Ba;e+:Bya;ea .ja;a;ya;tea Y;TRa;~ya va;a;.ca;kH Za;b.d;~ta:a;a;TRa:+.pa;a;tma;a .sa3 O;;vM
;}Bea;d;mua;pa;ga;,ah;a;ta (VPVr. 1.44 [102.57]). The second half of the verse says with
that arises from the divided sounds that here a speech
regard to the speech unit
unit is identical with a meaning, in that it now achieves a non-differentiation
between speech unit and meaning in the buddhi of the hearer.
D ERIVATION AND I NTERPRETATION
55
ings (arthes.u) which are to be signified, in the guise of what brings

meaning to light (prakasakabhavena), always subordinate to what
is to be signified (pratyayyaparatantrah.). The speech unit which
is devoid of sequence (labdhanusamh
arah.) is thus the cause (nimittam) of a unit that has acquired sequence (upajanitakramah.), and
the latter functions as a signifier (pratyayakah.). The speech unit,
moreover, is said to have the property of making a meaning understood (pratyayakatvam) in that it enters into cognitions (pratipattis.u praptasamavesasya) in a form that has lost any sequence (kramarupapratyastamayena). From the point of view of the listener,
then, what has sequence (kramavan) is the cause of that which has
no sequence (akramanimittam).4 The process from a speakers
stand is reiterated later in the Vakyapadya proper, where Bhartrhari draws a parallel with the light (jyotih.) inherent in a stick
rubs to make fire (aranisthah): as this is a cause of another,
one
.
.
outward, light (prakasa ntarakaran.am), similarly is a speech unit
(sabdah.) in the buddhi (buddhisthah.) separately a cause of sounds
(srutnam).5
A speakers perspective is portrayed in the Pan.inyasiks.a ,
4 v.ya;a;pa;a:=+a:ua :pra;a;ta;l+.b.Da;
a;va;
a;k+:ya;
a;va;Zea;SaH (ra;ea:a;a;nua;pa;a;ta;a :pra;k+:a;Za;k+:Ba;a;vea;na ;a;na;tyMa :pra;tya;a;yya;pa:=+ta:n:a;ea Y;TeRa;Sua :pra;yua:$ya;tea l+.b.Da;a;nua;sMa;h;a:=+ea ;a;na;a;ma:a;mua;pa:ja;a;na;ta;k+:ma;~tua :pra;tya;a;ya;k I+.tyea;ke
ta;~ya;a;
a;pa k+:ma:+.pa;pra;tya;~ta;ma;yea;nEa;va :pra;a;ta;pa; a:a;Sua :pra;a;a;sa;ma;a;vea;Za;~ya :pra;tya;a;ya;k+:tva;ma;a;.ca:a;tea ;pa:=
;a;h k+:ma;va;a;na;k+:ma;a;na;a;ma:a;m,a (VPVr. 1.44 [102.14]). Although eke some and
viewpoints (paksa) are represented from the
aparah. another are used, different
.
vrs.abha (VPP.
perspective of a speaker and listener. This is brought out by Sr
102.2122): k+:ma;BUa;ta;a;na;Ma . ca va;a;Ra;na;Ma bua;a:;d ;~Ta;ea ;a;na;a;ma:Ma na . ca .tea va;a;.ca;k+:a I+.a;ta :pa:a;a;nta:=+ma;a;h (ra;ea:a;pea:a;yEa;ta;de ;va ;
a;va;pa:=+ata;m,a Bhartrhari portrays the procedures again
in his commentary on the Mahabhas.ya (Dpika 3.1314); I do not enter here
into this passage. Nor does my exposition require considering here Bhartrha on
ris conception of word and meaning as mutually cause and effect, depending
the perspective in question, that of a listener or a speaker, as well as Buddhist
discussions of the same topic.
5 :=+ a;a;~TMa ya;Ta;a .$ya;ea;a;taH :pra;k+:a;Za;a;nta:=+k+:a:=+a;m,a ta;d;.cC+.b.d;ea Y;
a;pa bua;a:;d ;~TaH (rua;ta;a;na;Ma
k+:a:=+Ma :pxa;Ta;k, (VP. 1.47/46).
56
G. C ARDONA
where it is said that the a tman first apprehends meanings (arthan)

through the buddhi and, having done so (sametya), it associates/activates the manas with the desire to speak (vivaks.aya); the
manas then strikes the bodily fire, and this sets the breath (marutam) in motion (prerayati).6 A speakers perspective is also taken
in pratisa khyas, which deal with speech production but not speech
perception. For example, the Rgvedapratisa khya says the upward
moving breath (vayuh.) calledpran.a reaches a state (apadyate) of
being svasa (svasatam) or nada (nadatam
va) that is, unmodulated or modulated air flow depending on whether the glottal
aperture (kan..thasya khe) is open (vivrte) or close (samvr
te), when
(vaktrhayam).7
there is an effort on the part of a speaker
Derivation and a speakers perspective
2.1
Katyayana on speech being prompted by meaning
In consonance with viewing language from a speakers perspective, Katyayana states in his very first varttika that the grammar
(sa stren.a) serves as a means to establish a restriction for the sake
of merit (dharmaniyamah.) that is, to establish that only the use
of correct speech forms leads to merit given that correct speech
forms, their meanings, and the relation (sabdarthasambandhe) between the two are established (siddhe) and also that the use of such
speech forms (sabdaprayoge) is prompted by meaning (arthaprayukte), as is known from the world of communication (lokatah.).8
That is, as Nagesa points out, a speakers use of a speech form has
bua:;d;a;a .sa;mea;tya;a;Ta;Ra;n}.a;na;ea yua;*: e ;

a;va;va:a;ya;a ma;naH k+:a;ya;a;a;+;a;ma;a;h;
a;nta
/ / .sa :prea:=+ya;a;ta
ma;a:+:ta;m,a (PS. 3).

7 va;a;yuaH :pra;a;aH k+:ea;.a;ma;nua;pra;d;a;nMa k+:F+.~ya Kea ;
a;va;vxa;tea .sMa;vxa;tea va;a ;a;pa;d;a;tea (;a;a;sa;ta;Ma na;a;d;ta;Ma
va;a va; " +:a;h;a;ya;a;m,a (RPr. 13.1).
8 ;a;sa:;d e Za;b.d;a;TR
a;sa;}ba;nDea l+.ea;k+:ta;ea Y;TRa;pra;yua;e Za;b.d;pra;ya;ea;gea Za;a;~:ea;a ;Da;mRa;a;na;ya;ma;ea ya;Ta;a l+.Ea;
a;k+:k+:vEa;
a;d;ke+:Sua Patajali divides the varttika into sections for discussion.
6 ;a;tma;a
57
as its purpose that another person understand a particular meaning the speaker wishes to convey, so that the usage is due to this
purpose.9
2.2
Patajali on speech being caused by meaning
On several occasions, Patajali presents arguments which contrast

two stands concerning meanings and speech units.
2.2.1
Number distinction
For example, the Mahabhas.ya on A. 2.1.1 (.sa;ma;TRaH :pa;d;

a;va;a;DaH)
takes up the particular features (vises.a h.) due to meaning unification (ekarthbhavakrtah.) in forming derived units such as com
pounds.10 One of these
particular features concerns number distinctions. In strings such as
(1) a. rajah. purus.ah. a servant of a king
b. rajoh. purus.ah. a servant of two kings
c. rajam
purus.ah. a servant of several kings
there is a formal distinction of numbers: singular rajah., dual rajoh., and plural rajam. The compound
(2) rajapurus.ah.
on the other hand, shows no such distinction: the constituent rajancan correspond to any of the forms noted. The argument concerning number distinction proceeds as follows:
9 Ba;a;Syea
;TRa;pra;yua; I+.a;ta ;TRa;a;a;na;pra;ya;ea:ja;nea;na kx+:ta I+.tya;TRaH :pa:=+~ya ;

a;va; a;Za;;a;TRa;ba;ea;Da;ea Ba;va;a;tva;
/ a;ta Za;b.dH :pra;yua:$ya;tea :pra;ya;ea;gea;a;a;
a;va;BRUa;ta;Za;b.d;a;a;nea;na;a;TRa;a;a;na;a;
a;d;a;ta Ba;a;vaH (Ud. I.33).
10 I+.mea ta;h:aeR;k+:a;Ta;Ra;Ba;a;va;kx+:ta;a ;
a;va;Zea;Sa;aH .sa;* +;a;a;
a;va;Zea;Sa;ea v.ya;+:a;a;Ba;Da;a;na;mua;pa;sa:jRa;na;
a;va;Zea;Sa;Ma . ca
;ya;ea;gaH I+.a;ta (Bh. I.362.1214).
58
G. C ARDONA
I. In a string (vakye), there is a number distinction (sankhyavises.ah.), as in (1a), (1b), (1c), and this distinction does not
occur in a compound.11
II. There is a reason why this is so. What is the reason? Because the unit signifying the difference is not present. Just
you utter such a unit and that distinction will be understood.12
III. Response:13 This should not be thus. It should not be (bhavitavyam) that a meaning (arthena) is caused by a speech
unit (sabdakrtena produced by a speech unit). It should
unit is caused by a meaning. One should
be that a speech
thus view this in the following manner: it is due to a meaning aspect that in the present instance (2) as opposed to
(1a)(1c) the distinction is not understood. Moreover,
one must necessarily recognize this to be so. For, if one
considers that the distinction is not understood here because
the speech unit signifying the distinction is not present, then
conversely one would have to accept that the distinction
should be understood in the following examples:
11 .sa;* +;a;a;
a;va;Zea;Sa;ea
Ba;va;a;ta va;a;k+.ae .=+a;aH :pua:+:SaH .=+a;a;eaH :pua:+:SaH .=+a;a;Ma :pua:+:Sa I+.a;ta .sa;ma;a;sea na
Ba;va;a;ta .=+a:ja;pua:+:Sa I+.a;ta (Bh. I.362.1415).

12 ;a;~ta
/ / k+:a:=+Ma yea;nEa;ta;de ;vMa Ba;va;a;ta ;
a;kM k+:a:=+a;m,a ya;ea Y;sa;Ea ;
a;va;Zea;Sa;va;a;.ca;a Za;b.d;~ta;d;sa;a; a;a;Dya;a;t,a ;*: ;
a;h Ba;va;Ma;~ta;mua;a;a:=+ya;tua gMa;~ya;tea .sa ;
a;va;Zea;SaH (Bh. I.362.1517). The
respective endings nas,
os, and a m of rajah., rajoh., and rajam belong to the
subsets of nominal endings called ekavacana, dvivacana, and bahuvacana; these

occur when one, two, or several entities are signified: ba;hu;Sua ba;hu;va;.ca;na;m,a d;ae;k+:ya;ea;
a;dR;va;.ca;nEa;k+:va;.ca;nea (A. 1.4.2122, see PWT.: 157 241).
13 na;nua . ca .nEa;tea;nEa;vMa Ba;
a;va;ta;v.ya;m,a na ;
a;h Za;b.d;kx+:tea;na na;a;ma;a;TeRa;na Ba;
a;va;ta;v.ya;m,a ;TRa;kx+:tea;na na;a;ma
Za;b.de;na Ba;
a;va;ta;v.ya;m,a ta;de ;ta;de ;vMa dx ;Zya;ta;a;ma;TRa:+.pa;mea;vEa;ta;de ;va:*+;a;t
a;a;ya;kM yea;na;a:a ;
a;va;Zea;Sa;ea na ga;}ya;ta
I+.a;ta ;va;ZyMa . cEa;ta;de ;vMa ;
a;va;ea;ya;m,a ya;ea ;
a;h ma;nya;tea ya;ea Y;sa;Ea ;
a;va;Zea;Sa;va;a;.ca;a Za;b.d;~ta;d;sa;a; a;a;Dya;a;d:a
;
a;va;Zea;Sa;ea na ga;}ya;ta I+.ta;a;h ta;~ya ;
a;va;Zea;Sa;ea ga;}yea;ta ;psua;.ca:=H ga;ea;Sua;.ca:=H va;Sa;Ra;sua:ja I+.a;ta (Bh.
I.362.1721).
59
(3) a. apsucarah. (fish)

b. gos.ucarah. (rooster)
c. vars.a suja (the insect called indragopa).14
The arguments presented reflect contrasting positions with respect to what Pan.ini provides for. Pan.ini accounts derivationally
for the contrast and relation between (1a)-(1c) on the one hand and
(2) on the other hand by providing that a pada which terminates in
a sixth-triplet ending optionally combines with a semantically and
syntactically related pada to form a unit to which the class name
samasa (compound) is assigned:15
(1) a0 . (rajan-nas
purus.a-su)samasa
b0 . (rajan-os purus.a-su)samasa
c0 . (rajan-am purus.a-su)samasa
A compound is a derived nominal base (pratipadika), an ending
contained in which is generally deleted:16 (rajan-nas
purus.a-su)
(rajan- purus.a-), (rajan-os purus.a-su) (rajan- purus.a-), (rajan-am purus.a-su) (rajan- purus.a-). Although rajan- in such
constructs lacks a nominal ending, by convention an operation
which is conditioned by the presence of an affix still takes effect even if the affix is deleted.17 Accordingly, rajan- resulting
from the deletion of nas,
os, or a m still bears the class name pada,
which depends on a nominal base having a nominal ending after
14 I have glossed in accordance with commentators, e.g., Pr.. II.510: ga;ea;Sua;.ca:= H
ku+:*:u +:f o+..cya;tea . . . va;Sa;Ra;sua:ja I+.ta;a;nd+ga;ea;paH RaPr. V.22: ;psua;.ca:= I+.a;ta ma;t~ya;a;d;Ea .+.Q+ea
Y;ya;m,a ga;ea;Sua;.ca:= I+.a;ta ku+:*:u +:fe .+.Q+ea Y;ya;m,a va;Sa;Ra;sua:ja I+.a;ta I+.nd+ga;ea;pea .+.Q+ea Y;ya;m,a
15 :Sa; +a .sa;ma;TRaH :pa;d;
a;va;a;DaH :pra;a;
*: +:q+.a:=+a;tsa;ma;a;saH (A. 2.2.8, 2.1.1, 2.1.3); see PWT.:
140141 221, 205206 301.
16 A. 1.2.46: kx+.a;a:;d ;ta;sa;ma;a;sa;a;(
a (:pra;a;a;ta;pa;
a;d;k+:m,a 45), A2.4.71: .sua;pa;ea ;Da;a;tua;pra;a;a;ta;
a;d;k+:ya;eaH (lu +.k, 58); see PWT.: 22 48, 186 278.
17 :pra;tya;ya;l+.ea;pea :pra;tya;ya;l+a;a;m,a (A. 1.1.62); see PWT.: 6364 103.
60
G. C ARDONA
it; thereby, the -n of rajan- is dropped by virtue of being treated as

a pada-final segment.18 There are also compounds where endings
of first constituents are not absent; (3a, b, c) are instances of such
compounds. In (3a) and (3b), the second constituent contains the
derived nominal base cara-, formed with the krt suffix .ta after the
constituent converbal base car (go about); in (3c) the second
tains ja- from the base jan (be born) with the suffix d.a.19 Provision is made that in a tatpurus.a compound whose second constituent is a derivate with a krt suffix the ending of a preceding first
20 This accounts for the presence
constituent is variously retained.
of the locative plural ending su in the first constituents of (3a, b,
c).
The claim made under II assumes that the zero-replacement
(luk) which Pan.ini provides for (see note 16) actually serves to
eliminate an ending which would otherwise occur because the
number distinction associated with it (see note 12) is actually
present to begin with in the compound and that the elimination of
the ending results in this distinction not being understood.21 The
response in III, on the other hand, reverts to the point made in I,
namely that the absence of a number distinction in the first constituent of (2) is due to meaning unification: there is a separate
meaningful unit, the compound rajapurus.a, in which no number
distinction is made for the constituent raja-. Moreover, if the phys18 .sua;a;a;z+.ntMa
:pa;d;m,a na;l+.ea;paH :pra;a;a;ta;pa;

a;d;k+:a;nta;~ya (A. 1.4.14, 8.2.7; see PWT.: 23 49,
347 539).
19 A. 3.2.16: . ca:=e+H (;a;Da;k+.=+ea 15), A. 3.2.97: .sa;a;}ya;Ma .ja;nea;qR H. The suffixes belongs to the class called krt and are introduced on condition that the base be used
ta if this term denotes a locus (adhikarane) in which
with a co-occurring term:
.
.
the agent in question goes about, d.a if the term is a word terminating in a seventhtriplet ending referring to something where an agent is born. The derivates in -t.a
and -d.a obligatorily form a compound with these terms; see PWT.: 219 315.
20 A. 6.3.14: ta;tpua:+:Sea kx+:a;ta ba;h;l+.m,a (;lu +.gua:a:=+pa;de 1).
u
21 ta;d;sa;a; a;a;Dya;a;
a;d;a;ta lu +.k+:a ;
a;va;Ba;e+:a;nRa;va;a;tRa;ta;tva;a;
a;d;tya;TRaH (Pr. II.510), lu +.#Za;a;~:a;pa;ya;Ra;l+.ea;.ca;na;ya;a;TRa;sa:va;ma;nua;ma;a;ya;ta I+.a;ta Ba;a;vaH (Ud. II.510).
61
ical presence of singular, dual, or plural endings were the decisive

factor in the distinction between strings like (1a, b, c) and a compound like (2), then one would have to accept that the presence of
the plural ending su in apsu, gos.u, and vars.a su of (3a, b, c) would
determine ones understanding of plurality, so that apsucara and
gos.ucara would be understood to refer to things which go about in
many bodies of water and among many head of cattle; vars.a suja
would correspondingly be understood to refer to something born
at many seasons. But this is not true.
In sum, Pan.ini recognizes a relation between strings such as
(1a, b, c) and derived nominal bases such as rajapurus.a in (2),
and he accounts for these relations by introducing nominal endings to form theoretical strings such as (1a0 , b0 , c0 ), then allowing two possibilities: derive (1a, b, c) or optionally combine the
constituent padas to form a new nominal base, within which endings are dropped. This derivational procedure, however, merely
serves to account for the relations. It does not require that there
be a meaning distinction among three numbers that is eliminated. It recognizes instead that the distinction in question does
not obtain in such derived nominals.
2.2.2
Residual relations
A related discussion in the Mahabhas.ya concerns (1a) as an example of a genitive ending accounted for by A. 2.3.50: :Sa; +a Zea;Sea
(;na;a;Ba;
a;h;tea 1). This sutra states that a nominal ending of the sixth
triplet (s.as..th) follows a nominal on condition that a residual (ses.e
remainder) relation is to be expressed, provided it is not already
expressed (anabhihite). A residual relation is a relation that does
not hold between an action and a participant which plays a role
in bringing this action to accomplishment22 The relation between
a king and his servant spoken of in (1a) is a possessor-possessed
22 On
ses.a relations, see Cardona 2008, 2013.
62
G. C ARDONA
relation (svasvamibhavasambandha): the property of being a possessor (svamin) resides in the king, that of being a possession (sva)
in the man whom the king in effect owns by virtue of paying and
maintaining him. In addition, the king is here viewed as a qualifier (vises.an.a) with respect to the servant, who is the qualificand
(vises.ya), qualified as belonging to the king. Accordingly, the servant is said to be the principal (pradhana) entity referred to and the
king is considered a subordinate (gun.a) in the qualifier-qualificand
status. Since the possessor-possessed relation as formulated comprises two relational properties, it is considered to reside in both
(dvis..tha) of the related entities, although it is expressed as something additional (vyatiricyate) in the king, which now has a subordinate status, being intended to qualify a qualificand.23
The Bhas.yas discussion centers on the syntax of (1a): why
rajan- takes a sixth-triplet ending (rajah.) but purus.a- is followed
by an ending of the first triplet (purus.ah.).24 In his third varttika,
Katyayana brings up the issue: if one states A. 2.3.50, then it is
necessary also to prohibit (pratis.edhah. negation, prohibition) a
sixth-triplet ending for the term denoting a qualificand, and one
must provide for a first-triplet ending.25 Patajali clarifies, as fol23 ;
a;d;+ea
Y;pya;sa;Ea :pa:=+a;TRa;tva;a;u+ea;Sua v.ya;a;ta;a:=+.cya;tea ta:a;a;a;Ba;Da;a;ya;ma;a;na;(a :pra;Da;a;nea Y;pyua;pa;yua:$ya-
;tea (VP. 3.7.157). The second half of the verse says that the relation thus overtly
expressed as occurring in the subordinate also is associated (upayujyate is used)
as occurring in the principal entity. This reflects the Bhas.ya discussion. Kaiyat.a cites the karika (Pr. II.823) with upayujyate in the fourth pada, and this
is the reading in Raus edition, without variants. In his commentary, Helaraja
(VPH. 3.7.157 [358.2]) says upabhuktam, which he glosses spas..tam evabhihitam
(clearly expressed), and Subramania Iyers and Raghunatha Sarm

as editions
read upabhujyate.
24 Such an ending is introduced when there is to be signified nothing more than
a base meaning and a gender, a measure, or a number: :pra;a;a;ta;pa;
a;d;k+:a;TRa;a;l+.*: +.pa;a:=+ma;a;a;va;.ca;na;ma;a:ea :pra;Ta;ma;a (A. 2.3.46); see PWT.: 156 240.
25 :Sa; +a Zea;Sa I+.a;ta . cea;
a;d;Zea;Sa;~ya :pra;a;ta;Sea;DaH ta:a :pra;Ta;ma;a;
a;va;a;DaH (2.3.50 vt. 34). Kaiyat.a explains (Pr. II.821) that the problem arises under Pan.inis formulation. In
63
lows. Under the Pan.inian statement, a prohibition has to be stated.

In (1a), the king is a qualifier, the servant a qualificand. With respect to the latter, the meaning of the nominal base (pratipadikarthah.) is added to (vyatiriktah.): purus.a- here refers to a man as a
qualificand. Hence, an ending of the first triplet, used if the base
meaning alone is denoted (see note 24) does not obtain, so that a
sixth-triplet ending by A. 2.3.50 would occur, given that more than
the base meaning is involved and the man in question does not play
the role of a participant in an action.26 One must state a negation
of such an ending, and an ending of the first triplet must be provided (vidheya).27 Katyayana responds to this by alluding to what
had been in the second varttika on A. 2.3.46: what is addition to
the base meaning of purus.a- in (1a) is a meaning attributable to
the string (vakyarthah.), so that the base meaning remains without
addition, thus allowing a first-triplet ending.28
This response gives rise to the question: for what reason is the
additional meaning in the servant of (1a) said to be attributable to
the string? Why should not the additional meaning in the king be
such?29 At this point, Patajali enters into a discussion which in
part is parallel to the discussion shown in 2.2.1, as follows:
another grammar, the rule is worded gua;ea :Sa; +a, specifying that the ending occurs
with respect to a subordinate (gun.e): :Sa; +a Zea;Sa I+.a;ta . cea;
a;d;a;ta ;a;sma;nsU
/ / / a:a;nya;a;sea d;ea;SaH
.sa;}ba;nDa;~ya ;
a;d;+tva;a;tpra;Da;a;na;a;d;
a;pa :Sa; +a;pra;sa;*:H
/ /
v.ya;a;k+.=+a;a;nta:=e tua gua;ea :Sa; +a;a;ta va;.ca;na;a;a;a;a;~ta
26 Ba;a;Syea :Sa; +a .~ya;a;
a;d;a;ta k+:a:=+k+:pra;a;a;ta;pa;
a;d;k+:a;TRa;v.ya;a;ta;a:=++:tva;a;tSa;.ae;va .~ya;a;
a;d;tya;TRaH (Ud.
II.821). I do not take up a possible complication, namely that a form of a verb
meaning be such as asti is can be understood in (1a), signifying an act
in which the referent of purus.a- functions as a karaka.
27 :Sa; +a Zea;Sa I+.a;ta . cea;
a;d;Zea;Sya;~ya :pra;a;ta;Sea;Da;ea va;+:v.yaH .=+a;aH :pua:+:Sa I+.tya:a .=+a:ja;a ;
a;va;Zea;Sa;Ma
:pua:+:Sa;ea ;
a;va;Zea;SyaH ta:a :pra;a;a;ta;pa;
a;d;k+:a;Ta;eRa v.ya;a;ta;a:=+ I+.a;ta kx+:tva;a :pra;Ta;ma;a na :pra;a;pa;ea;a;ta ta:a :Sa; +a
.~ya;a;t,a ta;~ya;aH :pra;a;ta;Sea;Da;ea va;+:v.yaH (Bh. I.464.35), ta:a :Sa; +Ma :pra;a;ta;
a;Sa;Dya :pra;Ta;ma;a ;
a;va;Dea;ya;a
.=+a;aH :pua:+:Sa I+.a;ta (Bh. I.464.8).
28 o+.M :pUa;veRa;a (2.3.50 vt. 5). ;
a;k+:mua;+:m,a na va;a va;a;k+.a;a;TRa;tva;a;
a;d;a;ta ya;d:a;a;a;Da;k+.aM va;a;k+.a;a;TRaH
.saH (Bh. I.464.10).
29 ku+:ta;ea nua Ka;vea;ta;tpua:+:Sea ya;d;a;a;Da;k+.aM .sa va;a;k+.a;a;TRa I+.a;ta na :pua;na;a .=+a:ja;a;na ya;d;a;a;Da;k+.aM .sa
va;a;k+.a;a;TRaH .~ya;a;t,a (Bh. I.464.1012).
64
G. C ARDONA
I. The additional meaning in question is understood to apply
for the king even without the use of the term purus.a, but
this does not apply to the servant without the use of the term
rajan.30
II. There is a reason why this is so. What is the reason? You
utter a sixth-triplet ending after the term rajan. Just you utter
a sixth-triplet ending after the term purus.a and that meaning
will be understood.31
III. This should not be thus. It should not be that a meaning

is caused by a speech unit. It should be that a speech unit
is caused by a meaning. One should thus view this in the
following manner: it is due to a meaning aspect that this additional meaning is understood to apply for the king even
without the use of the term purus.a. And what is that? Being one who possesses (svamitvam). And what is this due to
(kinkrtam produced by what?)? To what he possesses (sva This has an analog. Distinctions in the meanings of
krtam).
nominal
bases are brought about by actions (kriyakrtah.), and
ah) come
particular names produced thereby (tatkrtas cakhy
.
to the fore (pradur bhavanti appear): karman (object),
karan.a (instrument), apadana, sampradana, adhikaran.a
(locus). Further, these names sometimes serve (upadyante
are taken, expressed) as causes (nimittatvena) for triplets
of nominal endings (vibhaktnam) coming into play, sometimes not. They do so when they deviate from the meaning
of a base. When they do not, they remain solely names.32
30 ;nta:=e+a;a;
a;pa
:pua:+:Sa;Za;b.d;pra;ya;ea;gMa .=+a:ja;a;na .sa;ea Y;Ta;eRa ga;}ya;tea na :pua;na:=+nta:=e+a .=+a:ja;Za;b.d;pra;ya;ea;gMa

:pua:+:Sea .sa;ea Y;Ta;eRa ga;}ya;tea (Bh. I.464.1213).
31 ;a;~ta
a;kM k+:a:=+a;m,a .=+a:ja;Za;b.d;a;a:;d Ba;va;a;nSa; +a;mua;a;a:=+ya;a;ta ;*:
;
a;h Ba;va;a;npua:+:Sa;Za;b.d;a;d;pyua;a;a:=+ya;tua gMa;~ya;tea .sa;ea Y;TRaH (Bh. I.464.1315).
32 na;nua . ca .nEa;tea;nEa;vMa Ba;
a;h Za;b.d;kx+:tea;na na;a;ma;a;TeRa;na Ba;
a;va;ta;v.ya;m,a ;TRa;kx+:tea;na
na;a;ma Za;b.de;na Ba;
a;va;ta;v.ya;m,a ta;de ;ta;de ;vMa dx ;Zya;ta;a;m,a ;TRa:+.pa;mea;vEa;ta;de ;va:*+;a;t
a;a;ya;kM yea;na;a:a;a;nta:=e+a;a;
a;pa
65
IV. For the king of (1a), his being a possessor is due to a possession, and a sixth-triplet ending occurs with rajan-. In the
same way, for the kings servant, his being a possession is
due to a possessor, so that a sixth-triplet ending obtains also
for purus.a-.33
V. That meaning is expressed by the sixth-triplet ending that
comes in after the term rajan-; hence, there will not be a
sixth-triplet ending after the term purus.a-.34
VI. Then is there not a string
(4) purus.asya raja the mans king?
There is, but then a first-triplet ending follows the term
rajan-.35 Then is there not a string
(5) rajah. purus.asya the kings mans?
There is such as string, but only if another (bahyam external) meaning is considered (abhisamks.ya).36
Under III, the point is made that rajah. of (1a) has a sixthtriplet ending conditioned by a meaning: the king in question bears
:pua:+:Sa;Za;b.d;pra;ya;ea;gMa .=+a:ja;a;na .sa;ea Y;Ta;eRa ga;}ya;tea ;
a;kM :pua;na;~ta;t,a .~va;a;a;ma;tva;m,a ;
a;k+:*:x +.tMa :pua;na;~ta;t,a
.~va;kx+:ta;m,a ta;d;a;Ta;a :pra;a;a;ta;pa;
a;d;k+:a;Ta;Ra;na;Ma ;
a;k+:ya;a;kx+:ta;a ;
a;va;Zea;Sa;a o+.pa:ja;a;ya;ntea ta;tkx+:ta;a;(a ;a;K.ya;aH
:pra;a;du;BRa;va;a;nta
/ / k+:mRa k+.=+a;ma;pa;a;d;a;nMa .sa;}.pra;d;a;na;ma;a;Da;k+.=+a;a;ma;a;ta ta;a;(a :pua;na;
a;vRa;Ba; +:a;na;a;mua;tpa:a;Ea
k+:d;a;a;.ca; a;a;a;ma:a;tvea;na;ea;pa;a;d ;a;ya;ntea k+:d;a;a;.ca;a k+:d;a . ca ;
a;va;Ba; +:a;na;a;mua;tpa:a;Ea ;a;na;a;ma:a;tvea;na;ea;pa;a;d ;a;ya;ntea ya;d;a v.ya;a;Ba;.ca:=+a;nta
/ / :pra;a;a;ta;pa;
a;d;k+:a;TRa;m,a ya;d;a ;
a;h na v.ya;a;Ba;.ca:=+ntya;a;K.ya;a;BUa;ta;a O;;va ta;d;a
Ba;va;a;nta
/ / k+:mRa k+.=+a;ma;pa;a;d;a;nMa .sa;}.pra;d;a;na;ma;a;Da;k+.=+a;a;ma;a;ta (Bh. I.464.1523).
33 ya;TEa;va ta;
a;hR .=+a:ja;a;na .~va;kx+:tMa .~va;a;a;ma;tvMa ta:a :Sa;.ae;vMa :pua:+:Sea Y;
a;pa .~va;a;a;ma;kx+:tMa .~va;tvMa ta:a
:Sa; +a :pra;a;pa;ea;a;ta (Bh. I.464.2324).
34 .=+a:ja;Za;b.d;a;du;tpa;d;a;ma;a;na;ya;a :Sa;.a;a;a;Ba;
a;h;taH .sa;ea Y;TRa I+.a;ta kx+:tva;a :pua:+:Sa;Za;b.d;a;tSa; +a na Ba;
a;va;Sya;a;ta (Bh. I.464.2425).
35 r
ajan-su rajan- rajan raja; see PWT.: 347 539.
36 na ta;h ;Ra;d;a;na;a;a;ma;dM Ba;va;a;ta :pua:+:Sa;~ya .=+a:jea;a;ta Ba;va;a;ta .=+a:ja;Za;b.d;a:ua ta;d;a :pra;Ta;ma;a na ta;h ;Ra;d;a;na;a;a;ma;dM Ba;va;a;ta .=+a;aH :pua:+:Sa;~yea;a;ta Ba;va;a;ta ba;a;h:a;ma;TRa;a;}Ba;sa;m
/ / /
a;a:ya (Bh. I.464.2527).
66
G. C ARDONA
a relation of possession to another entity. Since the other

thing is necessary for the relation to obtain, the kings property of
being a possessor is said to be caused by it. A parallel is drawn
to relations between karakas and the acts in which they participate
to bring them to accomplishment. These acts are said to bring
about distinctions in the meaning of nominal bases such that the
things denoted are given particular class names such as karman
and karan.a. Thus, a sickle may serve as the means for cutting
(lunati cuts), so that it is given the karaka class name karan.a.37
The denotation of a karan.a, moreover, can serve as a condition for
introducing a third-triplet ending after a nominal,38 as in datren.a
( datra-t.a ) of
(6) datren.a lunati . . . is cutting . . . with a sickle.
Here, the base meaning of datra has added to it the property of
being a means. One may, however, speak merely of a sickle itself
and say that it is a means, as in
(7) datram
karan.am.
Now the base meaning of datra has not been added to, so that the
base takes a first-triplet ending (datram datra-su).
Under IV, the point is made that the argument advanced for
rajah. of (1a) having a sixth-triplet ending can equally well be
made for purus.a- having the same ending. This is conceded, but
the possible fault is avoided, under V, by noting that introduction
of such an ending after purus.a- is ruled out because the relation is
signified by the ending which follows rajan-; in accordance with
A. 2.3.1, the ending is not introduced once more to denote the same
relation. Finally, under VI, the question is posed whether one does
not then have an expression (5) with the same ending after both the
37 .sa;a;Da;k+:ta;mMa
k+.=+a;m,a (A. 1.4.42); see PWT.: 137138 214.

(A. 2.3.18); see PWT.: 155156 240.
38 k+:tRxa;k+.=+a;ya;ea;~txa;ta;a;ya;a
67
bases in question. This is countered by noting that, although (5)

is possible, it is not a complete expression: the ending of rajah.
designates a relation between the king and his man, but the ending
of purus.asya now signifies a relation external to this, one which
requires a third relatum.
2.2.3
Gender distinction
The Bhas.ya on A. 7.1.33: .sa;a;ma ;a;k+:m,a (yua;Sma;d;sma;;d

;a;a;m,a 27) once
more brings up the same issue. The context in which the issue is
raised concerns whether an alternative formulation of A. 7.1.33,
with a mah. instead of samah. would wrongly allow the initial augment sut. to be added to the genitive plural ending a m, by A. 7.1.52:
;a;a;ma .sa;vRa;na;a;}:aH .sua;f, (;a;t,a 50), which serves to introduce the augment to a m if this follows a pronominal stem ending with an avowel. The conditions occur when the -d of yus.mad and asmad
is deleted before a m.39 In his third, fourth and fifth varttikas on
this sutra, Katyayana presents the following argument.40 In deriving yus.makam and asmakam from yus.mad-am and asmad-am,
one must disallow the introduction of the initial augment sut. to
the ending. For, deleting the final sound of these pronouns by A.
7.2.90 (see note 39) is acknowledged to apply before an ending
that has replaced another. If, then, a kam is allowed to replace a m,
one will allow, undesiredly, a final form with this augment (*yus.masam, asmasam).41 This problem is avoided by suggesting an
alternative formulation for A. 7.2.90: instead of letting this delete
the final sound of the stems, the rule is now stated so as to delete
39 Zea;Sea
l+.ea;paH (A. 7.2.90); see PWT.: 324 498, 315 478.

l+.ea;pa;
a;va;a;a;na;a;t,a na va;a ;a;f;l+.ea;pa;va;.ca;na;a;d;a;de ;Zea f;a;ppra;a;ta;Sea;d;a;TRa;m,a na va;a
;a;l+.*: +.a;Ba;a;va;a; a;*+:l+.ea;pa;va;.ca;na;a;na;TRa;k+.a;m,a (7.1.33 vt. 35).

+.a;ta;Sea;Da;~tua va;+:v.yaH ;
a;kM k+:a:=+a;m,a ;a;de ;Zea l+.ea;pa;
a;va;a;a;na;a;t,a yaH .sa Zea;Sea l+.ea;pa ;a;de ;Zea
41 .sua;*
;
a;va;a;a;ya;tea (Bh. III.253.23).
+.a;ta;Sea;Da;~tva;a;de ;Zea
40 .sua;*
68
G. C ARDONA
the segment which begins with its last vowel (t.ilopavacanat):42

yus.mad-am yus.m-am, asmad-am asm-am. Moreover, the
claim is made, this reformulation is necessary in order to avoid allowing the feminine suffix .ta p, by A. 4.1.4: :ja;a;d;a;ta;;a;p,a (;a;~:
/ a;ya;a;m,a 3)
43
after yus.ma- and asma-. This alternative is rejected, however, as
serving no purpose, because these pronouns have no gender. Nor
does one have to state this fact. It is known without being explicitly
stated.
At this point, arguments come up that parallel those presented
earlier (2.2.12.2.2).
I. For there is no distinction in yus.mad and asmad used in the
feminine, masculine, or neuter.44
II. There is a reason why this is so. What is the reason? Because the unit conveying the difference is not present. Just
you utter such a unit and that distinction will be understood.45
III. This should not be thus. It should not be that a meaning
has a speech unit as its cause (sabdanimittakena). On the
contrary, it should be that a speech unit has a meaning as its
va;a .sua;* +.a;ta;Sea;Da;ea va;+:v.yaH ;
a;kM k+:a:=+a;m,a ;a;f;l+.ea;pa;va;.ca;na;a;t,a ;a;de ;Zea yaH .sa Zea;Sea l+.ea;pa;
a;;l+.ea;paH .sa va;+:v.yaH (Bh. III.253.56). ;.ca;ea Y;ntya;a;
a;d ;a;f (A. 1.1.64); see PWT.: 18
41.
43 ;
a;kM :pra;ya;ea:ja;na;m,a f;a;ppra;a;ta;Sea;Da;a;TRa;m,a f;a;b.ma;a BUa;
a;d;a;ta (Bh. III.253.67).
44 .sa ta;
a;hR ;a;f;l+.ea;pa;ea va;+:v.yaH (Bh. III.253.7). na va;a va;+:v.ya;m,a ;
a;kM k+:a:=+a;m,a ;a;l+.
*: +.a;Ba;a;va;a;t,a ;a;l+.*:e yua;Sma;d;sma;d ;a ;
a;kM va;+:v.ya;mea;ta;t,a na ;
a;h k+:Ta;ma;nua;.cya;ma;a;nMa gMa;~ya;tea na h:a;a;~ta
/ /
;
a;va;Zea;Sa;ea yua;Sma;d;sma;d;eaH ;a;~:
/ a;ya;Ma :pMua;a;sa na;pMua;sa;ke va;a (Bh. III.253.911).
45 ;a;~ta
a;kM k+:a:=+a;m,a ya;ea Y;sa;Ea ;
a;va;Zea;Sa;va;a;.ca;a Za;b.d;~ta;d;sa;a; a;a;Dya;a;t,a ;*: ;
a;h Ba;va;Ma;~ta;mua;a;a:=+ya;tua gMa;~ya;tea .sa ;
a;va;Zea;SaH (Bh. III.253.1113). I have used
the unit conveying the difference to render vises.avac sabdah., with conveying
instead of signifying, because feminine suffixes are considered to be not signifiers (vacaka) but cosignifiers (dyotaka) of femininity, a meaning attributed to a
nominal base with which an affix is used; see note 47 and 2.3.
42 na
69
cause (arthanimittakena). One should thus view this in the

following manner: it is due to a meaning aspect that in the
present instance the distinction is not understood. Moreover,
one must necessarily recognize this to be so. For, if one considers that the distinction is not understood here because the
speech unit conveying the distinction is not present, then one
would have to accept that the distinction should not be understood in the following examples as well: drs.ad (stone
[fem.]), samidh (kindling wood [fem.]).46
That is, one cannot argue that a form such as tvam (you [nom.
sg.] yus.mad-su) or aham (I asmad-su) fails to make understood a feminine gender distinction because it lacks a feminine
suffix. Such reasoning would require that a term such as drs.ad,
which does make a feminine gender understood, should be accompanied by a feminine suffix; but it is not.47
2.3
Meaning conditions and derivation
A. 3.1.26: :he;tua;ma;a;ta . ca (; a;a;.c,a 25, ;Da;a;ta;eaH 7, :pra;tya;yaH 1, :pa:=+(a 2) introduces the unit n.ic, to which is assigned the class name pratyaya
(affix), after (parah.) a verbal base (dhatoh.). The sutra contains
the locative singular hetumati, referring to an action which has
a causal agent (hetu), that is, causation. This locative is subject
to two possible interpretations: (a) it is a locative absolute, with
the seventh-triplet ending introduced on condition that an action
46 na;nua
. ca .nEa;tea;nEa;vMa Ba;
a;h Za;b.d;a;na;a;ma:a;ke+:na;a;TeRa;na Ba;
a;va;ta;v.ya;m,a ;
a;kM ta;
a;hR ;TRa;a;na;a;ma:a;ke+:na na;a;ma Za;b.de;na Ba;
a;va;ta;v.ya;m,a ta;de ;ta;de ;vMa dx ;Zya;ta;a;m,a ;TRa:+.pa;mea;vEa;ta;de ;va:*+;a;t
a;a;ya;kM yea;na;a:a
;
a;va;Zea;Sa;ea na ga;}ya;ta I+.a;ta ;va;ZyMa . cEa;ta;de ;vMa ;
a;va;ea;ya;m,a ya;ea ;
a;h ma;nya;tea ya;ea Y;sa;Ea ;
a;va;Zea;Sa;va;a;.ca;a
Za;b.d;~ta;d;sa;a; a;a;Dya;a;d:a ;
a;va;Zea;Sa;ea na ga;}yea;ta I+.ta;a;h;a;
a;pa ta;~ya ;
a;va;Zea;Sa;ea na ga;}yea;ta dx ;Sa;tsa;a;ma;
a;d;a;ta
(Bh. III.253.1317).
47 It is on account of such usages that feminine suffixes such as ta
. p in terms
like aja (she goat aja-a) are treated as cosignifying a feminine, a meaning
attributed to the accompanying base; see PWT.: 441 685.
70
G. C ARDONA
serves to characterize another action;48 (b) it has a seventh-triplet

ending signifying a locus (adhikaran.a) in which an act occurs.49
Under (a), A. 3.1.26 provides that the affix n.ic is introduced (bhavati comes into being, is) after a verbal base if an action with a
causal agent is to be signified (abhidheye sati). Under (b), on the
other hand, the sutra provides that the affix is introduced following
a base which occurs in this meaning (hetumati vartamanat).50
Comparably, A. 4.1.3: ;a;~:
/ a;ya;a;m,a (za;a;ppra;a;a;ta;pa;
a;d;k+:a;t,a 1, :pra;tya;yaH
3.1.1, :pa:=+(a 2) can be a heading whereby: (a) the affixes .ta p and so
on are introduced by subsequent rules if the meaning feminine
is to be signified (striyam abhidheyayam
satyam); (b) the affixes
are introduced after a nominal base (pratipadikat) which occurs in
this meaning (striyam
vartamanat pratipadikat), so that they are
allowed to occur in a meaning attributed to the base (svarthe).51
The positions which Pan.inyas finally accept for these sutras
48 A.
2.3.37: ya;~ya . ca Ba;a;vea;na Ba;a;va;l+a;a;m,a (.sa;a;ma;a 36); see PWT.: 173174 265.
. ca (A. 2.3.36); see PWT.: 155156 240.
50 k+:Ta;a;ma;dM ;
a;va;a;a;ya;tea :he;tua;ma;tya;a;Ba;Dea;yea ; a;a:$Ba;va;ta;a;a;ta ;a;h;ea;a;~va:
/ / ;d e ;tua;ma;a;ta ya;ea ;Da;a;tua;vRa;tRa;ta
I+.a;ta (Bh. II.31.78). k+:Ta;a;ma;dM ;
a;va;a;a;ya;tea :he;tua;ma;tya;a;Ba;Dea;yea ; a;a:$Ba;va;a;ta ;a;h;ea;a;~va:
/ / ;d e ;tua;ma;a;ta ya;ea ;Da;a;tua;vRa;tRa;tea (:he;tua;ma;a;ta) va;ma;Ra;na;a;
a;d;a;ta (VPVr. 1.26 [73.12]). :he;tuaH .~va;ta:n:a;~ya
k+:tRuaH :pra;ya;ea:ja;kH ta;d ;a;ya;ea;v.ya;a;pa;a:=+ea :he;tua;ma;a;n,a ta;a;sma;a;

/ / / a;Ba;Dea;yea ;Da;a;ta;ea; a;Ra;.cpra;tya;ya;ea Ba;va;a;ta (Kas.
3.1.26). Bhartrhari alludes to the Bhas.ya passage I have cited in the context of
saying that allsuch optional procedures depend on the considerations of persons
who abstract putative constituent meanings from full utterances, which are in fact
unanalyzable wholes from the point of view of communication.
51 k+:TMa :pua;na;a:=+dM ;
a;va;a;a;ya;tea ;a;~:
/ a;ya;a;ma;a;Ba;Dea;ya;a;ya;Ma f;a;ba;a;d;ya;ea Ba;va;nta;a;a;ta ;a;h;ea;a;~va;t~:
/ / a;a;sa;ma;a;na;a;a;Da;k+.=+a;a;tpra;a;a;ta;pa;
a;d;k+:a;
a;d;a;ta (Bh. II.198.2021). ;a;sa:;d M tua ;a;~:
/ a;ya;aH :pra;a;a;ta;pa;
a;d;k+:
a;va;Zea;Sa;a;tva;a;t~va;a;TeRa f;a;ba;a;d;yaH (4.1.3 vt. 5). ;a;sa:;d ;mea;ta;t,a k+:Ta;m,a :pra;a;a;ta;pa;
a;d;k+:
a;va;Zea;Sa;a;tva;a;t,a
:pra;a;a;ta;pa;
a;d;k+:
a;va;Zea;Sa;Ma .~:a;a;g{a;h;Ma .~va;a;a;TRa;k+:a;;a;ba;a;d;yaH .nEa;vMa ;
a;va;a;a;ya;tea ;a;~:
/ a;ya;a;ma;a;Ba;Dea;ya;a;ya;a;a;ma;a;ta na;a;
a;pa .~:a;a;sa;ma;a;na;a;a;Da;k+.=+a;a;
a;d;a;ta k+:TMa ta;
a;hR ;a;~:
/ a;ya;Ma ya;tpra;a;pa;
a;d;kM va;tRa;tea ta;sma;a;*+:a;ba;a;d;ya;ea
Ba;va;a;nta
/ / k+:a;sma;a;Te
/ / / Ra .~va;a;TeRa (Bh. II.199.1922). ta;Ta;a ;a;~:
/ a;ya;a;a;ma;a;ta ;
a;kM .~:ya;a;Ba;Da;a;nea f;a;ba;a;d;yaH .~:ya;TRa;vxa:eaH :pra;a;a;ta;pa;
a;d;k+:a;t~va;a;TeRa :vea;a;ta (VPVr. 1.26 [73.24]). As shown, a third
after a base that is coreferential
position is considered: affixes are introduced
with one used in the feminine. The finally accepted view is (b). See Cardona
1983: 4449.
49 .sa;a;}ya;a;Da;k+.=+ea
71
are (a) for A. 3.1.26, (b) for A. 4.1.3. Under the accepted interpretation of the former, then, this sutra introduces an affix under a
meaning condition: if meaning M is to be denoted, the affix A is
introduced after base B.
Affixation meaning conditions
It is demonstrable that Pan.ini himself, not just later students of his

work, operated in the above manner, allowing affixes to occur on
condition that stated meaning conditions are met.
Three sutras are particularly interesting on this point: A.
3.3.156: :he;tua;he ;tua;ma;ta;ea;a;lR +.z, (;Da;a;ta;eaH 3.1.91), A. 3.3.139140: ;a;l+.
a;*:+.a;ma:ea
lx +.
a;*:" +.ya;a;a;ta;pa:a;Ea (Ba;
a;va;Sya;a;ta 136, ;Da;a;ta;eaH 3.1.91), BUa;tea . ca. According to the
first sutra, the l-affix lin follows a verbal base to signify an agent
on condition that the action this denotes is related to another act
as cause and effect (hetuhetumatoh.). A. 3.3.139140 introduce lrn
under a condition that determines that lin is introduced (linnimitte)

that is, if related verbs denote acts related as cause and effect
when the acts in question are referred to the future (bhavis.yati)
and also to the past (bhute), provided that these fail to take place
(kriyatipattau). The first rule serves to account for conditional sentences with optative forms, such as
(8) krs.n.am
namec cet sukham
yayat If one venerates Krs.n.a, one
attains
happiness.
The next rules account for the use of conditional forms in contrafactual conditions, for example
(9) yadi vrs..tir abhavis.yat subhiks.am abhavis.yat If there were to
rains (but there wont be), there would be plenty of
be good
food./ If there had been good rains, there would have been
plenty of food.52
52 See
PWT.: 168169 257258.
72
G. C ARDONA
Now, linimitte in A. 3.3.139 can be interpreted properly only

as a s.as.t.htatpurus.a form (linnimittam = lino nimittam).53 Other
interpretations are ruled out on various grounds. To begin with,
consider the possibility that linnimitte is a tatpurus.a of the subclass called karmadharaya, in which both terms are coreferential.
This compound, equivalent to liny eva nimitte (a condition that
is lin)
would refer to the l-affix lin as a unit serving to condition
something. Since this is a speech element, the locative is to be
interpreted as referring to a right context for an operation on a preceding element.54 However, the only element that could be said
to precede is a verbal base (dhatu), given that A. 3.3.139 comes
under the heading of A. 3.1.91. Further, since A. 3.3.139 is part
of a set of rules headed by A. 3.1.1, the only reasonable operation
is the introduction of an affix (pratyaya) after a verbal base. At
best, then, A. 3.3.139 under the supposed interpretation, would let
a verbal base receive an affix on condition that lin followed. But
the affix is now not specified. Consequently, interpreting linnimitte
as a karmadharaya leads to nonsense in terms of Pan.inis grammar.
There are also no contextual grounds for interpreting linnimitte as
a bahuvrhi (lin nimittam
yasya tasmin). The only thing lin serves
to condition by way of the affixes that replace it is the introduction of affixes. Thus, for example, in deriving namet of (8),
the ending that replaces lin conditions the introduction of the stem
suffix sap after the base.55 Only if lin of linnimitte refers to an
affix, is it possible for A. 3.3.139 to provide for an affix. But then
the affix can hardly be lin itself. Consequently, the only reasonable interpretation of linnimitte in this sutra is the one adopted by
53 For example, K
as. 3.3.139: :he;tua;he ;tua;ma;ta;ea;a;lR +.
a;z+.tyea;va;ma;a;
a;d;kM ;a;l+.z+.ea ;a;na;a;ma:a;m,a ta:a
;a;l+.
a;*:+.a;ma:ea Ba;
a;va;Sya;a;ta k+:a;le lx +*: +.tya;ya;ea Ba;va;a;ta ;
a;k+:ya;a;a;ta;pa:a;Ea .sa;tya;a;m,a
54 ta;a;sma;
/ / / a;a;a;ta ;a;na;
a;dR ;e :pUa;vRa;~ya (A. 1.1.66); see PWT.: 53 90.
55 nam-l nam-ti nam-t nam-y
ast nam-a-yast nam-a-yat nama-iyt nam-a-it namet; see PWT.: 152 235, 157 243, 168 257.
73
Pan.inyas, so that the condition in question is the relation between

actions stated in A. 3.3.156.
One must, then, accept that in Pan.inis derivational system a
meaning can serve as a condition (nimitta) for the occurrence of
an affix.
This does not mean, nevertheless, that one must reject the possibility that speech units are themselves causes. On the contrary,
both meanings and speech units indeed serve as causes. From a
speakers point of view, a meaning to be made understood to a listener is a cause of his using a speech unit in order to have that
listener understand what he wishes to convey. From a listeners
point of view, the sounds a speaker utters serve as conditions for
his understanding what the latter means to say, by converting those
sounds to the systematic units both speaker and listener share in a
common language. From a listeners viewpoint also, utterances
used by speakers are subject to analysis leading to abstracted constituents. In a derivational system like Pan.inis, such abstracted
units are then synthesized to account for strings which speakers of
the language would use.56 This derivational system proceeds by
introducing affixes after bases under stated conditions, and among
these are meaning conditions: an affix is allowed to occur if a given
meaning is to be signified.
A different approach suggested
Recently,57 Jan E. M. Houben has argued on several occasions

that Pan.inis grammar did not proceed in this manner. Instead,
56 Panini operates with rules of such synthesis and presupposes the analysis
.
that leads to the abstraction of units; see PWT.: 428449.
57 This section replicates, with modifications, section 7 of Panini and his pre.
decessors: tradition and innovation, the Professor N. N. Choudhury memorial
lecture presented at Delhi University on January 29, 2012, which is to appear in
a planned volume of the lecture series.
74
G. C ARDONA
he maintains, the derivational system operates with respect to an

actual linguistic string, characterized as a provisional statement.
Houben (1999: 45) also recognizes, in accordance with some other
modern linguists, discrete levels in Pan.inis derivational system
and speaks of what he calls level 1, which he labels not as semantics but as semantics, pragmatics, intentionality. Houben
(1999: 45) says of this level:
It is the interface between the user and the grammar.
Statements in the grammar belonging to this level ask
the user to make decisions and judgements regarding
different aspects of his provisional statement: semantic aspects or aspects concerning the referents (artha)
of his statement, pragmatic aspects or aspects concerning the context (prakaran.a) of his statement, and
intentional aspects or aspects of his intention (vivaks.a ).
The same author (Houben 2009a: 52324) expresses some wonder at what other scholars of Pan.ini seem to have missed: Can
it be that a simple point has been overlooked by most scholars of
Pan.inis As..ta dhyay, namely that there too the starting point and
the final aim of the user of grammar has always been the sentence
and never a word in isolation, that the starting point is a preliminary sentence that needs to be checked or that needs some little
extra refinement? In addition, he says (Houben 2009b: 14) the
following of Pan.inis system:
The system of Pan.inis grammar clearly requires a
user who wants to check and possibly improve a preliminary statement (Houben 2003: 161). The system implies the presence of a knowledgeable user, a
preliminary statement and the application of first analytic and next synthetic procedures to the words in it,
75
with the user keeping in mind the preliminary statement and its purport, and aiming at the best possible,
sam-skr
ta form of his preliminary statement.
Immediately preceding this, Houben notes what he calls a vital

point, namely that, Pan.ini was definitely aware of various substandard forms of the language, forms which from a modern perspective we would assign to an altogether different language such
as Prakrit. After the extended statement cited above, he also goes
on to emphasize that, The concrete starting point for a derivation in the synthetic phase of the consultation cycle of a user of
grammar in Pan.inis time will then never be pure meaning or an
autonomous level of semantic representation . . . . It is, instead, the
selection of a root such as bhu or a nominal stem. He continues,
saying (Houben 2009b: 14):
In the sociolinguistic context of Pan.inis time we can
suppose that the preliminary statement of the user of
the grammar contained not necessarily only perfectly
formed words but also substandard ones, for instance
honti or bhonti instead of bhavanti. The knowledge
of the user of grammar in Pan.inis time concerns not
only the basic outlines of the grammar and knowledge
of the language aimed at but also substandard forms
current in his time and area.
The expression we can suppose in Houbens statement is
telling. For in all of this discussion, as also in his earlier paper
of 2003, he does not offer concrete evidence from the As..ta dhyay itself demonstrating the impossibility of starting with semantics associated with grammatical contrasts that can serve as conditions, along with co-occurrence conditions, for introducing affixes
to bases. He repeatedly speaks simply of semantics in the first
76
G. C ARDONA
level of those scholars with whose views he disagrees,58 and comments on the semantic level accepted by two of these scholars,
remarking (Houben 1999: 2627):
One wonders how the semantic level can be placed as
the basis and, as far as derivations are concerned, at
the beginning of the sophisticated grammar of Pan.ini,
while it is admitted at the same time that this semantic
level is very sketchy, and that Pan.ini makes no attempt
to characterise the possible semantic representations
by any rules.
Houben does not, however, note that the semantics selected are
precisely those that are linked with grammatical and phonological
rules, not some vague pure semantics.
He also speaks of a users intentional aspects or aspects of his
intention (vivaks.a ), neglecting to note that Patajali makes a pertinent distinction between an individual users desire to speak (prayoktr vivaks.a ) in a particular manner in effect to use Prakrit
if he wishes as opposed to the communal vivaks.a (laukik vivaks.a ), that is the generally accepted usage in a community of speakers who serve as a model for the grammar.59
58 Houben
nevertheless remains comfortable with the view that there are discrete levels in Pan.inis system. He assumes (1999: 46) four levels, though
these differ in detail from those posited by earlier modern scholars. I think I
have demonstrated, however (Cardona 2009), that Pan.inis system proceeds from
meaning to initial strings subject to operations that result in final utterances, with
a set of karaka classification rules that mediate between semantics and expressions, and that one proceeds in a continuum, with possible appeal to meaning
conditions at later stages of a derivation. Scharf (2009) also speaks against the
anachronistic notion that Pan.ini operated with four discrete levels.
59 ;
a;va;va:a;a . ca d;ya;a ;~tyea;va :pra;a;ya;ea; +:a ;
a;va;va:a;a ;a;~ta
/ / l+.Ea;
a;k+:k
+:a :pra;a;ya;ea;" +:a ;
a;va;va:a;a :pra"
;ya;ea;+:a ;
a;h mxa;d;a;a :a;=+:
/a;gDa;ya;a (;;a;ya;a
;a:ja;+:ya;a mxa;dU;a;n=+:
/ / / a;gDa;a;V(;;a;a;VZa;b.
d;a;npra;yua;*: e l+.Ea;
a;k+:k
+:a
;
a;va;va:a;a ya:a :pra;a;ya;~ya .sa;}.pra;tya;yaH :pra;a;ya I+.a;ta l+.ea;k+:ea v.ya;pa;
a;d;Zya;tea (Bh. II.342.26343.3).
The Bhas.ya on A 6.3.109 (:pxa;Sa;ea;d:=+a;
a;d;na;a ya;Ta;ea;pa;
a;d;;m,a) takes up the issue of who
77
In connection with vivaks.a , Pan.inyas bring up examples such

as
(10) dhanus.a vidhyati . . . pierces . . . with a bow
(11) dhanur vidhyati The bow pierces . . . .
(10) calls for assigning a bow to the karaka category named karan.a
by virtue of its serving, more than any other participant in the act of
piercing (vyadh), as the means of bringing this act to accomplishment (sadhakatamam); such a karaka is denoted by a third-triplet
ending. (11) calls for the bow to have the category name kartr
by virtue of being spoken of as an independent (svatantra) participant in the act of piercing. The bow getting the category name
kartr is accounted for by stating the karaka classification rules in
question
in a particular order: A. 1.4.42: .sa;a;Da;k+:ta;mMa k+.=+a;m,a 1.4.54:
.~va;ta:n:aH k+:ta;Ra. In addition, these rules come under the heading of A.
1.4.12 (;a k+:q+.a:=+a;de ;k+:a .sa;Va;a ;
a;va;pra;a;ta;Sea;Dea :pa:=M k+:a;yRa;m,a), so that, of the
class names provided for, only one is allowed to apply to an entity at once and, if two rules come into conflict, what is provided
for by the later rule takes precedence. Concerning (10)(11)
his examples (1)(2) Houben says (1999: 28) he accepts the
position that vivaks.a continues to play a role in the choice of karaka-designations, especially if it is clear that a certain number of
alternative ontological-logical classifications are possible (the bow
as instrument in (1) and agent in (2)). In accepting the role of vivaks.a with respect to karaka classifications, he is accepting what
Pan.inyas have always maintained; there is nothing new here. On
those called sis..ta are and where they live. See, for example, PWT.: 550554
834. The Bhas.ya also considers upadis..ta of yathopadis..tam to mean uttered
(Bh. III.104.174.2: ;Ta ;
a;k+:a;ma;d;mua;pa;
a;d;;a;a;na o+.a;a:=+ta;a;a;na ), which assumes the use
of the words in question by native speakers. This position is in harmony with
Pan.inis formulating sutras of the type ;nyea;Bya;ea Y;
a;pa dx ;Zya;tea, providing that certain
operations apply in unspecified domains, which require observing native speakers to see where the operations actually apply; see Cardona 2004.
78
G. C ARDONA
the other hand, Houben glosses over an important detail. The Bhas.ya to 1.4.1 vt. 30 does not deal with (10) and (11) opposed to
each other.60 The bows classifications as karan.a and kartr in each
instance is opposed to its classification as apadana (A. 1.4.24:

;Drua;va;ma;pa;a;yea Y;pa;a;d;a;na;m,a). For a bow to be spoken of as an instrument
in piercing something, it has simultaneously to be spoken of as a
point of departure for arrows which are released from it and serve
as means of piercing:
(12) dhanus.o nirgatair ban.air vidhyati.
For a bow to be spoken of as an agent of piercing, the same must
also hold: the bow must also simultaneously be spoken of as a
point of departure. In brief, (10) and (11) are acceptable utterances
which Pan.ini accounts for. Pan.ini also knows that the semantics
associated with these utterances involve another, conflicting, classification, reflected in (12). That is, he accounts for acceptable
utterances and their semantics.
Houben also accepts that what he calls a provisional statement subject to modification may include Middle Indic forms like
honti. In considering this, let us start with a possible strong position under which the question arises: would such forms be allowed
in preliminary strings of a Pan.inian derivation? For this to be possible, one would require either that Pan.inis grammar have some
rules providing for sound shifts of the type bh- h- or include in
the set of verb bases not only items such as bhu but others such as
ho found in forms like Middle Indic hoti, honti. There is no evidence for either. Houbens assertion could be maintained also under the view that a speaker who considers a preliminary statement
and is familiar with Middle Indic in addition to Sanskrit could conceive of such a string with some Middle Indic elements. This,
however, runs contrary to the very conception of what sabdanusa sana and vyakaran.a mean to Pan.inyas: a work by which correct
60 See
Cardona 1974b: 235237, 283284 note 16.
79
usage is explained once this has been segregated from incorrect

usage.61 Thus conceived, Pan.inis grammar deals only with correct usage. This is noted at the beginning of the Mahabhas.ya,
where Patajali considers three alternatives of how the sa bdanusa sana should proceed: should it serve to teach correct speech forms
(sabdopadesa), incorrect speech forms (apasabdopadesa) or both
(ubhayopadesa)?62 The third option is set aside as redundant. Explaining correct forms would establish a restriction whereby incorrect forms are ruled out, and teaching that incorrect forms are not
to be allowed would result in limiting the grammar to correct usage. The decision to explain only correct speech forms is made on
the grounds of brevity (laghutvat). For any single correct speech
form such as gauh. there are several incorrect forms (apabhram
sa h.
corruptions) such as gav, gon., gopotalika. Consequently, Pan.inis sabdanusa sana is a grammatical treatise whereby are explained
speech forms that are desired and accepted.63 Moreover, Patajali remarks that augments, replacements, affixes, verbal bases, and
nominal bases recited in the As..ta dhyay are all given therein with61 See
PWT.: 564572 845848. This accepted view is not weakened by the

fact that according to varttika twelve on A. 1.3.1: BUa;va;a;d;ya;ea ;Da;a;ta;vaH (BUa;va;a;
a;d;pa;a;FH
:pra;a;a;ta;pa;
a;d;k+:a;a;pa;ya;tya;a;
a;d;a;na;vxa:ya;TRaH) one reason for listing bhu and so on in the dhatupat.ha is to exclude Middle Indic forms like a n.apayati. This statement occurs
in the context of suggestions that a term assigned the name dhatu be defined as
one which signifies an action as denoted by the term kriya (kriyavacana: Bh.
I.254.13: ya;
a;d :pua;naH ;
a;k+:ya;a;va;.ca;na;ea ;Da;a;tua;a:=+tyea;ta;+.a;Ma ;
a;k+:yea;ta ) or the term bhava (Bh.
I.256.18: ya;
a;d :pua;na;Ba;Ra;va;va;.ca;na;ea ;Da;a;tua;a:=+tyea;ta;+.a;Ma ;
a;k+:yea;ta ): a list of bases assigned the
name dhatu would still be required in order to exclude such items as a n.apayati. But Pan.ini himself does not define dhatu in either of the ways suggested; he
operates with a set of bases divided into subsets.
62 Za;b.d;a;nua;Za;a;sa;na;a;ma;d;a;na;Ma k+:tRa;v.ya;m,a ta;tk+:TMa k+:tRa;v.ya;m,a ;
a;kM Za;b.d;ea;pa;de ;ZaH k+:tRa;v.ya ;a;h;ea;a;~va;d;pa;Za;b.
/ /
d;ea;pa;de ;Za ;a;h;ea;a;~va;d
/ / u ;Ba;ya;ea;pa;de ;Za I+.a;ta (Bh. I.5.1214).
63 l+.Gua;tva;a;.cC+.b.d;ea;pa;de ;ZaH l+.Ga;a;ya;a;VZa;b.d;ea;pa;de ;ZaH ga:=+a;ya;a;na;pa;Za;b.d;ea;pa;de ;ZaH O;;kE+:k+:~ya Za;b.d;~ya ba;h;va;ea Y;pa;BrMa;Za;aH ta;d;a;Ta;a ga;Ea;a:=+tya;~ya ga;a;va;a;ga;ea;a;a;ga;ea;pa;ea;ta;a;l+.k+:a;d;ya;ea Y;pa;BrMa;Za;aH I+.;a;nva;a;K.ya;a;nMa Ka;va;
a;pa Ba;va;a;ta (Bh. I.5.2022).
80
G. C ARDONA
out any fault of pronunciation.64 At least from the time of Patajali, then, the As..ta dhyay and its ancillary texts were transmitted
without faults of pronunciation. Accordingly, unless one wishes
completely to divorce Pan.ini from his later tradition with respect
to this issue, one can rule out of court the possibility of unassimilated Middle Indicisms in the As..ta dhyay and its ancillaries.
Under Houbens thesis, a weaker claim is made: Pan.inis
derivational rules enter into play after a preliminary string has been
formed and serve to polish such a string. Even this position, however, fails to find concrete support in rules of the As..ta dhyay. If,
for example, a preliminary string could contain forms such as hoti, which required being corrected to bhavati, one should expect
a rule substituting bh- for initial h-. No such rule is to be found
in the As..ta dhyay. At most, one would have to say that a speaker
was made aware in some way that bhavati is the accepted form
corresponding to hoti. In Houbens words, the first level of Pan.inis derivational system serves a consultative purpose: Statements in the grammar belonging to this level ask the user to make
decisions and judgements regarding different aspects of his provisional statement (1999: 45). Houben does not, however, demonstrate with evidence from the As..ta dhyay precisely how Pan.inis
derivational system would interact with any provisional statement.
Moreover, the fact remains that Pan.inis very words show he considered meanings to be conditions (nimitta, see 3) for introducing
affixes.
Consequently, it is difficult to see how Pan.inis grammar would
allow and interact with preliminary sentences that Houben envisions, containing possible Middle Indic forms, which would require the intervention of the grammar to bring them to acceptable
ta;a;va;d;yMa :pra;;v.yaH :*:e +:mea .sMa;vxa;ta;a;d;yaH (rUa;yea:=+ a;a;a;ta ;a;ga;mea;Sua ;a;ga;ma;aH Zua:;d ;aH :pa;F:a;ntea ;
a;va;k+:a:=e+Sua ta;
a;hR ;
a;va;k+:a:=+aH Zua:;d ;aH :pa;F:a;ntea :pra;tya;yea;Sua ta;
a;hR :pra;tya;ya;aH Zua:;d ;aH :pa;F:a;ntea
;Da;a;tua;Sua ta;
a;hR ;Da;a;ta;va;ea Y;
a;pa Zua:;d ;aH :pa;F:a;ntea :pra;a;a;ta;pa;
a;d;ke+:Sua ta;
a;hR :pra;a;a;ta;pa;
a;d;k+:a;nya;
a;pa Zua:;d ;a;a;na
:pa;F:a;ntea (Bh. I.14.1518).
64 ;Ta;vea;dM
81
status. Such utterances would have to be creations of some speakers who possibly intended to utter something in accordance with
accepted norms reflected in the grammar but somehow slipped. Although Pan.inyas accept that Pan.inis grammar serves to contrast
accepted and unacceptable usages, there is no evidence of their
conceiving of the As..ta dhyay as a set of rules serving to touch up
unacceptable preliminary statements.
There is thus no evidence from Pan.inis work or the statements
of Pan.inyas to support the major claim Houben makes. Nor does
Houben himself offer much concrete evidence to back up his assertions. I find one repeated claim that is based on actual statements
in the As..ta dhyay, and this concerns Vedic rules. Houben remarks
(1999: 48):
It is further to be noted that if Pan.inis grammar were
to go basically from meaning to form, there would be
no good justification for the presence of rules dealing with finished utterances in transmitted Vedic literature. But if the starting point is not meaning
but a provisional statement the grammar may not
only serve to fine-tune (sam-skr
) our own statements
(mainly bhas.a , in the case of

u ha possibly also chandas), but also to check a given expression (chandas,
but also bhas.a ) and contribute where necessary to its
proper understanding and codification.65
65 At this point, Houben adds a note (1999: 48 note 31): Cf. Patajalis statement that the padakaras (apparently the authors of a padapa.tha) are to follow
the grammar rather than the other way around . . . . The Mahabhas.ya discussions
to which Houben then alludes concern the presence or absence of an avagraha
and the segmentation ! :!a;,aY;va;a;n,a, ! :!a;,aY;va;ntaH, found in the Rgvedapadapa.tha,
instead of ! :aY;va;a;n,a, ! :aY;va;ntaH. Houbens apparently is therefore inappropriate; obviously would have been better.
82
G. C ARDONA
In the same vein, Houben says (2003: 161), If the system is only
synthetic, why would so much attention have been paid to the finished utterances of Vedic texts with all their grammatical exception? More recently (2012: 16667), the same author asks
If placing semantics as the starting point is in conflict with the presence of a considerable number of
rules, namely the specifically Vedic ones (and moreover with the majority of the explicit aims mentioned
in the Paspasa hnika: raks.a preservation of the Veda,
u ha ritual modification etc.), how can it be that this
view has become so widespread among Pan.ini specialists (cf. Houben 1999: 2326)?
It is noteworthy that Houben leaves padapat.has out of consideration except to refer to padakaras in passing (see note 65). Now,
Pan.inian rules which provide for usages that apply only in the domain of Vedic are fully compatible with his derivational system.
Pan.inian rules which provide for Vedic usage, whether as a general
domain (chandasi) or in particular subdomains such as mantras or
verses (rc), should be considered to operate on the basis of the posited strings, as in padapathas. Consequently, Panini
oretically
.
.
can state the same sort of rules for Vedic usage as he does for the
language current in his time and area, differing from these only by
applying in specific limited domains. For example, in Vedic there
occur first person plural active forms of the type smas` (e.g., RV.
1.29.1b) instead of smah. and nominative plural forms of the type

devasàh. instead of devah. (e.g., RV. 1.19.6b). In the Rgvedapada
. To account for
these, Pani akalya gives devasàh., smas`
pa.tha, S
.
ni provides that in Vedic (chandasi) alone, masi with final -i

added to mas occurs and the nominal ending jas receives the final augment asuk if it follows a stem ending with an a-vowel.66 A
66 A.
7.1.46: I+.d;nta;ea ma;a;sa (C+.nd;a;sa 38), 7.1.50: ;a:*.a;sea:=+sua;k, (C+.nd;a;sa 38). The Kasika on each of these sutras refers to what is designated by chandasi as a domain
83
slightly more complex example of variation concerns third person

plural middle verb forms such as a sàte, duhrate, and duhre. The

first represents the norm in Sanskrit, with the a tmanepada ending

-ate. The second has the ending -ate, but with an initial r-, and the
last example lacks -t- (-e -ae). Pan.ini accounts for these starting
with a base followed by the ending jha, which replaces the l-affix
lat.. The -a of this ending is replaced by -e, since it derives from an
l-affix marked with .t. In a tmanepada endings following a stem that
does not end in -a, jh of this ending is replaced by at. Further, in
Vedic usage (chandasi), the ending can receive the initial augment
rut. and also have its t deleted.67 Stating that such forms occur in
a particular domain, Vedic usage, is comparable to saying that in
akat.a yana speakers use third plural imperfect forms
the dialect of S
with the ending us instead of ant ( an) after verbal bases in -a as
well as after dvis.: ayuh., advis.uh. instead of ayan, advis.an.68 Moreover, this system has a background in the procedures of padapat.ha
akalya, who posit structures to account for samhit
a
authors like S
69
texts.
Houben also brings into play the sociolinguistic situation at
the time of Pan.ini and early Pan.inyas and claims that the grammar will serve not only to eliminate imperfections such as possible
Middle Indic forms but also to fine-tune and check expressions.
Now, Patajali does indeed consider the coexistence of nonstandard and accepted standard usage. He also makes the important
(vis.aya): C+.nd;a;sa ;
a;va;Sa;yea ma;a;sa;tya;yMa Za;b.d I+.k+:a:=+a;nta;ea Ba;va;a;ta ;va;a;Ra;nta;a;d;*: +.a;du:a:=+~ya
.ja;sea:=+sua;ga;a;ga;ma;ea Ba;va;a;ta C+.nd;a;sa ;
a;va;Sa;yea
67 ;a;f;ta ;a;tmea;pa;d;a;na;Ma .fe:=e (A. 3.4.79); A. 7.1.5, 8: ;a;tma;nea;pa;de ;Sva;na;taH (H 3, ;t,a 4),
A. 7.1.8: ba;hu;lM C+.nd;a;sa (.+:f, 6); A. 7.1.41: l+.ea;pa;~ta ;a;tma;nea;pa;de ;Sua (C+.nd;a;sa 38). By A.
6.1.97: ;ta;ea gua;ea (;pa;d;a;nta;a;t,a 96, :pa:=;+.pa;m,a 94, O;;kH :pUa;vRa;pa:=+ya;eaH 84), -e is the single
replacement for -ae ( ate) once -t- has been dropped.
68 A. 3.4.111, 112: l+.z H Za;#f;a;ya;na;~ya (;a;taH 110, e +jRua;s,a 108), ;
a;d;Sa;(
a
69 I have mentioned S
akalya in particular, since it is certain that Pan.ini knew
his padapat.ha.
84
G. C ARDONA
point that eminent standard speakers could use nonstandard forms

with impunity, so long as this was not done in a ritual.70 In addition, Katyayana and Patajali lay stress on the question whether
the knowledge or use of correct speech leads to merit (dharma) and
felicity (abhyudaya). After an extended discussion, the conclusion
is reached that both the knowledge and use of correct speech forms
should be involved: the use of correct speech forms preceded by a
knowledge of the grammar results in felicity.71 As Kaiyat.a notes,
when Katyayana says sa strapurvake prayoge, he means the use of
correct speech forms preceded by a knowledge of the grammar,
in the sense that one who uses such speech forms knows them
to be correct as described in Pan.inis sabdanusa sana in terms of
bases and other constituents involved in derivation.72 That is, although a native speaker of the accepted speech also knows vernacular speech and could use either to express what he wishes to say,
it is not sufficient that he be able to use correct speech. This is meritorious if one also knows that the speech in question is granted its
status by virtue of being described as it is in Pan.inis grammar.
That speakers of Pan.inis time and place knew not only the accepted correct speech but also vernacular forms dubbed incorrect
(asadhusabda, apasabda) nevertheless does not require concluding that they made use of the As..ta dhyay to correct preliminary
utterances. Such a claim, made without conclusive evidence from
the As..ta dhyay itself, is mere speculation. It simply assumes that
because Pan.ini doubtless lived in a multi-lingual environment, his
grammar must serve to touch up utterances of one speech to yield
utterances of another.
Patajali also names the preservation of the Veda as the first
of the compelling reasons for studying grammar.73 To be sure,
70 See
PWT.: 550 833.

PWT.: 547548 831.
72 . . . Za;b.d;a;na;Ma :pra;kx+:tya;a;
a;d;
a;va;Ba;a;ga;a;a;na;pUa;vRa;kH :pra;ya;ea;ga I+.tya;TRaH (Pr. I.40).
73 See PWT.: 544545 828.
71 See
85
preservation of the Veda does involve checking Vedic expressions.

Consider again forms of the type duhrate, duhre. Someone who is

a native speaker of the accepted correct speech, where forms with

-ate are the norm, could possibly consider such forms deviant, but
one who knows grammar does not commit this error. As Kaiyat.a notes, such a person knows that Pan.inis grammar accounts for
such Vedic forms.74 This does not imply that Pan.inis grammar
must be a system that begins with preliminary utterances which
can be checked for correctness. It means simply that a grammarian conversant with Pan.inis derivational system knows that it accounts for such Vedic forms in a particular manner. In the same
vein, the fact that Patajali also invokes the modification (uha) of
mantras in particular contexts does not require that Pan.inis system
conform to what Houben claims. For example, if a mantra contains
the dative form agnayè in one context and another context calls for
a corresponding form of the stem su rya-, any native speaker knows

that the form in question is surya` ya.75 One who also knows Pan.inis grammar is capable of modifying a mantra with a knowledge
of how the grammar accounts for the forms. Both forms involve
nominal stems followed by the ending ne.
After a stem in -a, this
ending is replaced by ya, which then conditions the substitution of
a long vowel -a in the stem,76 but an ending marked with n conditions replacement by a gun.a vowel for the final -i of a stem that
belongs to the ghi class:77 surya-e surya-ya suryaya, agni-e
74 l+.ea;ke l+.ea;pa;a;d;a;dx ;M :vea;de dx ;;a Bra;a;}yea;d;vEa;ya;a;k+.=+aH ;vEa;ya;a;k+.=+a;~tua na Bra;ma;a;ta :vea;d;a;Ta . ca;a
;Dya;va;~ya;a;ta (Pr. I.8). Kaiyat.a illustrates with aduhra, the imperfect corresponding
to duhre, and cites the pertinent rules. In his Mahabhas.yadpika, Bhartrhari illustrates with several examples; it is not necessary to discuss these formshere.
75 Kaiyata (Pr. I.9) cites the two forms and Vedic texts in which they are used.
.
Bhartrhari goes into much more detail concerning u ha. Again, it is not necessary
to consider
all the details in the present discussion.
76 A 7.1.13: :ze+.yRaH (;taH 9, ;
*: +.~ya 6.4.1); 7.3.102: .sua;
a;pa . ca (;ta;ea d ;a;Ga;eRa ya;a;Va 101).
77 A 7.3.111: ;Gea;
a;zR+.a;ta (gua;aH 108).
86
G. C ARDONA
agne-e agnaye.78
Let me finally revert to the central issue of Pan.inian derivations
beginning with meaning conditions. Concerning modern scholars
who accept this basic point, although they differ in other respects,
Houben remarks (1999: 31),
However, if the models are taken seriously, it can
hardly be imagined that anyone was ever able to make
effective use of Pan.inis grammar: How can a potential speaker start with mere meanings or with semantic
elements, or with an abstract semantic level, and next
ask his grammar to provide him with the proper utterance to express it?
This remark betrays, I think, a misunderstanding of what Pan.inis
work is meant to be. As Pan.inyas from earliest times recognize,
this work is an anusa sanam, a means of explaining, through derivation, correct speech units as they occur in utterances of model
speakers (sis..ta). Pan.inyas do not conceive of the As..ta dhyay as
something that should provide a speaker with the proper utterance
to express what he wishes to convey. It is viewed as a means of
accounting for what speakers utter when they communicate with
each other. To be sure, Bhartrhari remarks that a speaker selects
m)
(upadatte) an appropriate (yogya
speech unit (sabdam) with respect to meanings to be conveyed (abhidheyes.u) that have been
brought to mind (-prapitasannidhanes.u made to be present) by
his wish to express (vivaks.a -), just as one applies the appropriate
organ of perception with respect to objects when one wishes to
perceive something; and he speaks of this vivaks.a as dependent
on appropriate speech units (yogyasabdanibandhanaya),79 that is,
78 A
6.1.78: O;;.ca;ea Y;ya;va;a;ya;a;vaH (;a;.ca 77).

;
a;h Za;b.dM :pra;ya;ea;+:a ;
a;va;va:a;a;pra;a;
a;pa;ta;sa; a;a;Da;a;nea;Sva;a;Ba;Dea;yea;Sua :pra;tya;TRa;mua;pa;a;d:ea ta;d;a;Ta;ea;pa;a;l+.psa;ma;a;naH :pra;a;ta;
a;va;Sa;yMa ya;ea;gya;mea;vea;a;nd
/ +ya;mua;pa;l+.b.Da;Ea :pra; a;a;Da:ea (VPVr. 1.13 [44.45]);
79 ya;ea;gyMa
87
on speech units which have the capacity to express the meanings

one wishes to convey. But the same author has no illusions of
Pan.inis or anyone elses grammar being a true model of what is
in a speakers mind or instructing speakers on how they may produce utterances. On the contrary, he explicitly denies this. For
example, he says that affix meanings (pratyayarthah.) which are
stated (nirdis..ta h.) as a condition (nibandhanam) for the introduction of affixes in a given grammatical tradition (anvakhyanasmrtau
an explanatory smrti) are said (udahrtah.) to be, in another gram
matical tradition (smrtyantare), meanings attributed to bases (prakrtyarthah.).80 In hisauto-commentary on the karika in question,
Bhartr
hari remarks that the division of meanings that is adopted by
different grammarians (paurus.eyah. pravibhagah. division made

by men) depends solely on what these grammarians wish to accept: in the grammars of some grammarians, all affixes are explained as redundant, serving to cosignify meanings attributed to
their bases.81 Neither Bhartrhari nor his predecessor or successor
Pan.inyas conceive of the grammar

as somehow directly playing
a role in a speakers producing an utterance, serving to provide
a potential speaker with the proper utterance to express meanings. Pan.inyas consider Pan.inis sabdanusa sana a description, a
cf. Za;b.d O;;va;ea;pa;de ;;a ta;tsa;a;ma;TyRa;mea;va;a;nua;ga;.cC+.nta;ea va;+:a:=+ea ya;ea;gya;Za;b.d;a;na;ba;nDa;na;yEa;va ;
a;va;va:a;ya;a
:pra;va;tRa;ntea (VPVr. 1.129 [209.12]).
80 ;nva;a;K.ya;a;na;smx
a;ta;Ea yea . ca :pra;tya;ya;a;Ta;Ra ;a;na;ba;nDa;na;m,a ;a;na;
a;dR ;;a;~tea :pra;kx+:tya;Ta;RaH .smxa;tya;nta:=
o+.d;a;&+.ta;aH (VP. 2.231).
81 .~vea;.cC+.a;a;na;ba;nDa;na O;;va;a;yMa :pa;Ea:+:Sea;yaH :pra;
a;va;Ba;a;ga;ea Y;Ta;Ra;na;Ma ba;h;Da;a;(ra;a;ya;tea ta;Ta;a ;
a;h :ke+:Sa;au
;a:*.a;tsma;tRXa;a;Ma .smxa;a;ta;Za;a;~:ea;Sua .sa;vRa O;;va :pra;tya;ya;aH .~va;a;a;TRa;k+:aH :pra;kx+:tya;Ta;Ra;nua;va;a;

a;d;na I+.tya;nva;a;K.ya;a;ya;ntea (VPVr. 2.231 [248]). Different views are in fact considered in the Masubsequent works concerning meanings to be attributed to bases
habhas.ya and
such as asva (horse). Some scholars attribute five meanings to the base: the
generic property (jati) of being a horse, individual (vyakti) horses, gender (lin ga), number (sankhya), and the possible role a referent plays in accomplishing
an action (karaka); others say the first three are base meanings, the last two affix
meanings. It is not necessary to discuss the issues here. See Deshpande 1992.
88
G. C ARDONA
means of accounting for how a community of speakers expresses

itself, intended for an audience of native speakers of the language
described.
In sum, Houben has not presented cogent evidence supporting
his claims. What he adduces in support of these claims is equally
well explained by understanding Pan.inis system in the way that
Pan.inyas have understood it from the times of earliest commentators onwards. His appeal to the sociolinguistic situation of Pan.inis
time does not suffice to counter the incontrovertible evidence that
Pan.inis rules treat meanings as conditions for the introduction of
affixes. Nor does it demonstrate that speakers used the As..ta dhyay to modify preliminary utterances which could include Middle
Indic forms.
It is proper to reject Houbens attempt to demonstrate that Pan.inis system does not operate in the manner accepted by Pan.inyas
and most modern scholars. It is appropriate to accept instead that
Pan.inis is a derivational system which starts with meanings to be
expressed by utterances that are accounted for by combining bases
and affixes under stated conditions.82
5
5.1
Interpreting Pan.inian sutras

Introduction
The As..ta dhyay is composed in Sanskrit to account for utterances

in this language by native speakers of Pan.inis time as well as
82 Scharf 2011 also argues against Houbens thesis. I have limited myself to
considering Houbens claims because he presents some arguments. S. D. Joshi
(2009: 2) merely asserts, Contrary to some Western misconceptions the starting
point of Pan.inis analysis is not meaning or the intention of the speaker, but word
form elements as shown in the initial stages of the prakriya. Here morphemic
elements obtained from analysis are put side by side in an order of purva and
para from left to right.
89
for archaic usage in Vedic texts. Accordingly, Pan.ini can expect

that students of his grammar, who are themselves native speakers,
knew conventions of usage common to him and all other native
speakers of the language he describes. For example, when he says
.~va:=+a;
a;d;a;na;pa;a;ta;ma;v.ya;ya;m,a (A. 1.1.37),83 he can assume of his audience
that they will understand a verb form asti or bhavati.84 In addition, he can assume his audience to understand an utterance like A.
1.1.37 in the same manner one understands
(13) ayam
gauh. (asti) This (is) a bull
in response to the question
(14) ko yam What is this?
Someone who sees a bull for the first time and asks (14) receives
(13) as an answer informing him that the animal present is one
called by the word go (nom. sg. gauh.). Similarly, A. 1.1.37
states that items of the set beginning with svar (heaven) and those
called nipata are named avyaya. The native command of a user of
the As..ta dhyay does not, however, suffice to interpret certain forms
in certain contexts. Pan.ini supplies metarules to allow unambiguous interpretations in such instances.
5.2
Contexts requiring metarules
In his first three varttikas on :Sa; +a .~Ta;a;nea;ya;ea;ga;a (A. 1.1.49), Katyayana

considers the purpose of this sutra and under what conditions the
rule comes into play.85 First, he remarks that the sutra is stated
83 See
PWT.: 5 11, 27 54.

formulates this convention in varttika 11 on ;na;a;Ba;
a;h;tea (A. 2.3.1).
See PWT.: 67 13.
85 On the issues considered in this section see Cardona 1974a for a fuller discussion.
84 Katyayana
90
G. C ARDONA
in order to establish a restriction (niyamartham).86 That is, this

metarule establishes that of all the possible relations which can
condition the use of a sixth-triplet ending, only the relation in
place of is to be understood in grammatical statements. The second varttika notes that this allows for an overly broad application
(atiprasangah.) in that the restriction thus established would apply
also where a sixth-triplet ending should be understood to designate
other relations, such as part-whole.87 Lastly, Katyayana notes that
the restriction does not obtain (apraptih.) for relations such as partwhole, because the relation (yogasya) in question is not doubted
(asandigdhatvat).88 As Patajali explains, a restriction is formulated for instances where there is doubt, and in the case of terms
with sixth-triplet endings signifying relations such as part-whole,
there is no doubt. Further, one does not have to say this explicitly,
since it is known from common usage. For example, a person A
going to another village may inquire of a person B, saying, Im
going to another village. Please tell me the way. B then says to A,
At that spot, you should take the right-hand path, at that spot the
left one. The direct diagonal path is not explained, because there
is no doubt about it. Similarly, in the grammar, the restriction applies in case of doubt, and there is no doubt where a relation such
as part-whole is understood.89
86 :Sa;.a;aH
.~Ta;a;nea;ya;ea;ga;va;.ca;nMa ;a;na;ya;ma;a;TRa;m,a (1.1.49 vt. 1).

Za;a;sa;ea ga;ea;h I+.a;ta (1.1.49 vt. 2). I deal here only with
the first rule alluded to, A. 6.4.34 (see 5.2.1).
88 ;va;ya;va;Sa;.a;a;d ;a;na;Ma . ca;a;pra;a;a;a;ya;eRa;ga;~ya;a;sa;a;nd;gDa;tva;a;t,
/
a (1.1.49 vt. 3).
89 ;va;ya;va;Sa;.a;a;d ;a;na;Ma . ca ;a;na;ya;ma;~ya;a;pra;a;a;aH ;
a;kM k+:a:=+a;m,a ya;ea;ga;~ya;a;sa;a;nd;gDa;tva;a;t,
/
a .sa;nde ;he
;a;na;ya;ma;ea na . ca;a;va;ya;va;Sa;.a;a;
a;d;Sua .sa;nde ;hH ;
a;kM va;+:v.ya;mea;ta;t,a na ;
a;h k+:Ta;ma;nua;.cya;ma;a;nMa gMa;~ya;tea
l+.Ea;
a;k+:k+:ea Y;yMa dx ;;a;ntaH ta;d;a;Ta;a l+.ea;ke k+:a:*.a;tk+: a;(a;tpxa;.cC+.a;ta g{a;a;ma;a;nta:=M ga;a;ma;Sya;a;a;ma :pa;nTa;a;nMa mea Ba;va;a;nua;pa;
a;d;Za;a;tva;
/ a;ta .sa ta;sma;a ;a;.ca;e ;mua;a;Sma;a;va;k+:
/ /
a;Zea h;~ta;d; a:a;a;ea g{a;h ;a;ta;v.ya;ea
Y;mua;a;Sma;a;va;k+:
/ /
a;Zea h;~ta;va;a;ma I+.a;ta ya;~ta:a ;a;ta;yRa;#pa;Ta;ea Ba;va;a;ta na ta;a;sma;nsa;nde
///
;h I+.a;ta kx+:tva;a na;a;sa;a;vua;pa;
a;d;Zya;tea O;;va;a;ma;h;a;
a;pa .sa;nde ;he ;a;na;ya;ma;ea na . ca;a;va;ya;va;Sa;F:a;a;
a;d;Sua .sa;nde ;hH (Bh.
I.118.1925).
87 ;va;ya;va;Sa;.a;a;
a;d;Sva;a;ta;pra;sa;
*:H
91
In brief, A. 1.1.49 is invoked only where the context of a sutra

does not suffice to show that a particular relation is concerned.
5.2.1 :Sa; +a .~Ta;a;nea;ya;ea;ga;a (A. 1.1.49)
Recall (see 2.2.2) that Patajali says (5) rajah. purus.asya is in
effect an incomplete string which creates an expectancy for something else. If, for example, one supplies bharya (wife), the result
is
(15) rajah. purus.asya bharya
saying that a woman is the wife of the servant of the king. Here, the
genitive purus.asya is left without any expectancy for something
related to the man. Each of the genitives rajah. and purus.asya is
now bound to another term.
A parallel to (5) and (15) in the As..ta dhyay is A. 6.4.34: Za;a;sa I+.d:* :+.l+.eaH (o+.pa;Da;a;ya;aH ;
a;*;+. a;ta 24).90 The genitive sa sah. (of sa s)
is linked to upadhayah., and the relation between the two entities
referred to is known: sa s is a base and upadha signifies a penultimate sound. In addition, upadhayah. can be linked to it, which
denotes a short vowel i, but the meanings of these terms cannot
make known the particular relation that holds between a penultimate sound and i, so that the relation signified by the sixth-triplet
ending (A. 2.3.50, see 2.2.2) in upadhayah. remains unspecified.
The metarule A. 1.1.49 therefore comes into play: a genitive not
susceptible of a single interpretation in a given sutra is understood
to denote one relation: in place of (sthane). Accordingly, sthane is understood and linked with upadhayah. in A. 6.4.34, which
thus provides that i (it) occurs in place of the penultimate sound
(upadhayah.) of sa s (command, instruct) before the suffix an or
the initial consonant of a suffix marked with k or n ; e.g., asis.at
90 See
Cardona 1974a: 30810.
92
G. C ARDONA
(3sg. aor.), sis..ta- (past participle).91 A. 1.1.49 does not serve to

interpret the sa sah., however, because this is already linked to upadhayah., so that the relation involved that of a part and a whole
(avayavavayavibhavasambandhah.) is contextually known.
In a similar vein, consider A. 7.1.8588:92 :pa;a;Ta;ma;Tyxa;Bua:a;a;ma;a;t,a
(.sa;Ea 82, ;*: +.~ya 6.4.1), I+.ta;ea Y;tsa;vRa;na;a;ma;~Ta;a;nea Ta;ea Y;nTaH Ba;~ya .fe;l+.eRa;paH,
which state operations for the stems pathin- (path, way), mathin(churning utensil), and rbhuks.in- (Rbhuks.in). Before endings of
ana (su,au, jas, am, aut), -a- substithe set called sarvanamasth
.
tutes for -i- of these stems, and -this replaced by -nth-; before the
first-triplet ending su in particular, the final sound -n of these stems
is replaced by a ; before vowel-initial endings, however, that part
of the stem that begins with its last vowel that is, -inis replaced by zero; e.g., panthah. (nom. sg. pathin-su), panthanau
(nom-acc. du. pathin-au/aut.), patha (instr. sg. pathin-t.a ).
The genitive plural pathimathyrbhuks.a m in A. 7.1.85 cannot be un A. 1.1.49 serves to interpret this
ambiguously interpreted, so that
as signifying the relation in place of. The same genitive is understood to recur in A. 7.1.8687, which also contains the genitive
forms itah. and thah.. Here, pathimathyrbhuks.a m is immediately
linked to itah. and thah., so that contextually

one understands the
first genitive to refer to the stems of which the vowel i and the consonant th are parts. Accordingly, here A. 1.1.49 serves to interpret
itah. and thah., not pathimathyrbhuks.a m. Similarly, in A. 7.1.88
this genitive as well as bhasyaare understood to be linked directly
with .teh.: they refer to the stems in question when they occur before vowel-initial endings, so that they bear the label bha. On the
other hand, .teh. is not bound. Hence, A. 1.1.49 comes into play:
that part of the stem which is called .ti is deleted.
91 See
92 See
PWT.: 298 447.

PWT.: 312 472.
93
The part-whole relation holds also between an item and a segment that is added to it as an augment. For example, A. 6.4.71
72: lu +.*: +.*: x +*+: +.qu+.d;a:aH (;*: +.~ya 1), ;a;q+ja;a;d ;a;na;a;m,a introduce at. and a .t
to stems which occur before verb endings that derive from the laffixes lun,
a .t for vowel-initial stems, at. for others.93
lan,
and lrn;
Now, at. and a .t bear the marker .t, and the sutra ;a;d;a;nta;Ea f;
a;k+:ta;Ea (A.
1.1.46) provides that entities marked with .t and k respectively are
the initial (adi) and final (anta) parts of the items to which they are
introduced. Accordingly, the genitive ending of anganam understood in A. 6.4.7172 is interpreted contextually as signifying the
part-whole relation, so that A. 1.1.49 does not come into play.94
There are also sutras with genitive forms naming various grammarians whose observations Pan.ini mentions. For example, A.
8.3.18, 19, 22: v.ya;ea;lR +.Gua;pra;ya;a;ta:=H Za;a;k+:f;a;ya;na;~ya l+.ea;paH Za;a;k+:ya;~ya
h;a;l .sa;veRa;Sa;a;m,a (; a;Za 17) concern different ways in which pada-final
-v and -y preceded by an a-vowel as well as bho, bhago, and agho
akat.a yana, they
are treated before voiced segments: according to S
are produced with very slight effort, as light glides; according to
akalya, they are deleted; but all agree that they are deleted before
S
a consonant.95 Clearly, sa kat.a yanasya, sa kalyasya, and sarves.a m,
used with reference to persons, do not refer to entities subject to
replacement. They refer to scholars who have described different
substitutions for -v and -y, so that the relation in question is that
of one by whom something has been stated and the one who has
akars.t (3sg. aor. kr-lun),
akarot (3sg. impf. kr-lan),
akaris.yat
(3sg. cond. kr-lrn);
aiks.ata (3sg. impf.
aiks.is..ta(3sg. aor. ks.-lun),
(3sg. cond. ks.-lrn)

ks.-lan),
aiks.is.yata
from the bases kr (do, make) and ks.
(look).
94 The singular a
ngasya of the major heading is contextually modified to the
plural anganam as required by the plural lunlanlrn ks.u, and this is coreferential
with ajadnam.
93 E.g.
95 See
PWT.: 356358 557558.
94
G. C ARDONA
stated this (proktapravacakabhavasambandha).96 Here again, A.

1.1.49 does not enter into play.
Consider now the sutra .sa:+.pa;a;a;a;mea;k+:Zea;Sa O;;k+:
a;va;Ba;+:Ea (A. 1.2.64).
This rule provides for one (eka-) of a group of homophonous terms
(sarupan.a m) to remain (-ses.ah. which is left over) when a single
ending is used.97 Here again, there is a bound genitive: sarupan.a m is linked with eka-, which is coreferential with -ses.a in the
compound ekases.ah.. The relation in question can be considered
that of part and whole: one of a group remains. In addition, however, the construction can be of the type vadatam
varah. best of
speakers, where a sixth-triplet ending follows a nominal denoting
a group from which (yatah.) a member (or members) is singled out
(nirdharan.am).98 Accordingly, A. 1.1.49 is not required for interpreting A. 1.2.64, so that it is not brought into play here. The rule
does not provide for substitution by a single unit.99
5.2.2 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)
5.2.2.1 Purpose of the rules. These sutras apply to interpret
locative and ablative forms in contexts where a native speaker
could not otherwise interpret such terms unambiguously.100 The
96 Accordingly, some commentators supply matena (according to the opinion
of . . . ), e.g., Kas. 8.3.18: va;k+:a:=+ya;k+:a:=+ya;ea;Ba;eRa;Ba;ga;ea ;Ga;ea ;va;Ra;pUa;vRa;ya;eaH :pa;d;a;nta;ya;ea;lR +.Gua;pra;ya;a;ta:= ;a;de ;Za;ea Ba;va;tya; a;Za :pa:=+taH Za;a;k+:f;a;ya;na;~ya;a;.ca;ayRa;~ya ma;tea;na
97 See PWT.: 1112 13, 260261 374.
98 A. 2.3.41: ya;ta;(
a ;a;na;Da;Ra:=+a;m,a (.sa;a;ma;a 36, :Sa; +a . ca 38). Cf., e.g., N. 1.2.64 (I.372374): .sa:+.pa;a;a;Ma Za;b.d;a;na;a;a;ma;a;ta ;a;na;Da;Ra:=+ea :Sa; +a .tea;na .~va:+.pa;a;a;Ma ma;Dyea .sa;ma;a;na:ja;a;ta;a;ya;ea yaH
.sa:+.paH .sa O;;k O;;va ; a;Za;Sya;tea na ;
a;va:+.paH .sa;ma;a;na:ja;a;ta;a;ya;~yEa;va ;
a;h ;a;na;Da;Ra:=+Ma Ba;va;a;ta na ;
a;va:ja;a;ta;a;ya;~ya ;Ta;va;a .sa;mua;d;a;ya;l+a;Ea;Sa;a :Sa; +a .sa:+.pa;sa;mua;d;a;ya;a;ntaHpa;a;ta;a . ca .sa:+.pa O;;va Ba;va;a;ta
na ;
a;va:+.paH .tea;na .sa:+.pa O;;va ; a;Za;Sya;tea
99 Pontillo (2013: 11824) attempts to show that the rule does provide for substitution, but does not consider the syntax of the genitive sarupan.a m.
100 See Cardona 1974a: 31226, PWT.: 53 90.
95
rules apply, as does A. 1.1.49,101 to restrict (niyamartham) what

one is to understand.102 For example,
(16) grame devadattah. Devadatta is in the vicinity of the village
has the locative grame referring to a place near where Devadatta is
situated. One does not know from this sentence whether this man
is at a place before or after the village. Similarly, if one says
(17) gramad devadattah.
the ablative gramat could ambiguously be connected with different
direction words:103 purvah. (before, east of), parah. (beyond).
Supplying terms such as purvah. in such instances eliminates the
ambiguity. Similarly, in rules such as A. 6.1.77 and 8.1.28 (see
below), where locative and ablative forms occur which a native
speaker cannot unambiguously interpret in context, A. 1.1.6667
enter into play.104
Patajali remarks that a substrate, to which the karaka class
name adhikaran.a is assigned, is of three kinds:105 an enveloping
(vyapaka) locus, such as sesame seeds in which (tiles.u [loc. pl.])
oil (tailam) occurs; a locus of contiguity (aupasles.ika); and a locus of domain (vais.ayika), which involves neither enveloping nor
101 See
5.2 with notes 8589.

:pUa;va;eRa:a:=+ya;ea;ya;eRa;ga;ya;ea:=+
a;va;Zea;Sa;a; a;a;ya;ma;a;Ta va;.ca;nMa d;Dyua;d;kM :pa;.ca;tya;ea;d;na;m,a
(1.1.6667 vt. 2).
103 It is assumed that (17) does not refer to the village as a karaka and is instead to be construed with a direction word to be supplied; A. 2.3.29: ;nya;a:=+a;
a;d;ta:=+teRa;
a;d;#Za;b.d;a:*.Ua:a:=+pa;d;a:ja;a;
a;h;yua;e (:pa:*.a;ma;a 28) accounts for fifth-triplet endings after
nominals linked with direction words; see PWT.: 165 250.
104 ta;a;smM
/ / / a;~ta;sma;a;
a;d;a;ta :pUa;va;eRa:a:=+ya;ea;ya;eRa;ga;ya;ea:=+
a;va;Zea;Sa;a; a;a;ya;ma;a;Ta;eRa Y;ya;ma;a:=+}BaH g{a;a;mea :de;va;d:aH
:pUa;vRaH :pa:= I+.a;ta .sa;nde ;hH g{a;a;ma;a;e +va;d:aH :pUa;vRaH :pa:= I+.a;ta .sa;nde ;hH O;;va;a;ma;h;a;pa;a;k+:ea ya;a;a;.ca d;Dyua;d;kM :pa;.ca;tya;ea;d;na;m,a o+.Ba;a;
a;va;k+:a;vua;Ba;a;va;.ca;Ea ;a;.ca :pUa;vRa;~ya;a;a;.ca :pa:=+~yea;a;ta .sa;nde ;hH ;a;ta;*:+.a;ta;z
I+.tya;a;ta;zH :pa:=+~yea;a;ta .sa;nde ;hH I+.Sya;tea . ca;a:a;a;a;.ca :pUa;vRa;~ya .~ya;a;d;a;ta;zH :pa:=+~yea;a;ta ta;a;a;nta:=e+a
ya;Ma na ;a;sa;Dya;ta;a;a;ta ;a;na;ya;ma;a;Ta va;.ca;na;m,a (Bh. I.172.2226).
105 ;a;Da;k+.=+Ma na;a;ma ;
a:a;pra;k+:a:=M v.ya;a;pa;k+:ma;Ea;pa:(;e+
a;Sa;kM ;vEa;Sa;a;ya;k+:a;ma;a;ta (Bh. III.51.89).
102 ta;a;smM
/ / / a;~ta;sma;a;
a;d;a;ta
96
G. C ARDONA
contiguity, for example, the ultimate happiness referred to as svarga (loc. sg. svarge) as the domain of ones wish (iccha). He
also notes that a speech unit cannot have any relation with another
speech unit which is a locus except contiguity.106
For example, the -i of dadhi- (yogurt) and -u of madhu
(honey) are replaced respectively by -y and -v before a vowel
such as the a- of atra (here): dadhy atra, madhv atra. The rule
which states this substitution, A. 6.1.77 (I+.k+:ea ya;a;a;.ca), consists of
three singular terms: a genitive ikah., a nominative yan., and a locative aci. The first two terms are linked through the intervention of
A. 1.1.49 (5.25.2.1): semivowels y v r l, denoted by yan., occur
in place of vowels i u r l, denoted by ik. The locative aci, however,
cannot be interpretedunambiguously
in the sutra as it stands. To
be sure, since aci is a locative and ac denotes vowels, so that the
rule concerns the replacement of vowels by semivowels, one understands a relation of contiguity. Nevertheless, the vowels to be
replaced could be understood to occur before or after other vowels.
A. 1.1.66 comes into play to specify that aci refers to vowels that
serve as contexts for an operation on what precedes (purvasya).
Analogously, A. 1.1.67 serves to interpret ablative forms such
as atinah. in A. 8.1.28: ;a;ta;*:+.a;ta;zH (;nua;d;a:Ma .sa;vRa;ma;pa;a;d;a;d;Ea 18). The
sutra provides that a pada which terminates with a verb ending has
all low-pitched vowels if it follows a pada with an ending other
than a verb ending (atinah.).107
5.2.2.2 Scope of 1.1.66. That A. 1.1.6667 establish restrictions such that locative and ablative forms are subject to specified
interpretations does not mean that they apply everywhere that a
sutra contains such a form.
106 Za;b.d;~ya . ca Za;b.de;na k+:ea Y;nya;a;a;Ba;sa;}ba;nDa;ea Ba;
a;va;tua;ma;hR ;tya;nya;d;ta o+.pa:(;e+Sa;a;t,a (Bh.
III.51.910). This is said by way of rejecting the need for A. 6.1.72 (.sMa;
a;h;ta;a;ya;a;m,a),
an issue that does not concern the present discussion.
107 See PWT.: 393 615.
97
Consider, for example, A. 7.2.1: ;a;sa;a;.ca vax ;a:;d H :pa:=+smEa;pa;de ;Sua (;*: +.~ya 6.4.1), where the genitive angasya is understood from a major heading and two locatives occur: sici, parasmaipades.u. Since
vrddhi is used, replacement applies with respect to sounds denoted
the abbreviation ik: the genitive ikah is understood to occur
by
.
here.108 In addition, ikah. is also understood to qualify angasya,
so that the operation in question applies to an anga that ends in a
vowel denoted by ik:109 a vrddhi vowel (a, ai, au) substitutes for
ends with a short or long vowel i,
the final sound of a stem which
u, r.110 The locative sici, interpreted by A. 1.1.67, designates the
context for this operation on a preceding stem. Now, Paniright
.
nyas generally consider not only sici but also parasmaipades.u to
be what is called a parasaptam. Thus, the Kasikas paraphrase of
A. 7.2.1 is :pa:=+smEa;pa;d;pa:=e ;a;sa;a;.ca :pa:=+ta I+.ga;nta;a;*: +.~ya vxa;a:;d ;BRa;va;a;ta. Further,
Jinendrabuddhi and Haradatta explicitly remark that sici and parasmaipades.u are both parasaptam, that sic is a following element
with respect to a stem while a parasmaipada ending is a following element with respect to sic.111 The locative sici can indeed
rightly be considered a parasaptam, as interpreted by A. 1.1.67, in
the sense of a locative designating a following item relative to an
immediately preceding item to which a stated operation applies.
On the other hand, parasmaipades.u cannot be interpreted appropriately by the same rule. For then A. 7.2.1 would be considered
to provide for vrddhi replacement for a stem that immediately pre
cedes parasmaipada
endings.112 In the present instance, the stem
108 I+.k+:ea
gua;a;vxa:;d ;a (A. 1.1.3); see PWT.: 56 97.

;
a;va;a;Da;~ta;d;nta;~ya (A. 1.1.72); see PWT.: 14 30.
110 ;l+.ea Y;ntya;~ya (A. 1.1.52); see PWT.: 55 95.
111 ;a;sa;.ca;a;a;ta :pa:=+smEa;pa;de ;a;Sva;
/ a;ta . ca :de ;pyea;tea :pa:=+sa;a;}ya;Ea ta:a ;a;sa;.ca;ea Y;*: +.a;pea:Ma :pa:=+tvMa :pa:=+smEa;pa;d;a;na;Ma tua ;a;sa:ja;pea:a;m,a (N. V.650). :a ;a;sa;.ca;a;a;ta :pa:=+smEa;pa;de ;a;Sva;
/ a;ta . ca :de ;
a;pa :pa:=+sa;a;}ya;Ea
ta:a ;a;sa;.ca;ea Y;*: +.a;pea:Ma :pa:=+tvMa :pa:=+smEa;pa;d;a;na;Ma ;a;sa:ja;pea:a;m,a (PM. V.650).
112 Where an operation may apply to an element separated (vyavahita) from the
conditioning context, Pan.ini specifies this; thus A. 8.4.1: .=+Sa;a;Bya;Ma na;ea .aH .sa;ma;a;na;pa;de
109 yea;na
98
G. C ARDONA
that immediately precedes parasmaipada endings ends in sic; e.g.,

aji-s-am ( ajais.am I was, have been victorious), asru-s-am
( asraus.am I [have] heard), akr-s-am ( akars.am I [have]
made). The operation would now
be comparable to the substitution of the vrddhi vowel a for the -r- of the stem mrj- (mrj-mi
A. 7.2.114: mxa:
marjmi I wipe)
as provided for by
jea;vRxa;a:;d HIn both
instances, the replacement affects a penultimate vowel of a stem
before an ending. However, if this were the case, one would expect A. 7.2.1 to have a genitive sicah. instead of the locative sici,
just as A. 7.2.114 has a genitive mrjeh.. Moreover, it is clear from
element which conditions rewhat Pan.ini says elsewhere that the
placement is sic, not the ending that follows this affix. Thus, in
order to preclude gun.a replacement for the penultimate vowel of
a base such as bhid split before sic in deriving abhitsi (1sg. aor.
mid.), he provides that this affix is marked with k if it is not augmented with initial it., precedes a tmanepada affixes, and follows a
base that ends in a consonant preceded by a vowel i, u, r or l.113
u in
In accordance with Pan.inis procedure, then, parasmaipades
.
A. 7.2.1 is more appropriately to be understood as a locative absolute (A. 2.3.37, note 48): vrddhi replacement applies to a base
before sic when parasmaipadaaffixes occur. Since these affixes are
substitutes for l-affixes, which are introduced after bases to begin
with, they do indeed occur after sic, which substitutes for cli after
verbal bases when these occur with suffixes that replace lun.
114
Pan.ini describes not only the accepted usage of his time and
area but also properties of speech in Vedic usage, referred to in
lets n. replace n following r or s. in the same pada, and A. 8.4.2: :*:u +.pva;a;*:u+.}v.ya;va;a;yea Y;
a;pa states that this replacement applies even if n is separated from r or s. by
particular elements; see PWT.: 366367 570.
113 ;a;l+.a;*;+.
.ca;a;va;a;tma;nea;pa;de ;Sua (A. 1.2.11); see PWT.: 335 515. Gun.a would apply by
A. 7.3.86: :pua;ga;nta;l+.gUa;pa;Da;~ya . ca; see PWT.: 79 127.
114 See also Cardona 1970: 6566 note 5.
99
general by the locative chandasi. As the Kasika notes,115 this

refers to a domain (vis.aya) of usage: the locative refers to a locus (adhikaran.a) of a particular kind, such that A. 1.1.66 does not
enter into play.
The following three sutras contain the locatives yus.madi,
asmadi, and ses.e: yua;Sma;d;au;pa;pa;de .sa;ma;na;a;a;Da;k+.=+ea .~Ta;a;a;na;nya;
a;pa ma;Dya;maH
;sma;d;au:a;maH Zea;Sea :pra;Ta;maH (A. 1.4.105, 107, 108). These rules provide
for the distribution of sets of verb endings which replace l-affixes
and to which are assigned the class names madhyama, uttama, and
prathama.116 Endings of the madhyama and uttama sets, respectively, are selected if the l-affix which has been replaced is coreferential respectively with a potentially used second and first person
pronoun yus.mad and asmad. Endings of the prathama group are
selected if neither of these conditions is met (ses.e remainder).
The term upapade (co-occuring term), stated in A. 1.4.105 and
understood in the next two rules mentioned, states a co-occurrence
condition. Accordingly, the locatives upapade, samanadhikaran.e,
and sthanini are not to be interpreted by A. 1.1.66. On the contrary,
these are locative absolutes, accounted for by A. 2.3.37.117
The same use of locative forms applies for rules in which
meaning conditions are stated for operations; see 2.3.
5.2.2.3 Scope of 1.1.67. The contexts in which ablative forms
are to be interpreted by A. 1.1.67 are comparably circumscribed.
Consider, for example, A. 3.1.97: ;.ca;ea ya;t,a (;Da;a;ta;eaH 91, :pra;tya;yaH 1,
:pa:=+(a 2) and A. 4.1.4: :ja;a;d;a;ta;;a;p,a (za;a;ppra;a;a;ta;pa;
a;d;k+:a;t,a 1, ;a;~:
/ a;ya;a;m,a 3,
:pra;tya;yaH 3.1.1, :pa:=+(a 3.1.2). Both come under the major headings
A. 3.1.12, whereby an item introduced by a subsequent rule is
called pratyaya, and is said to occur following (parah.) items to
115 See
4 with note 66.

PWT.: 150151 234.
117 See 2.3 with note 48.
116 See
100
G. C ARDONA
be specified. A. 3.1.97 is stated under the subheading A. 3.1.91, so

that the affix in question is introduced after an item which bears the
class label dhatu: the sutra introduces yat following a verbal base
(dhatoh.) that ends in a vowel (acah.).118 A. 4.1.4 comes under the
subheading A. 4.1.1: it provides for the suffix .ta p in derivates of the
set beginning with aja (goat) as well as after nominal bases that
end with short a (at).119 The ablative forms in these sutras are thus
unambiguously construed with the direction word parah. and there
is no question of A. 1.1.67 playing a role in their interpretation.
Similarly, there are many sutras which state the distribution of
a tmanepada and parasmaipada suffixes and in which Pan.ini refers
to verbal bases preceded by preverbs.120 For example, :pa;a:=+v.ya;vea;ByaH
;
a;k+:yaH (A. 1.3.18) provides for such affixes after kr preceded by
pari, vi or ava, as in pari krn.a ti buys, hires, buys back. The
ablative parivyavebhyah. is not to be interpreted by A. 1.1.67: the
sutra does not state an operation on kr immediately following pari, vi or ava. On the other hand, A. 1.4.80 (.tea :pra;a;gDa;a;ta;eaH) provides
that elements with the class names upasarga and gati occur before
a verbal base; pari, vi or ava construed with kr are members of the
upasarga set. Accordingly, ji follows these items, so that parivyavebhyah. of A. 1.3.18 is understood to be construed with a direction
word such as uttara (abl. sg. uttarasmat) following.121
Summary
Pan.inyas not only recognize the interaction between a speaker

(prayoktr) and a listener (srotr) in communication (1) but also
that Paninis derivational
emphasize
system operates from a
.
118 See
PWT.: 76 122.
2.3 with note 51.
120 See PWT.: 8990 141.
121 Cf. K
as. 1.3.18: :pa;a:=+v.ya;vea;Bya o+a:=+sma;a;tk
+:a;a;a;tea:=+a;tma;nea;pa;dM Ba;va;a;ta
119 See
101
speakers perspective (2). There is, moreover, incontrovertible

evidence that Pan.ini himself adopts this position: a derivation accounting for speech units in an utterance starts from meanings to
be expressed, and these meanings serve as does also the cooccurrence of other items as conditions (nimitta) for the introduction of affixes to bases (3). A claim to the contrary has not
been justified (4).122 In addition, Pan.ini takes a listeners perspective in connection with the interpretation of sutras in the As..ta dhyay. Those who are expected to understand these statements
are native speakers of the language described, of whom Pan.ini can
and does expect a knowledge of conventions and syntactic constructions known to any native speaker (5). In accordance with
this, he formulates particular metarules to allow interpreting rules
where for reasons of economy and generalization he leaves
out certain terms necessary for the unambiguous interpretation of
given genitive, locative, and ablative forms.
The As..ta dhyay is thus not simply a set of rules which are to
apply mechanically in deriving utterances. The sutras of Pan.inis
grammar account through derivation for utterances which native
speakers of the language described use, and these sutras require
for their interpretation and application the active intervention of a
native speaker of this language.
122 After I presented this paper, J. E. M. Houben spoke at some length, attempting to defend his thesis. In an email message of June 21st, he once more attempted to justify this thesis. I have profited from his discussions, though I still
consider his claims to be speculative and not based on probative evidence.
102
G. C ARDONA
Table 1
Abbreviations
As..ta dhyay.
Mahabhas.ya. See Kielhorn 19621972; Vedavrata
19621963.
Dpika Bhartrharis Mahabhas.yadpika.
See Bronkhorst
1987.
Kas. Kasikavrtti. See Vijayapala 1997.
the Kasikavrtti. See Dwarika Das Shastri
N.
Nyasa on
19651967.
and Kalika Prasad Shukla
PM. Padamajar on the Kasikavrtti. See Dwarika Das
19651967.
Shastri and Kalika Prasad Shukla
Pr.
Pradpa on the Mahabhas.ya. See Vedavrata 1962
1963.
PS.
Pan.inyasiks.a . See Ghosh 1938.
PWT. See Cardona 1997.
sivaramendrasarasvat.
RaPr. Ratnaprakasa of Sr
See
Narasimhacharya 1978.
RV.
Rgveda. See Sonatakke and Kashikar 19331951.
RPr. Rgvedapratisa khya. See Mangal Deva Shastri 1931.
Ud.
Uddyota on the Mahabhas.ya. See Vedavrata 1962
1963.
VP.
Vakyapadya. See Rau 1977.
VPH. Helarajas Prakrn.aprakasa on the Vakyapadya. See
Raghunatha Sarm
a 1979, Subramania Iyer 1963.
VPP. Srvrs.abhas paddhati, the Sphut.a ks.ara, on the Vakyapad

ya. See Subramania Iyer 1966.
VPVr. Vrtti on the Vakyapadya. See Subramania Iyer 1966,
Subramania
Iyer 1983.
A.
Bh.
R EFERENCES
103
References
Bronkhorst, Johannes, ed. and trans. 1987. Mahabhas.ya-Dpika
of Bhartr.hari: fascicule IV, a hnika I. Bhandarkar Oriental Research Institute Post-graduate and Research Department Series
28. Poona: Bhandarkar Oriental Research Institute.
Cardona, George. 1970. Some principles of Pan.inis grammar.
Journal of Indian Philosophy 1: 4074.
. 1974a. On Pan.inis metalinguistic use of cases. Charudeva
Shastri Felicitation Volume: presented to Prof. Charudeva
Shastri on the occasion of his seventy-fifth anniversary by
friends and admirers, ed. by Charudeva Shastri felicitation volume editorial committee, pp. 30526. Delhi: Charudeva Shastri Felicitation Committee.
. 1974b. Pan.inis karakas: agency, animation and identity.
. 1983. Linguistic analysis and some Indian traditions. The Pandit Shripad Sastri Deodhar Memorial Lectures, first series.
Pune: Bhandarkar Oriental Research Institute.
. 1997. Pan.ini: his work and its traditions; vol. 1, Background
and Introduction. Second edition, revised and enlarged. Delhi:
Motilal Banarsidass.
. 2004. Pan.inian sutras of the type anyebhyo pi drsyate. Jambujyoti: Munivara Jambuvijaya Festschrift, ed. by M. A.
Dhaky and J. B. Shah, pp. 91107. Ahmedabad: Shreshthi
Kasturbhai Lalbhai Smarak Nidhi, Shardaben Chimanbhai Educational Research Centre.
. 2008. Pan.ini and Pan.inyas on ses.a relations: first Kunjunni Raja memorial lecture. Aluva, Kerala: Kunjunni Raja
Academy of Indological Research. [Revised as Cardona 2013.]
. 2009. On the structure of Pan.inis system. Sanskrit computational linguistics: first and second international symposia,
Rocquencourt, France, October 2007; Providence, RI, USA,
104
G. C ARDONA
May 2008; Revised selected and invited papers, ed. by Grard

Huet, Amba Kulkarni, and Peter M. Scharf, pp. 132. Lecture
Notes in Artificial Intelligence 5402.
. 2012. A note on Vakyapadya 1.45/46: a tmabhedas tayoh.
kecid . . . Samskr
tasadhuta Goodness of Sanskrit: studies in
A. N. Aklujkar, ed. by Chikafumi Watanhonour of Professor
abe, Michele Desmarais, and Yoshichika Honda, pp. 1009.
Delhi: D. K. Printworld.
. 2013. Pan.ini and Pan.inyas on ses.a relations. Indological research: different standpoints, ed. by P. C. Muraleemadhavan,
pp. 99146. Revised edition. New Delhi: New Bharatiya Book
Corporation. [Revision of Cardona 2008.]
Deshpande, Madhav Murlidhar. 1992. The meaning of nouns: semantic theory in classical and medieval India Namarthanirn.aya of Kaun.d.abhat..ta translated and annotated. Studies of
Classical India 13. Dordrecht: Kluwer.
Dwarika Das Shastri and Kalika Prasad Shukla, eds. 19651967.
arya Jinendrabuddhipada and PaNyasa or Pacika of Ac
damajar of Haradatta Misra on the Kasikavr.tti [Commentary on the As..ta dhyay of Pan.ini] of Vamana-Jayaditya:
Bodhisattvadesyacaryajinendrabuddhipadaviracitaya Nyasaparaparyayakasikavr.ttipacikaya vidvadvaraharadattamisraviracitaya Padamajarvyakhyaya ca sahita Kasikavr.ttih. Pan.inyas..ta dhyayvyakhya. 6 vols. Prachya Bharati Series 27.
Varanasi: Prachya Bharati Prakashan.
. a or the Si
Ghosh, Manomohan, ed. and trans. 1938. Pan.inya Siks
ks.a Vedan ga ascribed to Pan.ini (being the most ancient work
on Indo-Aryan phonetics): critically edited in all its five recensions with an introduction, translation and notes together
with its two commentaries. Calcutta: University of Calcutta.
[Reprint: Delhi: Asian Humanities Press, 1986.]
R EFERENCES
105
Houben, Jan E. M. 1999. Meaning statements in Pan.inis grammar: on the purpose and context of the As.t.a dhyay. Studien
zur Indologie und Iranistik 22: 2354.
. 2003. Three myths in modern Pan.inian studies. Asiatische
Studien/tudes Asiatiques 57: 12174.
. 2009a. Bhartr.hari as a cognitive linguist. Bhartr.hari: language, thought and reality: proceedings of the international
seminar, Delhi, December 1214, 2003, ed. by Mithilesh
Chaturvedi, pp. 52343. Delhi: Motilal Banarsidass.
. 2009b. Pan.inis grammar and its computerization: a construction grammar. Sanskrit computational linguistics: third international symposium, Hyderabad, India, January 2009, proceedings, ed. by Grard Huet and Amba Kulkarni, pp. 625.
Lecture Notes in Artificial Intelligence 5406.
. 2012. Studies in Indias Vedic grammarians, 1: Narayan.a
Studies in SanBhat.t.as Prakriya-sarvasva and Pan.inis Se.
skrit Grammars: proceedings of the Vyakaran.a Section of
the 14th World Sanskrit Conference, ed. by George Cardona,
Ashok Aklujkar, and Hideyo Ogawa, pp. 16394. Delhi: D.K.
Printworld.
Huet, Grard and Amba Kulkarni, eds. 2009. Sanskrit computational linguistics: third international symposium, Hyderabad,
India, January 2009, proceedings. Lecture Notes in Artificial
Intelligence 5406. Berlin; Heidelberg: Springer-Verlag.
Joshi, S. D. 2009. Background of the As.t.a dhyay. Sanskrit computational linguistics: third international symposium, Hyderabad, India, January 2009, proceedings, ed. by Grard Huet
and Amba Kulkarni, pp. 15. Lecture Notes in Artificial Intelligence 5406.
Kielhorn, Lorenz Franz, ed. 19621972. The Vyakaran.a = Mahabhas.ya of Patajali: revised and furnished with additional
readings, references, and select critical notes, rev. by Vasudev
Kashinath Abhyankar. 3rd ed. 3 vols. Poona: Bhandarkar Ori-
106
G. C ARDONA
ental Research Institute. [A reissue labelled fourth edition

was published in 1985 and 1996.]
Mangal Deva Shastri, ed. and trans. 1931. The R.gveda-Pratisa khya with the Commentary of Uvat.a: edited from original
manuscripts, with introduction, critical and additional notes,
English translation of the text, and several appendices; vol. 2,
Text in Sutra-form. Allahabad: The India Press Limited.
Narasimhacharya, M. S., ed. 1978. Commentaires sur le Mahabhas.ya de Patajali et le Pradpa de Kaiyat.a: Mahabhas.ya
Pradpa Vyakhyanani; vol. 5, Adhyaya 2 Pada 14. Publications de lInstitut Franais dIndologie 51.5. Pondichry: Institut Franais dIndologie.
Pontillo, Tiziana. 2013. Where the sense is intended although
the corresponding speech unit is not employed: the ekases.a
case. Proceedings of the 15th World Sanskrit Conference; vol.
2, Vyakaran.a Across the Ages: proceedings of the vyakaran.a
section of the 15th World Sanskrit Conference, ed. by George
Cardona, pp. 97132.
Raghunatha Sarm
a, ed. and comm. 1979. Vakyapadyam [Part III,
Vol. II] [Bhuyodravya-Gun.a-Dik-Sadhana-Kriya-Kala-Purus.a-Sankhya-Upagraha and Lingasamuddesa]: with the commentary Prakasa by Helaraja and Ambakartr by Pt.
Raghunatha Sarm
a: Vakyapadyam [trtyam
kan.d.am, dvityo bhagah.] (Bhuyodravya-gun.a-dik-sadhana-kriya-kala-purus.a-sankhya-upagraha-lingasamuddesa tmakam) Helarajaviracitaya prakasavyakhyaya Pa. Raghunathasarman.a viracitaya Ambakartrvyakhyaya samalankrtam. Sarasvat Bhava
na Granthamala 91. Varanasi: Sampurnanand
Sanskrit Vishvavidyalaya.
Rau, Wilhelm, ed. 1977. Bhartr.haris Vakyapadya: die Mulakarikas nach den Handschriften herausgegeben und mit einem
Pada-Index versehen. Abhandlungen fr die Kunde des Morgenlandes Band XLII, 4. Wiesbaden: Steiner.
R EFERENCES
107
Scharf, Peter M. 2009. Levels in Pan.inis As..ta dhyay. Sanskrit

computational linguistics: third international symposium, Hyderabad, India, January 2009, proceedings, ed. by Grard
Huet and Amba Kulkarni. Lecture Notes in Artificial Intelligence 5406.
. 2011. On the semantic foundation of Pan.inian derivational
procedure: the derivation of kumbhakara. Journal of the
American Oriental Society 131: 3972.
Sonatakke, N. S. and C. G. Kashikar, eds. 19331951. R.gvedaSamhit
a: with the commentary of Sayan.a carya. 5 vols. Poona:
Vaidic Samshodhan Mandal.
Subramania Iyer, K. A., ed. 1963. Vakyapadya of Bhartr.hari with
the Commentary of Helaraja: Kan.d.a III, Part 1: critically
edited . . . . Deccan College Monograph Series 21. Poona: Deccan College.
. ed. 1966. Vakyapadya of Bhartr.hari with the Vr.tti and the Paddhati of Vr.s.abhadeva: Kan.d.a I: critically edited . . . Deccan
College Monograph Series 32. Poona: Deccan College.
. ed. 1983. Vakyapadya of Bhartr.hari (An ancient Treatise on
the Philosophy of Sanskrit Grammar) Containing the T.ka of
Pun.yaraja and the Ancient Vr.tti: edited ... with a Foreword by
Ashok Aklujkar, Kan.d.a II. Delhi: Motilal Banarsidass.
bhagavatpatajaliviracitam
Vedavrata, ed. 19621963. Sr
Vyaka
ran.a-mahabhas.yam: Srkaiyyat.akrtapradpena Nagojbhat..takrtena Bhas.yapradpoddyotena cavibhus.itam. 5 vols. Gurukul
anam.
Jhajjar
(Rohtak): Harayan.a -Sahitya-Samsth
vamanajayadityaviracita Pan.inyas..ta Vijayapala, ed. 1997. Sr
dhyaysutravrttih. Kasika. Sonpat (Haryan.a ): Ram Lal Kapoor
Trust.
108
G. C ARDONA
Extension rules and the syntax of

As..ta dhyay sutras with -vati
G EORGE C ARDONA
Abstract: Pan.inis As..ta dhyay includes sutras of a type called
atidesasutra (extension rules), which provide for extending to elements properties which are not theirs to begin
with or to deny them properties which they would otherwise have. The rules in question are of two general kinds:
(a) those which merely predicate of units that they have
properties extended to them; and (b) those which provide
that particular operations, rules, and so on apply to given
items as they would elsewhere. The latter involve the use
of terms with the suffix vati; the former do not. In turn,
rules of type (b) involve two distinct syntactic patterns. I
shall take up extension rules with particular emphasis on
their syntax and consider also pragmatics associated with
them.
Keywords: Pan.ini, grammatical rule syntax, operation extension
Introduction
Pan.inis As..ta dhyay includes extension rules (atidesasutra). Sabara characterizes an atidesa in general as letting properties which are
provided for a given entity X go beyond X and have an additional
locus Y.1 For example, one may provide that Devadatta should be
1 JMSS.
7.1.1.12 (V: 386387): ;a;ta;de ;Za;ea na;a;ma yea :pa:=;a ;

a;va;
a;h;ta;a ;Da;ma;Ra;~ta;ma;ta;a;tya;a;nya:a .tea;Sa;Ma :de;ZaH ya;Ta;a :de;va;d:a;~ya Ba;ea:ja;na;
a;va;a;DMa kx+:tva;a Za;a;a;l+.sUa;pa;ma;Ma;sa;a;pUa;pEa;deR ;va;d:a;ea Ba;ea-
109
110
G. C ARDONA
fed rice, curry, meat, and apupa cakes, then also provide that Yajadatta should be fed similarly:
(1) Za;a;a;l+.sUa;pa;ma;Ma;sa;a;pUa;pEa;deR ;va;d:a;ea Ba;ea:ja;a;ya;ta;v.yaH
Devadatta should be fed . . .
(2) :de;va;d:a;va;d;a;a;d:a;ea Ba;ea:ja;a;ya;ta;v.yaH
Yajadatta should be fed similarly to Devadatta.
Pan.inyas recognize several kinds of extension rules according
to the type of property of a given item that is extended to another.
There are, to begin with, four major types: those which serve to extend operations (karya to be done), rules (sa stra), physical shape
(rupa), and identity (tadatmya). Three minor types in addition
to these are also noted: extension of conditions (nimitta), a name
(vyapadesa), and a meaning (artha).2
:ja;a;ya;ta;v.ya I+.a;ta ta;mea;va ;
a;va;a;DMa ya;a;d:ea Y;a;ta;
a;d;Za;a;ta :de;va;d:a;va;d;a;a;d:a;ea Ba;ea:ja;a;ya;ta;v.ya I+.a;ta
(;+ea;k+:ma;pyua;d;a;h:=+a;nta
/ /
:pra;kx+:ta;a;tk+:mRa;a;ea ya;sma;a:a;tsa;ma;a;nea;Sua k+:mRa;sua
;Da;mRa;pra;de ;Za;ea yea;na .~ya;a;tsa;ea Y;a;ta;de ;Za I+.a;ta ;a;~Ta;
/ / a;taH
I+.a;ta .sa . ca na;a;}:a;a va;.ca;nea;na va;a ta:a na;a;ma ;

a:a;
a;va;Da;ma;a;a;ta;de ;Za;kM k+:mRa;na;a;ma .sMa;~k+:a:=+na;a;ma ya;Ea;a;ga;k+:a;ma;a;ta va;.ca;nMa :pua;na;
a;dR;
a;va;DMa :pra;tya:a;(rua;ta;ma;a;nua;ma;a;a;na;kM . ca Pun.yaraja (VPPu. 2.77 [37.9])
succinctly defines atidesa as making a property proper to one entity obtain for
another: ;nya:a;a;nya;Da;mRa;pa;a:=+pra;a;pa;a;ma;a;ta;de ;ZaH
2 Vasudeva Dksita (SKB. 274 [= 7.1.95, I.277]) quotes a verse naming seven
.
types: k+:a;yRa:+.pa;a;na;a;ma:a;a;TRa;Za;a;~:a;ta;a;d;a;t}ya;Za;a;b./ d;ta;aH v.ya;pa;de ;Za;(a .sa;Ea;ta;a;na;a;ta;de ;Za;a;npra;.ca:a;tea Pun.yaraja, Haradatta, Bhat.t.oji Dks.ita and Janendrasarasvat recognize six
:pra;k+:a;=E +
a;vRa;Zea;Sa;a;a;ta;de ;ZaH :pa;a:=+dx ;Zya;tea (VPPu. 2.77
types: Za;a;~:ea .+.pa;a;a;ta;de ;Za;a;
a;d;a;BaH :Sa;
a:*.H
[37.2425), :Sa;ea;Q+a;a;ta;de ;Za;ea ;a;na;a;ma:a;a;a;ta;de ;Za;ea v.ya;pa;de ;Za;a;a;ta;de ;ZaH Za;a;~:a;a;a;ta;de ;Za;ea .+.pa;a;a;ta;de ;Za;~ta;a;d;a;t}ya;a;a;ta;de ;ZaH k+:a;ya;Ra;a;ta;de ;Za I+.a;ta (PM. 1.1.56 [I.184]), ;a;na;a;ma:a;v.ya;pa;de ;Za;ta;a;d;a;t}ya;Za;a;~:a;k+:a;yRa:+.pa;
a;va;Sa;ya;k+:tvea;na;a;a;ta;de ;Za;~ya :Sa;
a;q+:Da;tvea Y;
a;pa :pra;a;Da;a;nya;a;
a;d;h .+.pa;mea;va;a;a;ta;
a;d;Zya.tea
(SKP. 274 [406407]), ;a;na;a;ma:a;v.ya;pa;de ;Za;ta;a;d;a;t}ya;Za;a;~:a;k+:a;yRa:+.pa;
a;va;Sa;ya;tvea;na;a;a;ta;de ;Za;~ya;a;nea;k+:
a;va;Da;tvea Y;
a;pa :pra;a;Da;a;nya;a;
a;d;h .+.pa;mea;va;a;a;ta;de ;Zya;tea (SKT. 274 [I.277]). Jinendrabuddhi
mentions five: ;a;ta;de ;Za;ea Y;nea;k+:pra;k+:a:=H ;a;na;a;ma:a;a;a;ta;de ;Za;ea v.ya;pa;de ;Za;a;a;ta;de ;ZaH Za;a;~:a;a;a;ta;de ;Za;ea
.+.pa;a;a;ta;de ;ZaH k+:a;ya;Ra;a;ta;de ;Za;(ea;a;ta (N. 1.1.56 [I.182183]). I omit comparable statements by commentators on non-Pan.inian grammars.
E XTENSION RULES
111
There are extension rules in which terms with the suffix vat
(vati, with the marker i), such as sthanivat like a sthanin are used,
and others in which such terms are not used. In addition, sutras
with terms in -vati are subject to different syntactic analyses.
2
2.1
Varieties of extension
Karyatidesa
A major example of a sutra providing for the extension of operations is

A. 1.1.56 .~Ta;a;a;na;va;d;a;de ;Za;ea Y;na;a;va;Da;E
// a
According to A. 1.1.56, a replacement Y for an item X is like X,
that is, behaves in the same way as does X, with respect to operations, except with respect to an operation that would thus apply
as determined by an original sound X. For example, consider the
derivations of the non-feminine forms kena (instr. sg.), kabhyam
(instr.-dat.-abl. du.), kais, (instr. pl.), kasmai (dat. sg.), and vrks.a ya (dat. sg.) of the interrogative pronoun kim and the noun
vrks.a
(tree) shown in Table 1.3

As shown, the sutras which provide for replacing kim- by ka-,
t.a with ina, and ne
with ya come under the heading of A. 6.4.1:
;*: +.~ya, so that the operations in question apply relative to an item

bearing the class name anga. According to
A. 1.4.13 ya;sma;a;tpra;tya;ya;
a;va;a;Da;~ta;d;a;
a;d :pra;tya;yea Y;*: +.m,a
this name is assigned to an item Y which begins with an item X after which an affix A is introduced, if X or Y is followed by A. In the
3 The rules whereby Panini accounts for kaih, which occurs before pause or
.
.
voiceless velar and labial stops, do not concern the present discussion. I also
omit rules that introduce the nominal endings .ta and so on in the first place.
112
G. C ARDONA
Table 1
Derivation of non-feminine forms of the interrogative pronoun
kim and noun vrks.a
kim-t.a
ka-a
A. 7.2.103: ;
a;k+:maH kH (;*: +.~ya 6.4.1)
ka-ina
A. 7.1.12: f;a;z+.a;sa;z+.sa;a;a;ma;na;a;t~ya;aH (;taH 9, *: +.~ya
6.4.1)
kena
A. 6.1.87: ;a;u +aH (;a;.ca 77, O;;kH :pUa;vRa;pa:=+ya;eaH 84)
kim-bhyam
ka-bhyam A. 7.2.103: ;
a;k+:maH kH (;*: +.~ya 6.4.1)
kabhyam A. 7.3.102: .sua;
a;pa . ca (;ta;ea d ;a;Ga;eRa ya;a;Va 101, ;*: +.~ya
6.4.1)
kim-bhis
ka-bhis
7.2.103: ;
a;k+:maH kH (;*: +.~ya 6.4.1)
ka-ais
A. 7.1.9: ;ta;ea ;a;Ba;sa Oe;;s,a (;*: +.~ya 6.4.1)
kais
A. 6.1.88: vxa;a:;d ;=e +a;.ca (;a;t,a 87, O;;kH :pUa;vRa;pa:=+ya;eaH 84)
kim-ne
ka-e
A. 7.2.103: ;
a;k+:maH kH (;*: +.~ya 6.4.1)
kasmai
A. 7.1.14: .sa;vRa;na;a;}:aH .smEa (:ze 13, ;taH 9, ;*: +.~ya
6.4.1)
vrks.a-ne
ksa-ya A. 7.1.13: :ze+.yRaH (;taH 9, ;*:+.~ya 6.4.1)

vr
.
ksa ya
vr
A.
7.3.102:
.
s
a
u
;
a
;pa
.
c
a
(;ta;e
a
d ;a;Ga;eRa ya;a;Va 101, ;*: +.~ya
.
6.4.1)
E XTENSION RULES
113
present examples, kim of kim-a, kim-bhyam, kim-bhis and kim-ne
is an anga relative to the endings .ta , bhyam, bhis, and ne,

which
have been introduced after the base kim. Once kim has been replaced by ka, however, the conditions for membership in the anga
set cease to apply, since the endings in question were introduced
after kim, not ka. According to A. 1.1.56 (p. 111), the replacement
ka is treated as though it were the original base kim, so that the
operations relative to the stem can apply: after a stem in -a, the
endings .ta and bhis are respectively replaced by ina and ais; and if
the stem in question is pronominal (sarvanaman),4 the ending ne
is
replaced by smai. After a non-pronominal stem ending in a, on the
other hand, ya replaces ne,
as in vrks.a-ya ( vrks.a-e). Now, the
by the long
vowel a before
final sound of a stem in a is replaced
5
a nominal ending (supi) that begins with one of the sounds denoted by the abbreviation ya, which includes y and bh. Thus the
-a of ka-bhyam is replaced by -a: kabhyam. Moreover, although
the ya of vrks.a-ya is not the original ending introduced after vr
ks.a-, by A.1.1.56 (p. 111) it has the status of the original ending,
so that the preceding item is treated as an anga relative to it and the
long-vowel replacement is allowed for: vrks.a-ya vrks.a ya.
In all these instances, an operation
which would apply to or
in connection with an item that is replaced is extended to that replacement.6
astratidesa
2.2 S
There is a set of rules stated under the heading
A. 4.2.92 Zea;Sea
4 By
.sa;va;Ra;d ;a;a;na .sa;vRa;na;a;ma;a;a;na (A. 1.1.27), items of the ordered set beginning with
sarva all, entire in which kim is included have the name sarvanaman.
5 The abbreviation sup (loc. sg. supi) refers to the set of twenty-one basic
nominal endings divided into seven triplets, from su au jas through ni
os sup.
6 .~Ta;a;a;na;k+:a;yRa;ma;a;de ;Zea Y;a;ta;
a;d;Zya;tea (Bh. 1.1.56 [I.133.2324]).
114
G. C ARDONA
which introduce taddhita affixes7 under conditions that constitute

a remainder (ses.a) with respect to conditions stated in prior rules
of the taddhita section. As do other groups of rules in this section,
sutras stated under A. 4.2.92 (p. 113) provide for affixes to follow padas which contain particular nominal bases8 and also state
meaning conditions under which such affixations apply. For example, consider the rules
A. 4.3.11 k+:a;l+.a;*+:V,a
A. 4.3.17 :pra;a;vxa;Sa O;;yaH
A. 4.3.54 ;
a;d;ga;a;
a;d;Bya;ea ya;t,a
A. 4.3.53 ta:a Ba;vaH .
A. 4.3.53 provides for an affix to follow a pada which is a value

of tatra,9 that is, which has a nominal base followed by a seventhtriplet nominal ending,10 to form a derived nominal base meaning
which occurs in or at X. A. 4.3.11 and A. 4.3.17 state suffixes
which follow padas containing nominals referring to time (kala):
in general, such a pada takes the suffix .tha,11 but if the base is
pravrs. (rain season) in particular, the suffix is en.ya. For example,
masika ( masa-ni-ika)
located in a month, pravrs.en.ya (
7 A. 4.2.92 (p. 113) heads a subsection of the Asta

. . dhyay within the section
headed by ta;a:;d ;ta;aH (A. 4.1.76); the latter section itself falls within the larger group
of rules under the heading of :pra;tya;yaH (A. 3.1.1). By A. 3.1.1, items introduced
by subsequent rules bear the name pratyaya (affix) and by 4.1.76, affixes introduced by following sutras are called taddhita. See PWT.: 40 70, 7 15.
8 In the absence of such a specification, the affix an applies for meaning con.
ditions by rules beginning with :pra;a;gd ;a;v.ya;ta;ea Y;,a (A. 4.1.83) up to the meaning
condition stated in .tea;na d ;a;v.ya;a;ta Ka;na;a;ta .ja;ya;a;ta ;a:ja;ta;m,a (A. 4.4.2).
9 Locative singular of the pronoun tad (that) used as a variable.
10 .sa;ma;Ta;Ra;na;Ma :pra;Ta;ma;a;d;a (A. 4.1.82) provides that affixes introduced by subsequent
rules are introduced optionally and, if a sutra gives more than one related pada, an
affix follows the first of semantically and syntactically related (samartha) padas.
11 th is a cover symbol which, by F+.~yea;kH I+.sua;sua;+:a;nta;a;tkH (A. 7.3.5051), is
.
replaced by ika or ka, depending on the shape of the preceding stem.
E XTENSION RULES
115
pravrs.-ni-en
. ya) located in the rain season. A. 4.3.54 (p. 114) in
troduces yat after a pada containing a base of the set beginning
with dis direction): dis-ni-ya
disya (located in a direction).
Another subsection headed by A. 4.2.92 (p. 113) comes under

the heading
A. 4.2.24 .sa;a;~ya :de;va;ta;a .
Rules of this section introduce affixes after a pada with a firsttriplet ending to form derivates denoting something whose divinity
is such and such. For example, aindra ( indra-su-an.) dedicated
to Indra, with the suffix an. (see note 8). Particular rules of the section provide for different affixes, depending on the nominal base
of the pada in question. Thus,
A. 4.2.30 .sa;ea;ma;a;*.a;,a
introduces .tyan. after a pada with the base soma: soma-su-t.yan.
saumya (dedicated to Soma).
There are also derivates that conform to the schema of A.
4.2.24 , which have nominal bases denoting times. For example,
masika (dedicated to a month), pravrs.en.ya (dedicated to the rain
season). To account for such derivates,

A. 4.2.34 k+:a;le +.Bya;ea Ba;va;va;t,a
lets affixes introduced under the condition stated in A. 4.2.24 apply
as they do under the conditions which hold for derivates formed
under the meaning condition stated in A. 4.3.53 (p. 114) (bhavavat
in the same manner as for located in). A. 4.2.34 requires that
two conditions hold: (a) affixes that are introduced in the meaning located apply for derivates meaning Y whose deity is X;
and (b) those affixes that are introduced in the meaning located
after padas whose nominal bases denote times apply for derivates
meaning Y whose deity is X. If A. 4.2.34 provided for (a) alone, it
116
G. C ARDONA
would allow any affix allowed under the meaning condition stated
in A. 4.3.53 (p. 114) to follow any base denoting a deity, so that
yat by A. 4.3.54 (p. 114) would wrongly be allowed to occur with
a time word.12 If, on the other hand, (b) alone were provided for
by A. 4.2.34 (p. 115), then any affix provided for by A. 4.3.53
(p. 114) would be allowed under the conditions stated in A. 4.2.34
(p. 115), so that .tha would wrongly be allowed, in accordance
with A. 4.3.11 (p. 114), after a base such as indra, forming *aindrika instead of the required aindra.13 The desired derivates are
accounted for only if both (a) and (b) hold. Both would not hold,
moreover, if A. 4.2.34 (p. 115) provided for extending merely the
operations that apply for A. 4.3.53 (p. 114) to derivations by A.
4.2.24 (p. 115). For both conditions to hold, this sutra should extend not operations but rules (sa stra) which provide for operations
under specified conditions.
This requirement is met because A. 4.2.34 (p. 115) states bhavavat, with the suffix vati, thus requiring that there be a similarity
in conditions for operations.14
1 (k+:a;le +.Bya;ea Ba;vea :pra;tya;ya;ma;a:a;pra;sa;*:H);
I have formulated (a) in accordance with Patajalis expansion of this varttika

(Bh. II.276.1617): ya;
a;d ;
a;va;a;a;ya;tea Ba;vea yea :pra;tya;ya;a ;
a;va;
a;h;ta;a;~tea Ba;va;a;nta
/ / k+:a;le +.ByaH .sa;a;~ya :de;va;tea;tya;a;sma;a;TR
/ / / a I+.a;ta :pra;tya;ya;ma;a:Ma :pra;a;pa ;ea;a;ta I have supplied an example of what
would wrongly obtain in accordance with Kaiyat.a (Pr. III.639): :pra;tya;ya;ma;a:a;a;ma;a;ta
;
a;d;ga;a;
a;d;Bya;ea ya;
a;d;a;ta ya;tpra;tya;ya;ea ;
a;va;
a;h;taH .sa k+:a;l+.va;a;a;.ca;ByaH :pra;a;pa;ea;ta;a;tya;TRaH
13 This possibility is envisioned in 4.2.34 vt. 2 (yaH k+:a;le +.b.ya;ea Ba;vea ta;~ya ;
a;va;Da;a;nea
:pra;a;a;ta;pa;
a;d;k+:ma;a:a;a;tpra;sa;*:H
) ; Bh. II.276.1920: ;Ta ;
a;va;a;a;ya;tea k+:a;le +.Bya;ea Ba;vea yea :pra;tya;ya;a ;
a;va;
a;h;ta;a;~tea Ba;va;a;nta
/ / .sa;a;~ya :de;va;tea;tya;a;sma;a;TR
/ / / a I+.a;ta :pra;a;a;ta;pa;
a;d;k+:ma;a:a;a;tpra;a;pau ;va;a;nta
/ / Pr.
III.639: :pra;a;a;ta;pa;
a;d;k+:ma;a:a;a;
a;d;a;ta k+:a;l+.a;*+:a;Va;a;ta F+.a;Va;nd+a;de :=+
a;pa :pra;a;pa;ea;a;ta
14 This is the explanation accepted in the K
asika for A. 4.2.34 (p. 115) and
comparable sutras: va;tk+.=+Ma .sa;vRa;sa;a;dx ;Zya;pa;a:=+g{a;h;a;TRa;m,a Later commentators explain
similarly. In his third varttika on A. 4.2.24 (p. 115) (;a;sa:;d M tUa;Ba;ya;a;na;deR ;Za;a;t,a ), on
the other hand, Katyayana says the correct results are assured because both the
conditions are stated. This is most straightforwardly understood as Patajali takes
it, reformulating the sutra to state kalebhyah. twice (k+:a;le +.Bya;ea Ba;va;va;tk+:a;le +.ByaH ) so
as to provide for affixes after terms denoting times under the conditions which
12 This possibility is envisioned in 4.2.34 vt.
E XTENSION RULES
2.3
117
atidesa
Rup
At a stage in deriving the bahuvrhi compound darsaniyabharya

(one whose wife is beautiful) from a string darsanya-su bharyasu, there is a complex darsanya-bharya, in which the feminine
term darsanya- occurs before the term bharya, with which it is
coreferential. In order to account for a first constituent such as
darsanya- in darsanya-bharya,
A. 6.3.34 ;a;~:
/ a;ya;aH :pMua;va;;
+a;
a;Sa;ta;pMua;~k+:a;d;nUa;*;+.
/ a;ya;a;ma;pUa ma;a;na;a;a;Da;k+.=+ea ;a;~:
:=+a;a;
a;pra;ya;a;
a;d;Sua
provides that a feminine term (striyah.) has the form of a corresponding masculine (pumvat
like a masculine) if, in the same
meaning, it has a masculine counterpart, the feminine suffix that

follows this term is one other than u n,
and it is followed by a coreferential (samanadhikaran.e) feminine term (striyam) which is neither a member of the set beginning with priya (dear) nor a term
containing an ordinal suffix.15
2.4
Tadatmyatidesa
The derivate devadatta is a name (saja) used with reference

to an individual so named because he is someone whose parents
concern such items in the section where A. 4.3.53 (p. 114) is stated: ;a;sa:;d ;mea;ta;t,a k+:Ta;mua;Ba;ya;a;na;deR ;Za;a;t,a o+.Ba;ya;a;na;deR ;ZaH k+:tRa;v.yaH k+:a;le +.Bya;ea Ba;va;va;tk+:a;le +.Bya I+.a;ta (Bh.
III.276.22277.l). Patajali goes on to reject this reformulation, remarking that
the function of vati is accomplished only if A. 4.2.34 (p. 115) lets those affixes
introduced to form derivates meaning located in/at X apply to form derivates
meaning . . . whose deity is X under the same distribution of bases: .sa ta;h:uaR;Ba;ya;a;na;deR ;ZaH k+:tRa;v.yaH na k+:tRa;v.yaH Ba;va;va;
a;d;a;ta va;a;ta;na;a ;a;na;deR ;Za;ea Y;ya;m,a ya;
a;d . ca ya;a;ByaH
:pra;kx+:a;ta;Bya;ea yea;na ;
a;va;Zea;Sea;a Ba;vea :pra;tya;ya;a ;
a;va;
a;h;ta;a;~ta;a;ByaH :pra;kx+:a;ta;Bya;~tea;na ;
a;va;Zea;Sea;a .sa;a;~ya
:de;va;tea;tya;a;sma;a;Te
/ / / Ra Ba;va;a;nta
/ / ta;ta;ea Y;ma;a Ba;va;va;tkx+:ta;aH .~yuaH (Bh. II.277.14).
15 I have simplified for presentation by not mentioning all three possible interpretations of str in striyah. and striyam, or how bhas.itapumsk
adanun is to be
interpreted.
118
G. C ARDONA
wished the gods should give him to them; comparably, vis.n.usruta

names an individual to whom one wishes Vis.n.u would harken.16
Such a derivate, which includes the expression of a wish (asis.i),
and in which datta (given) or sruta (heard) follows a term denoting a participant in an action (karakat), has high pitch on its
last vowel:17 devadatta-, vis.n.usruta-. If, however, such a term is
used to address someone, its first vowel is high-pitched: devàdatta,
vis.n.u` sruta.18 According to
A. 6.1.198 ;a;ma;a:/n:a;ta;~ya . ca (;a;
a;dH 197)
the first vowel of a nominal form with an ending given the class
name a mantrita that is, the ending of what is called a vocative
in western terminology is high-pitched. The same rule accounts
for the accentuation of a form such as vrscàn (cutting), vocative
singular of the present participle vrscat (nom. sg. m. vrscan).
Now, in a phrase such as :pa:=+Zua;na;a vxa;(a;n,a (o, you who are cutting
with an axe), the vocative vrscan is not so accented. Moreover, the
instrumental paràsuna hashigh pitch on its first vowel instead of
the stem-final -u (parasuna` ). In effect, the entire phrase is treated

as a single unit, with high pitch on its first vowel. Pan.ini provides
for this in
A. 2.1.2 .sua;ba;a;ma;a:/n:a;tea :pa:=+a;*: +.va;t~va:=e
according to which a form that terminates in a nominal ending
(sup) and occurs before a term with an a mantrita ending (see note
18) behaves like a part of the following item (paran gavat). Thus,
parasuna in :pa:=+Zua;na;a vxa;(a;n,a is treated as though it were part of the following participial form, the entire complex being treated as a unit.
16 :de;va;a
O;;nMa :de;ya;a;suaH :de;va;d:aH ;

a;va;Sua:=e+nMa (rUa;ya;a;t,a ;
a;va;Sua;(rua;taH (e.g., Kas. 6.2.148).
6.2.148: k+:a:=+k+:a;;a;(rua;ta;ya;ea;=e +va;a; a;Za;
a;Sa (;ntaH 143, .sa;Va;a;ya;a;m,a 129).
18 In the derivation of such vocative forms, the ending su (named sambuddhi)
is deleted. A first-triplet nominal ending used in addressing is called a mantrita.
For details, which need not be considered here, see PWT.: 39 68.
17 A.
E XTENSION RULES
119
Similarly, in the Rgveda pada (1.3.1b) d+va;tpa;a;a;!a Zua;Ba;~.pa;ta;a ([Asvins,] you of quickhands, masters of the good), both the compound
dravàtpan. and the phrase subhàs pat19 have a single high-pitched
vowel, the first one.
2.5
Nimittatidesa
The base d.ukr is marked with to show that an l-affix follow by a tmanepada suffixes if the result of the act
ing it is replaced
denoted is intended for the agent and by parasmaipada suffixes if
not:20 kurute (does, makes . . . for himself), karoti (. . . for someone else). The derived base ending cikrs.a, with the desiderative
suffix san, also takes a tmanepada and parasmaipada suffixes under
these conditions: cikrs.ate, cikrs.ati. Accordingly,
A. 1.3.62 :pUa;vRa;va;tsa;naH
provides that a tmanepada affixes occur after a derivate with san
as they would after the base which precedes this suffix (purvavat).
Under the view that A. 1.3.62 allows for the extension of conditions, this sutra is considered to extend to the derived base with
-san the marking that accompanies the primitive base which this
affix follows; this marking is treated as the cause (nimitta) for the
introduction of a tmanepada or parasmaipada suffixes. Although
A. 1.3.62 is a standard example given of nimittatidesa,21 it is also
admitted that the same result is achieved with an operation extension (karyatidesa 2.1). That is, A. 1.3.62 can be considered
to extend to a derived desiderative base the occurrence of a tmanepada suffixes under the same conditions as apply with respect
19 Padapatha
Zua;BaH :p!a;ta;!a I+.a;ta

k+.Ra;a;Ba;pra;a;yea ;
a;k+:ya;a;P+.le (A. 1.3.72); see PWT.: 8889 139.
21 E.g., SKT. 274 (= 7.1.95, I.277): ;a;na;a;ma:a;a;a;ta;de ;Za;ea ya;Ta;a ;a;.ca;k
+:a;SRa;a;ta ;a;.ca;k
+:a;SRa;tea
:a .sa;a;nta;~ya :pUa;vRa;va;tsa;na I+.tya;nea;na ;a;Va:va;a;a;ta;de ;Zea k+:tRxa;ga;a;a;ma;a;na ;
a;k+:ya;a;P+.l ;a;tma;nea;pa;dM
;a;sa;Dya;a;ta
.
20 .~va;a:=+ta;a;Va;taH
120
G. C ARDONA
to the primitive base in question. The reason for considering this

extension to be of a separate type, namely, an extension of conditions, is that the extension of the operation requires considering
first the condition involved.22 Moreover, under the view that A.
1.3.62 (p. 119) involves a nimittatidesa, sanah. is most appropriately considered a genitive form:23 a verbal base in -san is similar
to the base which precedes the affix by having the markers that
serve to show that the base takes a tmanepada affixes. However,
there are unambiguous forms, such as gatihims
arthebhyah. of na ga;a;ta;
a;hM ;sa;a;TeRa;ByaH (A. 1.3.15) showing that sanah. should be considered
an ablative form, so that A. 1.3.62 (p. 119) serves to extend a tmanepada affixes, letting these follow a derived base in -san as they
do a primitive base preceding the suffix.24
2.6
Vyapadesa tidesa
A vyapadesa extension extends a name to an item that otherwise

is not so named. A standard example of this type of extension involves extending the terms for the beginning and end of a string to
a single sound as occurs in the derivation of a bhyam (instr.-dat.abl. du. of idam this) and aupagava (descendant of Upagu).
The first is derived from idam-bhyam, with replacement of -m by
-a and deletion of the segment ida, after which long-vowel replacement should apply to a- of a-bhyam, as it does in vrks.a bhyam (
vrks.a-bhyam).25 The patronymic aupagava is derived

by intro
ducing the taddhita affix an. after the pada upagu-nas,
with the
sixth-triplet ending nas.

The affix of aupagava should be the only
high-pitched segment in the derivate. Now, the replacement of -a

22 na;nua
k+:a;ya;Ra;a;ta;de ;Zea;na;a;pyea;ta;+ta;a;TRa;a;ma;a;ta . cea;d:a;a;huH ;a;na;a;ma:a;pra;a;ta;sa;nDa;a;na;pUa;vRa;k+:tva;a:a;~yea;a;ta

;a;na;a;ma:a;a;a;ta;de ;ZaH :pxa;Ta;gua; I+.a;ta (SKT. 274 [I.277278]).
23 Janendrasarasvat uses (see note 21) the genitive sannantasya.
24 Cf. K
as. 1.3.62: .sa;naH :pUa;va;eRa ya;ea ;Da;a;tua:=+a;tma;nea;pa;d ;a ta;d;tsa;a;nta;a;d;a;tma;nea;pa;dM Ba;va;a;ta
25 See 2.1 with note 3.
E XTENSION RULES
121
by -a applies to the final vowel of a stem before its ending, but ain a-bhyam has a stem that consists of a single vowel. Similarly,
there is a rule whereby the first vowel of an affix is high-pitched;26
for example the first vowel of tavya in kartavya ( kr-tavya) has
high pitch. The affix an., although it consists of a single

vowel,
should similarly be high pitched. Accordingly,
A. 1.1.21 ;a;d;a;nta;va;de ;k+:a;sma;n,
/// a
provides that an operation takes effect with respect to a single
sound as it would with respect to one that is first or last (adyantavat).
It may be proposed that A. 1.1.21 serves to extend to single
sounds the property of being first or last, thereby letting them be
called first and last even though neither is part of a group of
sounds.27 However, this amounts to providing that a single sound
will be named a di and anta, hence referred by these terms in sutras, as in A. 3.1.3 (see note 26). A. 1.1.21 would then be a naming
rule (sajasutra) comparable to
A. 1.1.23 ba;hu;ga;a;va;tua;q+.a;ta .sa;* +;a;a
whereby sankhya (number [word]) is allowed to refer not only

to usual number words like s.as. six but also to bahu many, gan.a group and items that end with the suffix d.ati. If Pan.ini had
intended A. 1.1.21 to function in this way, however, one would
expect him to formulate it in a manner comparable to A. 1.1.23,
which he does not. In fact, Pan.inyas reject vyapadesa tidesa as
a distinct category28 and consider A. 1.1.21 an instance of a rule
26 ;a;d;a;d;a:a;(
a
(A. 3.1.3); see PWT.: 378 592.

for example, with respect to a sound being treated as the first of a
group, SKT. 274 (I.278): v.ya;pa;de ;Za;a;a;ta;de ;Za;ea ya;Ta;a ;a;d;a;nta;va;de ;k+:a;sma;
/ / / a;a;tya;nea;na k+:tRa;v.ya;a;ma;tya;a;d;Ea .sa;a;va;k+:a;Za;pra;tya;ya;a;d;au;d;a:a;tva;P+.l+.k ;a;
a;d;tva;v.ya;pa;de Za ;Ea;pa;ga;va;a;d;a;va;a;ta;
a;d;Zya;tea
28 For example, VPPu. 2.77 (38.78): v.ya;pa;de ;Za;a;a;ta;de ;Za;~tua v.ya;a;k+.=+ea .nEa;va .sa;}Ba;va;a;ta
.sa;Va;a;pa:a;a;d;
a;va;Zea;Sa;a;d;tk+.=+a;vEa;P+.ya;pra;sa;*: +.a;a
u
27 Thus,
122
G. C ARDONA
extending operations (karyatidesa): an operation applies (bhavati

is) for a single (sound) also in the same manner (iva) as for a first
or last (sound of a larger unit).29
2.7
Arthatidesa
A standard example of an Arthatidesa is

A. 1.2.66 .~:a;a :pMua;va;a (vxa:;d ;ea yUa;na;a ta;+.a;a;(ea;de ;va ;
a;va;Zea;SaH 65, O;;k+:Zea;SaH
64)30
which provides for the single remainder of one of a group of related terms. In this case, a term denoting an elder female of a
lineage in conjunction with a younger male descendant constitutes
a single remainder. Moreover, the female (str) is then treated as
male (pumvat).
For example, gargyau (nom. pl.) refers to the
elder woman (garg) and her younger male counterpart (gargyayan.a). The term str is interpreted as denoting a meaning female
which is said to behave as male. It is possible, nevertheless, to
maintain that the operations that apply for a male are here extended
to a female.31
29 For
example, Kas. 1.1.21: ;a;d;a;

a;va;va;a;nta I+.vEa;k+:a;sma;a;
/ / / a;pa k+:a;ya Ba;va;a;ta

example, SKT. 274 (I.278): ;Ta;Ra;a;ta;de ;Za;ea;d;a;h:=+Ma tua ga;a;ga;Ra . ca ga;a;gya;Ra;ya;a;Ea . ca ga;ga;RaH :a .~:a;a :pMua;va;ea;tya;nea;na ga;ea:a;pra;tya;ya;a;nta;~:a;a;va;a;.ca;k+:~ya vxa:;d ;ea yUa;nea;tyea;k+:Zea;Sea .sa;a;ta .~:ya;TRa;~ya
:pua;ma;Ta;eRa Y;a;ta;
a;d;Zya;tea .tea;na ;a;~:
/ a;ya;a;ma;pa;tya;kx+:ta;ba;hu;tvea ya;Va;Va;ea;(ea;a;ta lu +.a;#sa;Dya;
/ / / / a;ta Janendra
illustrates with the plural gargah., whose derivation involves omitting the affix
ya by A. 2.4.64; Jinendrabuddhi and Haradatta give the same example in their
comments on Kas. 1.2.66 (N., PM. I.382); see note 95.
31 Cf. K
as. 1.2.66: :pMua;sa I+.va;a;~ya;aH k+:a;ya Ba;va;a;ta .~:ya;TRaH :pua;ma;TRa;va;;
+va;a;ta SK. 932 (=
A. 1.2.66, II.208): yUa;na;a .sa;h;ea;+:Ea vxa:;d ;a .~:a;a ; a;Za;Sya;tea ta;d;TRa;(a :pMua;va;t,a
30 For
E XTENSION RULES
123
Extensions of the type X is Y
The extension rules discussed in 2 contain terms with the suffix

vati, but there are also extension rules that do not have such terms.
For example,
A. 1.2.1 ga;a;*:u +.f;a;
a;d;Bya;ea Y;a;Va;
/ / / a;nz+.
/ t,a
A. 1.2.4 .sa;a;vRa;Da;a;tua;k+:ma;
a;pa;t,a (;
a;z+.t,a 1)
are two of a series of rules32 whereby particular affixes are said to
bear the marker (it) n (nit). By the first, any affix not marked with
or n. is marked with n if it follows the verbal base gan (suppletive
with in study, learn) and bases of the set beginning with kut.a
(curve, bend). By the second, a sarvadhatuka suffix not marked
with p is marked with n . For example, the affix sic in aga-sic-ta (
ags..ta in adhy ags..ta [has] learned) is marked with n , so that it
conditions the replacement of stem-final -a by -.33 The endings tip
and tas differ in that the first bears the marker p from its inception
and the second does not. By A. 1.2.4, the latter is marked with n .
It is therefore barred from conditioning gun.a replacement for the
final vowel of a stem such as cinu-: cinutas (they two are heaping,
plucking) versus cinoti (3sg. pres. act.).34
Conversely,
A. 1.2.19 ;a;na;+a Za;a;a;*++.

a;d;a;ma;
a;d;a://va;
a;d;Dxa;SaH (na .sea;q, 18, ;
a;k+:t,a 5)
provides that an affix of the class named nis..tha, that is, kta or ktavatu, is not (na) marked with k (kit) if it is augmented with initial
it. and follows one of a particular set of verbal bases, including sn
(lie, sleep). The nis.t.ha suffixes are introduced bearing the marker
k showing that they do not condition gun.a and vrddhi replacements
32 On
these and comparable sutras, see PWT.: 332338 511521.

6.4.66: ;Gua;ma;a;~Ta;a;ga;a;pa;a:ja;h;a;a;ta;sa;Ma h;a;l (IR +.t,a 65); see PWT.: 300 451.
34 ;
a;*;+.
a;ta . ca (A. 1.1.5); see PWT.: 57 98.
33 A.
124
G. C ARDONA
which affixes of the general class to which they belong regularly

condition (see note 34); e.g., krta- (done, made) : kartum (to do,
(p. 123) negates this marking for
make [ kr-tumun]). A. 1.2.19
kta augmented with it. and following sn:

sayita.
A. 7.1.91 .a;lu +a;ma;ea va;a (; a;a;t,a 90)
on the other hand, provides for an optional marking: the ending
n.al of the uttama triplet what is called first person singular in
western terminology is optionally (va) marked with n.. An affix marked with n. conditions vrddhi replacement for a stem-final
singular perfect cakara (. . . did,
vowel;35 e.g., kar-a in the third
made [ kr-n.al]). The corresponding first-person form, on the
other hand,
is not only cakara but also cakara, without vrddhi re
placement.
Rules like A. 1.2.1 and A. 1.2.4 (p. 123) simply state X is Y.
A sarvadhatuka not marked with p is marked with n , and a negative
rule like A. 1.2.19 (p. 123) denies k marking. The latter does not
allow an original k marker to remain, any more than A. 1.2.4 allows
for an ending to remain unmarked with k. Accordingly, A. 7.1.91
has to state va; a negative rule using na would absolutely deny n.
marking for the ending.
A rule using a term with the affix vati, on the other hand,
merely provides that an entity Y is like an X; it does not remove
the original properties of Y. Thus, A. 1.1.56 (p. 111) lets a replacement be like the entity in whose place it stands, but does not make
it that entity. For example, ya substituting for ne
is allowed to behave like the replaced affix, so that it can condition an operation
on a preceding unit which is now treated as a stem with respect to
ya, but it does not make ya to be ne.
Similarly,
A. 3.1.87 k+:mRa;va;tk+:mRa;a;a tua;ya;
a;k+:yaH (k+:tRa;a:= 68)
35 A.
7.2.115: ;.ca;ea ;a;Va;

/ / / a;ta (vxa;a:;d H 114, ;*: +.~ya 6.4.1); see PWT.: 284 417.
E XTENSION RULES
125
provides that an agent (karta) which has the same status with respect to an action (tulyakriyah. whose action is the same) which
it would have as its object (karman.a ) behaves as though it were an
object (karmavat).36 The rule does not provide that such an agent
is classed as object. Hence, it allows for the operations which apply when an object is to be signified, but it does not deny the status
of an agent as agent, so that any operation which depends on this
(svasrayam depending on what it is properly) still applies. For
example, one says of a grain storage pot (kusula) that breaks apart
at a time that no one is striking it, there is no wind or great sunlight
to cause it to break:
(3) ;a;Ba;d;a;tea ku+:sUa;lH The kusula is breaking
just as one can say
(4) ;a;Ba;d;a;tea ku+:sUa;l+.ea :de;va;d:ea;na The kusula is being broken by Devadatta.
Both (3) and (4) have bhidyate, in which the base bhid is followed
by two affixes: the a tmanepada ending te ( ta) and the stemsuffix yak. In deriving the passive (4), the l-affix lat. is introduced
after bhid on condition that an object (karman) of the act of breaking causing something to break apart37 is to be signified, and
is then replaced by the sarvadhatuka ending ta ( te); yak follows
the verbal base which precedes such an affix.38 The same opera36 ya;a;sma;nk+:
///
mRa; a;a
k+:tRxa;BUa;tea Y;
a;pa ta;d;a;tk
/ +:ya;a l+ya;tea ya;Ta;a k+:mRa; a;a .sa k+:mRa;a;a tua;ya;
a;k+:yaH
k+:mRa;va;;
+va;ta;a;a;ta (Bh. 3.1.87 [II.66.78). By karman.a is meant an act which occurs
in an object of an action: k+:mRa; a;a ;
a;k+:ya;a k+:mRa k+:mRa;~Ta;ya;a ;
a;k+:ya;ya;a tua;ya;
a;k+:yaH k+:ta;Ra
k+:mRa;va;;
+va;a;ta (Kas. 3.1.87). The act in question for (3) and (5) is breaking apart
(dvidhabhava), which resides in the kusula that is breaking.
37 I use this wording merely for clarity. The uses of bhid in (3), (5) and (4), (6)
respectively correspond to the use of intransitive and transitive break. Sanskrit
differs from English and also other languages, such as Latin in allowing
the type (5), with an instrumental denoting the agent of the objectless verb.
38 A. 3.4.69: lH k+:mRa; a;a . ca Ba;a;vea . ca;a;k+:mRa;ke+:ByaH (k+:tRa;a:= 67), A. 3.1.67: .sa;a;vRa;Da;a;tua;ke
ya;k, (Ba;a;va;k+:mRa;a;eaH 66). See PWT.: 100 156, 148 232.
126
G. C ARDONA
tions apply in deriving (3), where, however, the ending of bhidyate

refers to an agent, not an object. In this instance, the base bhid
denotes an objectless (akarmaka) action, so that an a tmanepada
ending can follow it also to denote the act (bhava) of breaking.
The agent of this act is then signified by an instrumental form, as
in
(5) ;a;Ba;d;a;tea ku+:sUa;le +.na
This utterance contrasts with (4), which is the passive counterpart
to
(6) :de;va;d:aH ku+:sUa;lM ;a;Ba;na; a:a Devadatta is breaking the kusula.
By A. 3.1.87 (p. 124), an object such as the kusula of (3) and
(5) is said to be like an object (karmavat); that is, the operations
which apply where an object of an action is referred to are allowed
to apply. Since the kusula is not, however, said actually to be an
object of the action in question, it remains its agent. Thereby, the
base bhid remains objectless (akarmaka), so that A. 3.4.69 (see
note 38) can serve to introduce an l-affix on condition that an act
is signified.
Suppose now that, since A. 1.2.4 (p. 123) serves to extend n marking to a sarvadhatuka not marked with p, one were to interpret
n it here as equivalent to n idvat. The sutra would now provide that
an affix not marked with p behaves like one that is marked with
n . It would not, however, do away with the affixs property of not
bearing a marker n . An ending such as tas, then, would now be
said to be marked with n while not ceasing to lack this marking.
Consequently, one would allow not only acceptable forms such as
cinutas but also unacceptable ones like *cinotas. Similarly, if kit
of a A. 1.2.19 (p. 123), were interpreted as equivalent to kidvat,
the affix kta would be said to act as though it lacked the marker
k, while still retaining its original property of being so marked.
E XTENSION RULES
127
Obviously, such interpretations lead to contradiction. One could,

then, argue that it is not allowed to extend to an X a property of
a Y which is contradictory to what X is. This amounts, however,
to admitting simply that terms like n it and kit in the sutras at issue
may not be interpreted as equivalent to n idvat and kidvat.
In sum, there is an essential difference between extensions provided by sutras which state that an entity Y is like another entity
X in that it is treated as having properties of X, which it does not
have to begin with, and sutras which state that an X has certain
properties extended to it, making it different.
4
4.1
Syntactic structures
Predication without comparison
An utterance such as A. 1.2.4 (p. 123) is formally comparable to

utterances of the type
(7) ga;ea;ma;a;nde ;va;d:aH Devadatta has many cows
(8) :pa;vRa;ta;ea va;
a;+:ma;a;n,a The mountain has fire on it
in that A. 1.2.4 has a bahuvrhi compound n it (. . . that has a marker
n ) and (7)(8) use derivates with matup, which are equivalent to
such compounds. All three utterances also involve predication: it
is predicated of a sarvadhatuka affix not marked with p that it is
marked with n , of Devadatta that he possesses many cows, and of
a mountain that it has fire on it.
4.2
Comparisons
The comparisons at play in extension rules of the As..ta dhyay are of

two kinds: (a) one entity is comparable to another by virtue of the
two having in common an action in progress (kriya); (b) an entity
is comparable to another by virtue of being the locus or possessor
of a property shared by the other.
128
4.2.1
G. C ARDONA
Comparison involving a common action
Utterances such as A. 1.1.56 (p. 111), with terms that have the suffix -vati, involve a comparison, and such utterances can be paraphrased by sentences using tulyam (equal to, similar[ly]) or iva.
For example,
(9) a. .~Ta;a;a;na;va;d;a;de ;Za;ea va;tRa;tea A replacement is like its original (cf.
A. 1.1.56 (p. 111))
is equivalent to
(10) b. ;a;de ;ZaH .~Ta;a;a;na;na;a tua;yMa va;tRa;tea A replacement is (i.e., behaves) like the entity in whose place it stands39
and what is expressed in
(11) va;sUa;a;na ta;ea;yMa ;Ga;na;va;d;a;k+:a:=+a;t,a (Dasaratha) spread wealth as a
cloud (spreads) water (BhK. 1.3a)
can be expressed equivalently by
(12) va;sUa;a;na ta;ea;yMa ;Ga;nea;na tua;yMa v.ya;k+:a:=+a;t,a
or
(13) ya;Ta;a ;Ga;na;~ta;ea;yMa ;
a;k+.=+a;ta ta;Ta;a va;sUa;a;na v.ya;k+:a:=+a;t,a40
These utterances involve a comparison of actions, for example
the spreading of material wealth by Dasaratha is compared to the
shedding of water by a cloud. These are comparable in that there
is a shared act of spreading something beneficial with no expectation of a reward. Pan.ini accounts for derivates with vati in such
instances by
39 Cf.
40 Cf.
Kas. 1.1.56: .~Ta;a;a;na;na;a tua;yMa va;tRa;ta I+.a;ta .~Ta;a;a;na;va;t,a
BhKJ. 1.3: ta;ea;ya;mua;d;k+:m,a ;Ga;na;ea mea;GaH :P+.l+.a;na:=+pea:a;ya;a ya;Ta;a ;
a;va;
a;k+.=+a;ta ta;d;t,a
E XTENSION RULES
129
A. 5.1.115 .tea;na tua;yMa ;

a;k+:ya;a . cea;d;a;taH,
which can be summarized as follows:
schema: P1: Bn-3 + P2: tulya-241

affix: vati; meaning of affix: like X
condition: what is like X is an action (kriya)42
operation: the taddhita affix43 vati is optionally (va) introduced after a value of tena, the first of semanticallysyntactically related padas44 referred to by A. 5.1.115.
Accordingly, both
(14) a. .tea;na tua;yMa va;tRa;tea
b. ta;d;d;tRa;tea
41 A
second-triplet ending can be accounted for by: k+:mRa; a;a ;

a;d;ta;a;ya;a d;ae;k+:ya;ea;
a;dR;va;.ca;nEa;k+:va;.ca;nea (A. 2.3.46, 1.4.22) under the assumption that tulyam of A. 5.1.115
signifies a karman (A. 1.4.49: k+:tRua:=+a;a;psa;ta;ta;mM
/ /
a k+:mRa). Alternatively, tulya- is a nipata according to the gan.asutra o+.pa;sa;gRa;
a;va;Ba;a;+:~va:=+pra;a;ta:+.pa;k+:a;(a ;a;na;pa;a;ta;aH (under A.
1.4.57: . ca;a;d;ya;ea Y;s,a:vea, cited by Patajali, Bhas.ya on 2.2.24 vt. 22 [Bh. I.425.12]:
o+.pa;sa;gRa;
a;va;Ba;a;+:~va:=+pra;a;ta:+.pa;k+:a;(a ;a;na;pa;a;ta;sa;Va;a Ba;va;a;nta
/ / ) . tulya- then takes the ending
su (A. 2.3.46: :pra;a;a;ta;pa;
a;d;k+:a;TRa;a;l+.*: +.pa;a:=+ma;a;a;va;.ca;na;ma;a:ea :pra;Ta;ma;a 1.4.22: d;ae;k+:ya;ea;
a;dR;va;.ca;nEa;k+:va;.ca;nea). In the present discussion, there is no need to take up additional details.
42 That is, an action that is comparable to an act residing in the significand
of the base in Bn-3. Commentators bring this out. For example, the Nyasa on
A. 5.1.115 (IV.99) says, concerning b.ra;a;;ea;na tua;yMa va;tRa;tea b.ra;a;;a;va;t,a (see note 64),
that, although a word like brahman.a does not denote an action, this is understood
through its relation with what the term does denote: ya;d;a;
a;pa b.ra;a;;a;a;
a;d;Za;b.dEH ;
a;k+:ya;a na;ea;.cya;tea ta;Ta;a;pya;a;Ba;Dea;ya;sa;}ba;nDea;na :pra;ta;a;ya;tea Further details need not be discussed
here. Subramania Iyer (1974) summarizes both what the Mahabhas.ya has to say
on A. 5.1.115 and Bhartrharis lengthy exposition of the sutra and bhas.ya, in his
Vakyapadya.
43 ta;a:;d ;ta;aH (A. 4.1.76). See note 7.
44 .sa;ma;Ta;Ra;na;Ma :pra;Ta;ma;a;d;a (A. 4.1.82). See PWT.: 69 111.
130
G. C ARDONA
meaning . . . is (behaves) like X45 are derived as alternative expressions from the same basic string, as shown in Table 2.
Table 2
Derivation of (14a) .tea;na tua;yMa va;tRa;tea and (14b) ta;d;d;tRa;tea
taa-a48
ta-a51
ta-ina54
tena61
45 A
tad-t.a 46 tulya-am
tulya-am vrt-ta49
vrt-te52
vrt-sap-te55
tulyam57 vart-a-te58
tulyam
60 /
tulyav62 vartate
vrt-lat.47
tad-a-vat50
tad-vat53
tadvat-su56
tadvat59
vrt-ta
vrt-te
vrt-sap-te
vart-a-te
tadvad63
vartate
possible alternative to (14a) or (14b) would have bhavati is instead of

vartate. Comparably, one then speaks of an entity Y that has the property of
being like X: tadvadbhava.
46 tua;ya;a;TERa:=+tua;l+.ea;pa;ma;a;Bya;Ma
txa;ta;a;ya;a;nya;ta:=+~ya;a;m,a .sua;paH d;ae;k+:ya;ea;
a;dR;va;.ca;nEa;k+:va;.ca;nea (A.
2.3.72, 1.4.103, 1.4.22)
47 lH k+:mRa; a;a . ca Ba;a;vea . ca;a;k+:mRa;ke+:ByaH .~va;ta:n:aH k+:ta;Ra va;tRa;ma;a;nea l+.f, (A. 3.4.69, 1.4.54,
3.2.123).
48 tya;d;a;d ;a;na;a;maH (A. 7.2.102).
49 l+.~ya ;a;ta;a;a;~
/ / /
.0 ;nua;d;a:a;
a;z+.ta ;a;tma;nea;pa;d;m,a ta;z+.a;na;a;va;a;tma;nea;pa;d;m,a (A. 3.4.77
78, 1.3.12, 1.4.100).
50 A. 5.1.115 (p. 129).
51 ;ta;ea gua;ea (A. 6.1.97).
52 ;a;f;ta ;a;tma;nea;pa;d;a;na;Ma .fe:=e ;.ca;ea Y;ntya;a;
a;d ;a;f (A. 3.4.79, 1.1.64).
53 kx+.a;a:;d ;ta;sa;ma;a;sa;a;(
a .sua;pa;ea ;Da;a;tua;pra;a;a;ta;pa;
a;d;k+:ya;eaH (A. 1.2.46, 2.4.71).
54 f;a;z+.a;sa;z+.sa;a;a;ma;na;a;t~ya;aH (A. 7.1.12).
55 k+:tRa;a:= Za;p,a (A. 3.1.68).
+.pa;a:=+ma;a;a;va;.ca;na;ma;a:ea :pra;Ta;ma;a .sua;paH d;ae;k+:ya;ea;
a;dR;va;.ca;nEa;k+:va;.ca;nea (A.
56 :pra;a;a;ta;pa;
a;d;k+:a;TRa;a;l+.*:
2.3.46, 1.4.103, 1.4.22).
57 ;a;ma :pUa;vRaH (A. 6.1.107).
58 :pua;ga;nta;l+.GUa;pa;Da;~ya . ca o+=+'a;pa:= H (A. 7.3.86, 1.1.51).
E XTENSION RULES
131
The syntactically pertinent feature of constructions of type

(14) is, as Pan.ini observes, that a Y is likened to X by virtue of
an activity that both share: Y is like X, the two sharing the act of
being or behaving in certain ways, so that their actions are said to
be alike. For example,
(15) a. b.ra;a;;ea;na tua;yMa va;tRa;tea
b. b.ra;a;;a;va;d;tRa;tea
convey that someone, such as a ks.atriya, is acting like a brahman.a,
that is, that he is performing acts done by brahman.as, such as undergoing traditional study.64 Similarly, (11) and (12) convey that
Dasaratha shares with clouds the act of spreading about something
good without requiring rewards: material goods in the case of Dasaratha, water in the case of a cloud, and
(16) d;Ea;
a;h:a;ea Y;
a;pa h:a;mua:a .sa;nta;a:=+ya;a;ta :pa;Ea:a;va;t,a (Manu. 9.139cd) For the
child of a daughter also transports one (beyond hell) there (in
the afterworld), as a child of a son
compares a daughters sons transporting one (santarayati makes
to cross) to a sons son doing this.
4.2.2
Other comparisons
e :pra;a;k+:a:=H
(17) a. ma;Tua:=+a;ya;a;a;ma;va .~:ua.*+
59 ;v.ya;ya;a;d;a;psua;paH
.~va:=+a;
a;d;a;na;pa;a;ta;ma;v.ya;ya;m,a (A. 2.4.82, 1.1.37), va;t,a (gan.asutra under A. 1.1.37).
60 ma;ea Y;nua;~va;a:= H (A. 8.3.23).
+.aH (A. 6.1.87, 1.1.2).
61 ;a;+aH ;de ;*:
u
u
62 ;nua;~va;a:=+~ya ya;a;ya :pa:=+sa;va;RaH va;a :pa;d;a;nta;~ya (A. 8.4.5859).
63 +.l+.Ma .ja;Za;ea Y;ntea (A. 8.2.39).
64 Cf. K
as. 5.1.115: b.ra;a;;ea;na tua;yMa va;tRa;tea b.ra;a;;a;va;t,a The Nyasa explains that
these speak of a ks.atriya acting like a brahman.a in performing acts such as studying, which the latter performs: k+:a :pua;na:=;a ;
a;k+:ya;ab.ra;a;;ea;na tua;yMa va;tRa;tea a;
a:a;ya;a;
a;d;ga;ta;a;Dya;ya;na;a;
a;d;k+:a (N. IV.98).
132
G. C ARDONA
e :pra;a;k+:a:=H
b. ma;Tua:=+a;va;t~:ua.*+
(18) a. :de;va;d:a;~yea;va ya;a;d:a;~ya ga;a;vaH

b. :de;va;d:a;va;d;a;a;d:a;~ya ga;a;vaH
also involve comparison: the first pair speaks of a city wall in Srughna which is said to be like the one in Mathura,65 and the second
pair says that Yajadattas cattle are like Devadattas.66 These do
not, however, depend on a common action as the basis of comparison. Instead, they compare a city wall located in one city with a
wall located in another and cattle belonging to one man with cattle
that belong to another person.
To account for such usage, Pan.ini formulates
A. 5.1.116 ta:a ta;~yea;va
This rule also introduces vati, but now after a pada with an ending
of the seventh or sixth (tatra X-7, tasya X-6) triplet of nominal
endings. The taddhita affix vati is optionally (va) introduced after a value of tatra or tasya, the first of semantically-syntactically
related padas referred to in A. 5.1.116, to form a derived nominal
base meaning . . . like . . . in/at X or . . . like . . . related to X.
(16) has amutra (there), equivalent to the locative amus.min
(loke), as (18a) contains the locatives srughne and mathurayam.
This does not mean, however, that the two utterances are to be interpreted syntactically in the same manner. (17a) concerns a comparison of two entities, the city walls in Mathura and Srughna, but
(16) concerns a comparison of two actions, the transporting away
from hell by the son of a son and the son of a daughter, which take
place in the world beyond.
In a comparable vein, consider now
65 K
as.
66 K
as.
e :pra;a;k+:a:=H
5.1.116: ma;Tua:=+a;ya;a;a;ma;va ma;Tua:=+a;va;t~:ua .*+
5.1.116: :de;va;d:a;~yea;va :de;va;d:a;va;d;a;a;d:a;~ya ga;a;vaH
E XTENSION RULES
133
(19) :tEa;l+.pa;a;ke+:na tua;ya;ea ;Gax ;ta;pa;a;kH Cooking with sesame oil is like
cooking with ghee.
Although two sorts of cooking are compared, this is not an example of a comparison covered by A. 5.1.115 (p. 129), so that it is not
a possible domain for introducing vati by this rule after the pada
tailapaka-a.67 For A. 5.1.115 (p. 129) concerns an action (kriya) as
signified by a verbal base (dhatu), that is, an act considered as one
that is to be brought to accomplishment (sadhya) by participants
(karaka, sadhana). The suffix gha in a derivate like paka (
pac-gha)68 serves to signify an act as a single global entity, with
the status of a thing, as opposed to the act in progress signified by
the preceding base pac.69
4.3
Summary of syntactic structures
As a general diagnostic, one can distinguish between the following

two types of constructions:
I. B-vat is accounted for by A. 5.1.115 (p. 129). The derivate
in vat is used with a nominative form; the base B designates
an entity in which resides an act viewed in progress (kriya),
and another individuals act is compared to this.
67 Cf.
N. 5.1.115 (IV.99): na;nua . ca txa;ta;a;ya;a;sa;ma;TRa;

a;va;Ba;a;+:va;a;.cya;ya;a ;
a;k+:ya;yEa;va :pra;tya;ya;a;TRa;
a;k+:ya;a;ya;a;~tua;ya;ta;a ta;~ya;a;(a ;Ga;Va;nta;ea Y;va;k+:a;ZaH :tEa;l+.pa;a;ke+:na tua;ya;ea ;Gxa;ta;pa;a;k I+.tya;a;
a;d .nEa;Sa;ea
Y;~tya;va;k+:a;ZaH ;Da;a;tva;Ta;eRa ;
a;h ;
a;k+:ya;a .sa;a . ca :pUa;va;Ra;pa:=+a;BUa;ta;a
68 A. 3.3.18: Ba;a;vea (;Ga;V,a 16).
69 Cf. :pUa;va;Ra;pa:=+a;BUa;tMa Ba;a;va;ma;a;K.ya;a;tea;na;a;.ca;e v.ra:ja;a;ta :pa;.ca;ta;a;a;ta o+.pa;k+:ma;pra;Bxa;tya;pa;va;gRa;pa;yRa;ntMa mUa;ta .sa:va;BUa;tMa .sa:va;na;a;ma;a;BaH v.ra:$ya;a :pa;a;+:a:=+a;ta (Nir. 1.1 [28.12]), ;a;~ta
/ / Ka;va;
a;pa ;
a;va;Zea;SaH
kx+:d;a;Ba;
a;h;ta;~ya Ba;a;va;~ya ;a;ta;z+.a;Ba;
a;h;ta;~ya . ca kx+:d;a;Ba;
a;h;ta;ea Ba;a;va;ea d+v.ya;va;;
+va;a;ta (Bh. 3.1.67
[II.57.78]), ;nyaH .sa Ba;a;va;ea ba;a;h:aH :pra;kx+:tya;Ta;Ra;t,a ;nea;nea;d;a;na;a;ma;a;Bya;nta:=e Ba;a;vea .~ya;a;t,a kH
:pua;na:=e+ta;ya;ea;Ba;Ra;va;ya;ea;
a;vRa;Zea;SaH o+.+:ea Ba;a;va;Bea;d;ea Ba;a;Syea (Bh. 3.3.19 [II.145.1618]) kx+:d;a;Ba;
a;h;ta;ea Ba;a;va;ea d+v.ya;va;;
+va;ta;a;a;ta (Bh. 2.2.19 [I.417.1718]). Commentators on A. 5.1.115
(p. 129) and the Kasika and so forth thereon also take up the differences in detail;
it is not necessary to consider these here.
134
G. C ARDONA
II. B-vat is accounted for by A. 5.1.116 (p. 132). The derivate

in vat is used with a locative or genitive form; the base B
designates an entity in which or pertaining to which there is
something which occurs elsewhere or pertains to something
else, and the one is compared to the other.
The syntactic distinction is recognized by Pan.inyas, most
clearly and succinctly by Jinendrabuddhi in his commentary on the
Kasika to A. 3.1.87 (p. 124). As noted earlier (3, 4.2.1), karmavat contains vati by A. 5.1.115 (p. 129). Syntactically, this is seen
from the fact that the sutra contains a nominative tulyakriyah., not
a locative tulyakriye or a genitive tulyakriyasya which would accord with vati by A. 5.1.116 (p. 132).70 Similarly, the syntax of A.
1.1.56 (p. 111), with the nominative a desah. not a dese or a desasya accords with interpreting sthanivat as derived with vati
by A. 5.1.115 (p. 129). On the other hand,
A. 3.4.85 l+.ea;f;ea l+.z +.t,a
conforms to schema II. The term lot.ah., although formally ambiguous possibly nominative or accusative plural, ablative or
genitive singular contextually must be interpreted as a genitive
singular. A. 3.4.85 occurs in a section headed by
A. 3.4.77 l+.~ya
in which occurs the unambiguously genitive singular form lasya.
Rules of this section provide in the first instance for basic verb endings that occur in place of any l-affix, then particular replacements
for such verb endings if they substitute for specific l-affixes.71 According to A. 3.4.85, a replacement for an ending substituting for
70 k+:mRa;va;
a;d;a;ta .tea;na tua;yMa ;
a;k+:ya;a . cea;d;a;taH na tua ta:a ta;~yea;vea;a;ta tua;ya;
a;k+:ya I+.a;ta :pra;Ta;ma;a;a;na;deR ;Za;a;t,a ;nya;Ta;a tua;ya;
a;k+:yea tua;ya;
a;k+:ya;~yea;a;ta .sa;a;}ya;a :Sa;.a;a va;a ;a;na;deR ;ZMa ku+:ya;Ra;t,a (N. 3.1.87
[II.465466]).
71 See PWT.: 148149 232, 273279 395404.
E XTENSION RULES
135
lot. is like a replacement for an ending which substitutes for lan.

72
That is, the sutra provides for an extension of operations (karyatidesa), and lanvat is accounted for by A. 5.1.116 (p. 132).73
It is worthy of note, though not at all surprising, that Pan.inyas generally agree concerning the syntax of sutras, whether they
have forms in vati according to A. 5.1.115 (p. 129) or A. 5.1.116
(p. 132). One particular sutra stands out in this context:
A. 1.3.63 ;a;}.pra;tya;ya;va;tkx+:Va;ea Y;nua;pra;ya;ea;ga;~ya (;a;tma;nea;pa;d;m,a 12)
according to which kr used after a form with the suffix a m74 behaves like the base
which has the suffix with respect to the distribution of a tmanepada affixes.75 Given that A. 1.3.63 has the
genitive forms krah. and anuprayogasya, commentators appropri a mpratyayavat has vati according to A. 5.1.116
ately consider that
76
(p. 132). Nevertheless, this term was also interpreted as ending
with vati according to A. 5.1.115 (p. 129).77 This interpretation
depends on two factors. First, purvavat of A. 1.3.62 (p. 119) is
understood to recur in A. 1.3.63, so that the latter now makes two
separate provisions: (a) a positive provision (vidhi): d.ukr, which
would not take a tmanepada affixes if the result of an actis not intended for the agent participating in its accomplishment (see note
72 How lot and la
n here refer to endings that substitute for these l-affixes is not
.
crucial to my discussion.
73 ;a;ta;de ;Za;ea Y;ya;m,a l+.ea;f;ea l+.z+.tk+:a;ya Ba;va;a;ta (K
as. 3.4.85), l+.ea;f;ea l+.z I+.va k+:a;ya
.~ya;a;t,a (SK. 2198 = 3.4.85), l+.z I+.va l+.z +.t,a ta:a ta;~yea;vea;a;ta :Sa;.a;nta;a;d;a;taH ta;d;a;h l+.ea;f;ea
l+.z I+.vea;a;ta (SKB. 2198 [III.34]).
74 k+:a;~.pra;tya;ya;a;d;a;ma;ma:n:ea ;a;l+.a;f kx+.*.
a;a;nua;pra;yua:$ya;tea ;a;l+.a;f (A. 3.1.35, 40); see PWT.:
191 286.
75 For example, K
as. 1.3.63: ;a;}.pra;tya;ya;ea ya;sma;a;tsa;ea Y;ya;ma;a;}.pra;ta;yaH ;a;}.pra;tya;ya;~yea;va
;Da;a;ta;eaH kx+:Va;ea Y;nua;pra;ya;ea;ga;~ya;a;tma;nea;pa;dM Ba;va;a;ta
76 For example, ;a;}.pra;tya;ya;va;
a;d;a;ta ta:a ta;~yea;vea;a;ta va;a;taH kx+:Va;ea Y;nua;pra;ya;ea;ga;~yea;a;ta .sa;ma;a;na;a;a;Da;k+.=+ea :Sa;.a;Ea (N. 1.3.63 [I.460]), . . . ta:a ta;~yea;a;ta :Sa;.a;nta;a;d;a;ta;a:=+tya;a;h ;a;}.pra;tya;ya;~yea;vea;a;ta . . . (PM. 1.3.63 [I.460]).
77 SKB. 2240 = A. 1.3.63 (III.59]): ;a;}.pra;tya;ya;va;
a;d;a;ta txa;ta;a;ya;a;nta;a;d;a;taH This reflects
what Bhat.t.oji himself accepts; see note 79.
136
G. C ARDONA
20), does take such affixes if it is used with a term in -am following
an a tmanepadin base; e.g., ha cakre . . . exerted himself. (b) a
restriction (niyama): even if d.ukr is eligible for a tmanepada affixes because the act in questionis intended for its agent, it does
so only if the base with -am that it accompanies also takes such
affixes; accordingly, kr used with ubj be straight, straighten in
ud ubja cakara . . . straightened

. . . does not have an a tmane78
pada ending. Further, once purvavat is understood in A. 1.3.63
(p. 135), not only does one effectively have two statements,79 but
the syntax of A. 1.3.62 (p. 119) is imposed on A. 1.3.63 (p. 135).
As a consequence, moreover, the genitive kronuprayogasya is
80 Thus, it is not the
considered to have the value of an ablative.
case that Pan.inyas merely differ in how they interpret a mpratyayavat of A. 1.3.63 (p. 135). The differing interpretations of this
term are based on how the sutra itself is considered in connection
to the preceding and other related rules.
Semantic and pragmatic considerations
In addition to the syntactic and semantic factors that enter into

how extension rules use derivates in -vati, there are also pragmatic
factors, and these can give the appearance of a different syntactic
interpretation of the utterance with the term in -vati. Consider, for
78 ;Tea;h k+:sma;a;a Ba;va;a;ta o+.du;}Ba;a:*.
a;k+:a:= o+.du;b.ia;a:*.a;k+:a:= na;nua . ca;a;}.pra;tya;ya;va;
a;d;tyua;.cya;tea na
. ca;a:a;a;}.pra;tya;ya;a;d;a;tma;nea;pa;dM :pa;Zya;a;maH na b.rUa;ma;ea Y;nea;nea;a;ta ;
a;kM ta;
a;hR .~va;a:=+ta;a;Va;taH k+.Ra;a;Ba;pra;a;yea
;
a;k+:ya;a;P+.l ;a;tma;nea;pa;dM Ba;va;ta;a;a;ta .nEa;Sa d;ea;SaH I+.dM ;a;na;ya;ma;a;Ta Ba;
a;va;Sya;a;ta ;a;}.pra;tya;ya;va;de ;vea;a;ta ya;
a;d ;a;na;ya;ma;a;Ta ;
a;va;a;Da;nRa :pra;k+:pa;tea IR +.h;a:*.a;ke +:h;a:*.a;k I+.a;ta ;
a;va;a;Da;(a :pra;kw+:aH k+:Ta;m,a
:pUa;vRa;va;
a;d;a;ta va;tRa;tea ;a;}.pra;tya;ya;va;tpUa;vRa;va;ea;a;ta (Bh. 1.3.63 [I.290.16]).
79 ;a;}.pra;kx+:tya;a tua;ya;ma;nua;pra;yua:$ya;ma;a;na;a;tkx+:Va;ea Y;pya;a;tma;nea;pa;dM .~ya;a;t,a I+.h :pUa;vRa;va;
a;d;tya;nua;va;tyRa
va;a;k+.a;Bea:de;na .sa;}ba;Dya;tea (SK. 2240 [III.59]).
80 :pa:*.
a;}ya;TeRa :Sa; +a (SKB. 2240 [III.59]). This clarifies what Bhat.t.oji says (note
79).
E XTENSION RULES
137
example, a laukika example comparable to A. 1.1.56 (p. 111), such

as
(20) gua:+:va;u;+:pua:a;ea Ba;va;a;ta A teachers son is like a teacher.
In addition to (20), one may also say
(21) gua:+:va;u;+:pua:ea va;a;tRa;ta;v.ya;m,a One is to act towards a teachers student as one does toward the teacher.
Both convey that a teachers son who comes in place of his father
merits all the honoring acts due to the teacher; comparably, the operations proper to an original item are extended to its replacement
(adese).81 That (21) may be used properly to describe what is provided for in (20) does not mean that guruvat of the latter conforms
to schema II. Similarly, that one can appropriately say A. 1.1.56
(p. 111) provides for extending to a replacement using a locative form a dese an operation proper to the element this replaces
does not serve to show that sthanivat should be interpreted as containing -vati by A. 5.1.116 (p. 132) according to schema II.
A particularly interesting example of syntactic and pragmatic
criteria is afforded by
A. 7.1.95 txa:$va;tk+:ea;uH (.sa;vRa;na;a;ma;~Ta;a;nea 86, ;sa;}bua:;d ;Ea 92)
which provides that before a nominal ending of the sarvanamasthana class other than one called sambuddhi, the stem kros..tu(nom. sg. kros..tuh.) jackal behaves like the derivate in -trc from
the
the same verbal base krus. That is, the derivate in -tun takes
81 Cf.
1.1.56 vt. 1: .~Ta;a;nya;a;de ;Za;pxa;Ta;+:a;d;a;de ;Zea .~Ta;a;a;na;va;d;a;de ;Za;ea gua:+:va;u;+:pua:a I+.a;ta ya;Ta;a
Bh. 1.1.56 (I.133.2224): gua:+:va;u;+:pua:a I+.a;ta ya;Ta;a ta;d;a;Ta;a gua:+:va;d;a;sma;ngu

/ / / a:+:pua:ea va;a;tRa;ta;v.ya;a;ma;a;ta gua:=+Ea ya;tk+:a;ya ta;u;+:pua:ea Y;a;ta;
a;d;Zya;tea O;;va;a;ma;h;a;
a;pa .~Ta;a;a;na;k+:a;yRa;ma;a;de ;Zea Y;a;ta;
a;d;Zya;tea
138
G. C ARDONA
form of a derivate in -trc:82 kros..ta (nom. sg.), kros..ta rau (nom.

pl.), kros..to (voc. sg.). As Pan.inacc. du.), kros..ta rah. (nom.
yas recognize, trjvat in this sutra has -vati according to A. 5.1.115
(p. 129).83 Theeffect of letting kros..tu- behave as though it were
kros..tris, of course, to allow for one stem to occur instead of
the other,
just as a subsequent rule accounts for the alternation of
the stems before vowel-initial endings of the third and following
triplets, e.g., kros..tre/kros..tave (dat. sg.).84 In fact, after explicitly
interpreting trjvat in accordance with A. 5.1.115 (p. 129), Bhat.t.oji goes on tosay that what is intended by A. 7.1.95 (p. 137) (ity
arthah. thus what is meant) is that kros.tr is to be used instead
of kros..tu.85 As both Vasudeva Dks.ita andJanendrasarasvat are
quick to point out, however, Bhat.t.oji simply means to express the
82 k+:ea;u;Za;b.d;~tua;npra;tya;ya;a;ntaH
.sa;Va;a;Za;b.dH .sa;vRa;na;a;ma;~Ta;a;nea Y;sa;}bua:;d ;Ea :pa:=+ta;~txa:$va;;

+va;a;ta
txa:ja;nta;~ya ya;dU +pMa ta;d;~ya Ba;va;ta;a;tya;TRaH .+.pa;a;a;ta;de ;Za;ea Y;ya;m,a :pra;tya;a;sa:ea;(a ku+:Zea;=e +va txa:ja;nta;~ya
ya;dU+pMa ta;d;a;ta;
a;d;Zya;tea (Kas. 7.1.95). The suffix tun is provided by ;a;sa;ta;a;na;ga;a;ma;ma;a;sa;sa;.cya;
a;va;Da;a;Vku+: a;Za;Bya;~tua;n,a (US. 1.70).
83 txa:$va;;
+va;ta;a;a;ta txa;.ca;a tua;yMa va;tRa;ta I+.t,aI txa:$va;t,a (N. 7.1.95 [V.640]), k+:u;~txa:ja;ntea;na
tua;yMa va;tRa;tea (SK. 274 = 7.1.95 [I.277]), k+:ea;u;a:=+a;ta tua;a;nta;a;tpra;Ta;ma;a txa:$va;
a;d;a;ta txa;ta;a;ya;a;nta;a;d;a;taH (BS S 7.1.95 [I.510], SKT. 274 [I.277]), .tea;na tua;ya;a;ma;a;ta txa;ta;a;ya;a;nta;a;d;a;taH
(SKB. 274 [I.277]).
84 ;
a;va;Ba;a;Sa;a txa;ta;a;ya;a;
a;d;Sva;a;.ca (A. 7.1.97). It is not necessary here to consider feminine forms.
85 k+:ea;u;Za;b.d;~ya .~Ta;a;nea k+:ea;x;Za;b.dH :pra;ya;ea;+:v.ya I+.tya;TRaH (SK. 274 [I.277]). Much
earlier, Katyayana argued that if A. 7.1.95 (p. 137) is considered to provide a
rupatidesa, this would allow for tr to be a replacement for all of kros..tu-: .+.pa;a
;a;ta;de ;Za I+.a;ta . cea;tsa;va;Ra;de ;Za;pra;sa;*:H
(7.1.95 vt. 3), .+.pa;a;a;ta;de ;Za I+.a;ta . cea;tsa;va;Ra;de ;ZaH :pra;a;pa;ea;a;ta
.sa;vRa;~ya tua;na;nta;~ya txa;Za;b.d ;a;de ;ZaH :pra;a;pa;ea;a;ta (Bh. III.274.2526). Since the sutra in
question is not overtly a substitution rule, what this means is, as Nagesa remarks,
that extending the form of trc to the stem kros..tu- results in one occurring for
the other: .+.pa;a;a;ta;de ;Za;d;a:=e+a;a;de ;Za O;;va :P+.l+.ta;ea ;

a;va;Da;a;ya;ta I+.tya;TRaH (Ud. V.89). Other
parts of the Bhas.ya arguments, as well as Kaiyat.as comments thereon, cannot
be considered here.
E XTENSION RULES
139
ultimate result of the sutra.86

This is as it should be. For the syntax of A. 7.1.95 (p. 137)
is according to schema I: a nominative kros..tuh. along with trjvat,
which refers to the act that kros..tu- shares with the derivate in-trc,
namely having its form before sarvanamasthana endings other than

a sambuddhi ending. Despite the ultimate result noted by Bhat.t.oji, it is not immediately possible to interpret the rule in accordance
with schema II.87 If this were the case, one would expect a genitive
kros..toh. instead of the nominative kros..tuh..
In fact, however, the latter has been interpreted as a genitive
form in this sutra. Thus, Haradatta remarks that one should read
tasya after sajasabdah. of the Kasikas paraphrase of A. 7.1.95
(p. 137) (see note 82), because the genitive kros..tuh. is stated in the
sutra.88 He also goes on to say that trjvat contains vati introduced
after a genitive form, by A. 5.1.116 (p. 132),89 so that the sutra
conforms to schema II. The question nevertheless remains: how
can kros..tuh. be a genitive form of kros..tu-? Commenting on Rama krs.n.a confronts the
candras paraphrase of A. 7.1.95 (p. 137),90 Sr
question, noting that what Ramacandra says isbased on the following reasoning. To begin with, kros..tuh. in the sutra contains a
sixth-triplet ending after a citation form of the stem kros..tu, but the
form is stated as though it were from the stem kros..tr-. It would be
86 . . . I+.tya;a;Ba;prea;tya :P+.a;l+.ta;ma;a;h k+:ea;u;Za;b.d;~ya .~Ta;a;na I+.a;ta (SKB. 274 [I.277), :P+.a;l+.ta;ma;a;h k+:ea;u;Za;b.d;~ya .~Ta;a;nea k+:ea;x;Za;b.d I+.a;ta (SKT. 274 [I.277]). Cf. note 85.
87 Although some commentaries appear to do so. Thus, Dharmakrti says (RA.
7.1.95 [I.53]) k+:ea;u;Za;b.d;~ya txa:ja;nta;~yea;va .+.pMa Ba;va;tya;sa;}bua:;d ;Ea .sa;vRa;na;a;ma;~Ta;a;nea :pa:=+taH .+.pa;a;a;ta;de ;Za;ea Y;ya;m,a See also note 90.
88 .sa;Va;a;Za;b.d I+.a;ta tua;npra;tya;ya;~ya .sa;Va;a;TeRa ;
a;va;Da;a;na;a;t,a .ja;}bua;k+:~ya . cea;yMa .sa;Va;a va;~tua;~va:+.pak+:Ta;na;mea;ta;t,a :a;a;na;nta:=M ta;~yea;a;ta :pa;
a;F+.ta;v.yMa k+:ea;u;a:=+a;ta .sUa:ea :Sa; +a;a;na;deR ;Za;a;t,a (PM. 7.1.95
[V.640]).
89 txa:$va;;
+va;ta;a;a;ta txa;.ca I+.va txa:$va;t,a ta:a ta;~yea;vea;a;ta :Sa; +a;sa;ma;Ta;Ra;d;a;taH (PM. 7.1.95
[V.640]).
90 k+:ea;u;Za;b.d;~ya k+:ea;x;Za;b.d;va;d+pMa .~ya;a;d;sa;}bua:;d ;Ea .sa;vRa;na;a;ma;~Ta;a;nea (PK. 7.1.95 [I.168]);
U
similarly, Mit. 7.1.95.
140
G. C ARDONA
purposeless to have kros..tuh. be a citation of a form from kros..tr-.91
krs.n.a is aware of how difficult it is to maintain this interpretaSr

tion.For he immediately goes on to say that, alternatively, kros..tuh.
contains a first-triplet ending after the stem in -tun, that trjvat con by A.
tains vati introduced after a pada with a third-triplet ending,
5.1.115 (p. 129), and that Ramacandras paraphrase conveys only
the ultimate intent of the rule: kros..tu- behaves like the stem in -trc
by taking on a form identical with that of this stem.92

A final question remains: if this is indeed Pan.inis intent,
why did he formulate an extension rule and not a substitution rule
whereby kros..tu- would be replaced by kros..tr-, optionally in partic krostu, derived with
ular forms? The suggestion I can make is this:
..
tun, refers to a particular creature that shrieks, a jackal. No sutra
provides for a stem in trc similarly restricted. Stating A. 7.1.95
(p. 137) as an extensionrule allows for forms from a stem in trc
instead of kros..tu- but also allows for maintaining the semanticsof
the latter.
Summary
Pan.inis As..ta dhyay includes sutras, called atidesasutra, which

serve to extend properties or operations. Pan.inyas recognize a
fairly wide variety of extensions (2) up to seven types although there are four major kinds: extension of operations (ka91 .sUa:ea
. ca k+:ea;u;a:=+a;ta tua;a;nta;a;nua;k+.=+a;a;tSa; +a l+a;a;va;Za;a:ua txa:$va;;

+a;vea;na ;a;na;deR ;ZaH kx+:taH
txa:ja;nta;a;nua;k+.=+ea tua txa:$va;;
+a;va;
a;va;Da;a;na;ma;na;TRa;kM .~ya;a;
a;d;tya;a;Za;yea;na;a;h k+:ea;u;Za;b.d;~yea;a;ta (PKPr.
7.1.95 [I.261]). The published text has trjvadabhavena, which I have emended to
krs.n.a maintains that the form is

trjvadbhavena. Under the emended reading,
Sr
stated as though it were from the stem kros..tr-, in accordance with A. 7.1.97 (see
note 84).
92 ya;d;a ta;a;tpa;yRa;k+:Ta;na;mea;ta;t,a ta:va;ta;~tua k+:ea;u;a:=+a;ta tua;a;nta;a;tpra;Ta;ma;a txa;ta;a;ya;a;nta;a;d;a;taH txa;.ca;a
tua;yMa txa:$va;t,a ta;ta;(a k+:ea;u;Za;b.d;~txa:ja;ntea;na tua;yMa va;tRa;tea .tea;na tua;yMa .+.pMa :pra;a;pa;ea;ta;a;tya;TRaH (PKPr.
7.1.95 [I.261]).
E XTENSION RULES
141
rya), rules (sa stra), physical shape (rupa), and identity (tadatmya).
There are syntactically two major types of atidesasutras: those
which provide for an entity X to be Y through attributing particular properties to it, thereby depriving X of its original properties
(3, 4.1); those which provide that an entity X behaves or is like
an entity Y while still retaining its original properties. The second
type is itself twofold: X is comparable to Y in that the two share
an action (kriya), so that what is compared is that action relative to
X and Y (4.2.1); an X is said to be like a Y in that both are substantive entities sharing something that is located in or pertains to
an X with which Y is comparable (4.2.2). These two main types
of comparisons also differ syntactically. The first involves an act
in progress and a pertinent sutra contains either in the sutra itself or by anuvrtti a nominative form referring to the entity said
to be or behave as such and such. Sutras of the second type have
locative or genitive forms. There are also pragmatic factors which
can impinge on how sutras are understood (5).
A recent study (Katira, Dharurkar, and Kulkarni 2013) is devoted to considering, on the basis of a sample of rules, how modern translators differ in their interpretations of sutras containing
terms with vati, depending on whether they intend to convey literal meaning or the intent of a rule.93 It is worth emphasizing that
native speakers of Sanskrit themselves were expected to interpret
such sutras, basing their understanding on the conventions of their
own speech which Pan.ini describes with respect to the syntax of statements and how these would be paraphrased by native
speakers.
93 The
general study on extension rules by S. Lakshmi Narasimham (2008)

came to my attention after I had written this paper, so that I have not been able to
discuss this work here.
142
G. C ARDONA
7 As..ta dhyay sutras

in which terms with -vati
are used
7.1
List indicating extension type
Key to abbreviations in Table 3, column 3:

k karyatidesa
s sa stratidesa
r rupatidesa
t tadatmyatidesa
n nimittatidesa
v vypadesa tidesa
a arthatidesa
In notes accompanying sutras, I cite representative commentaries,
with no attempt to cover all.
Table 3
Sutras with terms in -vati
1.1.21
1.1.56
1.2.5194
1.2.6695
;a;d;a;nta;va;de ;k+:a;sma;n,
/// a
.~Ta;a;a;na;va;d;a;de ;Za;ea Y;na;a;va;Da;E
// a
lu +.
a;pa yua;+:va;d;a;a;+:va;.ca;nea
.~:a;a :pMua;va;a
k 2.6, p. 121
k 2.1, p. 111
k
ak
94 This s
utra includes a term with vati under one of two alternative interpretations proposed: (a) yuktavadvyaktivacane is a s.as.t.h-tatpurus.a the first constituent of which is yuktavat; this term derives from yuj join, connect with the
participial affix ktavatu and refers to a base meaning. (b) the sutra consists of
two padas, the first of which is yuktavat, with the suffix vati. Kas.: yua;+:va;
a;d;a;ta
;a;na;+a;pra;tya;yea;na +:va;tua;na;a :pra;kx+:tya;TRa o+..cya;tea .sa ;
a;h :pra;tya;ya;a;TRa;ma;a;tma;na;a yua;na;a; ta;~ya yua;+:va;ta;ea
v.ya;a;+:va;.ca;nea lu +.ba;TeRa ;
a;va;Da;a;yea;tea ;Ta;va;a yua;H :pra;kx+:tya;TRaH :pra;tya;ya;a;TeRa;na .sa;}ba:;d H ta;a;sma;
/ / / a;a;va
v.ya;a;+:va;.ca;nea lu +.ba;TeRa Ba;va;taH
95 This is one of a set of rules providing for one of a group of terms to be a
single remainder (PWT.: 260 374), as in gargyau (nom.-acc. du. masc.) equivalent to garg ca gargyan.as ca A senior female gotra descendant of Garga and a
E XTENSION RULES
1.2.6996
1.3.62
1.3.63
1.4.106
2.1.2
2.4.2697
2.4.27
na;pMua;sa;k+:ma;na;pMua;sa;ke+:nEa;k+:va;a;a;~ya;a;nya;ta:=+~ya;a;m,a
:pUa;vRa;va;tsa;naH
;a;}.pra;tya;ya;va;tkx+:Va;ea Y;nua;pra;ya;ea;ga;~ya
:pra;h;a;sea . ca ma;nya;ea;pa;pa;de ma;nya;tea:+.a;ma O;;k+:va;a
.sua;ba;a;ma;a:/n:a;tea :pa:=+a;*: +.va;t~va:=e
:pa:=+va;
a;+:*:M d;nd;ta;tpua:+:Sa;ya;eaH
:pUa;vRa;va;d:(;a;va;q+.va;Ea
143
k
k 2.5, p. 119
k 4.3, p. 135
k
t 2.4, p. 118
a
a
junior descendant of this line. As with other sutras where the term str is used,
the question arises whether this refers to (a) feminine affixes introduced under
the heading of A. 4.1.3: ;a;~:
/ a;ya;a;m,a (PWT.: 66 107), (b) the meaning feminine, or
(c) a feminine word. Under (a) or (c), a problem arises concerning the derivation
of gargah. equivalent to garg ca gargyayan.au ca A senior female gotra descendant of Garga and two younger descendants of this lineage, with gargya-n
p (
garg). According to A. 2.4.64: ya;Va;Va;ea;(a (PWT.: 308 465), ya of a derivate such
as gargya ( garga-nas-ya)
is omitted if it is used with reference to a plural
ity of descendants excluding a feminine. Under (a), A. 1.2.66 would prevent the
occurrence of a feminine suffix n
p (A. 4.1.16: ya;Va;(a) but the femininity of the
base which would take this suffix as a cosignifier of this meaning would not be
suppressed. Under (c), the rule would let the word in question be treated as masculine but again not eliminate its feminine meaning. A. 2.4.64, which excludes a
female, would then not apply. Moreover, the ending sas of an accusative plural
gargan ( garga-sas) equivalent to garg ca gargyayan.au ca would now not be
eligible for having its -s replaced by -n, since the rule in question (A. 6.1.103: ta;sma;a;.cC+.sa;ea naH :pMua;a;sa) requires a masculine referent. Hence, position (b) is adopted.
Thus, the Bhas.ya and later commentaries: ya;
a;d :pra;tya;ya;g{a;h;Ma va;a Za;b.d;g{a;h;Ma va;a ga;a;ga;Ra
. ca ga;a;gya;Ra;ya;a;Ea . ca ga;ga;RaH :ke+:na ya;Za;b.d;ea na (rUa;yea;ta ;a;~:
/ a;ya;a;a;ma;a;ta ;
a;h lu +.gua;.cya;tea I+.h . ca ga;a;ga;Ra . ca
ga;a;gya;Ra;ya;a;Ea . ca ga;ga;Ra;npa;Zya ta;sma;a;.cC+.sa;ea naH :pMua;sa;a;a;ta na;tvMa na :pra;a;pa ;ea;a;ta ;Ta;a;TRa;g{a;h;Ma na d;ea;Sa;ea
Ba;va;a;ta (Bh. I.247.23248.3). Accordingly, A. 1.2.66 can be considered to extend
masculinity. Nevertheless, it thereby extends operations proper to a masculine.
Cf. Kas. 1.2.66: .~:a;a vxa:;d ;a yUa;na;a .sa;h;va;.ca;nea ; a;Za;Sya;tea ta;+.a;a;(ea;de ;va ;
a;va;Zea;Sa;ea Ba;va;a;ta :pMua;sa
I+.va;a;~ya;aH k+:a;ya Ba;va;a;ta .~:ya;TRaH :pua;ma;TRa;va;;
+va;a;ta
96 . . . O;;k+:va;a;a;~ya k+:a;ya Ba;va;tya;nya;ta:=+~ya;a;m,a (K
as. 1.2.69).
97 O;;ta;ya;eaH :pa:=+pa;d;~yea;va ;a;l+.
*:M .~ya;a;t,a (SK. 812).
144
3.1.3998
3.1.87
3.3.131
3.3.132
3.3.135
3.4.85
4.2.3499
4.2.46
4.3.80
4.3.100100
4.3.156
5.1.96
5.4.22
G. C ARDONA
Ba;a;h;a;Bxa;hu;va;Ma (;u+va;a
k+:mRa;va;tk+:mRa;a;a tua;ya;
a;k+:yaH
va;tRa;ma;a;na;sa;a;ma;a;pyea va;tRa;ma;a;na;va;d;a
;a;ZMa;sa;a;ya;Ma BUa;ta;va;a
na;a;na;d;a;ta;na;va;a;tk
/ +:ya;a;pra;ba;nDa;sa;a;ma;a;pya;ya;eaH
l+.ea;f;ea l+.z+.t,a
k+:a;le +.Bya;ea Ba;va;va;t,a
. ca:=+ea;Bya;ea ;Da;mRa;va;t,a
ga;ea:a;a;d;*: +.va;t,a
.ja;na;pa;
a;d;na;Ma .ja;na;pa;d;va;tsa;va .ja;na;pa;de ;na .sa;ma;a;na;Za;b.d;a;na;Ma ba;hu;va;.ca;nea
k
+:a;ta;va;tpa;a:=+ma;a;a;a;t,a
ta:a . ca d ;a;ya;tea k+:a;ya Ba;va;va;t,a
.sa;mUa;h;va;a ba;hu;Sua
6.1.85101 ;nta;a;
a;d;va;a
6.1.129102 ;p;u+ta;va;du;pa;a;~Ta;te
// a
6.2.175
ba;h;ea;nRa;Vva;du:a:=+pa;d;BUa;a;}:a
k
k 3, p. 124
s
s
s
k 4.3, p. 134
s 2.2, p. 115
s
s
s
s
s
s
k
k
s
98 . . . O;;tea;Bya;ea ;a;l+.a;f :pa:=+ta ;a;}.pra;ta;ya;ea Ba;va;tya;nya;ta:=+~ya;a;m,a (;+a;

a;va;va . ca;a;a;sma;nk+:
///
a;ya Ba;va;a;ta
(Kas. 3.1.39).
99 A. 4.2.46; 4.3.80, 156; 5.1.96; 5.4.22 are comparable to A. 4.2.34.
100 .ja;na;pa;
a;d;na;ea yea ba;h;va;.ca;nea .ja;na;pa;de ;na .sa;ma;a;na;Za;b.d;a;~tea;Sa;Ma .ja;na;pa;d;va;tsa;va Ba;va;a;ta :pra;tya;yaH :prau
;kx+:a;ta;(a .sa;ea Y;~ya Ba;a;+:a:=+tyea;ta;a;sma;
/ / / a;nva;Sa;ye
/ a .ja;na;pa;d;ta;d;va;Dya;ea;(a (A. 4.2.124) I+.tya:a :pra;k+.=+ea
yea :pra;tya;ya;a ;
a;va;
a;h;ta;a;~tea .ja;na;pa;
a;d;Bya;ea Y;a;sma;a;Te
/ / / Ra Y;a;ta;
a;d;Zya;ntea (Kas. 4.3.100).
101 The s
utra is comparable to A. 1.1.56 (p. 111) (2.1): a single replacement
for two contiguous vowels behaves like the final and initial sound of the prior
and subsequent elements to which the two vowels replaced belong. Cf. SKP. 75
(162): .~Ta;a;a;na;va;tsUa:ea;Ea;va ga;ta;a;TRa;a;ma;dM .sUa:a;m,a
102 o+.pa;a;~Ta;tM
/ / a na;a;ma;a;na;a;SRa I+.a;ta;k+.=+aH .sa;mua;d;a;ya;a;d;va;a;./ c/ C+.d;a :pa;dM yea;na .~va:+.pea Y;va;~Ta;a;pya;tea
ta;a;sma;npa:
/ / / =+taH :p;u+ta;ea Y;p;u+ta;va;;
+va;a;ta :p;u+ta;k+:a;ya :pra;kx+:a;ta;Ba;a;vMa na k+.=+ea;a;ta (Kas. 6.1.129).
E XTENSION RULES
145
6.3.34103 ;a;~:
+a;
ma;a;na;a;a;Da;k+.=+ea ;a;~:
/ a;ya;a;ma;pUa:=+a;a;
a;pra;ya;a;
a;d;Sua
6.3.42
:pMua;va;tk+:mRa;Da;a:=+ya:ja;a;ta;a;ya;de ;Za;a;yea;Sua
6.3.68
I+..ca O;;k+:a;.ca;ea Y;}.pra;tya;ya;va;a
6.4.22104 ;a;sa:;d ;va;d:a;a Ba;a;t,a
6.4.62
.~ya;a;sa;.csa;a;yua;*+.
Ba;a;va;k+:mRa;a;ea:+:pa;de ;Zea
a;a;sa;Sua
Y;$+.na;g{a;h;dx;Za;Ma va;a ;a;.ca;va;
a;d:*:
7.1.74105 txa;ta;a;ya;a;
a;d;Sua Ba;a;
a;Sa;ta;pMua;~kM :pMua;va;+a;l+.va;~ya
7.1.95
txa:$va;tk+:ea;uH
7.2.61
;.ca;~ta;a;~va;tTa;ya;a;na;f;ea ;a;na;tya;m,a
7.4.93
.sa;nva;+:Gua;a;na . ca:*: +=e Y;na;gl+.ea;pea
8.1.9
O;;kM ba;hu;v.ra;a;
a;h;va;t,a
8.1.11
k+:mRa;Da;a:=+ya;va;du:a:=e+Sua
106
8.1.72
;a;ma;a:/n:a;tMa :pUa;vRa;ma;
a;va;d;a;ma;a;na;va;t,a
k
k
103 Similarly,
2.3, p. 117
r
k
k
k
r
r 4.3, p. 137
k
k
k
A. 6.3.42 provides that a feminine takes a masculine form in a

karmadharaya compound as well as when an affix jatyar or desya follows.
104 na ;a;sa:;d ;ma;a;sa:;d ;m,a ;a;sa:;d ;k+:a;ya;Ra;k+.=+a;sa;a;Da;}yeRa;a;a;a;sa:;d e ;na tua;yMa va;tRa;ta I+.tya;a;sa:;d ;va;t,a (N.
6.4.22 [V.363364]), na ;a;sa:;d ;ea Y;a;sa:;d H yaH :pua;naH ;a;sa:;d ;ea Y;
a;pa ;a;sa:;d ;k+:a;ya na k+.=+ea;a;ta .sa
.tea;na tua;yMa va;tRa;ta I+.tya;a;sa:;d ;va;t,a (PM. 6.4.22 [V.364]). The view is also entertained
that A. 6.4.22 provides for rule suspension (sa stratidesa); see PWT.: 423425
659661 and note 121 below.
105 Cf. A. 6.3.34; see note 103.
106 ;a;ma;a:/n:a;tMa :pUa;vRa;ma;
a;va;d;a;ma;a;na;va;;
+va;a;ta ta;a;sma;nsa;
/ / / a;ta ya;tk+:a;ya ta;a Ba;va;a;ta ;sa;a;ta ya:a;;
+va;a;ta (Kas. 8.1.72).
146
7.2
G. C ARDONA
Syntax of sutras
with terms in -vati
Pertinent forms are underlined.

7.2.1
Schema I (4.3)
Table 4
Ia. Sutras in which a nominative form is explicit
1.1.56107
1.2.66108
1.4.106
2.4.27109
3.1.87110
.~Ta;a;a;na;va;d, ;a;de ;Za;ea Y;na;a;va;Da;E

// a
.~:a;a :pMua;va;a
:pra;h;a;sea . ca ma;nya;ea;pa;pa;de ma;nya;tea:=, o+a;ma O;;k+:va;a
:pUa;vRa;va;d, :(;a;va;q+.va;Ea
k+:mRa;va;tk+:mRa;a;a tua;ya;
a;k+:yaH
6.3.68
I+..ca O;;k+:a;.ca;ea Y;m,a :pra;tya;ya;va;a
7.1.74111
txa;ta;a;ya;a;
a;d;Sua Ba;a;
a;Sa;ta;pMua;~kM :pMua;va;+a;l+.va;~ya
7.1.95112 txa:$va;t,ak+:ea;uH
8.1.9113
107 .~Ta;a;a;na;na;a
2.1, p. 111
3, p. 124
4.3, p. 137
O;;kM ba;hu;v.ra;a;
a;h;va;t,a
tua;yMa va;tRa;ta I+.a;ta .~Ta;a;a;na;va;t,a (Kas. 1.1.56), .~Ta;a;a;na;na;a tua;yMa va;tRa;ta I+.a;ta
ta:ua;ya;k+:a;yRa;tva;a:ea;na tua;ya;m,a (N. 1.1.56 [I.184]).
108 :pMua;sa I+.va;a;~ya;aH k+:a;ya Ba;va;a;ta .~:ya;TRaH :pua;ma;TRa;va;;
+va;a;ta (Kas. 1.2.66); see note 95.
109 The s
utra cites the dual asvavad.avau and states that its gender is like that
of the first constituent. Since the form given is nominative, purvavat is properly
interpreted according to schema I. On the other hand, this rule is an exception to
A. 2.4.26, which is formulated according to schema IIa.
110 See note 70.
111 :pMua;va;
a;d;a;ta :pMua;Za;b.de;na tua;ya;a;ma;tya;TRaH (PM. 7.1.74 [V.617]).
112 txa;.ca;a tua;yMa va;tRa;ta I+.a;ta txa:$va;t,a (N. 7.1.95 [V.640]).
113 A. 8.1.9 comes under the heading A. 8.1.1 .sa;vRa;~ya :de. Rules headed by the
latter provide for iterating padas under stated conditions. By A. 8.1.9, the term
eka (nom. sg. nt. ekam) when iterated behaves as does a bahuvrhi compound.
See PWT.: 257259 370372.
E XTENSION RULES
147
8.1.72114 ;a;ma;a:/n:a;tMa :pUa;vRa;m,a ;

a;va;d;a;ma;a;na;va;t,a
Table 5
Ib. Sutras in which a nominative form is understood
1.3.62115 :pUa;vRa;va;tsa;naH (;a;tma;nea;pa;d;m,a 12)
3.3.131116
3.3.132
3.3.135
4.2.34117
4.2.46
4.3.80
4.3.156
5.1.96
5.4.22118
2.5, p. 119
va;tRa;ma;na;sa;a;ma;a;pyea va;tRa;ma;a;na;va;d;a (:pra;tya;yaH 3.1.1)

;a;ZMa;sa;a;ya;Ma BUa;ta;va;a (:pra;tya;yaH 3.1.1)
na;a;na;d;a;ta;na;va;a;tk
/ +:ya;a;pra;ba;nDa;sa;a;ma;a;pya;ya;eaH (:pra;tya;yaH
3.1.1)
k+:a;le +.Bya;ea Ba;va;va;t,a (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH 2.2, p. 115
4.1.76)
. ca:=+ea;Bya;ea ;Da;mRa;va;t,a (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH
4.1.76)
g{a;ea:a;a;d;*: +.va;t,a (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH 4.1.76)
k
+:a;ta;va;tpa;a:=+ma;a;a;a;t,a (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH
4.1.76)
ta:a . ca d ;a;ya;tea k+:a;ya Ba;va;va;t,a (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH 4.1.76)
.sa;mUa;h;va;a ba;hu;Sua (:pra;tya;yaH 3.1.1, ta;a:;d ;ta;aH 4.1.76)
114 ;
a;va;d;a;ma;a;nea;na
tua;yMa va;tRa;ta I+.tya;

a;va;d;a;ma;a;na;va;t,a (N. 8.1.72 [VI.322]).
.tea;na tua;yMa ;
a;k+:ya;a . cea;d;a;taH (N. 1.3.62 [I.456]).
116 va;tRa;ma;a;na;sa;ma;a;pea BUa;tea Ba;
a;va;Sya;a;ta . ca va;tRa;ma;a;na;a:;d ;a;ta;eaH va;tRa;ma;a;na;va;tpra;tya;ya;a va;a Ba;va;a;nta
/ /
(Kas. 3.3.131). Jayaditya appropriately supplies the plural pratyayah., by contextual modification, for the singular pratyayah. in the major heading. He does the
same in his comments on A. 3.3.132: ta:a Ba;
a;va;Sya;a;ta k+:a;l ;a;ZMa;sa;a;ya;Ma ga;}ya;ma;a;na;a;ya;Ma
;Da;a;ta;ea;va;Ra BUa;ta;va;tpra;tya;ya;a Ba;va;a;nta
/ /
117 k+:a;l+.
a;va;Zea;Sa;va;a;a;.ca;ByaH Za;b.de;Bya;ea Ba;va;va;tpra;tya;ya;a Ba;va;a;nta
/ / .sa;a;~ya :de;va;tea;tya;a;sma;
/ / / a;nva;Sa;ye
/ a
(Kas. 4.2.34); comparably on A. 4.2.46.
118 ba;h;Sua :pra;kx+:tea;SUa;.cya;ma;a;nea;Sua .sa;mUa;h;va;tpra;tya;ya;a Ba;va;a;nta
/ / (Kas. 5.4.22).
u
115 :pUa;vRa;va;
a;d;a;ta
148
G. C ARDONA
6.1.85119 ;nta;a;
a;d;va;a (O;;kH :pUa;vRa;pa:=+ya;eaH 84)
6.1.129120 ;p; u +ta;va;du;pa;a;~Ta;te
// a
121
6.4.22
;a;sa:;d ;va;d:a;a;Ba;a;t,a
8.1.11122 k+:mRa;Da;a:=+ya;va;du:a:=e+Sua
7.2.2
Schema II
Table 6
IIa. B-vat = B6 iva
1.2.69
1.3.63
4.3, p. 135
2.4.26123
na;pMua;sa;k+:ma;na;pMua;sa;ke+:nEa;k+:va;a;a.~ya;anya;ta:=+~ya;a;m,a
;a;}.pra;tya;ya;va;t,a kx+:Va;ea Y;nua;pra;ya;ea;ga;~ya
:pa:=+va;
a;+:*:M d;ndta;tpua:+:Sa;ya;eaH
3.4.85
l+.ea;f;ea l+.z+.t,a
4.3, p. 134
119 See
note 101.
120 As shown in note 102, Vamana supplies plutah, which is appropriate.
For A.
.
6.1.129 has to be considered together with A. 6.1.125 :p;u+ta;pra;gxa;h:a;a ;a;.ca ;a;na;tya;m,a (:pra;kx+:tya;a 115), to which it is an exception: a pluta vowel obligatorily retains its original
form before a vowel, but before iti of a Vedic padapat.ha; it behaves as though it
were not a pluta element. The term pluta, included in the compound plutapragrhyah. is understood in the later rule, where it contextually has an ending of the
triplet.
first
121 Depending on whether A. 6.4.22 is considered to provide that rules or operations stated in the section that follows are treated as suspended, one supplies
sa stram or karyam. I consider that A. 6.4.22 provides for treating as suspended
operations that are stated in the section of rules it governs. See note 104.
122 A. 8.1.11 is a subheading under the heading A. 8.1.1 (see note 113). Rules
under the former state conditions for doubling such that the resulting sequence
behaves like a karmadharaya compound; that is, operations which apply to such
a compound apply also in these sequences. Accordingly, karyam is supplied.
Thus, for example, Kas. 8.1.11: I+.ta o+a:=e+Sua ;
a;d;vRa;.ca;nea;Sua k+:mRa;Da;a:=+ya;va;tk+:a;ya Ba;va;ta;a;a;ta
:vea;
a;d;ta;v.ya;m,a
123 O;;ta;ya;eaH :pa:=+pa;d;~yea;va ;a;l+.
*:M .~ya;a;t,a (SK. 812 = 2.4.26), :pa:=+va;
a;d;a;ta :Sa;.a;nta;a;d;a;taH
ta;d;a;h0 (SKB. 812 [II.119]).
E XTENSION RULES
149
4.3.100124 .ja;na;pa;
a;d;na;Ma .ja;na;pa;d;va;tsa;va .ja;na;pa;de ;na .sa;ma;a;na;Za;b.d;a;na;Ma ba;hu;va;.ca;nea
125
6.2.175
ba;h;ea;nRa;Vva;du:a:=+pa;d;BUa;a;}:a
6.3.34126 ;a;~:
+a;
ma;a;na;a;a;Da;k+.=+ea 2.3, p. 117
;a;~:
/ a;ya;Ma :pUa:=+a;a;
a;pra;ya;a;
a;d;Sua
6.3.42
:pMua;va;tk+:mRa;Da;a:=+ya:ja;a;ta;a;ya;de ;Za;a;yea;Sua (;a;~:
/ a;ya;aH 34)
124 See
note 100
6.2.175 says navat, so that it is connected with A. 6.2.172: na;Vsua;Bya;a;m,a (ba;hu;v.ra;a;h;Ea 162, ;ntaH 143), whereby a constituent in a bahuvrhi compound has
high pitch on its last vowel if it follows na or su. According to A. 6.2.175, a constituent in a bahuvrhi behaves similarly if it follows bahu used to signify a large
amount of what the following constituent denotes, as in b!a;hu!;y!a;vaH (. . . which has
a great deal of barley). Accordingly, the Kasika paraphrases supplying svarah.:
o+a:=+pa;d;ba;hu;tvea ya;ea ba;hu;Za;b.d;ea va;tRa;tea ta;sma;a;a;Va I+.va .~va:=+ea Ba;va;a;ta Under this interpretation,
navat is best considered to have vati by A. 5.1.115 (p. 129), so that the sutra fits
under schema I (p. 133). On the other hand, the use of navat obviously calls for
relating A. 6.2.175, where nais part of a dvandva with a fifth-triplet ending, an
ablative form. It is then difficult to account for vati by A. 5.1.115 (p. 129). Two
ways of solving this problem have been considered. First, the very use of navat
here serves to indicate that vati is allowed after a pada with a fifth-triplet ending.
Alternatively, navat here is considered to have vati introduced after a pada with
a sixth-triplet ending, by A. 5.1.116 (p. 132) under schema II (p. 134). Since
such an ending is used, according to A. 2.3.50 :Sa; +a Zea;Sea, to denote any non-karaka
relation in general (sambandhasamanya), it includes meanings attributed to any
nominal ending that does not signify a karaka, including a fifth-triplet ending;
in brief, navat has vati after a pada B6 equivalent to B5 construed with para
(following). Haradatta presents both alternatives: na;Va I+.va na;Vva;t,a ;sma;a;de ;va ;a;na;pa;a;ta;na;a;tpa:*.a;ma;a;sa;ma;Ta;Ra;d;a;taH ya;d;a .sa;}ba;nDa;sa;a;ma;a;nyea :Sa; +a ta;ya;a .sa;veRa ;
a;va;Ba;+.a;Ta;Ra .sMa;gxa;h:a;nta
I+.a;ta :pa:*.a;}ya;TeRa Y;
a;pa ta:a ta;~yea;vea;a;ta va;a;taH :Sa;.a;nta;a;de ;va (PM. 6.2.175 [V.166]). I think the
second position is the more plausible: navat = naa iva (as if related to na).
126 The term striy
ah. is to be interpreted as a genitive, in accordance with related
sutras containing unambiguously genitive forms: na k+:ea;pa;Da;a;ya;aH .sa;Va;a;pUa:=+a;ya;ea;(a
Thus, for example, Kas. 6.3.34: ta;~ya Ba;a;
a;Sa;ta;pMua;~k+:a;d;nUa;*;
, ;a;a;Za;b.d;~ya :pMua;Za;b.d;~yea;va .+.pMa
Ba;va;a;ta0
125 A.
150
G. C ARDONA
Table 7
IIb. B-vat = B7 iva
1.1.21127 ;a;d;a;nta;va;d, O;;k+:a;sma;n,

/// a
2.6, p. 121
128
1.2.51
lu +.
a;pa yua;+:va;d;a;a;+:va;.ca;nea
129
6.4.62
.~ya;a;sa;.csa;a;yua;*+.
a;a;sa;Sua Ba;a;va;k+:mRa;a;ea:+:pa;de ;Zea Y;$+.na;g{a;h;dx;Za;Ma va;a ;a;.ca;va;
a;d:*:
7.2.61
7.4.93
;.ca;~ta;a;~va;t,a Ta;ya;a;na;f;ea ;a;na;tya;m,a

.sa;nva;+:Gua;a;na . ca:*: +=e Y;na;gl+.ea;pea
Table 8
Abbreviations
A.
Bh.
BhK.
BhKJ.
BSS.
JMSS.
Kas.
Manu.
Mit.
As..ta dhyay.
Mahabhas.ya. See Kielhorn 19621972; Vedavrata
19621963.
Bhat..tikavya. See B. Misra 2004.
Jayamangalas Jayamangala on the Bhat..tikavya. See B.
Misra 2004.
Brhacchabdendusekhara. See Sitaram Shastri 1960.
abarabhas.ya on Jaiminis Mmams
S
asutra. See Abhyankar, Joshi, et al. 19761985.
Kasikavrtti. See Sharma, Deshpande, and Padhye
19691970.
Manusmrti. See Dave 19721984.
a. See Aryavaraguru
Mitaks.ar
and Bhat.t.anathaswamy
19031906.
127 .sa;a;}ya;TeRa
va;a;taH (Kas. 1.1.21).

va;a;taH (Kas. 1.2.51); see note 94.
129 .~ya;a;sa;.csa;a;yua;
*+.
/ a;ta .sa;a;}ya;a :pra;a;ta;ya;ea;a;ga;na;ea ;a;na;deR ;Za;a; a;a;a;a;va ;a;.ca;va;
a;d;a;ta .sa;a;ma;a;sa a;a;sa;a;Sva;
;ma;Ta;Ra;d;a;ta;
a;vRa;a;a;ya;tea ya;Ta;a ma;Tua:=+a;va;tpa;a;f;a;l+.pua:ea :pra;a;k+:a:= I+.a;ta (N. 6.4.62).
128 .sa;a;}ya;TeRa
R EFERENCES
N.
Nir.
PK.
PKPr.
PM.
PWT.
Pr.
RA.
SK.
SKB.
SKP.
SKT.
Ud.
US.
VPPu.
151
Nyasa on the Kasikavrtti. See Dwarika Das Shastri and

Kalika Prasad Shukla19651967.
Nirukta. See Sarup 19201929.
Prakriyakaumud. See Trivedi 19251931.
Prakasa on the Prakriyakaumud. See M. Misra 1977
1980.
Padamajar on the Kasikavrtti. See Dwarika Das
19651967.
Shastri and Kalika Prasad Shukla
See Cardona 1997.
Pradpa on the Mahabhas.ya. See Vedavrata 1962
1963.
Rupavatara. See Rangacharya 1916, 1917.
Siddhantakaumud. See Giridharasarma and Paramesvaranandasarma 19671975.
Balamanorama on Siddhantakaumud. See Giridharasarma and Paramesvaranandasarma 19671975.
Praud.hamanorama on Siddhantakaumud. See Sitaram
Shastri 19641992.
Tattvabodhin on Siddhantakaumud. See Giridharasarma and Paramesvaranandasarma 19671975.
Uddyota on the Mahabhas.ya. See Vedavrata 1962
1963.
Un.a disutra. See Aufrecht 1859.
Pun.yarajas T.ka on the Vakyapadya. See Subramania
Iyer 1983.
References
Abhyankar, Vasudev Kashinath, Ganeshashastri Ambadas Joshi,
majjaiminipran.te Mmams
et al., eds. 19761985. Sr
adarsane
. . . 7 vols. Anandasrama Sanskrit Series 97. Pune: Anand

asrama.
152
G. C ARDONA
arya Bhat.t.anathaswamy,
Aryavaraguru,
Jagannathaswamy and Ach
eds. 19031906. Vyakaran.amitakshara, a Gloss on Pan.inis
Annambhat..ta. Benares SanGrammatical Aphorisms by Sr
skrit Series 20. Benares: Braj B. Das.
Aufrecht, Theodor, ed. 1859. Ujjvaladattas Commentary on the
Un.a disutras. Bonn: Adolphus Marcus.
Cardona, George. 1997. Pan.ini: his work and its traditions; vol.
1, Background and Introduction. Second edition, revised and
enlarged. Delhi: Motilal Banarsidass.
Dave, J. H., ed. 19721984. Manu-Smriti with Nine Commentaries by Medhatithi, Sarvajanarayan.a, Kulluka, Raghavananda, Nandana, Ramacandra, Man.irama, Govindaraja, and
Bharuci. 6 vols. in seven parts. Bombay: Bharatiya Vidya Bhavan.
Dwarika Das Shastri and Kalika Prasad Shukla, eds. 19651967.
arya Jinendrabuddhipada and PaNyasa or Pacika of Ac
damajar of Haradatta Misra on the Kasikavr.tti [Commentary on the As..ta dhyay of Pan.ini] of Vamana-Jayaditya:
Bodhisattvadesyacaryajinendrabuddhipadaviracitaya Nyasaparaparyayakasikavr.ttipacikaya vidvadvaraharadattamisraviracitaya Padamajarvyakhyaya ca sahita Kasikavr.ttih. Pan.inyas..ta dhyayvyakhya. 6 vols. Prachya Bharati Series 27.
Varanasi: Prachya Bharati Prakashan.
Giridharasarma and Paramesvaranandasarma, eds. 19671975.
madbhat..tojidks.itaviracita Vaiyakaran.asiddhantakaumud
Sr
madvasudevadks.itapran.taya Balamanoramakhyavya. . . Sr
khyaya Srimajj
anendrasarasvatviracitaya Tattvabodhinyakhyavyakhyaya ca sanathita. 4 vols. Delhi: Motilal Banarsidass.
Katira, Dipesh, Chinmay Dharurkar, and Malhar Kulkarni. 2013.
Vat: a textual study. Proceedings of the 15th World Sanskrit
Conference; vol. 2, Vyakaran.a Across the Ages: proceedings of
R EFERENCES
153
the vyakaran.a section of the 15th World Sanskrit Conference,

ed. by George Cardona, pp. 14767.
Kielhorn, Lorenz Franz, ed. 19621972. The Vyakaran.a = Mahabhas.ya of Patajali: revised and furnished with additional
readings, references, and select critical notes, rev. by Vasudev
Kashinath Abhyankar. 3rd ed. 3 vols. Poona: Bhandarkar Oriental Research Institute. [A reissue labelled fourth edition
was published in 1985 and 1996.]
Lakshmi Narasimham, S. 2008. Pan.inye atidesa nuslanam. Ra tavidyapt.hagranthamala 187, Paramparikasa stras.t.riyasamskr
tavidyasamutkars.agrantham
ala 30. Tirupati: Ras.t.riyasamskr
pt.ham.
Bhat..tikavi: with
Misra, Bankelala, ed. 2004. Bhat..tikavyam of Sr
Jayamangala & Sathe commentaries Jayamangala by Sr
rvapathna by Sr Mallinatha Suri. 2 vols. SarasvatbhavanaGranthamala 147. Varanasi: Sampurnanand Sanskrit University.
Misra, Muraldhara, ed. 19771980. Prakriyakaumud by Rama kris.n.a edited with his Racandracarya: with Prakasa by Sr
smi by Pt. Muraldhara Misra. 3 vols. Sarasvartbhavana-Granthamala 111112. Varanasi: Sampurnanand Sanskrit Vishvavidyalaya.
Rangacharya, Rao Bahadur M., ed. 1916. The Rupavatara of Dharmakrti: edited with additions and emendations, for the use of
college students. Vol. 1. Madras: G. A. Natesan & Co.
. ed. 1917. The Rupavatara of Dharmakrti: edited with additions and emendations, for the use of college students. Vol. 2.
Bangalore: The Bangalore Press.
Sarup, Lakshman, ed. and trans. 19201929. The Nighan..tu and
the Nirukta: the oldest treatise on etymology, philology and
semantics: critical edition, translation, introduction, exegetical and critical notes, indexes and appendices. 1st ed. London: Oxford University Press; Lahore: University of the Pan-
154
G. C ARDONA
jab. [Reprinted with three parts in one volume: Delhi: Motilal

Banarsidass, 1967, 1984, 1998.]
Sharma, Aryendra, Khanderao Deshpande, and D. G. Padhye,
eds. 19691970. Kasika: a commentary on Pan.inis grammar by Vamana & Jayaditya. 2 vols. Sanskrit Academy Series 17, 20. Hyderabad: Sanskrit Academy, Osmania University. [Reprinted in one volume, 2008.]
Sitaram Shastri, ed. 1960. Brhat-Sabdendu

sekhara by Nagesa
Bhat..ta. 3 vols. Sarasvati Bhavana Granthamala 87. Varanasi:

Varanaseya Sanskrit Vishvavidyalaya.
Bhat..toji Dkshita:
. ed. 19641992. Praud.hamanorama of Sr
together with Br.ihachchhabdaratna and unpublished commen Hari Dks.ita and Laghusabdaratna of Sr
Nagetary of Sr
sa Bhat..ta. 2 vols. Continuous pagination. Varanasi: Banaras
Hindu University.
Subramania Iyer, K. A. 1974. Bhartr.hari on taddhita formations
involving comparison. Charudeva Shastri Felicitation Volume: presented to Prof. Charudeva Shastri on the occasion
of his seventy-fifth anniversary by friends and admirers, ed.
by Charudeva Shastri felicitation volume editorial committee,
pp. 24157. Delhi: Charudeva Shastri Felicitation Committee.
. ed. 1983. Vakyapadya of Bhartr.hari (An ancient Treatise on
the Philosophy of Sanskrit Grammar) Containing the T.ka of
Pun.yaraja and the Ancient Vr.tti: edited ... with a Foreword by
Ashok Aklujkar, Kan.d.a II. Delhi: Motilal Banarsidass.
Trivedi, Kamlashankar Pranashankar, ed. 19251931. The
Prakriykaumud of Rmachandra: with the commentary
Prasda of Vitt.hala and with a critical notice of manuscripts
and an exhaustive and critical introduction. 2 vols. Bombay
Sanskrit and Prakrit Series 78, 82. Poona: Bhandarkar Oriental Research Institute.
Vedavrata, ed. 19621963. Sr
Vyaka
ran.a-mahabhas.yam: Srkaiyyat.akrtapradpena Nagojbhat..ta
R EFERENCES
155
krtena Bhas.yapradpoddyotena ca vibhus.itam. 5 vols. Gurukul
anam.
Jhajjar
(Rohtak): Harayan.a -Sahitya-Samsth
156
G. C ARDONA
Voice, preverb, and transitivity

restrictions in Sanskrit verb use
Abstract: In the third pada of the As..ta dhyay Pan.ini states
rules that account for the occurrence of verbs in the active,
middle, and passive voices under semantic and coocurrence conditions. These rules include specific conditions
that specify not only which verbs occur in the active or
middle voice but also in which meanings they do so, with
which preverbs they do so, and whether they do so when
transitive or intransitive. Some of the information in these
rules is included in comprehensive bilingual lexical and
grammatical reference sources such as Bhtlingk and Roth
(18551875), and some reference sources such as Oberlies (2003) provide complementary information for specific genres and periods. Yet common reference works
such as Whitney (1885, 1945) do not include these specific conditions, and the information has yet to be used in
any systematic way to analyze Sanskrit syntax. The authors collect and analyze the provisions Pan.ini has made
concerning the semantic and cooccurrence conditions of
preverbs and verbal roots and utilize it in a computational
implementation of the third pada of the first adhyaya.
Keywords: voice assignment, verb usage, computational implementation, Pan.ini
157
158
S CHARF, G OYAL, A JOTIKAR, and A JOTIKAR
Introduction
Pan.ini introduces verbal terminations after verbal roots (dhatus) by

providing a cover symbol l marked variously under various conditions such as time and mood. For example,
A. 3.2.123 va;tRa;ma;a;nea l+.f, (;Da;a;ta;eaH 3.1.91)
The affix lat. occurs after a verbal root that denotes an action
occuring in present time.
These l-affixes are introduced to denote the agent (kartr), direct
object (karman), or the action itself (bhava) by

A. 3.4.69 lH k+:mRa; a;a . ca Ba;a;vea . ca;a;k+:mRa;ke+:ByaH
L-affixes occur to denote either an agent (kartr) or a direct
object (karman), or, after intransitive roots, todenote either
an agent (kartr) or the action itself (bhava).
These l-affixes are then replaced by verbal terminations indiscriminantly by
A. 3.4.78 ;a;ta;a;a;~/ / /
.+.a;sa;pTa;~Ta;a;ma;+;a;sma;~ta;a;ta;a;V+.
Ta;a;sa;a;Ta;a;nDva;a;ma;q+:
a;h;ma;
a;h;z,
The affixes tip etc. occur in place of an l-affix.
The first nine of the eighteen verbal affixes introduced by A. 3.4.78
are termed parasmaipada by A. 1.4.99, the second nine a tmanepada by A. 1.4.100.
A. 1.4.99 lH :pa:=+smEa;pa;d;m,a
Replacements for l-affixes are termed parasmaipada.
A. 1.4.100 ta;z+.a;na;a;va;a;tma;nea;pa;d;m,a
The latter nine affixes referred to by the abbreviation tan and
the (participial) affixes of the form a na (sa nac and kanac)
are termed a tmanepada instead.
VOICE , PREVERB , AND TRANSITIVITY RESTRICTIONS
159
Pan.ini accounts for what in European grammar is called the

voice of verbs by the intersection of two factors:
whether the verbal terminations are used to denote the agent
(kartr), the direct object (karman), or the action denoted by
the verbal
root itself (bhava).
whether the verbal terminations used are those termed parasmaipada or a tmanepada.
Subsequent Indian grammatical tradition refers to the usage of verbal terminations described in the first point as prayoga usage.
The phrases kartari prayoga, karman.i prayoga, and bhave prayoga and the corresponding compound terms kartr-prayoga, ka
rma-prayoga, and bhava-prayoga refer to the agentive,
passive,
and stative use of the verb respectively. By stative we refer to
the passive of an intransitive verb. The parasmaipada terminations
referred to in the second point are used in the active voice, and the
a tmanepada terminations are used in the middle voice and reflexive
usages as well as in the passive and stative. The possessive adjectives parasmaipadin and a tmanepadin are used to describe verbs
that use the former and latter affixes respectively, and the possessive adjective ubhayapadin to describe verbs that use both. The
term pada alone is currently used to refer to the general feature
that derscribes which terminations verbs use. These two factors,
prayoga and pada, align with and uniquely determine what in European grammar is termed active, middle and passive voice as
shown in Table 1.
In the third pada of the As..ta dhyay (A. 1.3.1293), Pan.ini
states restrictive rules that govern the occurrence of verbal terminations (tin)
under semantic and coocurrence conditions. These
rules range from general to particular. General rules utilize markers attached to verbal roots (n, , and nasalized anudatta and svarita vowels), and categorical semantic conditions in the Dhatupa artha) in order to restrict the occurrence of
.tha (gatyartha, hims
160

Table 1
Correspondence of voice with prayoga and pada
Voice
active
active reflexive
middle
passive
stative
Prayoga
kartr
kartr
kartr
karman
bhava
Pada
parasmaipada
a tmanepada
a tmanepada
a tmanepada
a tmanepada
a tmanepada terminations. Particular rules specify the occurrence

of a tmanepada terminations after certain roots when they occur
with certain preverbs, in certain meanings, or intransitively (akarmaka). Pan.ini mentions the coocurrence of certain preverbs with
certain roots throughout the third adhyaya and elsewhere as well
as in the third pada of the first adhyaya. Some of the linguistic
information contained in these rules is included in comprehensive
bilingual dictionaries such as Bhtlingk and Roths (18551875)
Sanskrit-Wrterbuch and Monier-Williams, Leumann, and Cappellers (1899) A Sanskrit-English Dictionary. Oberlies (2003)
pointed out departures from such restrictions in Epic Sanskrit in
his A grammar of Epic Sanskrit. Despite the partial and complementary information provided by these modern lexical and grammatical resources, the information provided by Pan.ini has yet to be
presented in a formal systematic way amenable to computational
use and has yet to be used to analyze Sanskrit syntax. After collecting and thoroughly analyzing the provisions Pan.ini has made concerning the cooccurrence conditions of preverbs and verbal roots
in the As..ta dhyay, the authors utilized it in a computational implementation of the third pada of the first adhyaya in order to generate
comprehensive data of the semantic and cooccurrence restrictions
Pan.ini details regarding the use of parasmaipada and a tmanepada
161
verbal terminations. The current paper describes the implementation of this module.
Conditions used to restrict the occurrence of parasmaipada and
a tmanepada terminations in (A. 1.3.1293) include root markers in
the Dhatupa.tha, prayoga, other semantic conditions, preverbs, and
other co-occurrence conditions.
1.1
Root markers
In the Dhatupa.tha, roots generally appear marked with phonetic

segments not part of the root itself. Pan.inis sutras A. 1.3.29 identify markers and remove them. Certain markers serve as conditions
that restrict the use of parasmaipada and a tmanepada terminations
in A. 1.3. Vowel-final roots may be marked with a final n or ,
and consonant final roots may be marked with an accented nasalized vowel. The accent is high-pitched (udatta), low-pitched (anudatta), or circumflex (svarita). Certain rules in A. 1.3 use these
markers to restrict the use of parasmaipada and a tmanepada terminations. For example,
A. 1.3.12 ;nua;d;a:a;
a;z+.ta ;a;tma;nea;pa;d;m,a
Atmanepada
affixes occur after a root marked with a lowpitched (anudatta) vowel or with n .
1.2
Prayoga
Some rules restrict the use of parasmaipada and a tmanepada terminations to when they are used to denote the agent, direct object
or the action itself. Pan.ini first accounts for the use of a tmanepada
terminations in the passive and stative generally.
A. 1.3.13 Ba;a;va;k+:mRa;a;eaH (;a;tma;nea;pa;d;m,a 12)
An a tmanepada affix occurs after a root when the action itself (bhava) or the direct object (karman) is to be denoted.
162
He then states a number of specific conditions under which a tmanepada terminations are used to denote the agent of the action, for
example,
A. 1.3.14 k+:tRa;a:= k+:mRa;v.ya;a;ta;h;a:=e (;a;tma;nea;pa;d;m,a 12)
An a tmanepada affix occurs after a root when the action is
mutual and the agent (kartr) is to be denoted.
Finally he states a remainder condition under which parasmaipada

terminations are used to denote the agent of the action.
A. 1.3.78 Zea;Sa;a;tk+:tRa;a:= :pa:=+smEa;pa;d;m,a
After a root for which an a tmanepada affix has not been provided by a previous rule, a parasmaipada affix occurs when
the agent (kartr) is to be denoted.
1.3
Other semantic conditions
Besides whether an agent, direct object, or the action itself is to

be denoted, other semantic conditions are also used as a basis to
restrict the use of parasmaipada and a tmanepada terminations.
A. 1.3.23 :pra;k+:a;Za;na;~Tea;ya;a;K.ya;ya;ea;(a (.~TaH 22, ;a;tma;nea;pa;d;m,a 12)
An a tmanepada affix occurs after the root stha when prakasana revelation of ones own intent or stheyakhya reference to an arbiter is to be denoted.
1.4
Preverbs
Many rules in A. 1.3 restrict the use of parasmaipada or a tmanepada terminations to when a root is used with or without a preverb or
with particular preverbs.
A. 1.3.17 nea;
a;vRa;ZaH (;a;tma;nea;pa;d;m,a 12)
An a tmanepada affix is employed after the root vis when
preceded by the preverb ni.
163
A. 1.3.43 ;nua;pa;sa;ga;Ra;d;a (k+:maH 38, ;a;tma;nea;pa;d;m,a 12)

An a tmanepada affix is optionally employed after the root
kram when not used with a preverb.
1.5
Other co-occurrence conditions
Many rules condition the use of parasmaipada or a tmanepada terminations when roots govern certain subordinate words (upapadas), for example,
A. 1.3.16 I+.ta:=e+ta:=+a;nya;ea;nya;ea;pa;pa;d;a;a (k+:tRa;a:= k+:mRa;v.ya;a;ta;h;a:=e 14, na 15,
;a;tma;nea;pa;d;m,a 12)
An a tmanepada affix does not occur after a root which has
as a subordinate co-occurring word itaretara each other or
anyonya one another when the action is mutual and the
agent (kartr) is to be denoted.
A. 1.3.85 ;
a;va;Ba;a;Sa;a;k+:mRa;k+:a;t,a (o+.pa;a;t,a 84, .=+maH 83, k+:tRa;a:= 78, :pa:=+smEa;pa;d;m,a 78)
A parasmaipada affix occurs only optionally after the root
ram preceded by the preverb upa when the root is intransitive.
1.6
Tabulation of voice conditions
As is evident from the rules exemplified above, the eighty-two

rules in A. 1.3 that specify the conditions under which parasmaipada and a tmanepada terminations are used are sufficiently complicated to make it difficult to sort out the exact conditions under
which Pan.ini accounts for the voice of each root. This difficulty
might be one of the reasons for the lack of a work that tabulates
these exact conditions. Another might be that the utility of such a
tabulation was not felt hitherto. Now, however, with the development of Sanskrit computational linguistics software, the utility of
164
such a tabulation is highly evident. For instance, a table that details

the voices permitted for roots when they occur with certain preverbs could add to the precision and completeness of the Sanskrit
Heritage system. Goyal and Huet (2013) discuss issues regarding
completeness of the Sanskrit Heritage system and particularly the
need to limit over-generation. In the Sanskrit Heritage platform
(http://sanskrit.inria.fr), root forms correspond to a
particular phase to which the preverbs are appended to form finite
verb forms that contain preverbs. Incidentally, the same mechanism is used to generate the nominal stems that contain preverbs.
Without any additional information regarding the dependence of
voice on the use of roots with particular preverbs, roots with preverbs would take exactly the same voice as defined for the roots
alone. For example, if the root vis were declared as taking only
parasmaipada affixes, vis would also take only parasmaipada affixes with any preverb, including with the preverb ni. Such forms
are incorrect according to A. 1.3.17 (p. 162) which states a restriction that only a tmanepada affixes occur after the root vis preceded
by the preverb ni. The Sanskrit Heritage system would thus be
incomplete because it would not be able to generate forms in the
middle voice such as nivisate and imprecise because it would generate erroneous forms in the active voice such as *nivisati. On the
other hand, if the root vis were declared as ubhayapadin, i.e. taking
both the parasmaipada and a tmanepada affixes, the system would
generate the correct active forms of the root by itself such as visati
and middle forms of the root with the preverb ni such as nivisate, but the system would also generate incorrect middle forms of
the root vis alone and incorrect active forms of vis preceded by the
preverb ni. Too narrow a restriction on the pada of the root leads
to imprecision and too loose a restriction leads to over-generation.
An exact tabulation of the conditions would enable the system to
be complete and at the same time not to over-generate.
Conversely, by incorporating the precise conditions under
165
which a root occurs in a particular voice, a syntax analyzer would

have access to syntactic information that would limit the number
of possible analyses of a sentence. For example, in the sentence
(1) h;tva;a .=+a;va;Ma .=+a;ma o+.pa:=+ma;tea
After slaying Ravan.a Rama desists.
the fact that the root ram preceded by the preverb upa occurs in
the middle voice indicates that it must be intransitive so that Ravan.a can only be the direct object of the action of slaying denoted
by the root han, not of the action of stopping denoted by the root
ram. For A. 1.3.85 (p. 163) permits the a tmanepada terminations
after the root ram preceded by the preverb upa only if it is used
intransitively.
As shown by these examples, finding the exact conditions for
voice has valuable applications, and yet these conditions are fairly
complex. The combination of the value and complexity motivated
the authors to produce a tabulation of conditions for the use of
proper voice using a computational implementation of the relevant
rules of Pan.inis grammar. The rest of the paper describes this
work in detail. Section 2 discusses existing efforts to develop a
computational implementation of the As..ta dhyay. Section 3 describes the steps taken to prepare the input data for this work. Section 4 gives the implementation details and discusses the notations
used in the derivation tree produced as output. Section 5 shows
the derivation tree produced for some of the roots and discusses
various related issues. Conclusions and future work are discussed
in section 6.
Related Work
Although the authors know of no effort to produce a detailed tabulation of the exact conditions for voice assignment for roots, there
166
have been numerous attempts to produce computational implementations of parts of Pan.inis As..ta dhyay and a few attempts to
produce computational implementations of the whole. Discussion
about how to formalize Pan.inis rules goes back to Staals (1965)
and Cardonas (1965) formal rendition of rules that specify replacement in left and right contexts in accordance with Pan.inis
metarules A. 1.1.49, A. 1.1.66, and A. 1.1.67. Cardona (1974) discusses these rules more explicitly, and Cardona (1969) utilizes the
formalization in a large number of phonetic rules.
In 1991, Scharf wrote a Pascal program that modeled Pan.inian sandhi rules. The program has since been reproduced in C,
Perl, and Java by Ralph Bunker, Malcolm Hyman, and Jim Funderburk respectively. Hyman (2009) described an XML vocabulary
for formalizing Pan.inian rules, and steps to translate it automatically into Perl executable code and to compile it into a finite state
transducer. Scharf (2009c: 11825) briefly describes Scharf and
Hymans joint production of XML formalizations of Pan.inian sandhi, nominal derivation, and verbal conjugation rules and their
automated translation into executable Perl code with rule tracking.
Scharf (2010) describes a similar XML module to derive participial stems from roots. M. Kulkarni (2009) illustrates problems of
phonological overgeneration that would result from a strict implementation of Pan.inian rules. Jha and S. K. Mishra (2009) consider difficulties in formalizing semantic categorization rules and
illustrate the formalization of karaka and vibhakti rules. Satuluri
and A. Kulkarni (2013) are currently working on a Sanskrit compound generator. Patel and Katuri (2014) are currently working
on nominal inflection software with rule-tracking. Swamiji (2010)
continues to expand his computerization of Pan.ini begun before he
hosted the World Sanskrit Conference in Bangalore in 1997.
A complete model of Pan.inis As..ta dhyay involves working
out how to represent the various stages of derivation (prakriya)
from the state of basic elements as they are initially taught (upa-
167
desa), both in components of the grammatical system supplementary to the rules such as the Gan.apa.tha and Dhatupa.tha as well
as in the rules themselves, to the final state of speech forms as
they are actually used. In computational terms, a complete model
requires determining a data structure to represent the derivation
tree from input to output. In the past several years, several articles have addressed fundamental theoretical issues directly relevant to creating a complete implementation of Pan.inis As..ta dhyay
and have taken practical steps to lay the foundation for such an
endeavor. Gillon (2007) demonstrates that Pan.inis grammar is a
generative grammar. Kiparsky (2009) provides an overview of the
organization of the grammar, types of rules, and principles of rule
application. Scharf (2009c: 95117) describes the components
of Pan.inian grammar, the extent of semantic conditions involved,
the categorization of semantic conditions, and their employment
in introducing and modifying speech forms. Scharf (2009a) denigrates a formal segregation into levels. Scharf (2011a) demonstrates that semantics and semantic categorization drive the introduction of basic elements at the foundation of Pan.inian derivational procedure. Scharf (2009b) reconstructs from the Madhavyadhatuvrtti the form of Sayan.as Dhatupa.tha expected as basic
elementsinput to the rules of the As..ta dhyay. Scharf and Hyman (2011) present a comprehensive and unambiguous phonetic
encoding scheme for Sanskrit, including Vedic, designed for linguistic processing and discuss relevant theoretical issues of which
Scharf (2014b) presents a concise overview. Scharf (2013a) describes a database of the As..ta dhyay containing a detailed analysis of the text on the sutra, pada, and component levels. Petersen
(2013) describes the launch of another initiative to create a Pan.inian database. Finally, A. Mishra (2009b,a, 2010, 2014) presents
a comprehensive computational model for representing the grammatical processes of Pan.inis grammar step-by-step from the in-
168
troduction of basic elements to the constitution of the final expression.

The production of a complete computational implementation
of Pan.inian grammar is not a monolithic task. Scharf (2008a,b)
demonstrates that different Pan.inian commentators held different
interpretations of rules and the order of their implementation. The
interpretation of each of these commentators results in a different
linguistic description which would require an alternative implementation.
The formalization of Pan.inian grammar has implications beyond the technological. Attempts to formalize Pan.inian procedure
often uncover theoretical issues hidden in the complexity of the
grammar and left unattended even by two thousand years of indigenous grammatical commentators and two hundred years of modern
scholarship. Several such issues encountered in their current work
of formalizing the entire As..ta dhyay in an XML framework are
discussed by T. Ajotikar, A. Ajotikar, and Scharf (2015), and by
Scharf (2014a). The latter discusses a fundamental gap in the procedure to derive nominal stems ending in taddhita affixes which is
not explicitly commented upon and is mentioned only briefly by
Cardona (1970: 4748).
An accurate representation of the As..ta dhyay involves designing a method of dynamic rule selection so that rules are triggered
on the state of the derivation without requiring an explicit predetermined rule ordering. For, although metarules (paribhas.a s)
stated in the As..ta dhyay itself and collections of them in other
texts such as the Vyad.iparibhas.a vrtti offer various principles of
rule selection, the As..ta dhyay itselfcontains no explicit rule selection mechanism that governs derivations from start to finish. On
the contrary, various principles are operative in various domains.
These include the following:
169
1. the priority of an exception over a general rule (utsargapavadabhava),

2. a substitutes inheritance of attributes and reversion to the
status of its substituend (sthanivadbhava),
3. the persistence of the effects of an affix even after its deletion
(pratyayalope pratyayalaks.an.am),
4. the priority of internally conditioned operations over externally conditioned ones, or the suspension of externally conditioned ones with respect to internally conditioned ones
(antarangatva),
5. the priority of the operation that bleeds the other over the
operation that feeds the other where two operations apply to
the same state (nityatva),
6. the priority of the later rule where two rules, each of which
applies to an exclusive domain, share a domain (vipratis.edha),
7. the suspension of rules (asiddha), and of their effects (asiddhavat),
8. the overriding of rule suspension where a rule with respect
to which a rule is suspended depends upon the effects of the
would-be suspended rule (asrayatva),
9. the priority of a rule that otherwise would have no scope
(anavakasatva).
Determining which of the above principles applies under what circumstances is not trivial and has been the subject of extended discussion and debate. Cardona (1997: 5264, 401427 89104,
623665), explained how most of these principles operate. Cardona (1970) discussed the proper domains of application of the
rule selection principles 1, 4, 5, 6, and 9. Scharf (2010) described
an implementation of forward-looking conditionsconditions that
depend upon a state brought about later in the derivation. Scharf
(2013b) proposed interpreting locatives in rules that introduce ver-
170
bal stem-forming affixes (vikaran.a) as vis.ayasaptams and utilizing forward-looking conditions to implement them in order to simplify accentual derivation. Scharf (2002) examined how the role of
a speakers intention (vivaks.a ) desired by Pan.inian commentators
transcends what can be captured by the operation of vipratis.edha
(6) in the karaka section. Scharf (2011b) examined cases of vipratis.edha (6) throughout the grammar and cast doubt on their consistent solvability by simple rule-selection principles. Kiparsky
and Joshi (2006) proposed that Pan.ini assumed an extended siddha principle according to which trajectories through the rules
that maximize rule-interaction are favored. Cardona (2011) explained how the asiddhatva principle (7) is limited by the principle
of a srayatva (8).
A complete model of the As..ta dhyay involves working out a
model of rule priority by mathematically representing conflict resolution techniques. Subbanna and Varakhedi (2009) discuss and
suggest formalizations for serveral of the principles listed above
and present a mathematical model of conflict resolution techniques
as well as of the asiddha and asiddhavat principles. Goyal, A.
Kulkarni, and Behera (2009) developed a model for automatically
triggering a rule in a prakriya structure as well, deciding rule priority based on the concepts of rule blocking (1), and the suspension
of rules and their effects (7).
The present work improves upon existing models in several
ways. First, it proceeds along the derivation path with the full
accentual inforamtion. Second, rules are triggered automatically,
and rule ordering is not decided a priori. Third, the implementation
not only uses all the semantic and co-occurrence conditions but
also keeps track of these conditions in the data structure so that
the end result contains all the conditions under which a particular
form is derived. Finally, the model proposed in this paper keeps
track of the entire history of derivation in order to implement the
principles of sthanivadbhava (2), asiddhatva and asiddhavattva (7).
171
In the next section, we will discuss the steps taken to prepare the
input data for this implementation.
3
3.1
Preparing inputs for the implementation

XML Dhatupa.tha database
As is well known, Pan.inis grammar assumes the existence of certain ancilliary lists that are expected as essential components of
the grammar. Scharf (2009c: 97, Figure 1) presents a diagram
of these components. Components include lists of nominal bases
(pratipadikas) recited in the Gan.apa.tha and, most prominently,
a list of verbal roots recited in the Dhatupa.tha. When Pan.inis
grammar introduces verbal terminations, as described for example
in A. 3.2.123 (p. 158), it does so after an item termed dhatu verbal
root in accordance with the heading
A. 3.1.91 ;Da;a;ta;eaH
After a verbal root
which recurs thoughout the remainder of the third adhyaya. Items
beginning with bhu listed in the Dhatupa.tha are termed dhatu by
A. 3.3.1 BUa;va;a;d;ya;ea ;Da;a;ta;vaH
The speech forms beginning with bhu are termed dhatu.
Roots in the Dhatupa.tha are expected by Pan.inian rules to have
particular phonetic shapes with certain markers and accents. The
introduction to Scharfs Madhavya Dhatuvrtti canonical index
(Scharf 2009b) described in detail the characteristics

expected, the
obliteration of many of these characteristics in normalized editions
of the Dhatupa.tha, and how these characteristics have been restored to exactly the form in which they are expected by Pan.inian
rules in his canonical edition of the Madhavyadhatupa.tha the
172
form of the Dhatupa.tha commented upon in Sayan.as Madhavyadhatuvrtti. Our computational implementation of A. 1.3 utilizes
Scharfs
XML database of the canonical Madhavyadhatupa.tha as
its source of verbal roots. The information there required to derive
forms in this implementation includes the following:
the root with its markers, accents, and nasalization,
the sense or senses of the root, and
the gan.a of the root.
As an example of how information is provided for a root in this
database consider the entry for the root d.ukr karan.e.
<sUtra sid="8.12" page="510-518"

preds="aaxu" predsrc="8.13">
<entry id="1670" preds="aaxu">
<fullDAtu>qukfY</fullDAtu>
<lemma>
<premarker>qu</premarker>
<root>kf</root>
<marker>Y</marker>
</lemma>
</entry>
<sense senses="karaRa">
<senseterm>karaRe</senseterm>
</sense>
</sUtra>
The root is given in two forms, with its markers in the fullDAtu
element, and with its markers removed in the root element. The
latter is a child of the lemma element and its sibling elements
premarker and marker contain the initial and final markers
attached to the root respectively. The root and marker accent information is provided in the value of the sUtra elements attribute
preds (short for predicates). Here four sequential characters
record information explicitly stated in sutras in the Dhatupa.tha
that describe characteristics of roots in the sections these sutras
173
conclude. The characteristics described include the following four

attributes of roots:
1.
2.
3.
4.
the root accent (u: udatta, a: anudatta),

whether the root conditions the augment it. (s: set., a: anit.),
the marker accent (u: udatta, a: anudatta, s: svarita), and
whether the root conditions parasmaipada or a tmanepada
verbal terminations (p: parasmaipada, a: a tmanepada).
The first attribute directly determines the second, and the third attribute, along with consonant markers n and , directly determines
the fourth. The root accent is used solely for the purpose of determining whether the root conditions the augment it.; for all roots are
given high pitch when introduced in a derivation by
A. 6.1.162 ;Da;a;ta;eaH (;ntaH o+.d;a:aH 159)
The last vowel of a root is high-pitched.
Where the condition is vacuous, for instance where there is no
vowel marker and hence no marker accent, the value is marked
with an x. The pada information indicated applies only to the root
alone; it does not include its pada with preverbs or other particular coocurring terms, nor in special senses. It is usually only this
general information regarding pada that is included in general lexical and grammatical reference sources as mentioned in 1 above
(p. 160). The sense of the root is given in two forms as well. The
senseTerm element gives the term that descibes the sense in the
exact form in which it appears in the Dhatupa.tha, including inflection and compounding. The value of the sense elements attribute senses lists the stems (pratipadikas) of each sense term.
The information regarding each root provided in Scharfs Madhavyadhatupa.tha XML database is converted to a form in another XML database that conforms more closely to the initial form
expected by Pan.inis rules (upadesa) and the initial state of the
174
datastructure to be used in our implementation. The entry of the

root kr above is converted to the following form:
<entry page="510-518" sid="8.12" xml:id="1670">

<slp1>qukf\Y</slp1>
<morpheme gaRa="tanAdi" type="upadeSa">qukfY
</morpheme>
<phones>
<phone>q</phone>
<phone>u</phone>
<phone>k</phone>
<phone svara="anudAtta">f</phone>
<phone>Y</phone>
</phones>
<sense senses="karaRa">
<senseTerm>karaRe</senseTerm>
</sense>
</entry>
Here the root is given in its full form in three different ways. The
accent information given in the value of the preds attribute is
indicated explicitly in the Sanskrit Library Phonetic basic encoding (SLP1) in the content of the slp1 element. The backslash
marks the preceding r, indicated by an f, as low-pitched. The
morpheme element contains just the phonetic segments of the
full root without using attributes to mark accent or nasalization.
Where SLP1 utilized ASCII characters to indicate phonetic features as well as phonetic segments,1 the phones element explicitly separates phonetic attributes from phonetic segments: the latter are given as the content of phone elements, and the former
are indicated in the value of attributes of the element it qualifies.
Our implementation employs a structure that similarly segregates
segments from attributes. We specify a string of characters each
1 See
Scharf and Hyman 2011 for a detailed analysis of the linguistic principles of encoding, and consistent segmental and featural encoding systems for
Sanskrit in addition to the serviceable ASCII-only SLP1.
175
of which represents a single phonetic segment and list attributes

keyed to sequences of one or more contiguous segments in this
string.
3.2
Preverb cooccurences attested by Pan.ini
The next task was to find out which preverbs are attested with each
root. Since we wanted our implementation to reflect Pan.inis linguistic description, we set out to determine which preverbs cooccur with which root as attested by him. We accept that Pan.ini
attests that a preverb occurs with a root if either of the following
two conditions holds:
A Pan.inian sutra explicitly mentions the preverb as a cooccurrence condition of the root.
A Pan.inian sutra uses a term that demonstrates the cooccurrence.
As an example of the former, consider
A. 1.3.19 ;
a;va;pa:=+a;Bya;Ma .jeaH
An a tmanepada affix occurs after the root ji conquer when
it is used with either of the preverbs vi or para.
This rule explicitly describes that the preverbs vi and para are used
with the root ji. As an example of where Pan.ini uses a term that
demonstrates that a preverb cooccurs with a root, consider
A. 8.3.59 ;a;de ;Za;pra;tya;ya;ya;eaH (I+.k+:eaH 57, ;pa;d;a;nta;~ya 55, .saH 56,
mUa;DRa;nyaH 55, nua;a;}va;sa:
/ / jRa;na;a;ya;Za;v.yRa;va;a;yea 58, ;
a;pa 58)
An s that is a replacement (adesa) or that belongs to an affix
(pratyaya), which follows a vowel of class i, u, r, or l and is
retroflex
non-final in a pada is replaced by its corresponding

sound (s.) even if the augment num, a visarga, or a sibilant
intervenes.
176
Here the sutra uses the form a desa, which is derived from the root
dis preceded by the preverb a , and the form pratyaya, which is
derived from the root i preceded by the preverb prati. Pan.inis
use of these derivates attests that in the language known to him the
preverb a was used with the root dis and the preverb prati was used
with the root i.
After collecting all the evidence of the cooccurences of preverbs with roots attested in the As..ta dhyay, we then had to determine with which particular root in the Dhatupa.tha to associate the
roots mentioned in the As..ta dhyay. For example, for the root ji
mentioned in A. 1.3.19, there are two possible candidates in the
Dhatupa.tha: ji jaye and ji abhibhave. For the root i in the derivate
pratyaya in A. 8.3.59, there are three possible candidates: in. gatau,
in adhyayane, and ik smaran.e. Likewise, in the sutra
A. 1.3.63 ;a;}.pra;tya;ya;va;tkx+:Va;ea Y;nua;pra;ya;ea;ga;~ya (k+:tRa;a:= 14, ;a;tma;nea:pa;d;m,a 12)

An a tmanepada affix occurs to denote the agent after the
root kr as it would after the root after which the affix a m
introduced and which kr follows in accordance
has been
with A. 3.1.40.
In this example, it was to be decided whether the root kr mentioned here corresponds with the root d.ukr karan.e or theroot kr
to which precise root
or
hims
ayam. The information concerning
roots a rule refers and with which precise roots a preverb is associated has to be provided to the system to avoid undesired ambiguities. We permitted roots mentioned or used in the As..ta dhyay to be
associated by default with whichever Dhatupa.tha roots they share
the phonetic form, markers, semantics, and cooccurrence conditions the As..ta dhyay mentions. In the case of derivates such as
pratyaya, etc., this means we associate them just on the basis of
the phonetic shape. However, where we have good reason to make
177
a more specific judgement, we do so, as for example in the case of

the root kr in anuprayoga, where we associate it solely with the
root d.ukr
karan.e.
Implementation details
Although the implementation was done with the specific purpose

of finding the pada information for various roots along with the exact conditions, the model was general enough to be applied to any
other section of Pan.inis As..ta dhyay. We describe various components of the implementation below.
4.1
Data Structure
For each stage of the derivation, the data structure includes a phonetic string in SLP1, minus any modifier characters that would
indicate phonetic features, and a list of attributes keyed to segments in the phonetic string. Attributes refer to the phonetic string
by indicating the offset of the segment in the complete string
which we term the start, and the length of the string. We refer to a derivation path using the term prakriya. At the ith point
in a derivation, ( f orm, list att)i denotes the status of the prakriya,
where att = (attribute, start, length) stores the relative position of
attribute in the string form. For example, the upadesa stage
(prakriy
a0 ) for the root d.ukr will be stored using the following
data structure:
prakriy
a0 = (d.ukr, {(gan
.a = tanadi | bhvadi, 0, 5),
(svara=anudatta, 3, 1),
(sense = karan.a, 0, 5)}
which gives the information that the phonetic form d.ukr has the
attributes gan
.a=tanadi|bhvadi and sense=karan.a for the com-
178
plete string (starting at index 0 and having length 5), while the attribute svara=anudatta belongs only to the root vowel r (starting
at index 3 and having length 1).
4.2
Rule formulation
The rules were formulated such as to divide the rule conditions

into various attributes which can be dealt with differently by the
program. The main attributes used for formulating a rule are described in the following subsections.
4.2.1
Phonetic conditions
The phonetic conditions expressed by a sutra are coded in this attribute using regular expressions. For example, the phonetic condition tusmah. specified in
A. 1.3.4 na ;
a;va;Ba;+:Ea tua;sma;aH (o+.pa;de ;Zea 2, h;l+.ntya;m,a 3, I+.t,a 2)
A dental stop, s, or m, occuring final in a vibhakti when
originally taught, is not termed it.
is encoded as phone-cond=([tTdDnsm]), where the phrase
[tTdDnsm] is a regular expression meaning any of the sounds t,
th, d, dh, n, s, or m.
4.2.2
Attribute conditions
These denote attributes in the prakriya including major and

subordinate classes of technical terms which we call saj
a, and varga, and a type attribute to distinguish
states of the prakriya. For example, the attribute conditions
for A. 1.3.4 are coded as att-cond = type=upadesa,saj
a=pratyaya,varga=vibhakti.

4.2.3
179
Semantic conditions
These conditions may or may not coincide with semantic conditions stated as meanings of roots in the Dhatupa.tha but might instead refer to shades of meaning that emerge in particular contextual uses. For example, the Dhatupa.tha lists the root stha in the
meaning cessation of locomotion (MDhP. 1.650 s.tha gati-nivr
ttau). A contextual meaning is stated as a semantic condition for
the use of a tmanepada terminations after this root by A. 1.3.23
(p. 162). The semantic conditions are coded as sem-cond =
prayoga=kartr,meaning=prakasana|meaning=stheyakhya.
4.2.4
Optionality
If a rule is optional, this information is encoded using a boolean

attribute optional=true in the rule structure.
4.2.5
Action
This attribute stores the operation prescribed by the rule. For instance, for the example A. 1.3.23 above, the operation is coded as
action = pada=atmane.
4.3
Control structure
The control structure of the implementation is depicted in Figure 1.

The starting point is the upadesa state, which we denote as prakriy
a0 . Rules are coded in the structure described in 4.2 and
are presented in the order in which they appear in Pan.inis grammar. The block Match conditions matches prakriy
ai for
the conditions specified in various rules, and the rules that match
the conditions are stored in the set triggered rules. The
rule priority principle is used to decide which among the triggered
rules is the winner rule. The winner rule is then applied
180
to prakriy
ai to obtain the next state prakriy
ai+1 . This process is applied recursively until no rules are triggered, which is the
termination condition and we obtain the final result.
Start
prakriya0
Match
Conditions
triggered
Rule 1
Rule 2
.
.
.
.
Rule n
rules
Rule
Priority
Rule
Exceptions
winner
rule
prakriyai
Apply
Rule
prakriyai+1
Figure 1
The control structure for the implementation.
This process is applied recursively until no rules are triggered.
In our implementation, we explicitly created pairs of rules
181
{rule1 , rule2 } such that rule1 forms an exception to rule2 . These

rule pairs are provided to the control structure. The resulting rule
priority function is thus quite straightforward.
The implementation proceeded along the following path:
1. Roots were taken from the Dhatupa.tha with their markers,
marker-like root accent, and marker accent.
2. Rules in section A. 1.3.29 of the As..ta dhyay were implemented to identify and delete markers to obtain the bare
roots.
3. A. 6.1.162 (p. 173) was applied to give the actual initial accent to root vowels.
4. The voice assignment for the roots was determined by implementing the rest of section A. 1.3 for all the roots.
In the next section, we will discuss the derivation trees produced
by the implementation and give examples for some of the roots.
5
5.1
Results and related discussions

Notations in the derivation tree
Let us first describe the notations used in the derivation tree by way
of an example. Figure 2 shows the derivation tree produced for the
root bhu. The nodes correspond to various states of the derivation.
Three different colors have been used as backgrounds for three different stages of the derivation. The upadesa state corresponds to
the top node of the tree in the figure which is shown with a bisque
background. The final states of the prakriya are shown with nodes
having rose background. The number of final states is equivalent
to the number of leaf nodes in the tree. In our current implementation, a final state corresponds to the state when a pada is assigned
to the root under certain conditions. All intermediate states of the
derivation are shown with nodes having olive background.
182
Inside each node, we first display the phonetic form of the output in Devanagar script with accentual information marked with
a vertical bar above a high-pitched vowel. Then we display the
attributes associated with that state of the derivation. For the upadesa state, all the attributes are shown, but for all subsequent states,
only those attributes that differ from the parent node are shown in
order to avoid cluttering the tree. A parent and a child node are
connected with an arrow going from the parent to the child node.
This arrow corresponds to the application of a certain rule and is
marked with the number of that rule. Again, two different colors
have been used to label the arrows. A rule is written with red color
if the rule has been applied on that particular path. However, if an
additional path or paths result from the implementation of a rule,
the rule number on the additional paths on which the rule did not
apply is written with blue color.
Let us explain the colors on the arrow labels with reference
to Figure 2. Consider A. 1.3.13. This rule gets applied to assign
pada=atmane to the root bhu, and the corresponding final state is
shown in rose color with the arrow labeled 1-3-13 in red. However, prayoga=bhava|karman indicates semantic conditions, but
since these are not exhaustive, we have to account for the derivation tree corresponding to prayoga=kartr as well. This prayoga is non-overlapping with the other two
prayogas mentioned in
A. 1.3.13. Therefore we find a bifurcation at the application of this
rule, with the other branch being labeled with the same rule but
in blue color. Note that this branch corresponds to the semantic
condition prayoga=kartr. A similar phenomenon occurs with
where three paths result due to comthe application of A. 1.3.16,
plementary semantic conditions and complementary cooccurrence
conditions at the same point.
The derivation tree for the root bhu contains four leaf nodes
denoting the final states. For any of these leaf nodes, the set of
final attributes can be obtained by tracing the derivation path from
183
Figure 2
Derivation tree for the root bhu
the root node to the leaf node. The derivation path is unique for
any node. We start from the attributes of the root node and traverse
down the unique derivation path from the root node to the leaf
node. With each node, we keep on adding the new attributes to the
existing list of attributes.
Now that we have discussed the graph notations, we will discuss certain issues that we faced in the computational implementation of this particular section of Pan.inis grammar and provide an
example demonstrating how we handled each issue.
184
5.2
Handling meaning conditions
Throughout his grammar, Pan.ini includes various semantic conditions in his sutras. As mentioned above (2), Scharf (2009c: 95
111) discussed these conditions in some detail. In a comprehensive
computational implementation, one needs to handle all possibilities, not just the mentioned ones; thus whenever a rule mentions
semantic conditions, action required for the complementary condition also needs to be worked out. As stated earlier (5), semantic
conditions were treated as bifurcations in our implementation; that
is, whenever a rule ordains a semantic condition, two branches of
the derivational tree will be created: one with the semantic condition, and one without it.
Consider the case of the root da shown in Figure 3 and the
application of A. 1.3.20 for example.
A. 1.3.20 ;a;z+.ea d;ea Y;na;a;~ya;
a;va;h:=+ea (k+:tRa;a:= 14, ;a;tma;nea;pa;d;m,a 12)
An a tmanepada termination is selected after the root da used
with the preverb a when the meaning to open the mouth
(asya-viharan.a) is not to be understood.
The arrow labeled 1-3-20 in red in the lower left in Figure 3
shows the application of this rule where the meaning condition is
satisfied. The terminal rose-colored oval shows the negation of
meaning=;a;~ya-;
a;va;h:=+a and the assignment :pa;d=;a;tma;nea. To handle the complementary semantic condition in the case when a syaviharan.a is to be understood, there is a bifurcation. Thus there
is another path labeled with the same rule number 1-3-20 in
blue pointing to an oval containing meaning=;a;~ya-;
a;va;h:=+a. The
derivation tree also shows a third path between these two which
corresponds to complementary coocurrence conditions, namely,
the cases where the root da is used without a preverb or with preverbs other than a alone. The right and left paths for A. 1.3.20 deal
only with the cases in which the root is used with the preverb a
185
as shown by the condition o+.pa;sa;gRa=;a written in the top line of the

ovals to which the arrows labeled 1-3-20 point.
5.3
Handling rule interaction
Determining how rules interact where more than one rule has
scope within the same domain is one of the most significant challenges to the computational implementation of Pan.inis grammar.
While all rules in this section are restrictions to A. 3.4.78 (p. 158),
the relations of being an exception (apavada), a negation (pratis.edha), and a remainder (ses.a) are also found. Principal among
these is the relation of being an exception. Of the eighty-two rules
in the section, twenty are exceptions, seven are negations, and one
is a remainder.
A. 1.3.78 (p. 162) is the remainder rule. This rule requires
that a parasmaipada affix be selected to denote the agent (kartr)
if none of the rules that provide an a tmanepada affix to denote

an agent, namely A. 1.3.12 and A. 1.3.1477, applies. Consider
the example of the root gam whose derivation tree is shown in
Figure 4. The two bottom rose-colored terminal nodes have arrows
labeled 1-3-78 indicating the application of the remainder rule.
Similar are the left and right bottom rose-colored terminal nodes
in Figure 2 and the bottom rose-colored terminal nodes in figures
3, 5, and 6. These nodes occur at the bottom precisely because
A. 1.3.78 is determined to apply only after considering all other
rules in the section, particularly those to which it serves as the
remainder.
A. 1.3.1516 (p. 163) are negations of A. 1.3.14 (p. 162). On
the branch marked with the label 1-3-13 in blue in Figure 4,
below the oval labeled :pra;ya;ea;ga=k+:tRxa, two rules, A. 1.3.14 (p. 162) and
A. 1.3.15, get triggered. Because the root gam occurs in the sense
of motion, A. 1.3.15 applies negating A. 1.3.14 as shown by the
label 1-3-15 in red at this point in the tree.
186
A. 1.3.15 na ga;a;ta;

a;hM ;sa;a;TeRa;ByaH (k+:tRa;a:= k+:mRa;v.ya;a;ta;h;a:=e 14, ;atma;nea;pa;d;m,a 12)
An a tmanepada termination is not used after roots having
the sense of motion or injury, when the action is mutual
and the (kartr) is to be denoted.
Rule1 is an exception to and overrides rule2 if rule2 would apply in the absence of rule1 . A. 1.3.83 is an exception to A. 1.3.12
(p. 161). Because the root rama krd.a yam is listed in the Dhatupa
.tha with a low-pitched marker vowel a, as shown in the first line of
the initial oval at the top of Figure 4, A. 1.3.12 (p. 161) would apply to select an a tmanepada termination. However, when the root
occurs with one of the three preverbs vi, a , or pari and an agent is
to be denoted, A. 1.3.83 applies to select a parasmaipada termination. The application of this rule is shown in the leftmost branch in
Figure 7, labeled 1-3-83, and the resulting attributes are shown
in the rose-colored oval that terminates that branch.
A. 1.3.83 v.ya;a:*: +.a:=+Bya;ea .=+maH (k+:tRa;a:= :pa:=+smEa;pa;d;m,a 78)
A parasmaipada termination is selected after the root ram
used with one of the preverbs vi, a , or pari.
As stated before (4.3), in our current implementation, we explicitly provide the system with information concerning rule exceptions. The file that pairs rules with exceptions handles negations and exceptions in exactly the same manner: the negation or
exception is selected as the winning rule over the negated or excepted rule.
When an exception applies, it applies to a domain that is a subset of the domain of the general rule to which it is an exception. In
the remainder of the domain of the general rule, the general rule
applies instead except where other exceptions carve out other domains. Hence an exception creates a bifurcation of the derivational
tree. Consider the derivation path of root ks.ip, as shown in Figure
5. Rules 1.3.72 and 1.3.80 get triggered during the derivation.
187
A. 1.3.72 .~va;a:=+ta;a;Va;taH k+.Ra;a;Ba;pra;a;yea ;

a;k+:ya;a;P+.le (k+:tRa;a:= 14, ;a;tma;nea;pa;d;m,a 12)
An a tmanepada termination is selected after a verb marked
with a svarita vowel or with n when the fruit of the action
accrues to the agent.
A. 1.3.80 ;a;Ba;pra;tya;a;ta;ByaH ; a:a;paH (k+:tRa;a:= 78, :pa:=+smEa;pa;d;m,a 78)
A parasmaipada termination is selected after the root ks.ip
used with one of the preverbs abhi, prati, or ati, even though
the fruit of the action accrues to the agent.
While A. 1.3.80 is an exception to A. 1.3.72, it applies only when
this root occurs with any of the preverbs abhi, prati, or ati and
not otherwise. Thus this exception applies only for the subset of
the derivational tree shown with the arrow labeled 1-3-80 in red
that terminates in the rose-colored oval containing o+.pa;sa;gRa=;a;Ba|:pra;a;ta|;a;ta. The complementary set for which A. 1.3.80 is not triggered is handled by another branch beginning with the arrow labeled 1-3-80 in blue pointing to the olive-colored oval containing o+.pa;sa;gRa=none|;va|:pa;a:=. On this branch, the general rule to which
A. 1.3.80 is an exception, namely, A. 1.3.72, applies, as is shown
by the arrow labeled 1-3-72 in red. In turn in the complementary
case when this rule does not apply, namely, when the fruit of the
action does not accrue to the agent, the complementary branch is
shown with the arrow labeled 1-3-72 in blue, pointing to the oval
containing ;a;Ba;pra;a;ya=;nya;a;a;Ba;Bra;a;ya. Finally this branch terminates in
the remainder rule A. 1.3.78.
Now returning to the derivation of the root ram shown in Figure 7, where A. 1.3.83 in exception to A. 1.3.12 creates a bifurcation of the tree, the arrows labeled 1-3-83 in blue show two paths
on which A. 1.3.83 does not apply, one for the complement of each
of the conditions included in A. 1.3.83. Notice that each of these
branches leads to a branch on which the general rule A. 1.3.12 applies. The middle branch is complicated by the fact that A. 1.3.84
188
(p. 189), which itself has an exception, is also an exception to

A. 1.3.12.
5.4
Handling praptavibhas.a
A rule is called a praptavibhas.a if it optionally prescribes an operation already prescribed by another rule. A praptavibhas.a is thus
an exception to the rule that already prescribed the operation. Consider how A. 1.3.77 interacts with A. 1.3.72 (p. 187).
A. 1.3.77 ;
a;va;Ba;a;Sa;ea;pa;pa;de ;na :pra;ta;a;ya;ma;a;nea (k+.Ra;a;Ba;pra;a;yea ;
a;k+:ya;a;P+.le 72,
k+:tRa;a:= 14, ;a;tma;nea;pa;d;m,a 12)
An a tmanepada termination is only optionally used when
the fact that the fruit of the action accrues to the agent is
indicated by an upapada.
A. 1.3.77 is a praptavibhas.a in exception to 1.3.72.
Consider the derivation tree for the root kr in Figure 6. There
are two branches corresponding to the application of A. 1.3.77. On
the branch with the arrow labeled 1-3-77 in red, we apply the
option that voids the application of A. 1.3.77, just as we would any
exception. On the other branch, A. 1.3.72 gets applied, just as the
general rule does in the domain where an exception does not apply.
Note that A. 1.3.72 applies only on the subset of the derivational
tree with the rose-colored oval containing ;a;Ba;pra;a;ya=k+.Ra;a;Ba;pra;a;ya and
not under the complementary condition shown in the olive-colored
oval containing ;a;Ba;pra;a;ya=;nya;a;a;Ba;pra;a;ya. The additional bifurcation
occurs because the domain of A. 1.3.72, to which A. 1.3.77 is an
exception, itself applies to a subset of the semantic space to which
the remainder rule A. 1.3.78 applies.
Now let us look at the praptavibhas.a that occurs in the derivation tree of the root ram shown in Figure 7. Consider the relation
of A. 1.3.85 (p. 163) with its preceding rule:
189
A. 1.3.84 o+.pa;a;a (.=+maH 83, k+:tRa;a:= 78, :pa:=+smEa;pa;d;m,a 78)

A parasmaipada affix is selected after the root ram used with
the preverb upa.
A. 1.3.85 is a praptavibhas.a in exception to A. 1.3.84.
There are two branches corresponding to the application of
A. 1.3.85, and the handling is exactly the same as in the case of
kr. On the branch with the arrow labeled 1-3-85 in red, we
apply the option that voids the application of A. 1.3.84. On the
complementary branch with the arrow labeled 1-3-85 in blue,
A. 1.3.84 applies on the subordinate branch with the arrow labeled
1-3-84 in red. Additional branches with arrows labeled 1-3-84
in blue occur because A. 1.3.84 is applicable only to a subset of the
semantic and cooccurence conditions.
Conclusions and Future Work
In this paper, we discussed the computational implementation of

As..ta dhyay 1.3 which deals with the voice assignment for roots
under various semantic and co-occurrence conditions. The present
work stands out in that it courses through all possible derivation
paths and thus presents an exhaustive table that associates the derived form with the conditions for rules and attributes assigned by
those rules. This information will be beneficial for computational
tools dealing with morphological generation and analysis as well
as with dependency parsing.
Future work involves evaluation of the correspondence of the
computational implementation to the coocurrences found in a
tagged corpus of the Pacakhyanaka and Mahabharata sentences.
The simulation described here is currently being extended in the
development of a complete computational model of the As..ta dhyay as described by Scharf (2015).
190
Figure 3
Derivation tree for the root da
Figure 4
Derivation tree for the root gam
191
192
Figure 5
Derivation tree for the root ks.ip
Figure 6
Derivation tree for the root kr
193
194
Figure 7
Derivation tree for the root ram
R EFERENCES
195
References
Ajotikar, Tanuja, Anuja Ajotikar, and Peter M. Scharf. 2015.
Some issues in the computational implementation of the As.t.a dhyay. Proceedings of the Sanskrit and the IT World section of 16th World Sanskrit Conference, ed. by Amba Kulkarni
and Vineet Chaitanya. New Delhi: D. K. Publishers.
Bhtlingk, Otto von and Rudolf von Roth. 18551875. SanskritWrterbuch. 7 vols. St. Petersburg: Kaiserlichen Akademie der
Wissenschaften.
Cardona, George. 1965. On translating and formalizing Pan.inian
rules. Journal of the Oriental Institute of Baroda 14: 30614.
. 1969. Studies in Indian grammarians I: the method of descrip
tion reflected in the Sivas
utras. Transactions of the American Philosophical Society, new series 59.1. Philadelphia: The
American Philosophical Society.
. 1970. Some principles of Pan.inis grammar. Journal of Indian Philosophy 1: 4074.
. 1974. On Pan.inis metalinguistic use of cases. Charudeva
Shastri felicitation volume: presented to Prof. Charudeva
Shastri on the occasion of his seventy-fifth anniversary by his
friends and admirers, ed. by S. K. Chatterji et al., pp. 30526.
Delhi: Charudeva Shastri Felicitation Committee.
. 1997. Pan.ini: his work and its traditions; vol. 1, Background
and Introduction. Second edition, revised and enlarged. Delhi:
. 2011. Purvatrasiddham and a srayat siddham. Studies in
Sanskrit grammars: proceedings of the Vyakaran.a section of
the 14th World Sanskrit Conference, 15 September 2009, Kyoto University, Kyoto, ed. by George Cardona, Ashok Aklujkar, and Hideyo Ogawa, pp. 12362.
Cardona, George, Ashok Aklujkar, and Hideyo Ogawa, eds. 2011.
Studies in Sanskrit grammars: proceedings of the Vyakaran.a
196
section of the 14th World Sanskrit Conference, 15 September

2009, Kyoto University, Kyoto. New Delhi: D. K. Printworld.
Gillon, Brendan S. 2007. Pan.inis As..ta dhyay and linguistic theory. Journal of Indian Philosophy 35: 44568.
Goyal, Pawan and Grard Huet. 2013. Completeness analysis
of a Sanskrit reader. Proceedings of the Fifth International
Sanskrit Computational Linguistics Symposium, ed. by Malhar
Kulkarni and Chaitali Dangarikar, pp. 13071.
Goyal, Pawan, Amba Kulkarni, and Laxmidhar Behera. 2009.
Computer simulation of As.t.a dhyay: some insights. Sanskrit
computational linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI,
USA, May 2008; Revised selected and invited papers, ed. by
Grard Huet, Amba Kulkarni, and Peter M. Scharf, pp. 139
61. Lecture Notes in Artificial Intelligence 5402.
Huet, Grard, Amba Kulkarni, and Peter M. Scharf, eds. 2009.
Sanskrit computational linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI, USA, May 2008; Revised selected and invited papers. Lecture Notes in Artificial Intelligence 5402. Berlin; Heidelberg: Springer-Verlag.
Hyman, Malcolm D. 2009. From Pan.inian sandhi to finite state
calculus. Sanskrit computational linguistics: first and second
international symposia, Rocquencourt, France, October 2007;
Providence, RI, USA, May 2008; Revised selected and invited
papers, ed. by Grard Huet, Amba Kulkarni, and Peter M.
Scharf, pp. 25365. Lecture Notes in Artificial Intelligence
5402.
R EFERENCES
197
Jha, Girish Nath, ed. 2010. Sanskrit computational linguistics: 4th

International Symposium, New Delhi, India, December 2010,
Proceedings. Lecture Notes in Artificial Intelligence 6465.
Berlin; Heidelberg: Springer-Verlag.
Jha, Girish Nath and Sudhir K. Mishra. 2009. Semantic processing in Pan.inis karaka system. Sanskrit computational linguistics: first and second international symposia, Rocquencourt,
France, October 2007; Providence, RI, USA, May 2008; Revised selected and invited papers, ed. by Grard Huet, Amba
Kulkarni, and Peter M. Scharf, pp. 23952. Lecture Notes in
Artificial Intelligence 5402.
Kiparsky, Paul. 2009. On the architecture of Pan.inis grammar.
Sanskrit computational linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI, USA, May 2008; Revised selected and invited papers, ed. by Grard Huet, Amba Kulkarni, and Peter M. Scharf,
pp. 3394. Lecture Notes in Artificial Intelligence 5402.
Kiparsky, Paul and S. D. Joshi. 2006. The extended siddhaprinciple. Annals of the Bhandarkar Oriental Research Institute 2005: 126.
Kulkarni, Malhar. 2009. Phonological overgeneration in Pan.inian
system. Sanskrit computational linguistics: first and second
international symposia, Rocquencourt, France, October 2007;
Providence, RI, USA, May 2008; Revised selected and invited
papers, ed. by Grard Huet, Amba Kulkarni, and Peter M.
Scharf, pp. 30619. Lecture Notes in Artificial Intelligence
5402.
Kulkarni, Malhar and Chaitali Dangarikar, eds. 2013. Proceedings
of the Fifth International Sanskrit Computational Linguistics
Symposium. New Delhi: D. K. Printworld.
Mishra, Anand. 2009a. Modelling the grammatical circle of the
Pan.inian system of Sanskrit grammar. Sanskrit computational
linguistics: third international symposium, Hyderabad, India,
198
January 2009, proceedings, ed. by Grard Huet and Amba

Kulkarni, pp. 4055. Lecture Notes in Artificial Intelligence
5406.
. 2009b. Simulating the Pan.inian system of Sanskrit grammar.
Sanskrit computational linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI, USA, May 2008; Revised selected and invited papers, ed. by Grard Huet, Amba Kulkarni, and Peter M. Scharf,
pp. 12738. Lecture Notes in Artificial Intelligence 5402.
. 2010. Modelling As.t.a dhyay: an approach based on the
methodology of ancillary disciplines (vedan ga). Sanskrit
computational linguistics: 4th International Symposium, New
Delhi, India, December 2010, Proceedings, ed. by Girish Nath
Jha, pp. 23958. Lecture Notes in Artificial Intelligence 6465.
. 2014. A formal re-presentation of the As.t.a dhyay. Unpublished.
Monier-Williams, Monier, Ernst Leumann, and Carl Cappeller.
1899. A Sanskrit-English dictionary: etymologically and
philologically arranged with special reference to cognate
Indo-European languages: new edition, greatly enlarged and
improved. Oxford: The Clarendon Press.
Oberlies, Thomas. 2003. A grammar of Epic Sanskrit. Indian
Philology and South Asian Studies. Berlin: Walter de Gruyter.
Patel, Dhaval and Shivakumari Katuri. 2014. Prakriyapradars.in.: an open source subanta generator. URL: http : / /
lanover.com/lan/sanskrit/subanta.html.
Petersen, Wiebke. 2013. Structure and implementation of a digital edition of the As.t.a dhyay. Proceedings of the Fifth International Sanskrit Computational Linguistics Symposium, ed.
by Malhar Kulkarni and Chaitali Dangarikar.
Satuluri, Pavan Kumar and Amba Kulkarni. 2013. Generation of
Sanskrit compounds. Proceedings of ICON 2013, the 10th International Conference on NLP, pp. 7786.
R EFERENCES
199
Scharf, Peter M. 2002. Pan.ini, vivaks.a , and karaka-ruleordering. Indian linguistic studies: festschrift in honour of
George Cardona, ed. by Madhav Murlidhar Deshpande and
Peter Edwin Hook, pp. 12149. Delhi: Motilal Banarsidass.
. 2008a. Pan.inian accounts of the class eight presents. Journal
of the American Oriental Society 128.3: 489504.
. 2008b. Pan.inian accounts of the Vedic subjunctive: let. krn.va
te. Indo-Iranian Journal 51.1: 121.
. 2009a. Levels in Pan.inis As..ta dhyay. Sanskrit computational linguistics: third international symposium, Hyderabad,
India, January 2009, proceedings, ed. by Grard Huet and
Amba Kulkarni. Lecture Notes in Artificial Intelligence 5406.
. 2009b. Madhavya Dhatuvrtti canonical index. URL: http:
//sanskritlibrary.org.
. 2009c. Modeling Pan.inian grammar. Sanskrit computational
linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI, USA, May 2008;
Revised selected and invited papers, ed. by Grard Huet, Amba
Kulkarni, and Peter M. Scharf. Lecture Notes in Artificial Intelligence 5402.
. 2010. Rule-blocking and forward-looking conditions in the
computational modeling of Pan.inian derivation. Sanskrit
Jha. Lecture Notes in Artificial Intelligence 6465.
. 2011a. On the semantic foundation of Pan.inian derivational
procedure: the derivation of kumbhakara. Journal of the
American Oriental Society 131.1: 3972.
. 2011b. Rule selection in the As..ta dhyay or Is Pan.inis grammar mechanistic? Studies in Sanskrit grammars: proceedings
of the Vyakaran.a section of the 14th World Sanskrit Conference, 15 September 2009, Kyoto University, Kyoto, ed.
200
by George Cardona, Ashok Aklujkar, and Hideyo Ogawa,

pp. 31950.
. 2013a. An analytic database of the As..ta dhyay. Proceedings
Symposium, ed. by Malhar Kulkarni and Chaitali Dangarikar.
. 2013b. Teleology and the simplification of accentuation in
Pan.inian derivation. Proceedings of the Vyakaran.a Section of
the 15th World Sanskrit Conference, ed. by George Cardona,
pp. 3153. New Delhi: D. K. Printworld.
. 2014a. Are taddhita affixes provided after pratipadikas or padas? Paper presented at the Troisime Atelier du Projet ANR
PP1617 (Pan.ini et les Pan.inens des XVIe XVIIe sicles)
accueilli par EFEO, IFP, EPHE, Pondichry, 1416 octobre
2014. In preparation.
. 2014b. Linguistic issues and intelligent technological solutions in encoding Sanskrit. Document numrique 16.3: 15
29. Paper presented at Gestions informatise des critures anciennes: tat des lieux & perspectives, 2122 May 2013, Centre dtudes Suprieures de la Renaissance (CESR), Universit
Franois-Rabelais, Tours.
. 2015. An XML formalization of the As..ta dhyay. Proceedings of the Sanskrit and the IT World section of 16th World
Sanskrit Conference, ed. by Amba Kulkarni and Vineet Chaitanya. New Delhi: D. K. Publishers.
Scharf, Peter M. and Malcolm D. Hyman. 2011. Linguistic issues
in encoding Sanskrit. Delhi: Motilal Banarsidass.
Staal, J. Frits. 1965. Context-sensitive rules in Pan.ini. Foundations of Language 1: 6372.
Subbanna, Sridhar and Srinivas Varakhedi. 2009. Computational
structure of the As..ta dhyay and conflict resolution techniques.
Sanskrit computational linguistics: third international symposium, Hyderabad, India, January 2009, proceedings, ed. by
R EFERENCES
201
Grard Huet and Amba Kulkarni, pp. 5665. Lecture Notes

in Artificial Intelligence 5406.
Swamiji, Shivamurthy. 2010. Gan.akas..ta dhyay. URL: http://
www.taralabalu.org/panini/.
Whitney, William Dwight. 1885. The roots, verb-forms, and primary derivatives of the Sanskrit language: a supplement to his
Sanskrit grammar. Leipzig: Breitkopf & Hartel.
. 1945. The roots, verb-forms, and primary derivatives of the
Sanskrit language: a supplement to his Sanskrit grammar.
American oriental series 30. New Haven: American Oriental
Society. [Reprint of Whitney 1885.]
202
Interrogatives and word-order

in Sanskrit
P ETER M. S CHARF
Abstract: Recent linguistic theory implies that subject
objectverb (SOV) languages place interrogatives at the
end of a sentence. This implies that it would hold true for
Sanskrit because it is an SOV language. It doesnt. Evaluation of the position of interrogatives in prose portions of
the Vis.n.upuran.a reveals that interrogatives locate in positions of focus and topic just as Hock (1989) showed relative pronouns do. Their position is thus influenced by discourse factors rather than grammatical relations. The analysis confirms the unimportance of word-order as compared
with inflection in embodying syntactic relations. The fact
that position is not fundamental to syntactic relations in
turn suggests that syntactic analysis be represented in an
abstract space that permits representation of more complex
structure than is permitted in the approach that imbues position in the one-dimensional space of linear word-order
with fundamental syntactic significance.
Keywords: Sanskrit, linguistics, word order, interrogative,
pronoun
The problem
At a Conference on Language Universals in 1961, Joseph Greenberg (1966) reported significant correlations in the ordering of constituents in various types of constructions within a language. Investigating a diverse sample of thirty of the worlds languages, he
203
204
P. S CHARF
observed, for instance, that the following sequences are mutually

correlated: objectverb, genitivenoun, adjectivenoun, adverb
verb, and nounpostposition and that these are correlated also with
the sentence-final occurrence of an interrogative particle in general
(i.e. yes/no) questions and with in situ occurrence of an interrogative in special (i.e. wh-) questions.
SOV/Postposition/GN/AN/AdvV/Q-final/SpQ-in situ
(1)
Conversely, the opposite sequences verbobject, noungenitive,

nounadjective, verbadverb, and prepositionnoun are correlated
with each other and with the occurrence of a sentence-initial interrogative particle in general (i.e. yes/no) questions and sentenceinitial interrogative in special (i.e. wh-) questions.
SVO/Preposition/NG/NA/VAdv/Q-initial/SpQ-initial
(2)
While Greenberg (1966: 104) considered the principles described ... as no more than suggestive, Theo Vennemann (1974:
80) formally stated two generalizations:
1. Any given operatoroperand relationship tends to be serialized unidirectionally in a language.
2. All operatoroperand relationships tend to be serialized
in the same direction in a language, namely that direction
which is defined by the serialization of object and verb.
He (1974: 81) then formulated these generalizations in the principle of serialization shown in equation 3,
[Operator[Operand]]
in OV languages
(3)
{Operator{Operand}} =>
[Operand[Operator]]
in VO languages
I NTERROGATIVES AND WORD - ORDER IN S ANSKRIT
205
noting that the objectverb relationship is itself an operator

operand relationship, with the object as operator and the verb as
operand.
Now Vennemann does not explicitly include interrogatives in
his discussion. However, Jackendoff (1977: 85) generalizes serialization over all phrase types stating, the evaluation measure for
phrase structure rules counts parallelisms of word order across Xn
as generalizations (where X is instantiated as a phrase type and
n as one of the three levels in his analysis). Comrie (1988: 454)
summarizes this development as follows:
Within X-grammar,
the basic schema specifying the
structure of different kinds of phrases (e.g. noun
phrase, verb phrase, prepositional phrase) is the same,
so that languages like Japanese have a basic phrasestructure rule specifying that the head is preceded
by its adjuncts [complement], while Welsh has the
mirror-image rule, specifying that the head is followed by its adjuncts [complements]. Vennemanns
terms operator and operand can simply be replaced
by adjunct [complement] and head, and languages
are either head-final or head-initial.
Chung and McCloskey (1987: 201) describe the serialization
of constituents in various relationships such as that of government
in their examination of Irish:
Much work (...) leads to the conclusion that government and other relationships such as Case assignment,
may be constrained by directionality conditionsthat
in a given language, for instance, governors must be
to the left of the phrases they govern, or Case assigners to the left of the NP to which they assign Case.
Now Irish is a massively regular head-initial language
206
P. S CHARF
[VSO p. 211] and in every instance of the government relation known to us the governor precedes the
governed phrase.
Recent work in generative grammar typically includes complementizer phrases (phrases such as subordinate clauses) among
phrases subject to the serialization constraints just described. Included among complementizer phrases are interrogative clauses.
In his introductory textbook on syntax, Radford (1997: 106) describes complementizers and complementizer phrases as follows:
Complementizers head a separate layer of functional
superstructure in clauses, which we termed a complementizer phrase with the head C position of CP being
filled by complementizers like that/for/if or auxiliaries
in yes/no questions.
In yes/no questions in languages that have explicit question markers, the question marker fills the specifier position within the complementizer phrase, and in special [wh-] questions in languages
which move the interrogative, the interrogative moves to the specifier position within the complementizer phrase. Radford (1997:
18, 130131) presents a binary wh-parameter in universal grammar that would either restrict interrogatives to remain in the canonical (i.e. usual) position associated with their grammatical function or allow them to be moved to the specifier position within the
complementizer phrase (spec-CP).
Complementizer phrases are at the highest level of the clausestructure. In general, movement of any constituent is raising it to
a higher position in the clause structure. Thus Jackendoff (1977:
75) acknowledges that the Extended Standard Theory does include a general uneasiness with lowering rules, and Chung and
McCloskey (1987: 195) assume a condition on possible adjunction sites ... that requires [that] elements to be adjoined always ...
207
move upward in a tree. Therefore, in a head-initial language such

as Irish, typically having VerbSubjectObject word order, an interrogative moves to the front of the sentence prior to the verb, as
it does also in English.
Where1 does an interrogative adverb go t1 ?
(4)
Now Sanskrit has overwhelmingly head-final word-order. Delbrcks (1878) description of the usual word-order found in the
Satapathabr
ahman.a [reported by Staal (1967: 51) may be summarized as follows:
subject ... dative ... accusativeverb
adjectivenoun
genitivenoun
nounpostposition
In accordance with Greenbergs observations of other languages
with objectverb order and with Vennemanns predictions, since
Sanskrit serializes the object and verb in the order stated, the or
dering of other operatoroperand pairs follows suit. In X-grammar
the general rule governing word-order in phrases of all types would
specify that the complement is followed by the head. In a headfinal language, if the language moves interrogatives, an interrogative particle in general (i.e. yes/no) questions and an interrogative
in a special (i.e. wh-) questions lodge at the head of the complementizer phrase, that is, after the verb at the end of the clause. We
do in fact find certain complementizers at the end of their clauses
in Sanskrit, for example, iti closes quotes and the relative pronoun
typically appears final in the clause in a commentators gloss of
bahuvrhi compounds. Therefore, as shown in (5),
kim
sabdah. tis..thati kutra?
(5)
208
P. S CHARF
since Sanskrit is a head-final language, according to the general

rule governing word-order, we would expect to find the interrogative particle final in yes/no questions and the interrogative pronoun
final in special [wh-] questions, unless in the latter it remains in situ
in the position associated with its grammatical function.
We dont.
The evidence
The data upon which the following conclusions are based consist
of all the interrogative clauses in prose in the fourth am
sa of the
Vis.n.upuran.a. The rather unadorned prose in the fourth am
sa of the
Vis.n.upuran.a, interrupted occasionally by verse, is probably about
as ordinary a Sanskrit word-order as one may find.
In the few general [yes/no] questions occurring, such as (1),
the interrogative particle api appears initial rather than final in the
clause.
(1) ;pyea;tea Y;sma;tpua:a;aH k+:l+.Ba;a;
a;Sa;aH :pa;;d
;a;Ma ga;.cCe +.yuaH (VP 4.2.43)
api ete asmatputrah. kalabhas.in.ah. padbhyam gaccheyuh.?
[yes/no q-marker] these our-sons softly-speaking by feet go?
Will these baby-talking sons of ours walk?
Of the forty special questions, none terminates in the interrogative,
that is, in a declined form of, or derivate of, kim. It is obvious
that Sanskrit does not move the interrogative to the sentence-final
position as the general rule serializing phrase constituents would
lead us to expect.
Where does the interrogative appear? Twenty-three of the forty
special questions begin with the interrogative, of which seven contain only a single word besides the interrogative, e.g. (2)(3).
(2) ;
a;kM ku+:mRaH (VP 4.2.26)
kim kurmah.?
What shall we do?
209
(3) *: ga;a;ma;Sya;Ta (MBh 4.329.22)

kva gamis.yatha?
Where will you go?
Seventeen of the forty have non-initial interrogative. In two of
these the interrogative occurs immediately following a subordinating conjunction e.g. (4);
(4) ta;Ta;a;
a;pa :ke+:na va;a .ja;n}.a;BUa;a;ma;nRa .sma;yRa;tea (VP 4.2.39)
tatha api kena va janmabhumih. na smaryate?
Nevertheless by whom or native-land no is missed?
Nevertheless, who doesnt miss his native land?
in two, after a vocative, e.g. (5);
(5) :pua;
a:a k+:sma;a;a .ja;a;ya;sea (VP 4.13.56)
putri kasmat na jayase?
Daughter, why not are you being borne?
Daughter, why arent you being borne?
in four, kim uta or kim punar begins a new clause, e.g. (6).
(6) ;sua:=+sua:=+ya:a;ga;nDa;vRa:=+a:a;sa;a;
a;d;a;Ba:=+pya;a;Ka;lE
+.BRa;ga;va;a;a .jea;tMua Za;k+.aH ;
a;k+:mua;ta;a;va;a;na;ga;ea;.ca;=E ;=+pa;va;a;yERa;nRa+:=EH . . . (VP 4.13.28)
asurasurayaks.agandharvaraks.asadibhih. api akhilaih. bhagavan na jetum sakyah.; kim uta avanigocaraih. alpavryaih.
naraih. . . . ?
By demons, gods, Yaks.as, Gandharvas, Raks.asas, etc. even
all your divinity not to conquer is able; what more earthrealm (instr.) small-strength (instr.) men (instr.) . . . .
Your divinity is not able to be conquered even by all the
demons, gods, Yaks.as, Gandharvas, Raks.asas, etc.; how
much less so by weak men whose realm is the earth . . . ?
This leaves eight embedded interrogatives and sixteen initial which
may suggest that the interrogative either occurs in situ or moves to
210
P. S CHARF
the initial position in the clause. Examples of sentences of more

than two words with initial interrogative are shown in (7)(8).
(7) k+:~yea;ya;ma;hR ;a;ta (VP 4.1.20)
kasya iyam arhati?
Of whom this one (fem.) is worthy?
Of whom is she worthy?
(8) :ke+:nEa;ta;n}.a:n:a;pUa;tMa va;a;a:= :pa;a;ta;m,a (VP 4.2.15)
kena etat mantraputam vari ptam?
By whom this mantra-purified water drunk?
by whom was this water, purified by mantras, drunk?
A closer look at the eight questions with embedded interrogative (9)(16) mostly corroborates the view that the interrogative
remains in situ or moves to the front of its clause. In (9), one may
say that the adverb anyatha otherwise and the pronoun tasya of
him, which refers to Krs.n.a about whom his soldiers speak, has
been fronted before the
initial interrogative adverb katham. Likewise in (10), one may say that svakam mataram own mother has
been topicalized. Similarly, in sentences (11)(12), one may say
that topicalized phrases have been fronted leaving the interrogative subject in the position usual for its grammatical function in
the clause, namely, initial. In (13), the particle nama indicates that
the participial clause has been fronted. The core clause then contains subjectobjectverb in the canonical order with the interrogative object in situ. Sentences (14)(15) are passive constructions
with the direct object in the nominative case in its canonical place
before the verb in (14) and before the negative adverb na in (15).
However, one does not seem to be able to explain the location
of the interrogative in sentence (16) as in situ or clause-initial after
fronted constituents. In (16), the interrogative dative intervenes
between the nominative direct object and the passive verb. Recent
generative grammar generally considers the direct object to be the
211
complement of the verb with which it forms a verb phrase prior

to being related to other phrases in the sentence. Moreover, there
does not seem to be any motivation for the object in sentence (16)
to move out of its verb phrase. There remains no explanation in
the serialization theories discussed for how the interrogative in this
sentence intervenes between the direct object and the verb.1
(9) ;nya;Ta;a ta;~ya k+:Ta;mea;ta;a;va;a;nta
/ / ;
a;d;na;a;a;na Za:ua:ja;yea v.ya;a:ea;pa;ea Ba;
a;va;Sya;a;ta
(VP 4.13.25)
anyatha tasya katham etavanti dinani satrujaye vyaks.epah.
bhavis.yati.
Otherwise, of him, how, for so many days, in the vanquishing
of the enemy, occupation will be?
Otherwise how could he be occupied in vanquishing the enemy for so many days?
(10) .~va;k+:Ma . ca ma;a;ta:=M ;
a;k+:a;ma;a;ta ;a;.ca:=M :*e :+:Za;ya;a;sa (VP 4.13.56)
svakam ca mataram kim iti ciram klesayasi.
Own and mother why for long you torment?
And why do you torment your own mother for a long time?
(11) Ba;ga;va;a;sma;a;k+:ma:a ;
a;va:=+ea;Dea k+:ta;maH :pa:a;ea .jea;ta;a Ba;
a;va;Sya;a;ta (VP 4.9.2)
bhagavan asmakam atra virodhe katamah. paks.ah. jeta bhavis.yati.
O lord, of us in this conflict which side winner will be?
O lord, in this conflict of ours, which side will be the winner?
(12) ;a;ta;.ca;pa;l+.a;.ca:a;a:a .~ya;nd;nea :ke+:ya;ma;a:=+ea;
a;pa;ta;a (VP 4.12.9)
aticapalacitta atra syandane ka iyam a ropita.
Whose-mind-is-exceedingly-agitated in this chariot which
1 If one argues that the nominative is the subject, not the object, then the position of the nominatives in sentences (14)(15) becomes problematic. As subjects,
which they would have to be considered by analogy with (16), they should not
intervene between the instrumental agents and the verbs.
212
P. S CHARF
this woman has been made to mount?
Who is this very mentally agitated woman you have made to
mount this chariot?
(13) .ja;a;ta;ea na;a;mEa;SaH kM ;Da;a;~ya;a;ta (VP 4.2.17)

jatah. nama es.ah. kam dhasyati.
Born, indeed, this one whom will suckle?
Once he has been born, whom will he suckle?
(14) ya;d;ae;vMa :pUa;vRa;mea;va tva;ya;a;hM . ca;ea;
a;d;taH .~ya;Ma ta;n}.a;ya;a tva;d;Ta ;
a;k+:ma;k+:tRa;v.ya;m,a
(VP 4.9.7)
yadi evam purvam eva tvaya aham coditah. syam tat maya
tvadartham kim akartavyam.
If thus at first only by you urged I had been, then by me for
you what would not have been done?
If you had urged me right at first, then what wouldnt I have
done for you?
(15) ;Ga;a;ta;a;ya;tvEa;nMa ta;n}.a;h;a:=+Ma tva;ya;a ;
a;kM na gxa;h:a;tea (VP 4.13.35)
ghatayitva enam tat maharatnam tvaya kim na grhyate.
is taken?
After slaying him, that great-jewel by you why not
After slaying him why dont you take that great-jewel?
(16) Ba;ga;va;ea;va;ma;va;a;~Ta;te
/ / a ma;yea;yMa k+:smEa :de;ya;a (VP 4.1.25)
bhagavan evam avasthite maya iyam kasmai deya.
O lord, thus situated by me this girl to whom is to be given?
O lord, when the situation is such, to whom should I give this
girl?
Parallels with other pronouns
Hock (1989) points out similar issues with relative and correlative pronouns. These pronouns, just as the interrogative pronouns,
213
locate either at the beginning of their clauses, e.g. the relative pronoun yas in (17) and the demonstrative pronoun tam in (18), or
just prior to the verb, e.g. the relative pronoun yo in (19) and the
demonstrative pronoun sa in (20). If one scrapped the serialization
hypothesis there is no problem explaining the initial position of
the pronouns in (17)(18) by movement to a higher position in the
syntactic tree. Yet one would still not be able to explain the position of the pronouns in (19)(20) because this would involve the
deplored movement to a lower position in the syntactic tree. The
relative pronoun yo and the demonstrative pronoun sa , which are
subjects, intervene between the finite verbs and their direct objects.
(17) ya;~te!a .sa;a;K
a;By!a ;a va:=+m,a (RV 1.4.4c)
yas te sakhibhya a varam

. . . , who is dearer to you than all friends (Hock 1989: 17)
(18) tMa tva;!a va;a:jea;Su!a va;a;a:ja;nMa va;!a:ja;ya;a;maH (RV 1.4.9ab)
tam
tva va jes.u va jinam
vajayamah.
We make victorious you, the victorious, in the victories
. . . (Hock 1989: 17)
(19) o+.du!;a;~:a;ya;!a .ja;a;na;ta;!a ya;ea .j!a:ja;a;na (RV 3.1.12c)
u d usrya janita yo jaja na

. . . who as creator created forth the cows (Hock 1989: 18)
(20) v!a;v.ra;<a ;n!a;nta;<a ;v!a .sa;a :pa;d ;a; (RV 7.104.17c)
vavra ananta a va sa pads..ta

May she fall down into limitless depths. (Hock 1989: 18)
Hock (1989: 2226) argues that stressed pronouns move to TOPIC
or FOCUS rather than to COMP. He concludes (1989: 26),
the landing site for RPs in Sanskrit relative-correlatives is not
COMP, but either TOPIC or preverbal FOCUS.
214
P. S CHARF
The evidence provided for interrogatives in this paper and

Hocks (1989) analysis of relative and correlative structures imply that the occurrence of interrogative, relative and demonstrative pronouns in initial and preverbal positions is due to movement
motivated by discourse factors rather than by grammatical ones.
Movement to either of these positions contradicts the predictions
of the serialization hypothesis for movement to final position in the
clause. Moreover, even if one abandons the serialization hypothesis, problems remain with syntactically motivated explanations of
movement altogether. While movement to initial position is unproblematic since it involves movement to a higher position in the
syntactic tree, movement to the preverbal position intervening between a verb and its direct object would involve lowering in the
syntactic hierarchy which is deplored. Syntactic serialization and
movement theories seem to be unable to account for the order of
constituents in such sentences adequately.
The solution
Theories of the serialization of word-order grow out of a tradition

of associating position with grammatical function. In English and
other Western European languages, word-order does indeed serve
grammatical function. However, highly inflected languages utilize
different mechanisms to indicate grammatical function. As Staal
(1967: 59) argued with great clarity more than thirty years ago, to
no avail it seems in the world of contemporary linguistic theory,
word order has no grammatical significance or value, in Sanskrit
and in other highly inflected languages, and (1967: 77) grammatical relations, ..., are always independent of order. He stated
plainly (1967: 71) even if free word order in this sense would not
obtain anywhere, the divergent arrangements of words in different
215
types of language would require the base of a system of universal

grammar to be unordered.
Comrie (1984: 13) argues that in highly inflected languages
such as Russian in which word-order is free, it is constrained by
the pragmatic structure of the sentence, that is, by factors such as
the topiccomment structure and the focus. Wallace (1984) discusses the interaction of word order and pragmatics in Sanskrit.
Hueckstedt (1985) points out numerous factors which influenced
the ordering of constituents in the style of Ban.a: (1985: 45) an
initial verb provides emphasis, (1985: 67) elements are ordered to
postpone the expected, and (1985: 78) sequence in time and space
determines the sequence of clauses, to name a few. It seems to
me that the behavior of interrogatives in Sanskrit conforms to this
functional sentence perspective. The interrogative is placed in positions at the beginning of the sentence and immediately before the
final verb for emphasis.
It is crucial to note that the placement of the interrogatives in
sentence (16) and of relatives and demonstratives in (19)(20) cannot be explained by movement from underlying positions associated with grammatical roles. An explanation that the pronouns
did move forward here, would have to explain how in general pronouns in Sanskrit could move to places of emphasis in two opposite directions: forward to the position just before the verb or
backward to the beginning of the clause where they often appear.
Movement in two opposite directions is incompatible with the theories that serialization is unidirectional in any given language.
In contrast, the theory that grammatical function is inherently
unordered does not conflict with evidence that shows usual orders.
It is quite clear that many languages utilize position for grammatical roles. It is also clear that there are elements of usual order
in Sanskrit and other highly inflected languages. Yet to claim that
such grammaticality is inherent to position and that movement is
unidirectional is ludicrous as is clearly shown regarding the inter-
216
P. S CHARF
rogatives in this paper. Rather than put linearity at the foundation

of grammatical relations and syntax, a more productive approach,
in agreement with Staal and as the author has argued elsewhere, is
to describe grammatical relations in an abstract multidimensional
space and to project these relations onto linear speech taking into
account discourse factors as well as grammatical roles.
References
Chung, Sandra and James McCloskey. 1987. Government, barriers, and small clauses in modern Irish. Linguistic inquiry 18.2:
173237.
Comrie, Bernard. 1984. Russian. Interrogativity: a colloquium
on the grammar, typology and pragmatics of questions in seven
diverse languages, Cleveland, Ohio, October 5th 1981 - May
3rd 1982, ed. by William S. Chisholm, pp. 746. Typological
Studies in Language 4.
. 1988. Linguistic typology. Linguistics: the Cambridge survey,
ed. by Frederick J. Newmeyer, vol. 1, pp. 447461. Cambridge: Cambridge University Press.
Delbrck, Bertold. 1878. Syntaktische Forschungen; vol. 3, Die al
tindische Wortfolge aus dem Satapathabr
ahman.a dargestellt.
Halle: Waisenhaus.
Greenberg, Joseph H. 1966. Some universals of grammar with
particular reference to the order of meaningful elements. Universals of language, ed. by Joseph H. Greenberg, pp. 73113.
Second edition.
Hock, Hans Henrich. 1989. Conjoined we stand: Theoretical implications of Sanskrit relative clauses. Studies in the Linguistic Sciences 19.1: 93126.
Hueckstedt, Robert. 1985. The style of Ban.a. Lanham, MD: University Press of America.
R EFERENCES
217
Jackendoff, Ray. 1977. X syntax: a study of phrase structure. Cambridge: The MIT Press.
Radford, Andrew. 1997. Syntax: a minimalist introduction. Cambridge: Cambridge University Press.
Staal, J. F. 1967. Word order in Sanskrit and universal grammar.
Foundations of Language, Supplementary Series 5. Dordrecht:
D. Reidel Publishing Co.
Vennemann, Theo. 1974. Analogy in generative grammar: the origin of word order. Proceedings of the eleventh international
congress of linguists, Bologna-Florence, August 28 - Sept 2,
1972, ed. by Luigi Heilmann, vol. II, pp. 7984.
Wallace, William D. 1984. The interaction of word order and
pragmatics in a Sanskrit text. Studies in the Linguistic Sciences 14.1: 16788.
218
P. S CHARF

MILIE AUSSANT
Abstract: This paper aims at briefly recalling the presuppositions the European and Indian grammatical approaches
present regarding word-classification. From its very beginning up to the most contemporary trends in linguistics,
the European approach knew essentially only one classification of words: the canonical list of eight parts of
speech which comes from Dionysius Thrax and which was
developed, restructured, sub-divided, etc., throughout the
course of history, according to the aims of particular scholars or schools. Within Vyakaran.a, native Sanskrit grammar, the situation is not so different: the Sanskrit grammarians generally alternated between only two classifications (the four-fold Yaskan division of parts of speech and
the two-fold division proposed by Pan.ini); some of them
like Nagesa having even tried to reconcile the two.
Keywords: Europe, India, history of linguistic ideas, epistemology of language sciences
Introduction
Even though the analysis of language into units appears to have

been essential in all traditions of language study, the classification
of words consists in an activity which is neither self-evident nor
homogeneous: the classifier (whether he is an individual scholar,
a school of thought or a trend) has an epistemological aim (that
219
220
. AUSSANT
is to say, he has something to explain) and one must consider the

regularities (hence the categories) he makes exist on this basis.
Epistemological aims have varied a lot in history, from one tradition of language study to another, but also within one and the same
tradition. The analysis and description of Sanskrit in the perspective of computational linguistics brings two different grammatical
traditions, as well as different classifications, face to face. Without calling into question their legitimacy (my approach is rooted
in the perspective of the history of ideas), I would like to recall, albeit briefly, the presuppositions these theories present. I will consider first the European approach and then the approach of Sanskrit
(Vyakaran.a perspective).1
2
2.1
The European approach

Parts of speech in computational linguistics
In computational linguistics, the parts of speech which are used

come, most of the time, from works realized in the field of formal linguistics. Very often indeed, taggers are statistical tools
trained on manually or semi-automatically annotated data. These
data were mostly annotated under the responsibility of formal linguists and according to the guidelines they elaborated on the basis
of their linguistic expertise, as well as on the basis of what results from the confrontation of the system with real data. Hence,
at first sight, there is no real norm: corpora can be tagged in various ways. However, some tag sets, such as the Brown Corpus
1 I wish to thank Aime Lahaussois, Clment Plancq and Benot Sagot for
their kind help.
T O CLASSIFY WORDS
221
tag set,2 the Penn Treebank tag set,3 and, to a lesser extent, the
French Treebank developed by the University of Paris Diderot,4
today constitute a reference. If one looks at these tag sets, one observes that the choices made remain mainly traditional it is even
explicitly given on the website of the French Treebank, For part
of speech, we made traditional choices [. . . ] and, once we subtract the numerous categories and sub-categories which divide or
complete each part of speech (see Table 1), and which are mainly
based on inflectional and lexico-semantic information, we find the
well-known list again: noun, verb, article, adjective, preposition,
pronoun, adverb, conjunction, and interjection. Even tools developed for languages other than English and French, such as the Natural Language ToolKit (NLTK) tag set for four Indian languages
(Bangla, Hindi, Marathi and Telugu),5 and the Stanford Natural
Language Processing Group (SNLPG) tag set for Chinese,6 resort
to this list. Where does this list come from? On what kind of
classification criteria is it based? How can one explain its stability
throughout the course of history?
2.2
Parts of speech in Graeco-Latin grammatical tradition
The list of eight parts of speech comes from the Graeco-Latin

grammatical tradition, and more specifically from the
2 http://www.comp.leeds.ac.uk/ccalas/tagsets/brown.
html
3 http://www.cst.dk/mulinco/filer/PennTreebankTS.
html
4 http://www.llf.cnrs.fr/Gens/Abeille/FrenchTreebank-fr.php
5 http://nltk.googlecode.com/svn/trunk/doc/book/
ch05.html#fig-tag-indian
6 http://nlp.stanford.edu/projects/chinese-nlp.
shtml#pos
222
. AUSSANT
Table 1
Sub-categories of noun in the Brown Corpus
NN
NN$
NNS
NNS$
NP
NP$
NPS
NPS$
NR
singular or mass noun

possessive singular noun
plural noun
possessive plural noun
proper noun or part of name phrase
possessive proper noun
plural proper noun
possessive plural proper noun
adverbial noun (home, today, west)
by Dionysius Thrax (2nd-1st c. BCE). The aim of

the Greek scholar was to provide, mainly for writers and poets,
an overview of the Alexandrian grammatical tradition, giving a
reference list for phonetics (letters-sounds, syllables) as well as
the list of eight parts of speech: noun, verb, participle, article,
pronoun, preposition, adverb and conjunction. This list itself is
the result of developments and restructurings of a more ancient
classification: that made by Aristotle (384322 BCE) in his Poetics (chapter XX), which distinguishes four parts of speech:
noun, verb, link (late conjunction),
articulation (late article) and which is mainly based on notional
criteria (semantic, ontological, psychological, and/or logical) (Lallot 1988). The four parts of speech list was quickly enlarged and
restructured on the basis of formal and semantic criteria. In the
by Dionysius Thrax, each part of speech is
defined, its accidents (e.g. everything which happens to it, that
is to say number, gender, inflection, diathesis and tense variations)
are listed, defined and illustrated, its morphological and/or seman-
T O CLASSIFY WORDS
223
tic sub-classes are listed and illustrated (31 for the noun, 13 for
the verb, 26 for the adverb, 9 for the conjunction) (Lallot 1998).
What are the criteria used for this classification in eight parts of
speech? Notional (see above) and formal (morphological) criteria were used very early; functional criteria (syntactical) appeared
a little bit later, with the Greek grammarian Apollonius Dyscolus
(2nd c. CE). But from the very beginning, parts of speech were
and are still today classified according to (a) their meaning
and/or their reference, (b) according to their form, and (c) according to their position and/or their relation to other parts of speech
in sentences (Lagarde 1988). The hierarchy of these criteria, the
number of their sub-categories as well as their relationship, varied
greatly in history. For instance, Ramus (Pierre De La Rame), a
16th c. French grammarian, tried to impose (in his Scholae Grammaticae) a model with two categories of words (lat. voces), using exclusively formal criteria: mots de nombre, i.e. nouns and
verbs, versus mots sans nombre, i.e. adverbs and conjunctions.7
For authors of universal grammars (which elaborate their own categories by linking them to cognitive structures), such as Beauze
and Condillac in the case of the Grammaire gnrale franaise,
notional criteria are much more relevant than the others (Auroux
1988: 81). As Lagarde (1988: 93) already highlights, the list of
eight parts of speech, though provoking strong debate,8 crossed
over the centuries and modeled grammatical consciousnesses. Every linguist must, explicitly or not, make a stand regarding this
model of analysis which has become an unavoidable reference.
7 See
Colombat 1988, 1998.

for instance, the debate opposing Haspelmath and Dryer, two leaders
of linguistic typology. Haspelmath (2010) argues that crosslinguistic grammatical comparison cannot be based on grammatical categories, because these are
language specific, while Brown and Dryer (2008) write, While there are often
ways in which languages differ from each other in terms of what word classes
they have, the differences are generally small compared with the similarities.
8 See,
224
. AUSSANT
The judgements made with regard to this model are generally negative. Despite endlessly repeated criticisms and the periodic attempts to develop universal categories or tag sets (such as the one
elaborated by Petrov, Das and McDonald in 2011),9 this classification remains. Auroux (1988: 82) observes quite rightly that, as
a word-classification tool used in various fields of language study
(language description, learning, etc.) which are not always theoretically well mastered, the list of eight parts of speech is stable,
much more so than the theoretical variations suggest. Simply for
communicability reasons, no theoretical change can lead to a fundamental change in the list of parts of speech. The question of
parts of speech cannot be reduced to a mere matter of word classes
because it is on the basis of this word-classification that European
grammar was elaborated.
The Indian approach (Sanskrit Vyakaran.a)
If we now move to the Indian side and, more particularly, to the

field of Sanskrit Vyakaran.a, what ways of classifying words do we
find? Two models were mainly used: the naman-akhyata-upasarga-nipata classification and the Pan.inian classification of padas.
3.1
The naman-akhyata-upasarga-nipata classification
One of the most ancient classifications of words seems to be the

one in four parts: naman, a khyata, upasarga and nipata, which are
traditionally translated as noun, verb, preverb and particle.
In section 13.9 of the Nirukta (N. hereafter) which is considered to
be an appendix (parisis..ta),10 while commenting on the Rg-vedic
verse
9 See
http://www.petrovi.de/data/lrec.pdf
(19201929) does not translate this section.
10 Sarup
T O CLASSIFY WORDS
225
Catva ri va k parimita pada ni

ta ni vidur brahman.a ye mans.n.ah.;
Guha trn.i nhita nen gayanti
turyam
vaco manus.ya` vadanti. (RV. 1.164.45)
Speech has been measured in foursteps. The Brahmins who have control over their mind know them.
The three of them which are deposited secretly do not
stir. The fourth [step] of the speech is that which men
speak. (Bhate undated: 2, revised).
Yaska mentions different readings of the expression catvari padani among which the following one:
Namakhyate copasarganipatas ceti vaiyakaran.a h..
(N. 13.9, Sarup 19201929: 226)
Grammarians [consider that these four kinds of pada
are]: the noun and the verb, as well as the preverbs
and the particles.
Yaska himself gives this classification of padas at the very beginning of his Nirukta:
Tad yani catvari padajatani namakhyate copasarganipatas ca tanmani bhavanti. (N. 1.1, Sarup 1920
1929: 27)
Now, what (are) the four classes of words? They are
the following: noun and verb; prepositions and particles. (Sarups translation)
This classification, which is found in other texts such as the Mahabhas.ya (Paspasa ), seems to come directly from the very early
grammatical analyses elaborated in India. Such analyses aimed
at deconstructing Vedic texts in their continuously recited version
(samhit
a-pa.tha), where sandhis are applied, to result in a word-byword version (pada-pa.tha) of the same texts, where any dissociable sequence was de facto isolated. This kind of analysis implied
226
. AUSSANT
an accurate study of morphology and syntax, and it is most likely

during this period that the four-part classification of words was
elaborated. The fact is that it is to be found in the Rgveda-Pratisa
khya:
Nama-akhyatam upasargo nipatas
catvary a huh. padajatani sa bdah.;
tan nama yenabhidadhati sattvam
tad a khyatam
yena bhavam
sa dhatuh.. (RPr. 12.5)
The noun, the verb, the preverb and the particle,

these
are the four classes of words say the grammarians.
The noun is that by which a substance is denoted; a
verb is that by which a process [is denoted]. It is a
dhatu.
...
kriya-vacakam a khyatam upasargo vises.a-krt;
uranah.
sattva-abhidhayakam
nama nipatah. pada-p
. .
(RPr. 12.8)
verb expresses an action; the preverb specifies
The
it. The noun expresses a substance; the particle completes the verse quater (pada).
Though we have access to neither the thought nor the method of
the authors of pada-pa.tha texts, it is nevertheless possible to infer
from their works some of the principles they adhered to. For instance, what they meant by pada, according to Bhate (undated: 4)
were isolated inflected forms (nominal, verbal), indeclinables, and
preverbs. As it appears in the first verse quoted previously, as well
as in the Nirukta, the criteria on which the distinction between the
noun and the verb is based is notional (ontological) criteria. Regarding the upasargas, two definitions are given by Yaska:
akat.a yana, says that upasargas do not ex One, ascribed to S
press meanings (na . . . arthan nirahur); they simply help the
T O CLASSIFY WORDS
227
emergence of the whole meaning by suggesting it (karmaupasamyoga-dyotak

a bhavanti) (N. 1.3)
The other, ascribed to Gargya, presents upasargas as units

expressing (vacaka) diverse meanings, meanings which
modify the meaning of verbs and nouns with which the upasargas are combined (uccavacah. padartha bhavantti gargyah.. tad ya es.u padarthah., prahur ime tam nama-akhyatayor artha-vikaran.am) (N. 1.3). The second verse
quoted from the Rgveda-Pratisa khya (upasargo vises.a-krt)
seems to refer tothis conception.

Concerning the particles, only one description is given in the
Nirukta: nipatas are said to express numerous meanings (nipata uccavaces.v arthes.u nipatanti): comparative (apy upamarthe), connective (api karma-upasamgrah
arthe), expletive (api pada-pura
n.a h.) (N. 1.4).
We note that the naman-akhyata-upasarga-nipata classification is presented as exclusively based on notional criteria: an ontological criterion for the noun and the verb and a semantic criterion
for the preverbs and the particles. This is surprising since, considering the context of the emergence of this classification, we would
have expected formal criteria.
3.2
The Pan.inian classification of padas
On a practical level, the As..ta dhyay (A. hereafter) provides guidelines for forming words with affixes. These affixes (pratyaya) are
directly taught in some of the 4000 sutras, in distinction from most
of the bases with which they combine and which are either verbal
roots (dhatu) or nominal bases. These nominal bases are:
either generally introduced by A. 1.2.45 arthavad adhatur apratyayah. pratipadikam (A meaningful [unit], which
228
. AUSSANT
is neither a verbal root nor an affix, [is called] pratipadika), which is completed by A. 1.2.46 krt-taddhita-samasa
s ca (Primary derivatives, secondary derivatives
and compounds [are] also [called] pratipadika);
or given in lists which gather together units sharing some
derivational features (cf. the Gan.a-pa.tha lists (GP. hereafter);
or directly introduced in some sutras, such as A. 3.1.16 bas.pa-us.mabhyam udvamane, which states that the (denominative) suffix kyan is used after the nominal bases bas.pa
tear and u s.man heat to express to emit.
Verbal and nominal bases constitute the two starting points of

the derivational process which occurs over the course of Pan.inian
sutras to generate more and more complex units up to sentences
(vakya).11 The steps of this derivational process consist in the application of operations to these units and, generally, it is these very
operations which determine the arrangement of the units in classes
or categories. For instance, a unit is called pratipadika, that is
to say, belongs to the pratipadika class because it undergoes (or
must undergo) the application of several operations. Sanskrit morphology being very rich, operations to be applied are numerous,
and then, the number of classes Pan.ini distinguishes is particularly
high. The main classes are those shown in Figure 1 (based on that
made by Bhate (undated: 7)).
At the top of the diagram, one finds the term pada: Pan.ini uses
it to designate the general class of words, but he also gives a definition of the term (probably the most ancient to have come down to
us) which clearly delimits its extension: the sutra A. 1.4.14 states
that the units which constitute the pada class are those that end
11 The
term vakya is used only twice in the As..ta dhyay, in A. 8.1.8 vakya-ader
a mantritasya . . . , and in A. 8.2.82 vakyasya .teh. . . . ).
T O CLASSIFY WORDS
229
Figure 1
The main classes of speech forms distinguished by Pan.ini
in a nominal (sup) or a verbal (tin)

ending, in other words, the
inflected words. One notes that in the Pan.inian system, neither naman nor a khyata are used to designate respectively the noun and
the verb: neither of these terms is used by Pan.ini in a metalinguistic way. One notes too that there is an empty square on the
diagram: it stands for the nominal stems which cannot be derived,
such as vrks.a, which are meaningful units but neither a verbal root,
pronoun, indeclinable, primary derivative, secondary
nor a suffix,
derivative, nor compound. As far as I know, Pan.ini does not give
a name other than pratipadika to this kind of unit. Patajali explains (on vt. 5 ad A. 7.1.2) that for Pan.ini, nominal stems
230
. AUSSANT
which cannot be derived are the un.a dis.12 Contrary to naman and
a khyata, upasarga is used in the As..ta dhyay, and, here too, it designates the preverbs (cf. A. 1.4.59 upasargah. kriya-yoge). However, it belongs to a larger class called pradi (GP. 154), which itself
belongs to the nipata class (the term is also used to designate the
particles). And this nipata class belongs to the avyaya class (indeclinables, cf. A. 1.1.37 svaradinipatam avyayam) which includes
also the svaradi class (GP. 254). The avyaya class constitutes a
sub-category of the pratipadika class, these units being meaningful. If upasarga and nipata are used in the As..ta dhyay with more or
less the same meaning as in the Nirukta, the Pan.inian system nevertheless establishes a difference in the level of analysis between
them. However that may be, the Pan.inian classification of padas
is firstly based on formal criteria. Semantics plays a role too, but
to a lesser extent; it comes as a complement (e.g. the definition of
nipata: A. 1.4.5657 . . . units which do not designate a substance
(asattve)).
3.3
Some remarks on a few other classifications
Still within the Vyakaran.a sphere one finds mention of a few

other classifications of words. I will not address the classifications which correlate linguistic categories with ontological categories, such as the jati-gun.a-kriya-yadrccha classification pro classification, as well
vided by Patajali in his Mahabhas.ya. This
as the a krti-kriya-gun.a-dravya list of padarthas, actually provides
a sub-classification
of pratipadikas, often illustrated by the sentence gauh. suklas calo d.itthah. [the] white moving cow D.ittha.
At the very beginning of the third kan.d.a of his Vakyapadya, Bhartrhari mentions three views on the classes of padas. It
12 Pr
atipadikavijanac ca paniner siddham.
Pratipadikavijanac ca bhagava.
tah. pan.iner a caryasya siddham. Un.a dayo vyutpannani pratipadikani. (A. 7.1.2,
vt. 5; MBh. 3.241.1921)
T O CLASSIFY WORDS
231
is very well-known that, according to Bhartrhari, padas exist only

in a philosophical
on the level of apoddhara [artificial] analysis;
perspective, only the sentence is relevant. But placing himself on
the apoddhara level, he says:
dvidha kaiscit padam
bhinnam
caturdha pacadhapi va;

apoddhrtyaiva vakyebhyah.
prakrtipratyayadivat. (VP. 3.1.1)
Some distinguish two [kinds] of padas, [some] four
[kinds] and [some] five kinds after having artificially
isolated them from sentences, just as [one would isolate] the stem and the suffix, etc.
These classifications are:
nama (which includes nipata) : a khyata (which includes upasarga and karmapravacanya);
naman : a khyata : upasarga : nipata;
naman : a khyata : upasarga : nipata : karmapravacanya.
A very similar classification is mentionned by Durga, in his
commentary on Nirukta 1.1:
Naikam
padajatam
yatharthah. padam aindran.a m iti.
Napi dve yatha subantah. tinantas ca. Napi trn.i nipatopasargav ekatah.krtva. Napi paca s.ad. va yatha
gatikarmapravacanyabhedeneti.
(Durga on Nirukta
1.1)
It is not the case that there is a single class of words
as, that which is meaningful is a word for the Aindras. Nor are there two [classes of words], for instance nouns and verbs. Nor are there three [classes of
words] combining in one particles and preverbs [along
with the two preceding ones]. Nor five or six [classes
232
. AUSSANT
of words], as with the distinction between gati and
karmapravacanya.
These classifications do not really improve upon the previous analyses: they can be seen as variants of the four-fold classification
given in the Nirukta and the Rgveda-Pratisa khya.
Concluding remarks
From its very beginning up to the most contemporary trends in linguistics, the European approach knew essentially only one classification of words: the canonical list of eight parts of speech which
comes from Dionysius Thrax. Throughout the course of history,
it has been developed, restructured, sub-divided, etc. according to
the aims of particular scholars or schools. The heterogeneity of
its criteria and its Indo-European focus did not prevent theoreticians from analyzing speech and from proposing more and more
elaborate descriptive or generative models.
Concerning the Sanskrit approach (within the Vyakaran.a
sphere), the situation is not so different. I would say following Deshpande (1992: 14) that Sanskrit grammarians care for
Pan.inis two-fold division of stems into nominal stems and verb
roots only as far as the derivational procedures in his grammar are
concerned. When the discussion deals with semantics, they generally follow the four-fold division mentioned by Yaska . . . . The
inclusion, in the Pan.inian model, of the avyaya class in the pratipadika class seems to have been counter-intuitive for most of the
vaiyakaran.as. Some, like Nagesa, nevertheless attempted to reconcile the four-fold Yaskan division of parts of speech with the twofold division proposed by Pan.ini. Commenting on the Mahabhas.ya passage catvari srn gan.i catvari padajatani namakhyatopasarganipatas ca, hesays: namasabdena subantam
. . . a khyatam
R EFERENCES
233
tinantam. upasarga-nipatayoh. prthag upadanam

gobalvardanya
yena.
As far as the tagging of parts of speech is concerned, computational linguists working on Sanskrit are not doing anything different from their predecessors: they try to establish more appropriate
models and have to hand a wonderful tool: the Pan.inian grammar, which matches their interests particularly well. More than
their 19th-20th c. colleagues who wrote Sanskrit grammars in European languages, . . . they are in a position to put into practice
one of the major insights of structuralist linguistics of the twentieth century, namely, that languages are best described in their own
terms (Haspelmath (2010: 664) referring to Boas (19111922)).
Table 2
Abbreviations
A.
As..ta dhyay.
GP. Gan.apa.tha.
MBh. Mahabhas.ya.
N.
Nirukta. See Sarup 19201929.
RV. Rgveda. See Sonatakke and Kashikar 19331951.
RPr.
Rgvedapratisa khya. See Rgnier 1858.
References
Auroux, Sylvain. 1988. La grammaire gnrale et les fondements
philosophiques des classements de mots. Langages 92: 79
92.
Bhate, Saroja. 2006. Classifications of terms. History of Science,
Philosophy and Culture in Indian Civilization, general editor
D. P. Chattopadhyaya; vol. 3.4, Philosophical Concepts relevant to Sciences in Indian Tradition, ed. by P. K. Sen, pp. 609
234
. AUSSANT
30. PHISPC-CONSSAVY Series. New Delhi: Centre for Studies in Civilizations.

. undated. Classification of terms. Unpublished.
Boas, Franz. 19111922. Handbook of American Indian languages. 2 vols. Washington, D.C.: U.S. Government Printing
Office.
Brown, La and Matthew S. Dryer. 2008. The verbs for and in
Walman, a Torricelli language of Papua New Guinea. Language 84: 52865.
Colombat, Bernard. 1988. Les parties du discours (partes orationis) et la reconstruction dune syntaxe latine au XVIe sicle. Langages 92: 5164.
. 1998. Ramus, Petrus. Corpus reprsentatif des grammaires
et des traditions linguistiques; vol. 1, ed. by Bernard Colombat
and Elisabeth Lazcano, pp. 9496. Paris: Socit dhistoire et
dpistmologie des sciences du langage.
Colombat, Bernard and Elisabeth Lazcano, eds. 1998. Histoire
pistmologie langage; hors-srie, vol. 2, Corpus reprsentatif
des grammaires et des traditions linguistiques. 2 vols. Paris:
Socit dhistoire et dpistmologie des sciences du langage.
Deshpande, Madhav Murlidhar. 1992. The meaning of nouns: semantic theory in classical and medieval India Namarthanirn.aya of Kaun.d.abhat..ta translated and annotated. Studies of
Classical India 13. Dordrecht: Kluwer.
Haspelmath, Martin. 2010. Comparative concepts and descriptive
categories in cross-linguistic studies. Language 86.3: 66387.
Lagarde, Jean-Pierre. 1988. Les parties du discours dans la linguistique moderne et contemporaine. Langages 92: 93108.
Lallot, Jean. 1988. Origines et dveloppement de la thorie des
parties du discours en grce. Langages 92: 1123.
. 1998. Denys le thrace. Corpus reprsentatif des grammaires
et des traditions linguistiques; vol. 1, ed. by Bernard Colombat
R EFERENCES
235
and Elisabeth Lazcano, pp. 1516. Paris: Socit dhistoire et

dpistmologie des sciences du langage.
Rgnier, Adolphe. 1858. tudes sur la grammaire vdique:
prtikhya du Rig-Vda. Paris: Imprimerie impriale.
Sarup, Lakshman, ed. and trans. 19201929. The Nighan..tu and
the Nirukta: the oldest treatise on etymology, philology and
semantics: critical edition, translation, introduction, exegetical and critical notes, indexes and appendices. 1st ed. London: Oxford University Press; Lahore: University of the Panjab. [Reprinted with three parts in one volume: Delhi: Motilal
Banarsidass, 1967, 1984, 1998.]
Sonatakke, N. S. and C. G. Kashikar, eds. 19331951. R.gvedaSamhit
a: with the commentary of Sayan.a carya. 5 vols. Poona:
Vaidic Samshodhan Mandal.
236
. AUSSANT

B RENDAN S. G ILLON
Abstract: While ellipsis and pronominal dependence occur
commonly in Classical Sanskrit, they have received attention neither in the Indian grammatical tradition nor by
contemporary scholars of Classical Sanskrit. Another phenomenon, distinct from ellipsis but often confused with
it, is the phenomenon of optional complementation, or
polyadicity. This too is widely attested in Classical Sanskrit and it too has received no attention. The aim of the
paper is to illustrate these phenomena for Classical Sanskrit and to show that they are best characterized in terms
of the constituency.
Keywords: anaphora, argument, complement, constituent,
context, cotext, ellipsis, lexical class, optional complement, phrase, proforms, pronominal dependence, setting.
Introduction
There are two assumptions on which I hope everyone with a serious interest in the grammar of Classical Sanskrit can agree. First,
some properties of the grammar of Classical Sanskrit remain unknown and that, not withstanding the superb work done by the Indian grammatical tradition in bringing to light many, if not most, of
the properties of Classical Sanskrit, it may be possible to improve
on that work. Second, the only way to investigate rigorously the
237
238
B. G ILLON
grammatical properties of Classical Sanskrit, or of any dead language for that matter, is through the construction, for the language
in question, of a repository of grammatically analyzed sentences,
or tree bank. After all, it is only through the study of attested
sentences in a language that one can either confirm or disconfirm
hypotheses about its grammar.
The point of this paper is to show that any tree bank of Classical Sanskrit which seeks to facilitate the investigation of ellipsis
and pronominal dependence, two phenomena which, though extremely common in Classical Sanskrit, were not addressed by Indian grammatical tradition and have hitherto been unaddressed by
contemporary scholars, must have some way of identifying constituents, including phrasal constituents, as well as the relationships between them.
No one will dispute that both words and clauses are constituents. I shall therefore discuss only briefly both lexical and
clausal classification in Classical Sanskrit. I shall take advantage
of my brief discussion of lexical classification to make an auxiliary point: choice of grammatical annotation, in this case, lexical
classification, may have a crucial impact on the utility of a tree
bank for the investigation of the grammatical features of Classical
Sanskrit. I shall then proceed to discuss ellipsis and pronominal
dependence in Classical Sanskrit, first explaining several relevant
concepts and then setting out some of their properties. As we shall
see, tracking these properties requires one to acknowledge a variety of constituents, including not only words and clauses but also
phrases. Further evidence for phrasal constituents arises from considerations pertaining to optional complements, a topic which I
shall also discuss below. I shall conclude with some evidence to
support the claim that, as important as semantic roles, or karakas,
might be, they are best viewed as restrictions upon the kinds of the
denotation of complements.
C ONSTITUENCY AND C OTEXTUAL D EPENDENCE
239
Parts of speech
The most obvious place to start in providing a grammatical analysis of the expressions in a language is to assign to them parts of
speech. In the case of Classical Sanskrit, three lexical classifications are available, two from the Classical Indian grammatical tradition and one from the European tradition. (See Aussants paper
in this volume for detailed discussion of the various forms of lexical classification.) The earliest classification is one which divides
words into four notional categories: words for substances (dravyavacana, also called sattva-vacana, or words for entities), words
for qualities (gun.a-vacana), words for actions (kriya-vacana, also
called bhava-vacana, or words for changes) and proper names
(yadrccha-sabda, literally, chance words). This classification an Panini and continued long after his grammar became stantedates
.
dard. The second classification is Pan.inis. It divides words into
those which take nominal inflection (subanta, also called naman,
or names), those which take verbal inflection (tinanta, also called
a khyata) and those which take no inflection and so are invariant
(avyaya), which subdivide into prefixes (upasarga) and particles
(nipata). Finally, there is the classification, traditional in European
grammar and modern linguistics, and dating back to the classical
Greek thinkers, which includes: noun, verb, adjective, preposition,
adverb, among others.
As one can easily see, there are trade-offs. Classifying words
in a tree bank with the categories of either of the Classical Indian
classifications will render difficult its use by those presupposing in
their work the European categories. Inversely, classifying words
in a tree bank with the European categories will not be helpful to
those who presuppose in their work either of the Classical Indian
classification schemes.
As we all know, words are constituents which themselves may
have constituents. Those constituents may be either derivational
240
B. G ILLON
suffixes or they may be words. This lexical structure gives rise to

ambiguities. Here is an example of a structurally ambiguous compound recently discussed on the Indology listserv: a-dharma-ja.
It has two possible meanings, corresponding to two constituent
analyses: a-(dharma-ja) (someone who does not know what the
dharma is) and (a-dharma)-ja (someone who knows what is not
the dharma). (For further details, see Gillon 1987, 1995.)
Clausal constituents
How to distinguish clauses from non-clauses is not a trivial matter.

For the purposes of this discussion, I shall take clauses to include
constituents which, in some rough way, can express a statement,
question or command. Such constituents include, on the one hand,
finite clauses, and, on the other non-finite clauses, which themselves include gerundial, infinitival and participial clauses as well
as genitive and locative absolutes. The classification just given depends, in part, on the occurrence of a verb, since only verbs are
distinguished into finite and non-finite and only verbs are either
gerunds, infinitives or participles. This means that all clauses have
verbs, but, as is well known, there are constituents which behave
like clauses they express statements or questions but they
have no verb. For the most part, such a constituent is well paraphrased by the clause which results from adding a suitable form of
the copula to the constituent in question. Henceforth, I shall refer
to such constituents as verbless clauses. Clearly, verbless clauses
should be identified as such in a tree bank.
Supra-lexical and sub-clausal constituents
Do Classical Sanskrit clauses have constituents intermediate between words and clauses? In other words, do Classical Sanskrit
241
clauses contain phrases? The answer is clearly yes. They include,

at the very least, noun phrases and prepositional phrases, as we
shall see below. They also probably include adjective phrases and
adverbial phrases, and they may even include verb phrases.
A range of data which requires that one hypothesize phrasal
constituents comes from the fact that human languages have words
or constructions which are context dependent and the construal
of these contextually dependent words or constructions depend on
phrasal constituents for their full construal. Words of this kind are
known as proforms and constructions are known as ellipsis.
4.1
Context Dependence
Before explaining how the existence of proforms and of ellipsis

require the identification of phrases, let me say something about
the nature of context dependence. It is useful to distinguish two
kinds of contexts: cotext and setting. The cotext of an utterance
is the text either preceding or succeeding the utterance; its setting
is the situation in which it is uttered. Two pairs of sentences will
make this distinction clear.
First, consider this pair of sentences:
(1) a. Ren DesCartes died here.
b. Ren DesCartes died in Stockholm.
The second sentence is true, as anyone with an adequate knowledge of Ren DesCartess life can judge, whereas even someone
with a thorough knowledge of his life cannot judge whether or not
the first sentence is true without knowing the circumstances of its
utterance. Clearly, then, a complete understanding of the first sentence requires knowledge of the situation in which it is uttered.
This is confirmed by the following. Holding the sentence in (1a)
fixed and changing the circumstances in which it is uttered, one
sees that its truth value can change. Said in Paris, the sentence is
242
B. G ILLON
false; said in Stockholm, the city to which he moved shortly before

he died, the sentence is true. The first sentence, then, is context
sensitive insofar as its full understanding requires knowledge of
the setting in which it is uttered.
Now, consider another pair of sentences:
(2) a. It is even.
b. Two is even.
These sentences, though syntactically indistinguishable, differ in
one crucial respect. Anyone who knows elementary arithmetic
knows that the second sentence is true, but not even the greatest
mathematician can say whether or not the first sentence is true.
To appreciate the point better, consider two other sentences,
each of which is true:
(3) a. Two is a prime number.
b. Three is a prime number.
Conjoin each of the sentences in (3) with the sentence in (2a),
thereby obtaining the following pair of biclausal sentences.
(4) a. Two is a prime number and it is even.
b. Three is a prime number and it is even.
Even in the absence of all knowledge of the situation in which
they are uttered, one judges the second clause of the first sentence
true and one judges the second clause of the second sentence false.
But the second clause is the very same for both sentences in (4).
Clearly, a complete understanding of the second clause in (4a) depends on an understanding of its preceding text, or cotext. In particular, a full understanding of what the second clause in the sentence in (4a) expresses depends on an understanding that what the
third person pronoun it in its second clause denotes is precisely
243
what the noun two in the first clause denotes. Similar considerations apply to the sentence in (4b).
The smallest expression in the cotext required for the determination of the value to be associated with the pronoun is the pronouns antecedent and the relation which a pronoun bears to its
antecedent is the antecedence relation. Whereas in many cases the
antecedent of a pronoun does, in fact, antecede the pronoun, it may
on occasion succeed the pronoun. Nonetheless, as these terms are
now customary, I shall continue to use them.
While the etymology of the word pronoun suggests that the
pronoun stands for (pro) a noun (noun), in fact, the antecedent for
the pronoun it, at least here, is a noun phrase. This is made clear
by the fact that, were the word two in the sentence in (4a) replaced
by the expression the first positive integer greater than one, clearly
a noun phrase, it would be this expression which would serve as
the antecedent of the pronoun it.
(5) The first positive integer greater than one is a prime number
and it is even.
The etymology of the word pronoun also suggests that a correct paraphrase of the clause containing the pronoun is given by the
clause obtained by replacing the pronoun with its antecedent. As is
well known, while this sometimes is true, it is not true in general.
Finally, not all words which are context dependent like pronouns
are nouns. The adjective such and the adverb so are also context
dependent much in the same way. For this reason, early generative grammarians coined the term proforms, thereby recognizing a
broader class of expression.
I shall indicate the antecedence relation by enclosing the two
terms of the relation in square brackets and labelling the left brackets with the label for the appropriate phrasal category and the right
brackets with the same indexing numeral. This annotation is illustrated in (6) with respect to the sentences in (4).
244
B. G ILLON
(6) a. [NP Two 1] is a prime number and [NP it 1] is even.

b. [NP Three 1] is a prime number and [NP it 1] is even.
Not only are there words which are context dependent like proforms, but there are constructions which are too. Such constructions are said to be cases of ellipsis. Before giving a characterization of ellipsis, let us consider a paradigmatic example.
(7) a. Dan a Porche.
b. Bill bought a BMW and Dan a Porche.
c. Bill sold a BMW and Dan a Porche.
The expression in (7a) is not a constituent of English. Moreover,
by itself, it fails to convey either a proposition, a question or a
command. Yet, when preceded by a clause which expresses a
proposition, say the clause Bill bought a BMW, it too conveys a
proposition. Moreover, the proposition which the non-constituent
in (7a) conveys when preceded by the clause Bill bought a BMW,
is expressed by a clause obtained from the non-constituent by inserting a word from the preceding clause. And finally, what the
non-constituent conveys varies as the preceding clause varies, as
illustrated by the sentences in (7b) and (7c) respectively.
Generalizing this example, we can characterize ellipsis as follows: An expression is elliptical just in case: (1) it itself is not
a constituent; (2) on its own, it fails to convey what any single
constituent conveys; (3) yet, in the presence of an appropriate cotextual constituent it does convey the same sort of information as
what the cotextual constituent conveys; (4) what it does convey is
expressed by a constituent constructed by supplementing the nonconstituent with words from the relevant cotextual constituent; and
(5) what the non-constituent expression conveys varies as the relevant cotextual constituent varies.
245
Now a clause, such as the one in (2a), and an elliptical expression, such as the one in (7a), are similar, in that, with the appropriate cotext, both are judged to be acceptable and both convey
propositions. Yet they differ inasmuch as, when taken in isolation,
the former is not judged to be defective, whereas the latter is.
I shall call an elliptical expression any expression whose
largest constituents fail themselves to form a constituent. Thus,
in the example in (7b) above, Dan a Porche is an elliptical expression. It contains two constituents, the noun phrase Dan and the
noun phrase a Porche, yet they fail to form a constituent. I shall
call the antecedent of the elliptical expression the smallest constituent in the cotext required for the correct construal of the elliptical expression. In the example in (7b), the antecedent of the elliptical expression is the constituent Bill bought a Porche. Both the
elliptical expression and its antecedent will be enclosed in square
brackets and the right bracket for the elliptical expression will bear
the same category label as its antecedent. This is illustrated in the
annotation in (8a) below. The largest constituents within the elliptical expression and their counterparts in the antecedent expression
will be enclosed in square brackets with the phrasal labels. As we
shall see, the constituents from the antecedent expression and the
constituents from the elliptical expression are in bijective correspondence and the pairs in correspondence have the same category.
Next, the expressions within the antecedent expression which are
not counterparts to the largest constituents within the elliptical expression and yet which serve to assist in the construal of the elliptical expression will be put in boldface. In (7b), this is a single
word, the verb bought. It is not unusual for the expressions in
boldface not to form a constituent. Finally, I shall mark with an
underscore the position in the elliptical expression where a copy
of the expressions in boldface in the antecedent can be placed, perhaps slightly altered, to yield a constituent of the same kind as the
elliptical expressions antecedent and which expresses what the el-
246
B. G ILLON
liptical expression conveys relative to its cotext. I shall call this

position the point of ellipsis and the expressions in boldface its
antecedent. (Notice that both elliptical expressions and points of
ellipsis have antecedents.)
(8) a. Bill bought a BMW and Dan a Porche.
[S [NP Bill ] bought [NP a BMW ] 1] and [S [NP Dan ]
[NP a Porche ] 1].
b. Bill sold a BMW and Dan a Porche.
[S [NP Bill ] sold [NP a BMW ] 1] and [S [NP Dan ]
[NP a Porche ] 1].
The expression Dan a Porche is defective. It comprises two
constituents, the noun phrases Dan and a Porche, with a point of
ellipsis between them marked with the underscore. The antecedent
is the entire preceding clause, which is enclosed by a pair of square
brackets, the left bracket labelled with S and the right bracket
with the numeral 1. The two constituents comprising the ellipsis
and the point of ellipsis are enclosed in square brackets, labelled
with the same labels as the antecedent clause. The string from the
antecedent which corresponds to the point of ellipsis is in bold.
4.2
Proforms
The principal proforms in English are the third person personal

pronouns, the relative pronouns, the demonstrative pronouns, the
special pronoun one, as well as demonstrative adjectives and the
definite article. Indeed, at least one English adverb so and one
English adjective such are also liable to having antecedents. Antecedents for the personal pronouns are either noun phrases or
clauses, while those for the adverb so are either verb phrases or
clauses.
247
Classical Sanskrit has a similar range of proforms: the third

person personal pronouns, the relative pronouns, the demonstrative pronouns as well as adjectives or adverbs such as evam (so).
Although no one has studied what sorts of expressions can serve
as antecedents, no one, I think, would dispute the claim that the
personal pronouns, like English personal pronouns, have as antecedents noun phrases, that is, constituents whose head is a noun
and which may, but need not, include other expressions, typically,
modifiers and complements.
Under what conditions can one expression serve as an antecedent to a proform? This question has been the focus of much
of linguistic research since the advent of generative linguistics. As
generative linguistics has made clear for languages such as English
and French, precedence and constituency play central roles in the
formulation of the conditions under which the antecedence relation obtains. This same question has yet to be investigated with
respect to Classical Sanskrit.
4.3
Ellipsis
English has a variety of forms of ellipsis. It includes gapping, interrogative ellipsis (or sluicing), appended coordination (or stripping), verb phrase ellipsis, copular complement ellipsis, nominal
ellipsis, adjectival ellipsis and prepositional ellipsis. As stated
above, ellipsis requires an antecedent. And there are constraints
in English on what can serve as antecedents to points of ellipsis.
Without going into details, suffice the following minimal pair of
sentences to illustrate the fact that such constraints exist. (See
Kehler and Ward 2004: 3.1 for discussion.)
(9) a. Although Bill bought a BMW and Dan a Porche, neither
was happy with his purchase.
b. Although Bill bought a BMW, he was unhappy with it;
*and Dan a Porche.
248
B. G ILLON
Classical Sanskrit has various forms of ellipsis, many forms

very similar to those found in English, including interrogative
ellipsis (sluicing) and appended coordination (stripping), among
others. Let us look at the various forms of ellipsis in some detail.
4.3.1
Interrogative ellipsis (sluicing)
Consider the following two-person dialogue, where the first person makes a statement and the second asks a question about what
the first person stated. The question typically comprises a single
interrogative constituent, which manages to convey the same thing
as the single interrogative constituent integrated into the preceding
statement, as shown below. The Sanskrit translation of the English
sentences equally illustrates interrogative ellipsis.
(10)
A: [S Devadatta teaches grammar 1]

B: [S Where 1] ?
i.e. Where does Devadatta teach grammar?
A: [S Devadattah. vyakaran.am sa sti 1]
B: [S kutra 1] ?
i.e. kutra Devadattah. vyakaran.am sa sti?
(For the sake of clarity and simplicity, the sentences in Classical

Sanskrit given as examples are not written observing the rules of
external sandhi.)
I postpone the presentation of a case of interrogative ellipsis
involving a single interrogative phrase until later, since the example, taken from Patajalis Mahabhas.ya, involves other forms of
ellipsis which must be explained first.
Interrogative ellipsis may also arise with indirect questions, as
illustrated by the next sentence.
(11)
[S Devadatta teaches grammar 1] but

[S Yajadatta does not know where 1] .
249
i.e. Yajadatta does not know where Devadatta teaches

grammar.
[S Devadattah. vyakaran.am sa sti 1] kimtu
[S Yajadattah. na janati kutra 1] .

i.e. Yajadattah. na janati kutra Devadattah. vyakaran.am
sa sti.
However, I have found no such instances in my repertoire of Classical Sanskrit sentences.
4.3.2
Appended coordination (stripping)
Appended coordination is where a non-clausal constituent is appended to a clause, typically with a coordinator. Here are two
examples from English.
(12) a. [S Devadatta teaches grammar 1], but
[S only in Pat.aliputra 1].
i.e. Devadatta does so only in Pat.aliputra.
b. [S Devadatta goes to Pat.aliputra 1], but
[S never with friends 1].
i.e. Devadatta does so never with friends.
Here is a case found in Classical Sanskrit.
(13) prajabhih. tu bandhumantah. rajanah. na jatibhih.. (Apte
30.2.2: HC 79.30)
[S [VP 0 [AP1 [NP3 prajabhih. [CNJ tu ] ] bandhumantah. ]
] [NP1s rajanah. ] ] [S [ADV na ] [ NP3 jatibhih. ] ]
But kings are bound to their subjects, not to their relatives.
(The numeral appended to the phrasal category NP labelling the
left bracket indicates the case of the noun which is the head of the
noun phrase. The numeral refers to the Pan.inian numbering of the
250
B. G ILLON
cases. The letter s, when appended to the label NP1 indicates

that the first case noun phrase is a subject noun phrase.)
4.3.3
Gapping
We now turn to gapping, illustrated above for English, in its simplest form, by the sentences in (7). To begin with, consider this
illustration:
(14) [S The family may be said to be more unified than the state
1],
and [S the individual [PP than the family ] 1].
i.e. the individual may be said to be more unified than the
family.
(Benjamin Jowetts English translation of Aristotles Politics
1261a2022)
Notice that the antecedent for the point of ellipsis, may be said to
be more unified, does not form a constituent. It looks as though
gapping may be a form of ellipsis to which constituency is irrelevant. However, that conclusion would be too hasty. Consider
another example:
(15) [S [NP The towns ] they attacked [PP with airplanes ] 1] and
[S [NP the villages ] [PP with tanks ] 1] .
i.e.: the villages they attacked with tanks.
(Quirk et al. 1985: 975)
The unit containing the ellipsis comprises two constituents,
namely, the noun phrase, the villages, and prepositional phrase,
with tanks. The noun phrase the villages corresponds to the noun
phrase the towns in the initial clause and the prepositional phrase
with tanks corresponds to the prepositional phrase with airplanes
also in the initial clause. The antecedent is the non-constituent expression in the initial clause between the noun phrase the towns
251
and the prepositional phrase with airplanes. In other words, the

antecedent expression is the string complement, as it were, of the
constituent whose initial subconstituent corresponds to the first
constituent in the unit containing the gap and whose final subconstituent corresponds to the final constituent in the unit containing
the gap. Though this is not a perfect description of the configuration of gapping, it is the best I know of.
Classical Sanskrit has many cases of gapping where the antecedent to the point of ellipsis is a simple verb.
(16) a. sa utpadyamanena yatha gava sambadhyate tatha mrd
gavakena api.
[S sa ] [NP3 utpadyamanena [ADV yatha ] gava ] [V
sambadhyate 1] [S [ADV tatha ] [NP3 mrd-gavakena api
] 1]
NS 2.2.63, Varttika thereto
Just as it is connected with an existing cow, so (it is connected) with a clay cow.
b. yavat uktam anityah. sabdah. anityatvat iti tavat anityah.
krtatvat iti.
yavat uktam [S anityah sabdah anityatvat iti ] 1] tavat
[RC
.
.
[S [S anityah. krtatvat iti ] 1]
PVSVT. 108.67
Just as it is said that sound is non-eternal because it is noneternal, so (it is said that it is) non-eternal because it is a
product.
(RC labels relative clauses.) I have found no cases in Classical
Sanskrit where gapping involves what otherwise would be considered a non-constituent.
We now turn to various forms of ellipsis within phrases, beginning with verb phrase ellipsis. English has two forms of ellipsis
within verb phrases, one is so-called verb phrase ellipsis, the other
is copular complement ellipsis.
252
4.3.4
B. G ILLON
Verb phrase ellipsis
English verb phrase ellipsis is where an entire verb phrase is ellipted, except for an auxiliary verb. Here is an example.
(17) Devadatta [VP likes milk 1] and Yajadatta [VP does
too.
1]
I do not recollect ever seeing a case of ellipsis in Classical

Sanskrit where the complements of a verb have been ellipted and
an auxiliary verb stands in for the verb in the antecedent. What
does occur in Classical Sanskrit are cases where the complements
of a verb have been ellipted, but the verb, identical with the verb
of the antecedent, remains.
4.3.5
Copular complement ellipsis
The other form of ellipsis pertaining to English verb phrases is

copular complement ellipsis. As the name suggests, the antecedent
for the point of ellipsis is a phrasal complement to the copula in
the clause containing the antecedent. Here are examples.
(18) a. Devadatta [VP is generous 1], but Yajadatta [VP is
too.
1]
b. Devadatta [VP is a brahmin 1], but Yajadatta [VP is

1] too.
c. Devadatta [VP is in Pat.aliputra 1] and Yajadatta [VP is
1] too.
It is difficult to determine on the basis of only a few examples whether or not Classical Sanskrit has a comparable form of
ellipsis. The following example contains a point of ellipsis whose
antecedent is a noun phrase, but it also has no copula.
(19) yatha dhumena pratyaks.en.a a-pratyaks.asya vahneh. grahan.am
anumanam evam gava pratyaks.en.a a-pratyaks.asya gavayasya
253
grahan.am.
NS 2.1.46 Bhas.ya
yatha dhumena pratyaks.en.a a-pratyaks.asya vahneh. grahan.am
[VP anumanam 1] evam gava pratyaks.en.a a-pratyaks.asya
gavayasya grahan.am [VP 1]
Just as the grasping of a (previously) unperceived fire
through (previously) perceived smoke is an inference, so is
the grasping of a (previously) unperceived gaval through a
(previously) perceived cow (an inference).
4.3.6
Nominal ellipsis
Nominal ellipsis in English is fairly complex and has been much

less studied than verb phrase ellipsis. In English nominal ellipsis,
the head can be ellipted, constituents preceding the head can be
ellipted and constituents following the head can also be ellipted.
Examples of each are given below.
(20) a. nominal head ellipsis:
Carol picked up [NP Bills coat 1] and Joan picked up [NP
Eds 1] .
b. nominal prehead ellipsis:
Bill will put [NP this knife 1] and [NP
table.
fork 1] on the
c. nominal post head ellipsis:

[NP An article on this topic 1] is more likely to be accepted than [NP a book 1] .
(Payne and Huddleston 2002: 424)
The sentence next below contains an example both of nominal head ellipsis as well as interrogative ellipsis. Notice that the
interrogative constituent which follows the first sentence, kes.a m
sabdanam (in which kinds of words) is a case of nominal head
254
B. G ILLON
ellipsis and its antecedent is anusa sanam (instruction). The English translation exhibits the same form of nominal head ellipsis.
What is interesting is that, in the Sanskrit sentence, the antecedent
for the point of nominal head ellipsis is the head of a compound
sabda-anusa sanam (word instruction), whereas, in the case of its
English translation, the antecedent is the head of a noun phrase
instruction in words. Unlike Classical Sanskrit, English does not
permit antecedents either for ellipsis or for proforms to be proper
constituents of words.
(21) sabda-anusa sanam sa stram adhikrtam veditavyam. kes.a m
sabdanam? laukikanam vaidikana
m ca.
MBhP 1.1
[S [NP1s sabda-(anusa sanam) 1] sa stram adhikr.tam veditavyam 2].
[S [NP1s [NP6 kes.a m sabdanam 3]
1]
2]
[S [NP1s [NP6 [AP6 laukikanam vaidikanam ca ] 3] 1]

2]
Instruction in words should be understood as the discipline
under discussion. In which (kinds of) words? In ordinary
ones and Vedic ones.
The same example contains a second case of nominal head
ellipsis found in the last expression, which serves as an answer to
the interrogative constituent. The point of ellipsis is rendered in
English by the pronoun one.
While English has cases both of ellipsis within adjectival
phrases and within prepositional phrases, such cases are cases
where the ellipted expression is the head.
(22) a. Ptolemy thought that Venus is [AP farther from the Earth
1] and Mercury (is) [AP closer 1] .
b. You may buy a car [PP with airbags 1] or (one) [PP without 1] .
255
The situation is unknown in Classical Sanskrit.
Further evidence for Phrases
The lesson of the foregoing is that cotextual dependence, the sort

accruing to the use of proforms and ellipsis, requires the recognition of constituents intermediate between the word and the clause,
typically noun phrases. Every phrase must have a head and may
have other phrases as well, distinguished either as modifiers or as
complements.
While modifiers are always omissible from a phrase without
disturbing the acceptability of the clause in which the phrase occurs, complements are often not omissible. Below, I furnish examples of English sentences containing words of each of the principal
lexical categories whose complements are inomissible. I also provide their translations into Classical Sanskrit.
(23) a. Adjective complement
Devadatta is averse *(to study).
Devadattah. adhyayanat apavrttah. vartate.
b. Noun complement
There is a lack *(of water) in the well.
Asti jalasya abhavah. kupe.
c. Preposition complement
Rama left Ayodhya without *(Sta).
Ramah. vina Stam Ayodhyayah. prayat.
d. Verb complement
Devadatta placed an offering *(on the altar).
Devadattah. balim sthan.d.ile nyadadhat.
(The asterisk preceding the left parenthesis indicates that the omission of the portion enclosed in parenthesis results in an unacceptable expression.)
256
B. G ILLON
It is easy to identify verb phrases, prepositional phrases and

adjective phrases in Classical Sanskrit sentences. Moreover, as illustrated from the following sentences, the subconstituents of any
of these kinds of phrases are often found in one contiguous sequence.
(24) a. AP complement (Apte 3.1.4: S 3.5.2)
es.a [AP1 [NP6 me ] manoratha-priyatama ].
This (woman), most dear to me among my desires.
b. NP complement (Apte 1.1.1: V 1.13)
[NP1 sa [NP6 nah. ] priya-sakh ].
[NP This dear friend [PP of ours ] ].
c. PP complement (Apte 24.1.5: Mu 1.5.11)
[PP [NP2 (candra-uparagam) ] prati ] tu kena api vipralabdha asi.
You have been deceived by someone [PP about [NP the
eclipse of the moon ] ].
d. VP complement (Apte 7.1.9: K 219.18)
Candrapd.ah. [VP [NP2 tam ] pran.anama ].
Candrapd.ah. [VP bowed [PP to her ] ].
And though it is not unusual for a phrase to be discontiguous, when
it is, there is a definite pattern: the part of the discontiguous phrase
which is not with the head is found at the end of the clause in which
the phrase occurs. (See Gillon 1996 as well as Gillon and Shaer
2005 for details.)
One of the confounding factors in distinguishing modifiers
from complements is that complements are sometimes optional.
Here are examples of words with optional complements both from
English and from French.
(25) a. Peter ate (something).
Pierre a mang (quelque chose).
257
b. Peter arrived (here) this morning.

Pierre est arriv (ici) ce matin.
c. Peter took a plane to Mumbai. He arrived (there) this
morning.
Pierre a pris lavion pour Mumbai. Il (y) est arriv ce
matin.
d. Peter and I are similar (to each other).
Pierre et moi sommes similaires (lun lautre).
e. Peter dressed (himself).
(The optional complements in the examples have been placed in
parentheses.)
There are several kinds of evidence which confirm that such
optional constituents are indeed complements and not modifiers. I
shall mention only two kinds. First, whether or not a phrase serving as a complement is obligatory depends on the specific lexical
item to which it is appended as a complement. Many words with
optional complements have near synonyms for which the corresponding constituent is not optional. Here are some examples for
the English words with optional complements illustrated by the
sentences in (25).
(26) a. Peter ate (something) .
Peter devoured *(something)
b. Peter arrived (here) this morning.
Peter reached *(here) this morning.
c. Peter took a plane to Mumbai, He arrived (there) this
morning.
Peter took a plane to Mumbai. He reached *(there) this
morning.
d. Peter and I are similar (to each other).
Peter and I resemble *(each other).
258
B. G ILLON
e. Peter dressed (himself).
Peter clothed *(himself).
Moreover, when optional complements are omitted, they can

be paraphrased with a canonical paraphrase, since the omission
of the complement gives rise to a construal characteristic of the
omission. For example, when the complement of the verb to eat is
omitted, it is construed as to eat something. When the complement
of the verb to arrive is omitted, it is construed as to arrive in some
definite place, contextually determined, be it from the setting or
from the cotext, as in the sentences in (26b). The adjective similar
yields a reciprocal construal when its complement is omitted. On
the omission of its complement, the verb to dress yields a reflexive
construal. (See Gillon 2012 for further discussion of English.) The
situation in English is summarized in Table 1.
Table 1
English optional complementation
E NGLISH contextual indefinite reciprocal reflexive
Verbs:
yes
yes
yes
yes
Adjectives:
yes
yes
yes
no
Nouns:
yes
yes
yes
no
Prepositions:
yes
no
no
no
This phenomenon of optional complementation is well attested
across languages, although little attention has been devoted to what
the range of variability of construal is when the complements are
omitted. Thus, Chinese and other East Asian languages certainly
abound in optional complements whose omission yields a contextual construal. Here is an example of a Chinese (Mandarin) sentence where the cotext determines the relevant construal.
(27) Wnt sh, xinzi sh b sh d [NP laoz 1] de shhu.
Women b nng ynwi wigu d [NP c 1] ji jude gai
259
d [NP c 1] .
Gang 1997, Rng women li d laoz, Jiu Sh Nin Di v.
7 (cited in Ling 1997: 9).
The question is: is now the time for us to read Laozi? It
cannot be that because foreigners read *(it) that we feel we
ought to read *(it).
(I indicate the omitted complement by putting the letter c enclosed in brackets. The left bracket is annotated with the label for
the phrasal category to which the complement would belong, were
it present. The letter c signals that the value of the constituent
is determined contextually, that is, it is determined either by the
cotext or by the setting. If the value is determined by the cotext,
then the right bracket is annotated with the numeral indicating its
antecedent in the cotext.) Notice that the English translation would
be unacceptable if the pronoun it were omitted.
Words in Classical Sanskrit also permit optional complements.
The examples below are all instances of nouns whose complements, when omitted, give rise to a contextual construal, where
the relevant context is provided by the cotext.
(28) prayojana-abhavat an-upacarah.. Na sarva-dharmi-dharmapratis.edha-artha-tvat.
PVSV 1.1213
[NP5 [NP6 c 1] (prayojana)-abhavat ] [NP1 (an-(upacarah.
1))].
na [NP5 [NP6 c 1] ((sarva-dharmi)-dharmapratis.edha-artha)-tvat].
It might be argued that there is no [NP synecdoche 1] (here)
since [NP it 1] would have no point. No, because [NP it 1]
has the purpose of prohibiting (from being a ground) a property property-possessors of which are everything (else).
Whereas the first English sentence, which translates the first
Sanskrit sentence, requires the pronoun it to express what it is
260
B. G ILLON
which has no point, namely, synecdoche, the Sanskrit sentence has

no pronoun, yet it conveys no less equivocally that what it is which
has no point is synecdoche. In other words, though the complement of prayojana (point) is omitted, nonetheless what would be
its complement is clearly understood.
The next example shows an instance of an omitted complement
whose antecedent is split between the subject noun phrase and the
object noun phrase of the main clause.
(29) svabhava-pratibandhe hi sati arthah. artham na vyabhicarati.
PVSV 2.19
[AC [NP7 [NP6 c 1,2] svabhava-pratibandhe ] hi [PC7 sati ]
] [NP1s arthah. 1] [VP [NP2 artham 2] na vyabhicarati ].
For, [NP a thing 1] does not deviate from [NP a thing 2],
when there is a natural relation *(between [NP them 1,2]).
(AC denotes absolutive clause. Its counterpart to a subject appears as a seventh case noun phrase and its counterpart of a verb
appears as a participle. The splitting of antecedents requires double indexing.)
The next sentence is an example of where three complements
are omitted, one of the two complements of the noun an-utpatti
(non-arising), one of the two of the noun utpatti (arising) and the
one complement of the nominalization of the adjective pratibhasin
(appearing). The cotextual constituent providing the construal of
each of these omitted complements is the subject noun phrase etau
dvau anumeya-pratyayau (these two kinds of cognition of what is
inferred).
(30) etau dvau anumeya-pratyayau saks.a t an-utpatteh. a-tatpratibhasi-tve api tat-utpatteh. tat-avyabhicarin.au.
PVSV 3.5
[S [NP1s etau dvau (anumeya1)-pratyayau 2] [NP5 [AC
261
[NP7 [NP5 [NP6 c 2] [ADV saks.a t ] an-utpatteh. ] [NP6 c

2] (a-(tat1)-pratibhasi)-tve api ] [PC7 0 ] ] [NP6 c 2] (tat1)utpatteh. ] [NP1 (tat1)-avyabhicarin.au ] ].
[NP These two (kinds) of cognitions of [NP what is to be
inferred 1] 2] do not deviate *(from [NP it 1]) because [NP
they 2] arise from [NP it 1], although, because they do not
arise directly *(from [NP it 1]), they do not manifest *([NP
it 1]).
It is interesting to note that, in addition, the sentence contains pronouns whose antecedent is the subordinate word anumeya (what
is to be inferred) in the compound anumeya-pratyayau (cognition
of what is to be inferred), which is the head of the subject noun
phrase.
So far, I have provided examples of optional complements
to nouns, however verbs too admit optional complements whose
omission triggers a contextual construal. This observation seems
to have first been noted by the great philosopher and grammarian
Bhartrhari. (See Gillon 2001 for further examples and detailed
discussion.)
Here is an example.
(31) sadhye pratisandhaya dharmam udaharan.e ca pratisandhaya tasya sadhanata-vacanam hetuh..
NS 1.1.34 Bhas.ya
[GC sadhye pratisandhaya [NP2 dharmam 1] ] [GC
udaharan.e ca [NP2 c 1] pratisandhaya ] [NP6 tasya 1]
sadhanata-vacanam hetuh..
The ground (of an argument) is the statement of [NP a property 1] as an establisher, having related [NP it 1] to what is to
be established and having related [NP it 1] to the corroborating instance.
(GC denotes gerund clause. Gerund clauses have nothing which
corresponds to a subject.) Here, dharma (a property) serves as the
262
B. G ILLON
complement of the first occurrence of the gerund pratisandhaya

(having related) and the cotextual determinant of the second occurrence of the gerund.
Whether or not a word takes a complement, and if so, whether
the complement is obligatory or optional, is something which must
be specified for each word. The past half century of generative
linguistics has shown that, at least for languages such as English
and French, whose finite clauses in the declarative mood require
subjects, the notion of a complement must be broadened to take
account of subjects. The term employed, argument, taken from
logic, includes complements and subjects. Thus, complements and
subjects are all argument positions. The upshot is that a grammar
must specify, for each word, which arguments it takes, in other
words, to use another term from logic, the words adicity. While
evidence for the utility of this broader notion has been almost exclusively syntactic, yet semantic considerations are decisive: one
can not provide a model for natural language expressions without
a specification of each words adicity.
It might be thought that semantic roles, karakas, could do the
same job of adicity, or argument specification. This is not so. To
begin with, some arguments have no semantic role associated with
them, while some modifiers do.
(32) a. Sta bowed to Rama in the courtyard.
Sta Rama angane pran.anama.
b. Yajakah. hutam agnau nyadadhat.
The priest placed an oblation on the fire.
In English, the prepositional phrase in the sentence in (32a) is a
modifier, but the one in the sentence in (32b) is a complement.
In addition, the complements of various nouns, adjectives,
prepositions and verbs often take a complement, but they are not
plausibly assigned any semantic role at all. For example, nouns
263
denoting individuals through relations of kinship (mother, father,

parent, daughter, son, child, cousin, etc.), nouns denoting individuals in social or institutional roles (husband, wife, neighbor,
enemy, etc.), nouns denoting things which bear an informational
relation (sign, evidence, picture, etc.), adjectives denoting intentionally based states (angry, cross, ready, sure, afraid), adjectives
denoting spatial ordering and temporal ordering (close, far, near
earlier, later, near), prepositions (before, after, soon, for, against,
etc.), verbs denoting states (to stand, to sleep, to remain, etc.).
Indeed, it seems that semantic roles are best understood as very
general selection restrictions.
Conclusion
We have seen that the treatment of the antecedence relation,

whether in the case of ellipsis or in the case of proforms, requires
the identification of constituents, not only constituents within compounds but also constituents within clauses, in particular, noun
phrases. Essential to the concept of a phrase are the concepts of a
phrases head and the heads modifiers and complements. We then
saw that, Classical Sanskrit, like many other languages, has optional complements, including ones which have antecedents. We
also saw that it is useful to generalize the notion of complement
to that of adicity, a concept borrowed from logic. I pointed out
that, while the complements of prepositions, those of many nouns
and those of some verbs fall within the purview of adicity, such
complements are not properly characterized in terms of karakas,
or semantic roles. The point here is not to dispense with karakas,
or semantic roles, rather the point is that the more general notion
of adicity is needed within which the notion of karaka, or semantic
role, may be subsumed.
264
B. G ILLON
Table 2
Abbreviations
Apte
Apte 1885. Reference to: chapter . exercise set . example sentence

HC
Harsacarita. Kane 1918. Reference to: page . line
K
Kadambar. Peterson 1883. Reference to: page . line
MBhP Mahabhas.ya Paspasa hnika. Joshi and Roodbergen
1986. Reference to: page . line
Mu
Mudraraks.asam. Telang 1884. Reference to: act .
verse . line after verse
NS
Nyayasutra. Taranatha and Amarendramohan 1936
1944. Reference to: adhyaya . a hnika . sutra
PVSV Praman.avarttika. Gnoli 1960. Reference to: page .
line
PVSVT. Praman.avarttika. Sam
. kr.tyayana 1943. Reference to:
page . line
S
Sakuntal
a. Godabole 1905. Reference to: act . verse .
line after verse
V
Vikramorvasyam. Pandit 1879. Reference to: act .
verse . line after verse
ACKNOWLEDGEMENTS :
The research in this paper was supported in part by the Social

Sciences and Humanities Research Council (SSHRC 412-20101007), whose support I am pleased to acknowledge.
References
Apte, Vaman Shivaram. 1885. The students guide to Sanskrit composition: being a treatise on Sanskrit syntax for the use of
schools and colleges. Second edition . . . enlarged. Pune: H. N.
Gokhale. Printed at Aryabh

us.an.a Press.
R EFERENCES
265
Gang, Sh. 1997. rng wo-mn li d Lao Z (Let us read Laozi).

The independent reader, ed. by Vivian Ling, pp. 914.
Gillon, Brendan S. 1987. Two forms of negations in Sanskrit:
Prasajyapratis.edha and paryudasapratis.edha. Lokapraja 1.1:
8189. Puri: Lokabhas.a Pracara Samitih., Bada Odiya Matha.
. 1995. Autonomy of word formation: evidence from Classical
Sanskrit. Indian Linguistics 56.14: 1552.
. 1996. Word order in Classical Sanskrit. Indian Linguistics
57.14: 136.
. 2001. Bhartrharis rule for unexpressed karakas: the problem of control in Classical Sanskrit. Indian linguistic studies:
festschrift in honour of George Cardona, ed. by Madhav Deshpande, pp. 93111. Delhi: Motilal Banarsidass.
. 2012. Exocentric (bahuvrihi) compounds in Classical Sanskrit. Samskrta-sadhuta. Goodness of Sanskrit. Studies in
Honour of Professor Ashok N. Aklujkar, ed. by Chikafumi Watanabe, Michele Desmarais, and Yoshichika Honda,
pp. 24057. D. K. Printworld.
Gillon, Brendan S. and Benjamin Shaer. 2005. Classical Sanskrit,
wild trees and the properties of free word order languages.
Universal grammar in the reconstruction of ancient languages,
ed. by Katalin . Kiss, pp. 45794. Berlin: de Gruyter.
Gnoli, Raniero, ed. 1960. Praman.avarttikam. Serie Orientale
Roma. Rome: Istituto Italiano per il Medio ed Estremo Oriente.
Godabole, Narayan.a Balakr.shn.a, ed. 1905. Sakuntal

a, rev. by Wasudeva Laxmana Shastri Pan.skar. 4th ed. Bombay: Nirn.aya
Sagara Press. [Reprint: 1960.]
Joshi, Shivram Dattatray and J. A. F. Roodbergen, eds. 1986.
Patajalis Vyakaran.a-Mahabhas.ya Paspasa hnika, introduction, text, translation and notes. Publication of the Centre of
Advanced Study in Sanskrit, Class C 15. Pune: The University
of Poona.
266
B. G ILLON
Kane, Pandurang Vaman, ed. 1918. The Harshacarita of

Ban.abhat..ta: text of uchchhvasas IVIII. Bombay. [Reprint:
Delhi: Motilal Banarsidass, 1965.]
Kehler, Andrew and Gregory Ward. 2004. Constraints on ellipsis and event reference. The handbook of pragmatics, ed. by
Laurance R. Horn and Gregory Ward, pp. 383403. Blackwell.
Ling, Vivian, ed. 1997. The independent reader. Taipei: SMC Publishing.
Pandit, Shankar Pandurang, ed. 1879. Vikramorvasyam, a drama
in five acts by Kalidasa, edited with English notes. Bombay
Sanskrit Series 16. 2nd edition, 1889. Bombay: Government
Central Book Depot.
Payne, John and Rodney Huddleston. 2002. Nouns and noun
phrases. The Cambridge grammar of the English language,
ed. by Rodney Huddleston and Geoffrey K. Pullum, pp. 323
524. Cambridge: Cambridge University Press.
Peterson, Peter, ed. 1883. Kadambar by Bana and his son. Bombay Sanskrit Series 24. Bombay: Government Central Book
Depot. [Reprint: Delhi: Motilal Banarsidass, 1965.]
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan
Svartik. 1985. A comprehensive grammar of the English language. 2nd edition (revised). London: Longman.
arya-Dharmakrteh. Praman.aSam
. kr.tyayana, Rahula, ed. 1943. Ac
varttikam (svarthanumana-paricchedah.) Svopajavr.ttya Karn.akagomi-viracitaya tat..tkaya ca sahitam. Allahabad: Kitab
Mahal.
Taranatha and Amarendramohan, eds. 19361944. Nyayadarsanam with Vatsyayanas Bhas.ya, Uddyotakaras Varttika,
Vacaspati Misras Tatparyat.ka and Visvanathas Vr.tti. Calcutta Sanskrit Series 1819. Calcutta: Metropolitan Printing
and Publishing House Limited. [Reprint: Delhi: Munshiram
Mahoharlal Publishers Limited.]
R EFERENCES
267
Telang, Kashinath Trimbak, ed. 1884. Mudrarakshasa written by

Visa khadatta, with Commentary of D
. hund.hiraja (written in
1713 AD) edited with critical and explanatory notes. Bombay
Sanskrit Series 27. Bombay: Government Central Book Depot.
[Reprint: Delhi: Nirn.aya Sagara Press.]
268
B. G ILLON

Abstract: Sanskrit being inflectionally rich, the conventional
wisdom about Sanskrit word order is that it is free. The
concept of sannidhi (proximity), one of the necessary factors in the process of verbal cognition, provides a constraint on the word order of Sanskrit. We study the free
word order of Sanskrit in the light of the dependency
framework. The weak non-projectivity condition on dependency graphs captures the sannidhi constraint. Gillon
worked within the framework of phrase-structure syntax
and noted that the freeness is constrained by clause boundaries. In an examination of the cases of dislocation observed by Gillon and all verses of the Bhagavadgta, we
notice that two relations, viz. adjectival and genitive, are
more frequently involved in sannidhi violation. We conclude that the relations involved in sannidhi violation correspond to utthapya-akan ks.a (expectancy which is to be
raised) barring a few exceptional cases.
Keywords: word order, a kan ks.a , utthita a kan ks.a , utthapya
a kan ks.a , sannidhi, projectivity, dislocation, phrase structure, dependency structure
Introduction
Sanskrit being inflectionally rich, the conventional wisdom about

Sanskrit word order is that it is free. It is also a common understanding among linguists that the free-ness in Sanskrit is confined
269
270
A. K ULKARNI
by clause boundaries. That is, it is not possible to freely interleave

the constituents from subordinate clauses with elements from the
main clause or other subordinate clauses. Thus the mixing of elements from different clauses is not allowed. In order to build a
parser, this conventional wisdom needs to be formalised. It is necessary to know whether there are any exceptions to it, and if there
are what their nature is.
The first systematic work on Sanskrit word order is by Staal
(1967). He discussed the distinction between sambandha and
abhisambandha in the beginning of his monograph. The former
indicates grammatical relations while the latter refers to word order or arrangement. Pan.ini was very much aware of the fixed order
between certain elements such as the order between a stem and a
suffix,1 the order between a prefix and a verb, and the order between the components of a compound. For example he designates
a certain component an upasarjana and then specifies the position
of the component so labeled in a compound. From the classical
Sanskrit literature, we are aware of the cases where the position
governs the meaning of the particles such as api. Speijer (1886)
lists several of them.
Staal concluded that almost all Indian theorists regarded word
order as free either implicitly or explicitly while the western Sanskritists rejected the free word order theory and classified the types
of word order into preferential, traditional, habitual, etc. based on
probabilities of occurrence and frequencies. For building a parser,
though specifying preferential word order is useful in prioritizing
multiple analyses produced, more precise constraints in terms of
what is not allowed in grammar would help in pruning out impossible parses.
1 pratyayah
. A. 3.1.1, and parasca A. 3.1.2, a suffix is placed after the root,

nominal base, or item ending in a feminine affix in relation to which it is introduced.
H OW FREE IS FREE WORD ORDER ?
271
In the same monograph, Staal presented his model of free word

order in Sanskrit. He noticed the analogy between the movement
of sister nodes in a Calder mobile and the free movement of elements within a phrase in a sentence. Staal called a tree that allowed
the free movement of its constituents a wild tree. In this model,
the sisters under each node can be freely transposed provided that
such transposition induces no tangling. Although Staal provided a
model characterising free word order, still he did not give a precise
definition that can be used to test the model empirically. Gillons
(1996: 7) empirical study led him to formalise Staals notion of
wild trees. Based on this empirical study, he noticed a pattern in
the discontinuities and hence modified Staals conjecture considering the discontinuity as a movement. He sums up his observations
as follows:
A moved constituent is either
1. a complement of the verb;
2. a complement or modifier of the subject noun
phrase; or
3. a complement or modifier of a complement of
the verb.
These observations provide useful relaxations for building a constituency parser.
While phrase structure grammar is suitable for understanding the constituency structure of a sentence, dependency grammar
helps us in understanding the relations between various components and the semantics associated with the sentence. In the field of
Natural Language Processing (NLP), recent years have also seen
a growing trend towards producing dependency output in addition
to constituency trees. The dependency format is preferred over the
constituency not only from the point of view of evaluation (Lin
2003) but also because of its suitability for a wide range of NLP
272
A. K ULKARNI
tasks such as Machine Translation (MT), information extraction,

question answering etc. (Marneffe, MacCartney, and Christopher
D. Manning 2006). The importance of dependency parsing has
been well recognised by computational linguists in the recent past
(Culotta and Sorensen 2004; Haghighi, Ng, and C. D. Manning
2005; Quirk, Menezes, and Cherry 2005).
Gillons observation that discontinuities are clause-bounded,
implies that all the dependents of the head of the clause are still
within the clause boundary, though they are free to move within
a clause. Dependency parsing has an adjacency principle, also
known as the projectivity principle, which is similar to this ban on
discontinuous structures in phrase structure grammar. We would
like to see how various cases of dislocation in phrase structure
grammar translate into the dependency framework. While phrase
structure analysis reveals extraposition from various positions, the
facts observed so far raise some questions which lead us to take up
this study further. Does extraposition lead to violation of the adjacency principle? Does it lead to any tangling in the dependency
structure? Do poetic constructions bring in any tangling? Is verse
more free than prose?
In what follows, we explain the concept of a kan ks.a (expectancy) and sannidhi (proximity) as discussed in the Indian
grammatical tradition. Sannidhi imposes certain restrictions on
word order. We compare the concept of sannidhi with the projectivity principle and the weak non-projectivity conditions of dependency trees. We attempt to formalise this notion of sannidhi
and the violation of sannidhi with regard to two types of a kan ks.a
(expectancy), viz., utthita (risen) and utthapya (that which is to be
raised) in order to test the constraint empirically. All the examples
of dislocation discussed by Gillon are studied from the point of
view of sannidhi. Further, we test this constraint on all the verses
of the Bhagavadgta (BhG.).
273
Word order in Sanskrit
Let us look at a few Sanskrit examples of possible and impossible

word orders. For the sentence
(1) ramah. gramam gacchati.
Rama{nom.} village{acc.} go{3p. sg.}.
Rama goes to a village.
with 3 words, we get 3! (= 6) possible word orders as shown below:
(1.1)
(1.2)
(1.3)
(1.4)
(1.5)
(1.6)
ramah. gramam gacchati.

ramah. gacchati gramam.
gacchati ramah. gramam.
gramam ramah. gacchati.
gramam gacchati ramah..
gacchati gramam ramah..
These sentences convey the same overall meaning. But they differ
from each other with respect to some additional meaning such as
topicalisation, focus, etc. In a given context, one of them may be
more suitable than the others. Some orders may be less frequent
than the others. For example, consider the sentence
(2) svetah. asvah. dhavati.
White{nom.} horse{nom.} run{3p. sg.}.
White horse runs.
The six possible word orders with the same overall meaning are
(2.1)
(2.2)
(2.3)
(2.4)
(2.5)
svetah. asvah. dhavati.

svetah. dhavati asvah..
asvah. svetah. dhavati.
dhavati svetah. asvah..
asvah. dhavati svetah..
274
A. K ULKARNI
(2.6) dhavati asvah. svetah..

Though all these are acceptable, and convey the same overall
meaning that a white horse runs, (2.1) and (2.4) are more probable than others. In (2.2) and (2.5) the modifier or the modified is
added as an after-thought. In (2.3) and (2.6) the modifier appears
after the modified, as if it is an after-thought.
Let us look at another example (3).
(3) ramah. dugdham ptva sa lam gacchati.
Rama{nom.} milk{acc.} drink{absolutive} school{acc.}
go{3p. sg.}.
Rama goes to school after drinking milk.
This sentence has 5 words. Are all the 5! (= 120) combinations
meaningful?
(4) *ramah. sa lam dugdham gacchati ptva.
*Rama{nom.} school{acc.} milk{acc.} go{3p. sg.}
drink{absolutive}.
*Rama to school milk goes drinking.
Sentence (4) which is obtained by permuting the words in sentence
(3) does not lead to any verbal cognition, and thus it shows that all
permutations need not be meaningful. So the question is, which
permutations are meaningful and which are not? We look at the
Indian grammatical theories that deal with these problems.
Indian theories of expectancy and proximity
The process of verbal cognition involves analysis of a sentence,

and this analysis typically involves non-determinism. The prob2
lem of non-determinism was well recognised by the mmamsakas

2 Kumarila
Bhat.t.a in his Tantravarttika (Sastri 1903: 5056) mentions 3 factors necessary for the understanding of the correlation between words:
275
who proposed three factors, viz., a kan ks.a (expectancy), yogyata

(mutual compatibility), and sannidhi as necessary conditions for
proper verbal cognition. We discuss here a kan ks.a and sannidhi,
which are relevant for studying the problem of dislocation and
word order.
anks
. a (expectancy)
3.1 Ak
We first come across the term a kan ks.a in the definition of a sentence in Jaiminis Mmams
asutra 2.1.46.
Arthaikatvat ekam vakyam sakamks
. am cet vibhage
syat.
A group of words forms a sentence i) if when separated the words
have mutual expectancy, and ii) the words serve a single purpose.
Thus a kan ks.a or syntactic expectancy among words is a necessary condition for a group of words to form a sentence. Literally a kan ks.a is the desire on the part of a listener to know (jatum
iccha) other words in a sentence for complete understanding. Now
if a kan ks.a is the curiosity (jijasa) on the part of a listener then
after listening to a verbal form such as a nayati brings, a listener
will have a curiosity to know who brings, what s/he brings, how
s/he brings, and so on. Further if the object of bring is, say, a cow,
the listener may have further curiosities to know what the color of
the cow is, what the purpose of bringing her is, and so on. There
is no end to such curiosities. These curiosities are more of a psychological nature than a syntactic one.
The Naiyayikas made a clear distinction between psycholog an ks.a according to Naiyaikas
ical and syntactic expectancy. Ak
is the syntactic expectancy a word has in order to correlate with
an ks.a sannidhanam
Ak
ca yogyata ceti ca trayam;
sambandhakaran.atvena klptam
nanantarasrutih..
276
A. K ULKARNI
another. For example, in a word dvaram to the door, the stem

dvara door denotes an object in the real world, and the am suffix
(an accusative marker) marks an expectancy of a verb whose karman (object) can be dvara door. This expectancy which arises
from the knowledge of suffix is a syntactic one, and it allows one
to connect the word dvaram with pidhehi close. This expectancy
is not one way, but mutual. It is also not psychological. It is based
on the usages of the verbs in a sentence and is thus syntactic in
nature. As an another illustration, consider two verbs gaml and
va. The verb gaml is used in the sense of motion (gaml gatau(SK.
982). The verb v

a is used in two meanings, viz., gati motion
and
gandhana pointing out (SK. 1050). Though both gaml and va are
used in the sense of motion, gaml is sakarmaka (transitive)

while
va is akarmaka (intransitive). This requirement of a karman (object) for gaml is not psychological but is based on the usage of the
verb. The expectancies

which are mutual, direct and natural are
termed niyata or utthita a kan ks.a (restricted or risen expectancy)
(Raja 1963). The expectancy between a verb and the words denoting karakas or between relational words falls under this category.3
In contrast to mutual expectancy, the expectancy that is unilateral is called aniyata or utthapya a kan ks.a (unrestricted or to be
raised). This is aroused only if necessary. So it is potential. For
example, in a phrase such as white cow, the a kan ks.a of white for
a substantive is natural, but the a kan ks.a of cow to have an adjective is potential. It gets aroused in the presence of an adjective such as white. Even a noun in apposition may arouse an expectancy. The example discussed in the Ramarudr commentary
on the Dinakar commentary on the Nyayasiddhantamuktaval is
udayati candrah. kumudabandhavah. (Jere 2002) Rises the moon,
the friend of a lotus (that opens its petals during night and closes
3 Niyat
akan ks.a : yatha kriyakarakapadanam
parasparakan ks.a . (Jere 2002:
23).
277
them in the morning). After hearing udayati candrah. the moon

rises all the expectancies are fulfilled. And thus understanding is
complete. Now when one hears kumudabandhavah. friend of a lotus then this word needs to be related to one of the words uttered
earlier because there can not be freely hanging words in a meaningful sentence. This word has an expectancy of a substantive, and
thus it gets related to candrah.. Both these cases are examples of
unilateral expectancy.
3.2
Sannidhi (proximity)
Sannidhi is defined in the Tarkasangraha as an utterance of

words without any gap (padanam avilambena uccaran.am.), or
as the presentation of word meanings without any intervention (avyavadhanena padajanya padarthopasthitih.). From the
text processing point of view, the important point is the presentation of word meanings without any intervention. In other
words, if the related words are intervened by some unrelated
words, then such an utterance does not produce any verbal cognition. To make the point clear, Visvanatha Pacanana in his
Nyayasiddhantamuktaval (Joshi 1985: 194), gives the following
example.
(5) girih. bhuktam agniman devadattena.
Hill is eaten fiery by Devadatta.
The words girih. hill and agniman fiery have mutual expectancy so do bhuktam eaten and devadattena by Devadatta.
Bhuktam, being a past participle in the neuter, expects a kartr in
the instrumental case and a karman in the neuter nominative.
But
this group of four words does not have a neuter nominative. So we
interpret either the verb bhuj to eat in this context to be an intransitive, or consider the example as a case of an ellipsis of the karman. In either case, this group of four words corresponds to two
278
A. K ULKARNI
independent sentences, whose arguments are intertwined. Bhuktam intervenes between the words girih. and agniman which have
mutual expectancy. Similarly agniman intervenes between the related words bhuktam and devadattena. This intervention forms an
obstacle to verbal cognition (sa bdabodha).
The condition of not having intervention is only a necessary condition in the process of sa bdabodha. For, even the nonintervention may give rise to more than one sa bdabodha one of
them as a true cognition (pramatmaka jana) and the other one as
a false cognition (bhramatmaka jana) as explained by Visvanatha
Pacanana (Joshi 1985: 194) with the example in (6).
Consider the group of words
(6) nlo ghat.ah. dravyam pat.ah..
blue pot thing cloth.
This may lead to two cognitions, viz.:
1. The pot is blue, and the cloth is a thing.
2. The cloth is blue, and the pot is a thing.
Visvanatha Pacanana argues that the first one, in the given situation, leads to a true cognition and the second one to a false cognition. The notion of true cognition and false cognition is thus
context dependent. The notion of sannidhi on the other hand depends only on the expectancies and the compatibility of meanings
of the words involved.
In the next section we compare the notions of sannidhi
and a kanks.a with the projectivity principle and the weak nonprojectivity conditions.
Dependency parsing and word order
Dependency analysis dates back to Pan.ini. The computational implementation of a dependency parser for Indian Languages based
279
on Pan.inian grammatical formalism is described by Bharati, Chaitanya, and Sangal (1994). In the recent years, the seminal work
of Tesnire (1959) became the basis for work on dependency
grammar. Meaning-text theory (Melcuk 1988), Word Grammar
(Hudson 1984), and Functional generative description (Segall, Hajicov, and Panevov 1986) are some of the flavours of dependency
grammar. Bharati, Bhatia, et al. (1998) extended their Pan.inian
grammatical formalism to English. However, the first full-fledged
computational implementation of a dependency grammar for English is the Link parser (Sleator and Temperley 1993).
The dependency parse of a sentence establishes relations between the morphemes in the form of nominal and verbal bases
(pratipadikas and dhatus), through the morphemes in the form
of nominal and verbal suffixes, through positional information,
or through concord. In the case of Sanskrit, it is predominantly
nominal and verbal suffixes which mark relations. In sentence (1)
above, the verbal suffix ti in gacchati establishes the relation between a person whose name is Rama and the activity of going represented by the verbal root gaml. The nominal suffix am represents
the relation between the entity
denoted by the nominal stem grama
and the activity of going denoted by gaml. Dependency structure
is also termed governance, since it tellsus which words govern
which others.
4.1
Tree traversal and possible word orders
The dependency structure of a sentence is typically represented as

a graph whose nodes correspond to the words and whose edges
correspond to the relation between nodes. This structure is characterised by two properties: a) the graph is acyclic, and b) every
node, except the root node, has exactly one incoming arrow. Thus
this structure is best represented in a tree. Tree traversal is the
process of visiting each node in a tree structure exactly once in a
280
A. K ULKARNI
systematic manner. Thus with every tree traversal is associated a

word order. The traversal is defined in the context of binary trees
by the order in which the nodes are visited. There are three distinct
traversals:
1. Post-order traversal,
2. In-order traversal, and
3. Pre-order traversal.
If the root node (or the top node), left node and right node are
denoted by T, L, and R respectively, then the post-order traversal
is L-R-T, the in-order traversal is L-T-R and the pre-order traversal
is T-L-R. At each visit of the node, all the sub-nodes under it are
descended recursively till all the leaf nodes under that node are
visited.
The trees in Figure 1 show these three traversals for sentence
(1) in the first row and the traversals after the leaf nodes are transposed in the second row.4 The dotted lines show relations between
words, while numbered thick lines (in red) show the order of tree
traversal. As is clear from the figures, the tree traversal orders
shown correspond to the six possible word orders listed above in
(1.1) through (1.6).
4.1.1
Generalising tree traversal
For a tree with more than 2 leaf nodes, we generalise the traversal
as follows:
Let A1 be the root node, and A2 , ... , An be the leaf nodes. Let
{A2 , ... , An } stand for all possible permutations of n 1 nodes.
Then the only possible traversals with n nodes are the following:
1. Pre-Order Traversals A1 {A2 , ... , An }.
4 For ease of reading, we label the node by a word including the suffix, and
not with a stem.
(a) Post-order (1.1)
(b) In-order (1.2)
(d) Post-order (1.4)
(e) In-order (1.5)
281
(c) Pre-order (1.3)
(f) Pre-order (1.6)
Figure 1
Traversals for sentence (1) with and without transposition
2. In-Order Traversals {A2 , ... , Ak } A1 {Ak+1 , ... , An }, where
k varies from 2 to n 1.
3. Post-Order Traversals {A2 , ... , An } A1 .
When any of the A2 to An is itself a tree, then at each of these
nodes, we recursively traverse the sub-trees in any of these three
orders to get possible word orders.
Now we look at the example (3) above. The dependency graph
for this sentence is shown in Figure 2. We notice that the word
order in (4) can not be produced from this dependency tree by
any of the above traversal methods. While the different traversals
give us a computational device to generate all possible word orders, they do not give an efficient mechanism to decide whether a
given word order can be obtained through one of the traversals or
not. The governance relation does not capture the word order or
the precedence. In dependency structures, the precedence is captured by projecting the nodes of the governance structures onto the
linear representation of the words in a sentence form. There have
282
A. K ULKARNI
Figure 2
Dependency graph for sentence 3
been several efforts in the domain of dependency framework to
study the relation between governance and precedence (Bodirsky,
Kuhlmann, and Mhl 2005; Havelka 2005; Nivre 2006). These
studies have resulted in various types of constraints on dependency
graphs. These constraints ban some word orders. The strongest
among these is the projectivity principle. The less stringent ones
are weak non-projectivity and well-nestedness.
4.2
Projectivity principle
The principle of projectivity states a constraint on the dependency

tree which bans certain dependency structures. There have been
various characterisations of the projectivity principle. Marcus
(1967) has shown the equivalence of some of the earlier characterisations e.g. of the ones by Harper and Hays, Lecerf and Ihm,
and Fitialov.
A sentence is projective if and only if we can draw a
dependency tree whose every node can be projected
283
by a vertical line onto its word form in the surface

string without crossing another projection or a dependency edge.
Hudson (1984: 98) calls this projectivity principle the adjacency
principle.
Figure 3 shows the projections for sentences (1.1) to (1.6). We
notice that the projections shown by dotted lines do not cross the
dependency relations shown by thick lines. And hence all the 6
sentences in (1.1) through (1.6) follow the projectivity principle.
(a) 1.1
(d) 1.4
(b) 1.2
(e) 1.5
(c) 1.3
(f) 1.6
Figure 3
Dependency structure with projections for sentences 1.1 to 1.6
Figure 4 shows the projections of sentences (2.1) to (2.6). We
notice that sentences (2.2) and (2.5) have the projection line crossing the governance relation showing the violation of the projectivity principle.
Sentence (4) also violates the projectivity principle. Figure 5
shows the governance relation between ptva and dugdham being
crossed by the projection of gacchati.
284
A. K ULKARNI
(a) (2.1)
(d) (2.4)
(b) (2.2)
(e) (2.5)
(c) (2.3)
(f) (2.6)
Figure 4
Projections for sentences 2.1 to 2.6
Figure 5
Projection for sentence (4)
285
Though both sentence (4) as well as sentences (2.2) and (2.5)

show violation of the projectivity principle, there is a difference
between the violation in (4) and the violation in (2.2) and (2.5).
The violation in (4) is an example of sannidhi violation of utthita
a kan ks.a while (2.2) and (2.5) are examples of sannidhi violation
of utthapya a kan ks.a . Thus there is a need to distinguish between
these two types of violation, the distinction between which is not
captured by the projectivity principle.
4.3
Weak non-projectivity (planarity)
It is possible to draw the dependency graphs for (2.2) and (2.5) by

rearranging the nodes to avoid the crossing of projection lines by
the dependency relations. Figure 6 shows possible graphs for (2.2)
and (2.5) that avoid crossing. However, there is no rearrangement
of the nodes of (4) that will avoid crossing. We capture this differ-
(a) (2.2)
(b) (2.5)
Figure 6
Projection with Rearrangement of nodes
ence by relaxing the projectivity constraint. Instead of considering
the crossing between two types of relations, viz., projection and
dependency, we consider only the crossing between dependency
relations with a further constraint that the nodes of the dependency
286
A. K ULKARNI
structure be represented in a linear order that reflects the surface

order of the words in the sentence they represent. Such a depencency graph is weakly non-projective if there are no crossing of
edges. If all the edges are drawn on the same side of a sentence
(either below or above), such a graph results in a planar graph.
Definition: A graph is weakly non-projective or planar, if it
does not have two edges wi w j and wk wl with i < k <
j < l.
Thus every projective structure is weakly non-projective, but the
reverse is not true.
Let us look at the sentences (2.2), (2.5) and (4) above. All these
were found to be non-projective. But as we notice from Figure 7,
the sentences (2.2) and (2.5) do not have crossing edges in their
linear projection. Hence these are weakly non-projective or planar.
But Figure 8 shows the crossing of edges making sentence (4) nonplanar.
svetah. dhavati asvah.

(2.2)
asvah. dhavati svetah.

(2.5)
Figure 7
Planar dependency graph for sentences (2.2) and (2.5)
Planarity (or weak non-projectivity) is the precise characterisation of the sannidhi constraint. Sannidhi violation leads to nonplanar graphs which correspond to the dislocation of constituents
in the phrase structure tree. But the dislocation need not correspond to sannidhi violation, as we shall see below.
Consider the following sentence analyzed by Gillon (1996:
12):
287
ramah. sa lam dugdham gacchati ptva

Figure 8
Planar dependency graph for sentence (4)
tena ca praman.ena sadhya-dharmasya tat-matraanubandhah. khyapyate. (PV 18.1)
And the dependence of the provable property merely
on it (i.e., the proving property) is made known by
that epistemic cognition.
In this example, Gillon observes that the third (instrumental) case
noun phrase is at the clauses left periphery, having been extraposed from the passive verb khyapyate (made known), which
is at the right periphery.
Figure 9 shows the dependency graph for this sentence which
is a planar (weakly non-projective) one. Here the dislocation does
not lead to a sannidhi violation.
tena .. praman.ena ... tat-matra-anubandhah. khyapyate.

Figure 9
Dislocation without sannidhi violation
288
A. K ULKARNI
Empirical evaluation
Now that we have a precise mathematical criterion for evaluating

the sannidhi violation, we test it on a real corpus consisting of both
prose as well as verse. For prose, we take the same corpus as studied by Gillon. Since sannidhi violation always leads to dislocation,
we need not look at the sentences which do not have dislocation.
We study only the cases of dislocation discussed by Gillon. For
the verse, we chose the Bhagavadgta (BhG.).
5.1
Cases of sannidhi violation from Gillons data
Gillons corpus consists of about a thousand sentences, approximately half of them from the Praman.varttika by the Buddhist
philosopher Dharmakrti and the rest from Aptes (1885) broadly
representative selection of examples of classical Sanskrit literature. Gillon observed dislocation in about 160 sentences which he
categorised into three classes:
1. extraposition from subject position,
2. extraposition from verb complement position, and
3. verb complement topicalisation.
We noted earlier that not all dislocation leads to sannidhi violation. Only those cases which have crossing edges in their planar
dependency graph are cases of sannidhi violation. Almost 75%
of the examples discussed by Gillon 1996 do not involve crossing
edges, hence do not lead to sannidhi violation. Among the examples of dislocation that involve sannidhi violation, we noted that
mainly two relations, viz., those of the adjective and the genitive,
are involved in crossing.

5.1.1
289
Dislocation of a genitive
We give below two examples of dislocation of a genitive discussed

by Gillon (1996).
(7) tayoh. baddhayoh. kim-nimittah. ayam moks.ah. (ASG. 14.1.2)
What basis does the release of the two prisoners have?
A copulative verb asti is supplied in the dependency stucture. The
two Indian schools Vyakaran.a (grammar) and Nyaya (logic) differ
in the analysis of sentences with missing copulative verbs. The
Vyakaran.a school supplies the missing copulative verb asti5 and
then establishes relations between the substantive and its predicative adjective through this verb, while the Nyaya school establishes
the relation between the substantive and the predicative adjective
directly. We follow the Vyakaran.a school. The non-planar and
planar dependency graphs are shown in Figure 10. As one can
see, the crossing links involve the verb asti, and the position of
this verb is crucial in the planarity of the dependency graph. If the
verb asti is placed in juxtaposition with kim-nimittah. then there is
no crossing. The verb asti in this sentence has only two arguments.
In example (8) (Figure 11) below, where there are more than two
arguments, the linear representation of the sentence has crossing
of links no matter where the copulative verb is placed.
(8) sarvatra audarikasya abhyavaharyam eva vis.ayah.. (ASG.
1.1.2)
In every case, a gluttons object is only food.
5 astr
11)
bhavantparah. prathamapurus.o prayujyamano pyasti. (A.2.3.1 vt.
290
A. K ULKARNI
tayoh. baddhayoh. kim-nimittah. ayam moks.ah. asti.

Figure 10
sarvatra audarikasya abhyavaharyam eva vis.ayah. asti.

Figure 11

5.1.2
291
Dislocation of a vises.an.a
Example (9) (Figure 12) shows the evidence of sannidhi violation

due to dislocation of an adjective. We supply the missing copulative verb bhavanti following the Vyakaran.a school.
(9) ete hi hrdaya-marma-bhidah. sams
ara-bhavah. (ASG. 8.1.3)
For, these worldly things are heart breakers.

Here the position of the missing copulative verb bhavanti dictates
the crossing of links. Certain positions of bhavanti lead to crossing
of links; others do not.
ara-bhavah. bhavanti.
ete hi hrdaya-marma-bhidah. sams
Figure 12
Dislocation of a vises.an.a
5.1.3
Other relations
Example (10) is a case where the relations involved in sannidhi

violation are other than vises.an.a and genitive (Figure 13).
(10) aham manda-autsukyah. asmi nagara-gamanam prati (ASG.
3.1.3)
I am (one who is) little eager about going to the city.
292
A. K ULKARNI
aham manda-autsukhah. asmi nagara-gamanam prati.

Figure 13
Dislocation of an argument
The edge marking the non-karaka relation between the words
manda-autsukyah. and prati crosses the edge between the kartr re
lation between aham and asmi.
5.2
Sannidhi violation in the Bhagavadgta
The word order in verse, though free, is constrained by metrical

considerations. In order to study the effect of metrical considerations on sannidhi, we analysed all the verses of the BhG. We did
not consider the slokas which have either conjunctive or disjunctive particles. Among the remaining 344 slokas, 300 instances of
sannidhi violations were found. Out of these, 193 cases involved
either an adjectival or genitive relation. The remaining cases had
other non-karaka relations such as negation, vocative, precedence,
293
simultaneity, etc. A total of 9 cases of anomalous behaviour were

found. Of these, in 7 cases both of the relations involved in crossing are karaka relations. In two cases there is crossing of an adverb
with a karaka relation. We discuss all these cases below.
5.2.1
Sannidhi violation involving kartr and karman
Five among the nine anomalous cases had the kartr relation of one
verb crossing with the karman relation of anotherverb. These instances are examined below.
(11) . ca:*.a ;lM ;
a;h ma;naH kx+:Sa :pra;ma;a;a;Ta ba;l+.va;x+Q+m,a
ta;~ya;a;hM ;a;na;g{a;hM ma;nyea va;a;ya;ea;a:=+va .sua;du;Sk+.=+m,a
cacalam hi manah. krs.n.a pramathi balavat drd.ham;
tasya aham nigrahammanye vayoh. iva sudus.karam. (BhG.
6.34)
O krs.n.a, the mind is fickle, turbulent, obstinate and strong,
I think it is as difficult as to control the wind.
hence
In the second line of this verse the main verb is manye whose
kartr is aham. The karman of the verbal noun nigraham is the
pronominal
tasya, which refers to manah. in the first sentence.
Thus the word sequence tasya aham nigraham manye produces
two crossing edges involving the relations of kartr and karman
(Figure 14).
(12) ;DUa;ma;ea .=+a;
a:a;~ta;Ta;a kx+:SaH :Sa;ma;a;sa;a d; a:a;a;a;ya;na;m,a
ta:a . ca;a;nd+ma;sMa .$ya;ea;a;ta;ya;eRa;ga;a :pra;a;pya ;a;na;va;tRa;tea
dhumah. ratrih. tatha krs.n.ah. s.an.masah. daks.in.a yanam;
tatra candramasam jyotih

. yog prapya nivartate. (BhG.
8.25)
The yog on travelling the path of smoke, night, the dark half
of the month and the six months of the southern path (of the
sun) after death, obtains the lunar light and returns (to this
world).
294
A. K ULKARNI
tasya aham nigraham manye vayoh. iva sudus.karam.

Figure 14
Analysis of BhG. 6.34
The second example with similar crossing of kartr and karman
sequence
was found in verse BhG. 8.25. In this verse, in the word
jyotih. yog prapya nivartate, jyotih. is the karman of the absolutive
verb prapya and yog is the kartr of the verb nivartate resulting in
the crossing (Figure 15).

(13) va;u+:m,a ;hR ;~ya;Zea;Sea;a ;
a;d;v.ya;a ;
a;h ;a;tma;
a;va;BUa;ta;yaH
ya;a;a;Ba;
a;vRa;BUa;a;ta;a;Ba;l+.eRa;k+:a;n,a I+.ma;Ma tvMa v.ya;a;pya ;a;ta;+a;sa
vaktum arhasi ases.en.a divyah. hi a tmavibhutayah.;
yabhih. vibhutibhih. lokan iman tvam vyapya tis..thasi. (BhG.
10.16)
You could tell me about your divine manifestations by which
you exist pervading these worlds.
The third example is from the verse BhG. 10.16. Here the word
sequence lokan iman tvam vyapya tis..thasi results in crossing edges
295
tatra candramasam jyotih. yog prapya nivartate.

Figure 15
with lokan being the karman of vyapya and tvam being the kartr
of tis..thasi (Figure 16).

(14) ;(ra;+Da;a;na;aH :pua:+:Sa;a ;Da;mRa;~ya;a;~ya :pa:=+nta;pa
;pra;a;pya ma;Ma ;a;na;va;tRa;ntea mxa;tyua;sMa;sa;a:=+va;tmRa;a;na
asraddadhanah. purus.a h. dharmasya asya parantapa;
aprapya mam nivartante mrtyusams
aravartmani. (BhG.
9.3)
O Parantapa, those who have no faith in this dharma return
to the circle of death and rebirth without attaining me.
The fourth example is from BhG. 9.3, where the dependency
arrow corresponding to the karman of the krdanta a-sraddadhanah.
of the main verb nicrosses the arrow corresponding to the kartr
vartante (Figure 17).

(15) ta;a;n,a .sa;ma;a:ya .saH k+:Ea;ntea;yaH .sa;va;Ra;n,a ba;nDUa;na;va;a;~Ta;ta;a;n,
//
a
xk+:pa;ya;a :pa:=+ya;a;
a;va;;ea ;
a;va;Sa;a;d; a;a;d;ma;b.ra;va;a;t,a
296
A. K ULKARNI
yabhih. vibhutibhih. lokan iman tvam vyapya tis.t.hasi.

Figure 16
asraddhadhanah. purus.ah. dharmasya ... nivartante.

Figure 17
297
tan samks.ya sah. kaunteyah. sarvan bandhun avasthitan.

(BhG. 1.27cd)
krpaya paraya a vis..tah. vis.dan idam abravt; (BhG. 1.28ab)
Seeing
all these relatives present there (on the battle-field)
Kaunteya filled with compassion uttered these words in dejection.
In BhG. 1.2728 also, we find the crossing of the dependency
arrows pointing to a kartr and a karman. Bandhun is the karman
of the absolutive samks.
ya and kaunteya is the kartr of the verb
abravt, and the edges marking these relations cross (Figure

18). In
tan samks.ya sah. kaunteyah. ... bandhun ... abravt.

Figure 18
this example, if we choose tan, which is an adjective of bandhun as
the karman of samks.ya, then this crossing vanishes! This, in fact,
happens to be an example of cataphora, and would be analysed as
in Figure 19.
298
A. K ULKARNI
tan samks.ya sah. kaunteyah. ... bandhun ... abravt.

Figure 19
Modified Analysis of BhG. 1.27
5.2.2
Sannidhi violation involving two kartr relations
Verse BhG. 8.19 has an instance of crossing edges involving two

kartr relations.
(16) BUa;ta;g{a;a;maH .sa O;;va;a;yMa BUa;tva;a BUa;tva;a :pra;l +.a;ya;tea

.=+a:ya;a;ga;mea Y;va;ZaH :pa;a;TRa :pra;Ba;va;tya;h:=+a;ga;mea
bhutagramah. sah. eva ayam bhutva bhutva pralyate;
ratri-agame avasah. partha prabhavati ahar-agame. (BhG.
8.19)
O Partha, these living beings merge and re-emerge at the
onset of night and on the coming of the day.
Here sah. is the kartr of the verb pralyate and bhutagramah. is the
a for the absolutive bhutva, and these relakartrsamanadhikaran
.
tions cross (Figure 20).
299
bhuta-gramah. sah. eva ayam bhutva bhutva pralyate ...

Figure 20
5.2.3
Sannidhi violation involving karman and kriyavises.an.a
In BhG. 1.37, the relation of karman of the absolutive hatva is

crossed by the kriyavises.an.a of the main verb syama (Figure 21).
(17) ta;sma;a;a;a;h;Ra va;yMa h;ntMua ;Da;a;tRa:=+a;";a;n~va;ba;a;nDa;va;a;n,a
.~va:ja;nMa ;
a;h k+:TMa h;tva;a .sua;a;Ka;naH

.~ya;a;ma ma;a;Da;va
tasmat na arhah. vayam hantum dhartaras..tran svabandhavan;
svajanam hi katham hatva sukhinah. syama madhava. (BhG.
1.37)
Therefore O Madhava, it is not suitable on our part to kill
our relatives, for by killing our own kinsmen, the sons of
Dhrtaras.t.ra, how can we remain happy?
A similar example is found in BhG. 11.32 where the relation of

kartr crosses with that of the adjective.
5.2.4
Sannidhi violation involving karman and apadana
The last example is from the last chapter of BhG. Here the karman
of the non-finite verb kathayatah. crosses the relation of apadana
300
A. K ULKARNI
sva-janam katham hatva sukhinah. syam.

Figure 21
between srutavan and krs.n.a t (Figure 22).
(18) v.ya;a;sa;pra;sa;a;d;a;.cC"u+.ta;va;a;nea;ta;u+h:a;ma;hM :pa:=+m,a

ya;ea;gMa ya;ea;gea:(;a:=+a;tkx+:Sa;a;tsa;a:a;a;tk+:Ta;ya;taH .~va;ya;m,a
vyasaprasadat srutavan etad guhyam aham param
yogam yogesvarat krs.n.a t saks.a t kathayatah. svayam. (BhG.
18.75)
By the grace of Vyasa I have heard this supreme secret yoga
directly from the lord of yoga, Krs.n.a.
Conclusion
We conclude with two observations. The first observation is that

the number of violations is much greater in verse than in prose.
This may be attributed to metrical considerations. Out of 344
verses, 154 verses had sannidhi violation, while Gillon reported
only 160 instances of violations in about 1000 prose sentences.
The second observation is with respect to the relations involved. The two major relations involved in sannidhi violation in
301
srutavan yogam ... krs.n.a t ... kathayatah..
Figure 22
both prose and verse are the vises.an.a and the genitive. Some of the
other relations involved in the sannidhi violation are sambodhana
(vocative), negation, precedence, and simultaneity. All these relations have unilateral expectancy and thus correspond to utthapya
a kan ks.a .
Empirical study thus reveals that mutual expectancies are
tightly coupled, and the words that have mutual expectancy are
always in close proximity (sannidhi). In contrast, a word that
has unilateral expectancy (utthapya a kan ks.a ) may be moved away
from its relatum and its relational path may be interrupted by one
or more words unrelated to them.
Table 1
Abbreviations
A.
As..ta dhyay. Reference to: chapter . section . verse
ASG. Apte 1885. Reference to: chapter . exercise set . example sentence
302
A. K ULKARNI
BhG. Bhagavadgta. Reference to: chapter . verse

SK. Siddhantakaumud. See Govindacharya 2010. Reference to: dhatusankhya
References

us.an.a Press.
Bharati, Akshar, Medhavi Bhatia, Vineet Chaitanya, and Rajeev
Sangal. 1998. Paninian grammar framework applied to English. South Asian Language Review 8.1: 123.
Bharati, Akshar, Vineet Chaitanya, and Rajeev Sangal. 1994. Natural language processing: a Paninian Perspective. New Delhi:
Prentice Hall of India Pvt Ltd.
Bodirsky, M., M. Kuhlmann, and M. Mhl. 2005. Well-nested
drawings as models of syntactic structure. Tenth Conference
on Formal Grammar and Ninth Meeting on Mathematics of
Language, pp. 195203. University of Edinburgh. Edinburgh.
Culotta, A. and J. Sorensen. 2004. Dependency tree kernels for
relation extraction. 42nd Annual Meeting of the Association
for Computational Linguistics: proceedings of the conference,
pp. 42329. Stroudsburg, Pennsylvania: acl.
Gillon, Brendan S. 1996. Word order in Classical Sanskrit. Indian Linguistics 57.14: 136.
Govindacharya. 2010. Vaiyakaran.asiddhantakaumud. 2 vols.
Varanasi: Chaukhambha surabharati prakashan.
Haghighi, A. D., A. Y. Ng, and C. D. Manning. 2005. Robust
textual inference via graph matching. Human Language Technology Conference (HLT) and Conference on Empirical Meth-
R EFERENCES
303
ods in Natural Language Processing (EMNLP), pp. 38794.

Stroudsburg, Pennsylvania: acl.
Havelka, J. 2005. Projectivity in totally ordered rooted trees.
Prague Bulletin of Mathematical Linguistics 84: 1330.
Prague: Charles University.
Hudson, R. 1984. Word grammar. Oxford: Basil Blackwell.
Jere, Atmaram Narayan. 2002. Karikaval. Varanasi: Chowkamba
Krishnadas Academy.
Joshi, K. R. 1985. Nyayasiddhantamuktaval. Bhandarkar Oriental
Series 21. Pune: Bhandarkar Oriental Research Institute.
Kulkarni, Amba. 2013. A deterministic dependency parser with
dynamic programming for Sanskrit. Proceedings of the Second International Conference on Dependency Linguistics
(DepLing 2013), pp. 15766. Prague, Czech Republic: Charles
University in Prague, Matfyzpress, Prague, Czech Republic.
URL: http://www.aclweb.org/anthology/W133718.
Kulkarni, Amba, Sheetal Pokar, and Devanand Shukl. 2010.
Designing a constraint based parser for Sanskrit. Sanskrit
Lin, Dekang. 2003. Dependency-based evaluation of MINIPAR.
Treebanks: building and using parsed corpora, ed. by Anne
Abeill, pp. 31729. Text, speech, and language technology
20. Berlin; Heidelberg: Springer.
Marcus, S. 1967. Algebraic linguistics: analytical models. Mathematics in Science and Engineering 29. New York: Academic
Press.
Marneffe, Marie-Catherine de, Bill MacCartney, and Christopher
D. Manning. 2006. Generating typed dependency parses from
phrase structure parses. Proceedings of LREC-06, pp. 44954.
Paris: elra.
304
A. K ULKARNI
Melcuk, I. 1988. Dependency syntax: theory and practice. Albany:

State University of New York Press.
Nivre, J. 2006. Constraints on non-projective dependency parsing. Eleventh Conference of the European Chapter of the Association for Computational Linguistics (EACL), pp. 7380.
Trento, Italy: acl.
Quirk, C., A. Menezes, and C. Cherry. 2005. Dependency treelet
translation: syntactically informed phrasal SMT. 43rd Annual
Meeting of the Association for Computational Linguistics: proceedings of the conference, pp. 27179. Stroudsburg, Pennsylvania: acl.
Raja, K. Kunjunni. 1963. Indian theories of meaning. The Adyar
Library Series 91. Reprints, 1969, 1977. Madras: The Adyar
Library and Research Center.
Sastri, Gangadhara. 1903. Tantravartikam. Varanasi: Chowkamba
Sanskrit Book Depot.
Segall, P., E. Hajicov, and J. Panevov. 1986. The meaning of the
sentence in its semantic and pragmatic aspects. Berlin; Heidelberg: Springer.
Sleator, D. and D. Temperley. 1993. Parsing English with a link
grammar. Third International Workshop on Parsing Technologies (IWPT), pp. 27792. Tilburg, The Netherlands; Durbuy,
Belgium.
Speijer, J. S. 1886. Sanskrit syntax. Leiden: E. J. Brill. [Reprint:
Tesnire, L. 1959. lments de syntaxe structurale. Paris:
Klinksieck.

preliminary results
P ETER S CHARF, A NUJA A JOTIKAR, S AMPADA
S AVARDEKAR, and PAWAN G OYAL
Abstract: It is well known that the syntax of verse departs
from that of prose. Sanskrit commentators generally reorder verse texts by putting them in the more usual order
of prose. With the availability now of syntactically tagged
texts it is possible to analyse how the features of poetic
syntax differ from those of prose in a systematic manner. Ingalls (1991) pioneered the employment of computational techniques in Sanskrit poetic syntax by systematically exploring the polarization of speech forms in the Mahabharata. Sellmer (2013) demonstrated the polarization
of particular strings in the Mahabharata. In the current
study, we investigate the polarization of certain parts of
speech in the Ramopakhyana (an epic verse text), and Pacatantra (mixed verse and prose) and explore the significance of different patterns between prose and verse. We
tag morphology and dependency relations in each text, use
graphing software to create dependency trees, and collect
statistics on the position of various parts of speech in their
governance structure. Finally we describe the significant
differences that emerge between prose and poetic syntax.
Keywords: dependency, user interface, prose syntax, poetic
syntax, prose-poetry divergence
305
306
S CHARF, A JOTIKAR, S AVARDEKAR, and G OYAL
Motivation
For more than three millennia the Vedic traditions of India have
passed down large collections of hymns composed in verse. The
tradition of oral epic verse composition produced enormous works
about two thousand years ago. These strong oral traditions undoubtedly inspired the perseverance of verse composition in classical Sanskrit even long after the use of writing became widespread.
Thus the heritage in the form of verse in Sanskrit is enormous
and spans all disciplines from literature and drama to mathematics,
medicine, and grammar.
Sanskrit commentary on Vedic and classical Sanskrit poetry
such as Sayanas commentary on the Rgveda, and Mallinathas
commentary on Kalidasas plays has
for over a thousand years
reordered words in verse in a more natural and logical order, an
order more typical of Sanskrit prose, while providing glosses of
obscure terms and additional explanation. The reordered sentence
is typically referred to by the term anvaya. The prevalence of
anvaya to help explain verse testifies to the fact that the syntax
of verse and the syntax of prose differ markedly. While it is common knowledge that the syntax in verse differs from the syntax in
prose and that the former is freer than the latter, only now with
the development of computational techniques and the compilation
of morphologically and syntactically tagged corpora of Sanskrit
texts is it convenient to study these differences systematically, collect statistical data and draw conclusions based upon quantified
results. Thus, the research reported in this paper was motivated by
the objective to procure sufficient statistical data to support definitive conclusions regarding whether, to what extent, and in what
respects the syntax of verse departs from the syntax of prose.
P OETIC SYNTAX
307
Expectations
Staal (1967) was the first to deal with Sanksrit syntax in a generative linguistic framework (Gillon and Shaer 2005: 12). He argued
that Sanskrit word order was free to the extent that the order of constituents that occur as daughters of the same phrasal constituent is
unconstrained. He (1967: 15) called governance structures that
tolerated such a lack of constraint wild trees. It is known that
Sanskrit and other languages exhibit word orders that violate even
the constraints of wild trees. Words intervene between words that
are construed together. These discontinuities are attempted to be
accounted for by transformations, that is, by motivated movement.
Gillon (1996) formalized and tested Staals hypothesis against
a corpus of Sanskrit sentences and characterized the types of discontinuities found. He argued that discontinuity is confined to the
extraposition of elements from subject position and verb complement positions, and to topicalization of verb complements. However, the sentences that violate these constraints, which he classifies as anomalous, demonstrate additional discontinuities which
are perfectly acceptable. While there are some obvious limits to
word order like the order of compound elements, particle placement, and clausal integrity as recounted by Gillon (1996: 12)
Sanskrit still exhibits a freedom of word order beyond the limits
Gillon (1996) describes.
Gillon and Shaer (2005) recognize that Staal (1967)s concept
of wild trees supplemented by the extractions Gillon (1996) describe still does not capture all of the attested word orders in Sanskrit. They extend the freedom of order not only by allowing
leftward and rightward movement but in addition by eliminating
the verb phrase as a cohesive constituent in opposition to the subject noun phrase. While including the subject noun phrase as an
equal partner among other unordered immediate constituents of
the sentence, they permit word order to emerge from information-
308
packaging functions. Such an approach is more in line with the

Pan.inian concept that considers action to be the principal component in relation to which participants in that action are dependent.
A. Kulkarni, Pokar, and Shukl (2010) designed a dependency
parser that utilizes Pan.inian concepts and the concept of expectancy (akan ks.a ) to label relations between words in a sentence
and employs the concept proximity (sannidhi) as a prioritization
criterion. For example, they use the relation of participants in action (karakas) to label relations between verbs and their complements. In the preceding paper in this volume, A. Kulkarni proposes criteria that improve upon Gillons description of the degree of freedom of Sanskrit word order. She (6) (p. 300) concludes that dislocations involving unilateral expectancy (utthapya-akan ks.a ) are permissible while dislocations involving mutual expectancy (utthita-akan ks.a ) are not. She found that all of the dislocations in Gillons prose sample conformed to the permissible type
as did 97.4% of her verse sample. These results indicate that the
characterization of freely ordered relations and dislocations so far
incorporated in her parser are highly accurate.
Although Hellwig (2009: 106) is dismissive of the applicability of grammatical and syntactic rules to real texts and hence denigrates the utility of descriptive studies of Vedic prose and computational implementations of rule-based taggers, on the opposite pole
he (2009) reports the utter failure of taggers based solely on statistical collocations and demonstrates their significant improvement
when supplemented with rules. Thus in spite of spurning (2009:
106) Staals (1967) concept of wild trees, he demonstrates in its
support, by the very fact of showing the failure of the unaided collocation approach, that word-order plays a relatively unimportant
role in comparison with morphology in Sanskrit.
Despite the fact that the work of Staal, Gillon, Shear, and
Kulkarni convincingly argues for the preferability of fundamentally unordered syntactic relations and the fact that such an ap-
P OETIC SYNTAX
309
proach is in consonance with the approach of traditional Indian

grammar, and despite the well-known fact of the acceptability of
a wide variety of word orders in Sanskrit, still it cannot be denied
that some orders are more usual than others. Although Scharfs
article, Interrogatives and word-order in Sanskrit, earlier in this
volume argues against a rigid application of the serialization principle, it is generally agreed that Sanskrit is a head-final language
and that most of the general observations regarding the serialization of parts of speech discussed there (1) (p. 203) are undoubtedly accurate. Thus in accordance with the expectations in a headfinal language and with the observations of descriptive linguists
regarding the syntax of Sanskrit over the past hundred and fifty
years regardless of whether they described Vedic or classical
we firmly expect to find the following word orders in prose in
our study:
Words denoting agents, objects, and instruments precede the
verbs denoting the actions in which they participate.
Adverbs, qualifiers and genitives precede what they limit.
Agents precede objects of the verbs of which they are agents.
To what extent these serializations hold and to what extent they
vary in verse this study aims to determine. Thereby the study expects to be able to characterize definitively the difference between
the syntax of prose and verse in these respects.
While we take inspiration from the pioneering work on polarization of speech forms in the Mahabharata conducted by Ingalls (1991) and Sellmer (2013), our work differs markedly in the
degree of abstraction. We do not measure statistics on particular speech forms. We do not measure the serialization of specific
sandhi variants of particular inflected forms of particular lexical
items. Rather we measure the serialization of annotations of parts
of speech in a syntactic tree. We describe our methodology in the
next section.
310
3
3.1
Methodology
Corpus preparation
In 2001, Scharf completed preparation of a morphologically

tagged database of the Ramopakhyana which appeared in the Sanskrit Librarys digital Kramapat.ha reader (Scharf 2000), and in
print (Scharf 2002). The episode narrates the story of Rama in
728 verses in Anus.t.ubh meter within the great Indian epic Mahabharata. During the year 20122013, Ajotikar morphologically
tagged two thousand sentences in the Sanskrit Librarys digital edition of the Pacakhyanaka (Scharf 2011) based on Hertels (1908)
critical edition, and Sharon Ben-Dor morphologically tagged 623
prose sentences in the Mahabharata. The morphological tags utilized are published under Tagsets on the Reference page at
sanskritlibrary.org and are a slightly revised version of
the morphological tags described by Scharf (2002: 2831).
Morphological tagging of the Pacakhyanaka proceeded by
utilizing a platform developed by Ralph Bunker, Grard Huet, and
Peter Scharf distributed between the Sanskrit Library and the Sanskrit Heritage site and described by Goyal et al. (2012: 1023
24). The platform allows machine-assisted human-validated tagging and can operate on local installations of the two servers. Sentences in the Sanskrit Librarys digital editions of the Pacakhyanaka and Mahabharata were fed to the Sanskrit Heritage parser
which summarized possible solutions in a user-friendly singlepage interface. The interface allows a user to select among words,
stems, and morphological tags presented, deprecates solutions in
the solutions summary as elements are validated and copies unique
tags to the candidate solution. The interface also allows the scholar
to edit and resubmit the sentence for re-analysis by the Sanskrit
Heritage parser; to edit, add, or delete words, stems, and tags; or
to tag the sentence manually. The scholar can construct tags using
P OETIC SYNTAX
311
a convenient dialogue box to ensure ease and validity of tagging.

Results are saved in XML files that can be reviewed with the same
interface. The first version of this morphological tagging platform,
limited by the practicalities of transferring and summarizing large
amounts of HTML output, motivated Huet and Goyal to develop
the summary interface, described in their (2013) article, as an integral part of the Sanksrit Heritage Reader. Later versions of the
platform coordinate with the summary interface to permit interactive pruning of solutions prior to transfer to the editing interface.
Ajotikar, Savardekar, and Ben-Dor used the results of the morphological tagging to tag dependency relations in 98 prose sentences and 170 verses in these three texts. Goyal developed an
HTML interface that displays each morphologically tagged word
in sequence in a row supplemented by two empty text fields in
which the words relation and the relations target could be added.
The interface saves the dependency data in an XML file. We then
utilized software previously prepared by Amba Kulkarni to produce dependency graphs of each sentence to allow taggers to assess the appropriateness of their tagging visually and to revise tagging as necessary. The XML data of the sentences with their words
in sequence along with their tagged morphology and dependency
relations was subsequently queried regarding the syntax of various
parts of speech in their governance structures. We will describe
the parameters of our syntactic analysis after describing the dependency tagset used.
3.2
An Indian cognitive linguistics dependency relations tagset
Based upon the fine-grained analysis of speech components undertaken by Pan.ini and presupposed in his grammar, upon discussions
in Patajalis Mahabhas.ya (second century BC), Bhartrharis Vakyapadya (fifth century CE), and their commentators,and upon
312
a, Indian cognitive
debates with proponents of Nyaya and Mmams
linguists of the seventeenth century summarized conclusions regarding the structure of the cognition that results from understanding speech. These linguists detailed the correspondence between
components of verbal cognition, including relations between substantive components, and the speech forms that bring about cognition of them. Using these correspondences, one can transpose the
analysis of verbal cognition onto syntactic relations among speech
forms to determine categories of syntactic relations. By doing just
this, Ramakrishnamacharyulu (2009) described a detailed set of
one hundred and three tags labeling categories and subcategories
of syntactic relations based upon the Indian cognitive linguistic
tradition.
Many of the distinctions among these categories require
knowledge of the suitability of objects for particular relations.
While this sort of information might be built into a detailed lexicon or word-net, it is not presently available to computational
linguists. Hence from the comprehensive, fine-grained set of relations described by Ramakrishnamacharyulu (2009), A. Kulkarni
and Ramakrishnamacharyulu (2013) distilled a set of thirty-one
dependency relations that could be distinguished solely on the basis of visible textual evidence, namely, on the basis of morphology
and syntax alone, and developed a tagset to label them. We utilized
this tagset to tag dependency relations in our corpus.
The principal feature of the structure of verbal cognition as
analyzed by the grammatical tradition in India is that the action
denoted by the verbal root is the principal element in the cognition and that all participants in the action, including the agent in
an active sentence and direct object in a passive sentence, are immediately subordinate to it. Hence, unlike typical phrase structure
analysis that recognizes a subject as a sibling of a verb phrase, the
subject of the sentence is a sibling of verbal complements. While
the school of logic in India (Nyaya) differed from the grammarians
P OETIC SYNTAX
313
in situating the subject of the sentence as the principal element in

its cognition, the school of ritual exegesis sided with the grammarians. Kulkarnis tagset adopts the convention of the grammarians
in this regard.
3.3
Queries
Once we tagged a number of sentences in prose and poetry with

dependency labels, we collected statistics on the relative position
of various pairs of relata, in particular, the following:
1.
2.
3.
4.
5.
6.
7.
8.
agentverb
objectverb
instrumentverb
adverbverb
qualifierqualified
genitivewhat it limits
objectabsolutive
agentobject
These pairs were chosen because the syntax for these pairs is assumed to be rigid in prose but not so in poetry. After collecting
statistics of the relative position of relata for each of these pairs,
we compared the statistics of the prose and poetry data. Below,
we display the results in comparison graphs and discuss the results
obtained for each of these pairs.
Before discussing these results, let us explain the notations
used in these comparison graphs. The statistics for the poetry sentences are drawn in red color, and those for the prose sentences are
drawn in blue color. The data points are marked with bold dots.
The x axis labels the various positions in which the dependent
relatum is found relative to the relatum on which it is dependent
in the dependency tree, and the y axis shows the probability that
314
the dependent relatum is found in each such position. For example, the probability that the agent (kartr) is found three words prior
it is the agent (kriya) is
to the verb denoting the action of which
shown at position 3 on the x axis in Figure 1; the probability
of the agent being found there in poetry and prose sentences is approximately 0.08 and 0.17 respectively as shown by the positions
of the red and blue lines relative to the y-axis at that point on the
x-axis. The graph conveys the fact that for a sentence in our poetry
corpus the agent occurs 3 padas before the verb with probability
0.08, while for a sentence in our prose corpus, this probability is
0.17. The number of samples, i.e. the number of pairs (agent,
verb) obtained from the data are shown at the top left of the figure. We also show the probability of the agent occurring in any
position before the verb at the top right. This probability value
corresponds to adding all the probability scores on the left hand
side of the y axis in the graph.
4
4.1
Results
Agent after verb
Figure 1 shows the comparison results for the relative position of

an agent (kartr) with respect to the verb denoting the action of
which it is theagent (kriya) in prose and poetry sentences. The
comparison results in Figure 1 indicate that the agent occurs after
the verb in 63 (34%) of the 187 poetic samples. On the other hand,
only 6 (9%) of the 69 prose samples have the agent after the verb.
It is clear that the agent occurs after the verb significantly more
often in poetry than in prose.
P OETIC SYNTAX
4.2
315
Object after verb
Figure 2 shows the comparison results for the relative position of

the direct object (karman) with respect to the verb denoting the
action of which it is the direct object (kriya) in prose and poetry
sentences. The probability scores for different positions indicate
that the direct object occurs after the verb more often in poetry
(40%) than in prose (15%).
4.3
Instrument after verb
Figure 3 compares the relative position of the instrument (karan.a)

with respect to the verb denoting the action of which it is the instrument (kriya) in prose and poetry sentences. While the number
of samples with the usage of an instrument are very few in our corpus (18 and 10 for poetry and prose respectively), nevertheless it is
significant that only 1 of the 10 prose sentences had the instrument
after the verb. On the other hand, 11 of the 18 poetic sentences
did.
4.4
Adverb after verb
Figure 4 compares the relative position of an adverb (kriyavises.an.a) with respect to the verb it qualifies in prose and poetry sentences. Quite strikingly, of the 36 prose samples with the pair verbadverb, none had the adverb occurring after the verb. On the other
hand, 27 (36%) of the 75 poetic samples had the adverb after the
verb.
4.5
Qualifier after qualified
Figure 5 compares the relative position of a qualifier (vises.an.a),

whether it is an adjective or noun, with respect to the noun it qualifies in prose and poetry sentences. A qualifier occurs after what
316
Figure 1
Relative position of an agent with respect the verb that governs it:
comparison of prose and poetry syntax
Figure 2
Relative position of a direct object with respect to the verb that
governs it: comparison of prose and poetry syntax
P OETIC SYNTAX
317
Figure 3
Relative position of an instrument with respect to the verb that
Figure 4
Relative position of an adverb with respect to the verb that
318
it qualifies more often in poetry (54% of the 319 samples) than in

prose (only 16% of the 143 samples).
Figure 5
Relative position of a qualifier with respect to what it qualifies in
a sentence: comparison of prose and poetry syntax
4.6
Genitive after what it limits
Figure 6 compares the relative position of a genitive (word ending

in a s.as..th vibhakti) with respect to what it limits in prose and
poetry sentences. The results were quite striking again. Only 1 of
the 21 prose samples had the genitive occurring after the nominal
it limits. On the other hand, 40% of the poetic samples had the
genitive after the nominal.
4.7
Object after absolutive
Figure 7 compares the relative position of a direct object (karman)

with respect to the absolutive (form ending in the affixes ktva or
lyap also called a gerund) that denotes the action of which it is
P OETIC SYNTAX
319
the direct object. In prose, the direct object almost always (with
probability of 0.84) occurs immediately before the absolutive, and
if not there, one or two words before that; it never occurs after the
absolutive. In poetry, while the direct object occurs immediately
before the absolutive 40% of the time, it also occurs immediately
after the absolutive 20% of the time and occurs widely distributed
elsewhere a good 40% of the time as well.
4.8
Agent after object
Figure 8 compares the relative position of the direct object and

agent of the verb that directly governs them both. The agent occurs
after the object more often in poetry than in prose. While in prose
the agent almost always (with a probability of 0.92) occurs before
the direct object, in poetry the most common position of the agent
is two words after the direct object.
4.9
Summary
Table 1 summarizes the statistics obtained for the positions of various pairs of syntactically related items in poetry and prose sentences. The statistics presented for a pair (x, y) are the probability
of x occurring before y in a sentence in poetry and in prose. As is
evident from the statistics in Table 1, prose syntax is much more
rigid for all these pairs. The probability of x occurring before y in
prose is greater than or equal to 0.84 for any of the eight pairs of
relata considered. In contrast, in poetry, the probability of x occurring before y is less than or equal to 0.66. Thus, poetic syntax
gives more freedom in choosing the location of any element of the
pair. In particular, there is at least a 21% greater probability of a
dependent relatum occurring after its head in poetry than in prose.
Although it is well-known to Sanskrit scholars and linguists in
general that the word order in poetry is far more liberal than the
word order in prose, the results of our research provide statistical
320
Figure 6
Relative position of a genitive with respect to what it limits in a
sentence: comparison of prose and poetry syntax
Figure 7
Relative position of a direct object with respect to an absolutive
that governs it: comparison of prose and poetry syntax
P OETIC SYNTAX
321
Figure 8
Relative position of a direct object with respect to an agent of the
same action governing them: comparison of prose and poetry
syntax
Table 1
Probability of x occurring before y in a relation pair (x,y):
comparing poetic sentences with prose
Pair (xy)
agentverb
objectverb
instrumentverb
adverbverb
qualifierqualified
genitivewhat it limits
objectabsolutive
agentobject
Poetry
0.66
0.6
0.39
0.64
0.46
0.6
0.64
0.51
Prose
0.91
0.85
0.9
1.0
0.84
0.95
1.0
0.92
322
proof of this understanding. The statistics we present not only clarify that ordering of the sentences constituents is far more liberal
in poetry, but also quantify to what extent word order is liberal in
prose and poetry and to what extent the positions of constituents
vary.
References
Gillon, Brendan S. 1996. Word order in Classical Sanskrit. Indian Linguistics 57.14: 136.
Goyal, Pawan, Grard Huet, Amba Kulkarni, Peter Scharf, and
Ralph Bunker. 2012. A distributed platform for Sanskrit processing. Proceedings of COLING 2012, pp. 101128. Mumbai: The COLING 2012 Organizing Committee. URL: http:
//www.aclweb.org/anthology/C12-1062.
Hellwig, Oliver. 2009. Extracting dependency trees from Sanskrit texts. Sanskrit computational linguistics: third international symposium, Hyderabad, India, January 2009, proceedings, ed. by Grard Huet and Amba Kulkarni, pp. 10615. Lecture Notes in Artificial Intelligence 5406.
Hertel, Johannes, ed. 1908. The Panchatantra: a collection of ancient Hindu tales in the recension, called Panchakhyanaka,
and dated 1199 A.D., of the Jaina monk, Purnabhadra. Harvard Oriental Series 11. Cambridge, Mass.: Harvard University.
Huet, Grard and Pawan Goyal. 2013. Design of a lean interface
for Sanskrit corpus annotation. Proceedings of ICON 2013,
the 10th International Conference on NLP, pp. 17786.
R EFERENCES
323
Ingalls, David. 1991. The Mahabharata: stylistic study, computer
analysis, and concordance. Essays on the Mahabharata, ed.
by Arvind Sharma, pp. 1956. Leiden: Brill.
Kulkarni, Amba, Sheetal Pokar, and Devanand Shukl. 2010.
Designing a constraint based parser for Sanskrit. Sanskrit
Kulkarni, Amba and K. V. Ramakrishnamacharyulu. 2013. Some
relation-specific issues. Proceedings of the Fifth International
Sanskrit Computational Linguistics Symposium, ed. by Malhar
Kulkarni and Chaitali Dangarikar, pp. 191212.
Kulkarni, Malhar and Chaitali Dangarikar, eds. 2013. Proceedings
Symposium. New Delhi: D. K. Printworld.
Ramakrishnamacharyulu, K. V. 2009. Annotating Sanskrit texts
based on sa bdabodha systems. Sanskrit computational linguistics: third international symposium, Hyderabad, India,
January 2009, proceedings, ed. by Grard Huet and Amba
Kulkarni, pp. 2639. Lecture Notes in Artificial Intelligence
5406.
Scharf, Peter M. 2000. Ramopakhyana the story of Rama in the
Mahabharata: an independent-study reader in Sanskrit. URL:
http://sanskritlibrary.org.
. 2002. Ramopakhyana the story of Rama in the Mahabharata: an independent-study reader in Sanskrit. London: RoutledgeCurzon.
. ed. 2011. The Pacakhyanaka of Purn.abhadra: first XML edition. URL: http://sanskritlibrary.org.
324
Sellmer, Sven. 2013. Words in metrical milieux: a computational

approach to a neglected field. Proceedings of the Fifth International Sanskrit Computational Linguistics Symposium, ed.
by Malhar Kulkarni and Chaitali Dangarikar, pp. 30120.

K ESHAV M ELNAD, PAWAN G OYAL, and P ETER S CHARF
Abstract: A significant portion of Sanskrit literature composed over more than three millennia beginning with the
Vedic hymns is composed in metrical verse. Discussion
of particular types of Sanskrit meter appears even in the
oldest extant Vedic text, the Rgveda, and the science of
meter documenting various poetic

meters is mentioned in
the oldest lists of disciplines. Based on an analysis of the
standard classical works of the science of poetics, Velankar
(1949) compiled an exhaustive list of more than six hundred different meters which was included by Apte, Gode,
and Karve (19571959) as an appendix. The present paper
presents Web-based software that analyzes Sanskrit metrical patterns and identifies meters. While using a precise
phonetic encoding it yet allows numerous input methods
and accepts either accented or unaccented text. The software has been successfully tested on a database of 1031
verses in the Pacakhyanaka, including 291 verses in 23
different types of meters besides Anus.t.ubh. The software
should be widely useful to Sanskrit students and scholars,
especially those who focus on poetics.
Keywords: prosody, metrics, poetics, Sanskrit meter, meter
identification, syllable parsing
Introduction
A major portion of Sanskrit literature is in the form of poetry. The

proportion of Sanskrit literature composed in verse over the period of more than three millennia from which works are extant is
325
326
M ELNAD, G OYAL, and S CHARF
significant. The earliest Vedic works, Rgveda, Samaveda, Atha composed in verse. The
rvaveda, and much of the Yajurveda, are
great epics Mahabharata and Ramayan.a are composed almost exclusively in verse. Works in every genre of classical Sanskrit literature from mathematics, linguistics, medicine, and philosophy to
the dramatic and literary arts are composed in verse. There are two
main purposes behind versification:
1. Composing a text in verse makes it easy to memorize.
2. Versified text is suitable for melodic and rhythmic chanting.
Much of Indic verse literature was intended for oral transmission,
ritual, and public performance.
The large body of Sanskrit poetry is composed in specific metrical patterns. Several metrical patterns are identified and discussed even in the Rgveda itself. Among the six ancillary disciplines called Vedan
gas associated with the study and understanding of the Vedas is the science of metrics (chandas). The Pan.inyasiks.a calls this science the feet of the Veda, chandah. padau tu
vedasya (PS. 41).
Identification of meter presents a difficult task for students and
scholars of Sanskrit. To offer assistance in this task, the present
work develops a software tool to recognize them automatically.
Recently Mishra (2007) developed software to analyze metrical
patterns. His software, which he deemed a test version, recognizes 1,352 metrical patterns (http://sanskrit.sai.uniheidelberg.de/Chanda/HTML/). Although the work is
prodigious, a few deficiencies detract from it:
1. It recognizes only meters with a fixed number of syllables
per verse quarter (pada) and cannot recognize meters based
upon the number of mor (matras).
2. It requires special treatment of contiguous vowels since the
Kyoto-Harvard encoding which it requires as input fails to
P OETIC SYNTAX
327
distinguish sequences of the contiguous simple vowels a

and a from the diphthongs ai and au.
3. It does not handle accented text, again due to limitations of
the Kyoto-Harvard encoding.
This paper presents a Sanskrit metrical analyzer, which we call
Meter Identifying Tool (MIT), that improves upon Mishras work
in some respects. It analyzes a wide range of meters, uses a precise phonetic encoding that accommodates accented text, and allows numerous input methods. While at present our tool recognizes only 661 metrical patterns, these include several types of
meters based upon the number of mor. Processing is done in
the Sanskrit Library Phonetic Basic encoding (SLP1) documented
by Scharf and Hyman (2011: 15158). SLP1 distinguishes diphthongs from sequential simple vowels and provides methods to indicate accented texts.
The paper is organized as follows. Section 2 describes the
source texts on which this work is based. Section 3 describes the
structure of Sanskrit meters. Section 4 describes MIT including
our database of metrical forms (4.1), the form of input and ouput
(4.2), and our algorithm for the identification of metrical patterns
(4.3). Section 5 evaluates the results of testing our tool. Conclusions and future work are discussed in section 6.
Sources describing Sanskrit meters
Ollett (2013) recently summarized the history of the Indian science

of poetics. The standard classical works include Pingalas Chandah.sa stra (c. 200 BCE) (Joseph 2011), Jayadevas Jayadevachandas (c. 600 CE), Jayakrtis Chandonusa sana (11th c.), Hemacandras Chandonusa sana (12th c.) (Velankar 1949), and Kedarabhat.t.as Vrttaratnakara (1112th c.) (Kedarabhat.t.a 1942) among
ngalas Chandahsa stra provides definitions of various
others. Pi
.
328
kinds of meters in a sutra text consisting of eight chapters. After

a general introduction to prosody in the first chapter and before
discussing the origin of meters in the closing chapter, Kedarabhat.t.a describes four types of meters in the central four chapters of
his Vrttaratnakara. Jayadeva introduced the brilliant mnemonic
technique
of composing the definition of a metrical pattern in the
very metrical pattern to be defined. For example, the definition
of the Indravajra meter given in (2) below (p. 332) consists of a
verse quarter in that meter. Such a definition is said to be endowed
with the object to be defined as well as its definition and is thus
termed in Sanskrit laks.yalaks.an.asamyukta.
Later poetic works on
Sanskrit meter employ this technique of definition as well.

Apte, Gode, and Karve (19571959) incorporated in Appendix
A Sanskrit prosody the list of Sanskrit metrical patterns compiled by Velankar (1949) from several of the classical Sanskrit poetic treatises. The second part of this appendix, called A classified list of Sanskrit meters, contains 769 metrical definitions.
MIT presently recognizes 661 of these, including all of the fixedsyllable (varn.avrtta) meters except those of the dan.d.aka variety.
Although MIT includes

several meters based upon the number of
mor, it has yet to include the bulk of these.
Sanskrit prosody
Sanskrit prosody is metrical. Numerous metrical patterns of several general types are based upon varying sequences of light and
heavy syllables that constitute a verse quarter or a line constituting
half a verse. After explaining the factors that determine syllable
weight, we describe basic units for the two major types of meter,
that based upon numbers of syllables (varn.avrtta) and that based
upon number of mor (matravrtta).
P OETIC SYNTAX
3.1
329
Syllable weight
Phonetically, a syllable consists of a single sonorous peak surrounded by less sonorous elements. In Sanskrit, syllables consist
of a single vowel or diphthong possibly preceded by up to five
consonants and possibly followed by a coda consisting of one or
two consonants. Indic scripts orthographically represent syllables
by any initial consonants, the vowel, and possibly an anusvara or
visarga. For metrical purposes, syllable weight is determined by
vowel length and the presence or absence of a subsequent consonant cluster. There are two weights for a syllable, viz. light (laghu)
and heavy (guru). Pan.ini in his As..ta dhyay dedicated 3 aphorisms
to explain the weight of vowels:
A. 1.4.10 :h;~vMa l+.Gua A short vowel is termed laghu.
A. 1.4.11 .sMa;ya;ea;gea gua: (:h;~va;m,a 10) A short vowel immediately
followed by a consonant cluster is termed guru.
A. 1.4.12 d ;a;Ga . ca (gua: 11) A long vowel is also termed guru.
In classical Sanskrit poetry, the conditions for determining a
light or heavy vowel, and a convention for marking them in writing, are described by Kedarabhat.t.a in the following verse:
.sa;a;nua;~va;a:=+ea ;
a;va;sa;ga;Ra;nta;ea d ;a;Ga;eRa yua;+:pa:=+(a yaH
va;a :pa;a;d;a;ntea tva;sa;Ea gva;k+:ea ea;ya;eaY;nya;ea ma;a;
a:a;k+:ea lx +juaH
(VR. 1.9 Kedarabhat.t.a 1942: 6)
A syllable is heavy if it has an anusvara, a visarga, or
a long vowel, or is followed by a consonant cluster,
and optionally if it occurs at the end of a pada. A
heavy syllable is denoted by a curly symbol (Y) and a
light by a straight line ( ).
Apte, Gode, and Karve (19571959: Appendix A, p. 1b) adds,
The consonant clusters :pra & :h, as also b.ra & k are said to be exceptions, before which the vowel may be short by a sort of poetical
330
license. In this paper, instead of the avagraha and dan.d.a, we use

the modern notations macron for heavy and breve for light. We
use va in our database to indicate the optional heaviness of the
final syllable of a pada. For example, consider the Vasantatilaka
verse quarter
(1) ;a;na;nd;ntua na;a;a;ta;a;na;pua;a;a ya;
a;d va;a .~tua;va;ntua.
Here the last syllable u is laghu. But the definition of the Vasantatilaka meter requires the final syllable of each verse quarter to be
heavy. In accordance with the rule concerning optional heaviness
of the final syllable of a pada (va padante), however, it may be
considered heavy and thus, the verse quarter matches the Vasantatilaka metrical pattern.
3.2
The basic unit of varn.avrtta meter: gan.a
The term gan.a is a technical term in metrics for a sequence of

three syllables. Every possible pattern of light and heavy syllables
is designated by a compound beginning with a single character and
ending in this term. Since there are two possible weights, laghu or
guru, for three syllables, there are eight (23 = 8) types of gan.as. A
popular verse of unknown origin describes the eight possible trisyllabic weight patterns, designated by the same number of terms,
by indicating where in the sequence of three syllables a light or
heavy syllable appears.
;a;
a;d;ma;Dya;a;va;sa;a;nea;Sua ya:=+ta;a ya;a;a;nta
/ / l+.a;Ga;va;m,a
Ba:ja;sa;a ga;Ea:=+vMa ya;a;a;nta
/ / ma;na;Ea tua gua:+:l+.a;Ga;va;m,a
At the beginning, middle, and end respectively, y, r,
and t go to lightness; B, j, and s go to heaviness, but
m and n go to heaviness and lightness.
P OETIC SYNTAX
331
Table 1 presents the eight patterns described in the verse.

Table 1
Gan.a patterns
stands for guru and for laghu
No.
1
2
3
4
5
6
7
8
3.3
ga;a
m,a
y,a
.=,
t,a
B,a
.j,a
.s,a
n,a
gan.a
m
y
r
t
bh
j
s
n
pattern
The basic unit of matravrtta meter: caturmatrika
Groups of syllables measuring four mor (matra) constitute the

basic unit of meters based on the number of mor. A laghu syllable is counted as one matra and a guru as two. Four matras constitute a caturmatrika. There are five possible patterns of a caturmatrika as shown in Table 2.
3.4
Types of meters
As mentioned above, there are two major types of meters: varn.avrtta and matravrtta, the first of which is based upon patterns of
as and the second
gan
of which is based upon patterns of caturma.
trikas. The first type has three subtypes. Hence the four types of
meters are as follows:
1. varn.avrtta
332

Table 2
Matra patterns
No.
1
2
3
4
5
Caturmatra
sarvaguru (G)
bhagan.a (B)
jagan.a (j)
sagan.a (s)
sarvalaghu (L)
Pattern
a. samavrtta
b. ardhasamavrtta
c. vis.amavrtta
2. matravrtta
A samavrtta meter has the same gan.a pattern in each of its four
ardhasamavrtta meter has two gana patterns, one in
padas. An
.
its first and third padas,and a different gan.a pattern in its second
and fourth padas. A vis.amavrtta meter may have a different gan.a
A matravrtta meter has a different
pattern in each of its four padas.
lines.
pattern of caturmatrikas in its first and second
Metrical patterns are defined by describing the sequences of
gan.as or caturmatrikas in a verse quarter or line. For example,
the samavrtta meter Indravajra is defined by stating that each verse
quarter consists
of the pattern of gan.as t t j g g. Apte, Gode, and
Karve (19571959: Appendix A, p. 4a) provides the following
laks.yalaks.an.asamyukta
definition of the Indravajra meter:
(2) .~ya;a;
a;d;nd+va:j"a;a ya;
a;d ta;Ea .ja;ga;Ea gaH
syadindravajra yadi tau jagau gah.
| | |
That is called indravajra if it consists of t t j g g.
P OETIC SYNTAX
333
We provide additional examples of the patterns of gan.as or caturmatrikas in verses while describing the output of MIT in 5 below.
Meter Identifying Tool (MIT)
MIT consists of a database of metrical definitions and a Java program. We describe the structure of the database, the input and
output, and our algorithm for conducting metrical analysis in the
following three subsections.
4.1
Meter database
Our meter database contains metrical definitions in an easily readable text file. The meter pattern conforms to a different prototype
for each of the four types of meters described in 3.4. Because a
samavrtta meter has the same pattern in all four padas, it is suffi the database to contain just one pattern for a pada to decient for
fine all four padas of a verse. Besides the pada pattern, a Boolean
variable indicates whether the metrical definition permits optional
heaviness of the final syllable of a pada (va padante); the value
is indicated as true when it does. Thus, a samavrtta metrical def
inition is stored in a prototype containing the meter
name and a
single pattern plus the Boolean value. Because an ardhasamavrtta
meter has one pattern for its first and third padas but a different
pattern for its second and fourth padas, the database must include
two patterns in its metrical definition each of which is followed by
the Boolean value. Because each of the four padas of a vis.amavrtta
meter may have a different pattern, the database must include inits
metrical definition four patterns each with its Boolean value. For
matravrtta meters, the patterns of caturmatrikas for the first and
secondlines are different. Thus a matravrtta meter is stored with
two patterns each of which is followed bythe Boolean value. Table 3 shows the prototypes of each of the four types of meters and
334
the number of metrical definitions of each type of meter contained

in our database as of the date of publication.
Table 3
Prototypes of each of the four types of meters
and the number of meters of each type in our database
Meter type
samavrtta
Prototype
<meter name>
<pattern1> <vA>
ardhasamavrtta <meter name>
<pattern1> <vA>
<pattern2> <vA>
vis.amavrtta
<meter name>
<pattern1> <vA>
<pattern2> <vA>
<pattern3> <vA>
<pattern4> <vA>
matravrtta
<meter name>
<pattern1> <vA>
<pattern2> <vA>
Total
4.2
Quantity
587
48
14
12
661
Input and output
MIT is an interactive program that accepts a full verse, line, or

verse quarter in Sanskrit and returns a five-part analysis. If the
input consists of a line or verse, the first line must end in a single dan.d.a and the second in a double dan.d.a. While at present
the input and display is in SLP1 (the encoding in which processing is performed), the tool will soon be provided with the Sanskrit
Library transcoding preferences that include input from and out-
P OETIC SYNTAX
335
put to popular meta-encodings such as Kyoto-Harvard, ITrans, Titus, Velthuis, and Hyderabad-Tirupati (WX), as well as to Unicode
Roman, and the Unicode representation of major Indic scripts Devanagari, Gurmukhi, Kannada, Bengali, Telugu, Malayalam, Gujarati and Oriya. The analysis returned consists of the following
five parts:
1. name and type of meter linked to a definition of the meter
2. the string parsed into orthographic syllables
3. the scansion (prastara) of the string showing the pattern of
light and heavy syllables.
4. the pattern of gan.as or caturmatrikas
5. the number of syllables or matras
4.3
Algorithm
After transcoding the input text to SLP1, the first task is to remove
any non-phonetic characters, except the periods that represent dan.d.as at the end of lines. The second step is to divide the string into
syllables. The definition of an orthographic syllable is CV X?,
where C stands for a consonant, V stands for a vowel, X stands
for the set {anusvara, visarga, jihvaamulya, upadhmanya}, and
and ? are regular expression quantifiers representing zero or more
and zero or one respectively. The regular expression used to match
an orthographic syllable in SLP1 is therefore:
[yvrlYmNRnJBGQDjbgqdKPCWTcwtkpSzsh]*
[aAiIuUfFxXeEoO][HMZV]?
Consonants final in a line are grouped with the last orthographic

syllable.
Once the syllables are found, we find the weights of the syllables, and then attempt to determine the pattern of gan.as or caturmatrikas. More restrictive metrical patterns are checked before
336
less restrictive ones. The input verse is checked for the samavr
tta, ardhasamavrtta and vis.amavrtta metrical patterns in that order.
Only when the input verse doesnot match any of these patterns,
do we search for a matravrtta meter pattern.
If the input text lacksline-end markers, it is assumed to be a
single pada and to belong to the samavrtta type of meter. If line
markers are present then an attempt ismade to divide the lines
into padas since these are not demarcated in the input. The pada
boundary is required for the meters of the varn.avrtta types, not for
to patterns per
the matravrtta meters which are defined according
line.
We first check the number of syllables per line and proceed to
search for the appropiate type of meter based upon the following
conditions:
1. If the number of syllables in the first line is the same as in
the second line and each line contains an even number of
syllables, then we divide the lines in half, generate the gan.a
patterns for each pada, and check whether the patterns of all
padas are the same. If so, we search for the pattern among
the samavrtta meter definitions before proceeding to search
for the pattern

among the ardhasamavrtta and vis.amavrtta
types.
2. If the number of syllables in the first line is the same as in
the second line, but each line contains an odd number of
syllables, then the verse cannot be of the samavrtta type. In
to those
this case, the search for a pattern proceeds directly
among the ardhasamavrtta type and, if not found, to those
among the vis.amavrtta
type.
3. If the number of syllables in the first line is different from

the number in the second line, then the verse cannot be of the
samavrtta or ardhasamavrtta types. In this case, the search
P OETIC SYNTAX
337
for a pattern proceeds directly to those among the vis.amavr

tta type.
Now, it is easy to find the pada boundaries in samavrtta meters
because the number of syllables is the same in each pada and we
can divide the set of syllables in each line into two equal parts. For
the ardhasamavrtta meters containing different numbers of syllables in each pa
da, or vis.amavrtta meters, on the other hand, we
there are numerous possible dineed a different method because
visions of each line into padas. If {x, y} represents the number of
syllables in padas one and two respectively, then, for example, for
an input line containing 17 syllables, it is possible to divide a line
into padas containing various numbers of syllables, {8, 9}, {9, 8},
{7, 10}, {10, 7}, etc. In general, for an input line containing x syllables, we will have x 1 possible pairs. This corresponds to the
fact that the first pada can contain any number of syllables from 1
to x 1, and the second pada will contain the remaining syllables.
For each possible division of a meter into padas as described
above, we search for matching gan.a patterns of the padas in our
database to determine whether there exists a meter definition that
matches these patterns. If we find a match, we stop searching and
output the result.
If the input verse does not match any of the samavrtta, ardha it with
samavrtta or vis.amavrtta type meters, we attempt to match
the matravrtta meters. The patterns for matravrtta meters can be

using a regular expression search.For example, the
implemented
a meter is: J4J4JjJg or J4J4JLJg, where
pattern for the Ary
j corresponds to jagan.a as listed in Table 2, g corresponds to a
guru syllable, 4 corresponds to any of the caturmatrika patterns,
J corresponds to any caturmatrika pattern other than jagan.a, and
L corresponds to a sarvalaghu. Hence we first convert each of the
meter definitions for matravrtta meters into an equivalent regular
expression automatically. Then,

we try to match the input verse
338
with the regular expressions corresponding to the various definitions.
Results
In this section, we will discuss one example verse from each meter
type and report the output produced by our algorithm when the
verse was given as input to MIT. The original verse is shown in Devanagar, the input below it, and the output next in SLP1 encoding.
5.1
A samavrtta verse in Vasantatilaka meter
;a;na;nd;ntua na;a;a;ta;a;na;pua;a;a ya;

a;d va;a .~tua;va;ntua
l+ma;aH .sa;ma;a;
a;va;Za;tua ga;.cC+.tua va;a ya;Tea;;m,a
;d;aE;va va;a ma:=+a;ma;~tua yua;ga;a;nta:=e va;a
nya;a;yya;a;tpa;TaH :pra;
a;va;.ca;l+.a;nta
/ / :pa;dM na ;Da;a:=+aH
nindantu nItinipuRA yadi vA stuvantu
lakzmIH samAviSatu gacCatu vA yaTezwam.
adyEva vA maraRamastu yugAntare vA
nyAyyAtpaTaH pravicalanti padaM na DIrAH..
The verse is in vasantatilakA meter which is a
samavftta.
The metrical patterns are...
......pAda 1:::::::::
Parsed String
: ni nda ntu nI ti ni pu RA ya
di vA stu va ntu
Syllable weights : g g l g l l l g l l g l g g
Gana Pattern
: t B j j g g
No. of Syllables : 14
......pAda 2:::::::::
Parsed String
: la kzmIH sa mA vi Sa tu ga cCa
tu vA ya Te zwam
Gana Pattern
: t B j j g g
P OETIC SYNTAX
339
......pAda 3:::::::::
Parsed String
: a dyE va vA ma ra Ra ma stu yu
gA nta re vA
Gana Pattern
: t B j j g g
......pAda 4:::::::::
Parsed String
: nyA yyA tpa TaH pra vi ca la
nti pa daM na DI rAH
Gana Pattern
: t B j j g g
5.2
An a rdhasamavrtta verse in Vegavat meter
.sma:= :vea;ga;va;ta;a v.ra:ja:=+a;ma;a :ke+:Za;va;vMa;Za:=+vEa:=+a;ta;mua;gDa;a

.=+Ba;sa;a;a gua:+.nga;a;ya;nta;a :ke+:a;l+.a;na;ku+.*+;g
xa;h;a;ya .ja;ga;a;ma
smara vegavatI vrajarAmA
keSavavaMSaravEratimugDA.
raBasAnna gurUngaRayantI
kelinikuYjagfhAya jagAma..
The verse is in vegavatI meter which is a arDasamavftta.
......pAda 1:::::::::
Parsed String
: sma ra ve ga va tI vra ja rA
mA
Syllable weights : l l g l l g l l g g
Gana Pattern
: s s s g
......pAda 2:::::::::
Parsed String
: ke Sa va vaM Sa ra vE ra ti mu
gDA
Syllable weights : g l l g l l g l l g g
Gana Pattern
: B B B g g
......pAda 3:::::::::
340

Parsed String
: ra Ba
ntI
Syllable weights : l l g
Gana Pattern
: s s s
......pAda 4:::::::::
Parsed String
: ke li
ma
Syllable weights : g l l
Gana Pattern
: B B B
5.3
sA nna gu rU nga Ra ya
l l g l l g g
g
ni ku Yja gf hA ya ja gA
g l l g l l g g
g g
A vis.amavrtta verse in Lalita meter
na;yua;gMa .sa;k+:a:=+yua;ga;lM . ca Ba;va;a;ta . ca:=+Ma txa;ta;a;ya;k+:m,a

ta;du;
a;d;ta;mua:+:ma;a;ta;a;Ba;lR +.a;l+.tMa ya;
a;d Zea;Sa;ma;~ya Ka;lu :pUa;vRa;tua;ya;k+:m,a
nayugaM sakArayugalaM ca
Bavati caraRaM tftIyakam .
taduditamurumatiBirlalitaM
yadi Sezamasya Kalu pUrvatulyakam .
The verse is in lalitam meter which is a vizamavftta.
......pAda 1:::::::::
Parsed String
: na yu gaM sa kA ra yu ga laM
ca
Syllable weights : l l g l g l l l g g
Gana Pattern
: s j s g
......pAda 2:::::::::
Parsed String
: Ba va ti ca ra RaM tf tI ya
kam
Syllable weights : l l l l l g l g l g
Gana Pattern
: n s j g
......pAda 3:::::::::
P OETIC SYNTAX
341
Parsed String
: ta du di ta mu ru ma ti Bi rla
li taM
Syllable weights : l l l l l l l l g l l g
Gana Pattern
: n n s s
......pAda 4:::::::::
Parsed String
: ya di Se za ma sya Ka lu pU
rva tu lya kam
Syllable weights : l l g l g l l l g l g l g
Gana Pattern
: s j s j g
5.4
a meter
A matravrtta verse in Ary
kx+:SaH ; a;Za;ZuaH .sua;ta;ea mea va;+:va;ku+:l+.f;a;a;Ba:=+a;&+.ta;ea na gax ;he

a;a;ma;
a;pa va;sa;tya;sa;a;
a;va;a;ta .ja;ga;a;d ga;ea;.a;Ma ya;Za;ea;d;a;ya;Ra
kfzRaH SiSuH suto me
vallavakulawABirAhfto na gfhe .
kzaRamapi vasatyasAviti
jagAda gozWyAM yaSodAryA ..
The verse is in AryA meter which is a mAtrAvftta
......pUrvArdha:::::::::
Parsed string
: kf zRaH Si SuH su to me va lla
va ku la wA Bi rA hf to na gf he
Syllable weight : g g - l g l - g g - g l l - l l
g - l g l - g l l - g
Pattern
: J 4 J 4 J j J g
No. of morae
: 30
......uttarArdha:::::::::
Parsed String
: kza Ra ma pi va sa tya sA vi ti
ja gA da go zWyAM ya So dA ryA
Syllable weight : l l l l - l g l - g l l - l g l
- g g - l - g g - g
Pattern
: J 4 J 4 J l J g
No. of morae
: 27
342
Recall that the pattern shown for matravrtta meters is different from the varn.avrtta meters; the pattern is explained in Table 2
of 4.3 (p. 332).
(p. 332) and at the end
5.5
Performance
To determine how useful the tool would be on a real dataset, we

evaluated the programs performance on all of the verses in Purn.abhadras Pacakhyanaka (Hertel 1908) extracted from the Sanskrit
Librarys digital edition of the work (Scharf 2011). The database
of these verses consists of 1031 verses, 291 of which contain 23
types of metrical patterns other than Anus.t.ubh. This database,
which also includes the name of the meter of each verse, served
as the gold standard for our evaluation. We used our tool to identify the metrical pattern corresponding to each verse. MIT correctly recognized 1018 out of the 1031 verses (98.7%) and 287 out
of the 291 non-Anus.t.ubh verses (98.6%). There were no cases in
which a meter was recognized incorrectly. On the contrary, our
tool discovered several data-entry errors in our digital text and in
the annotations of meter types that we then corrected by reference
to the original printed edition.
We analyzed the 13 verses in our database whose meters were
not identified to discover the reasons behind the failure. Our goldstandard identified 9 of the 13 verses not recognized as Anus.t.ubh,
ardulavikrd.ita. For the last and for one
3 as Upajati, and 1 as S
of the Anus.t.ubh verses, our meter definition did not include the
exceptional parameter described in 3.1 that permits syllables to be
considered short even if they occur before one of the four conjunct
consonants kr, pr, br, or hr (that is, r preceded by k, p, b, or h).
Our tool included only the most restrictive definition of Upajati
meter Apte, Gode, and Karve (19571959: Appendix A, p. 4a) list
according to which the meter consists of any combination of padas in the two meters Indravajra and Upendravajra. However, other
P OETIC SYNTAX
343
sa and
definitions they mention permit combinations of Indravam
sastha, Smrti and Sruti, and indeed combinations of any samaVam

tool did identify combinations of Indravam
sa and
vrtta meters. Our
sastha in two of the unidentified meters, and a combination of

Vam
three samavrtta meters in one of the unidentified meters. Of the
remaining 8Anus.t.ubh meters not recognized by our tool, 3 have
an extra syllable in one pada, 3 contain a prohibited sagan.a pattern
in the 5th through 7th syllables of the first or third pada, 1 contains
a prohibited nagan.a pattern in the 2nd through 4th syllables in the
first pada, and the last unidentified verse lacks a required jagan.a
pattern in its second pada. In every case in which the meter of a
verse was not identified, our tool correctly identified the metrical
pattern of each pada individually.
Conclusions and future work
In this paper, we described a tool for the analysis of Sanskrit

prosody and identification of meter type. The tool covers a wide
range of meters and has been successfully tested on a considerable
number of verses. The evaluation of this tool over a database of
1031 verses from the Pacakhyanaka gave impressive results with
greater than 98.6% accuracy.
At present the tool is limited by the ability to recognize just
the 661 meters in our database. Future work involves extending
our database to cover all kinds of meters and to deal with certain
peculiarities. As an example of one such peculiarity, various meter
definitions specify restrictions on the position of a caesura (yati)
and require that the preceding Sanskrit word end before it. For
ardulavikrd.ita meter has 19 syllables in each pada.
example, the S
An additional condition is that there is a caesura after 12 syllables
and that the preceding Sanskrit word end prior to it. To recognize whether these conditions are satisfied requires a parser able
344
to determine word boundaries. In future work we would like to

integrate the output of a text segmentation tool such as the Sanskrit Heritage Reader (Huet 2005; Huet and Goyal 2013) with our
system in order to check this condition.
Among meters not adequately handled in our tool at present
are Vedic meters. Generally Vedic meters conform to classical
definitions for meters bearing the same names regarding the number of syllables per pada but not regarding the patterns of light and
heavy syllables. In other cases, the definition of a meter in Vedic
differs from its classical definition. For example, the Gayatr meter
in Vedic has three padas each of which consists of eight syllables
whereas in the classical definition it has four padas of six syllables
each.
As mentioned in the previous section, our tool is currently over
restrictive in the realm of Upajati definitions. Likewise, only a few
a meters have been implemented. There are sevof the several Ary
a meter such as Gti, Vaitalya, Vaktra and
eral subdivisions of Ary
Matrasamaka with different metrical patterns. We plan to include
a subdivisions and
Vedic meters, additional Upajati patterns, Ary
other metrical patterns in future work.
Although the Web version of the program currently handles
just one verse at a time, a command-line version of the program
run locally is able to analyze any number of verses in a file. We
intend to provide this facility on the Web as well.
At present, if a string of text without line markers is submitted
for analysis, it is assumed to be a pada of a samavrtta meter and
the program quits if a matching samavrtta metricalpattern is not
found. We intend to add procedures to subdivide the string into
lines and padas to check for possible matches to other types of
meters as well. In this way one could submit any string to discover
whether it is metrical or not. After implementing this feature, we
plan to test MIT against a known database of non-metrical text to
see whether it falsely identifies metrical patterns. We also have
R EFERENCES
345
plans to extend our tool to highlight probable spelling mistakes in

the input verse.
The tool has many potential applications other than identifying
the metrical pattern of an input verse. The feature of discovering
whether an input text contains any metrical pattern could have applications to syntactic parsing. As Scharf, Goyal, Ajotikar, and
Savardekar describe in the preceding paper in this volume, poetry
and prose differ significantly in their syntax. If a tool could determine whether the input text is prose or poetry, it would improve
the performance of dependency parsers if they provide a different
set of penalties for poetry than they do for prose.
References
Apte, Vaman Shivaram, Parshuram Krishna Gode, and Cintamana
Ganesa Karve. 19571959. Revised and enlarged edition of
Prin. V. S. Aptes The practical Sanskrit-English dictionary.
3 vols. Pune: Prasad Prakashan.
Hertel, Johannes, ed. 1908. The Panchatantra: a collection of ancient Hindu tales in the recension, called Panchakhyanaka,
and dated 1199 A.D., of the Jaina monk, Purnabhadra. Harvard Oriental Series 11. Cambridge, Mass.: Harvard University.
Huet, Grard. 2005. A functional toolkit for morphological and
phonological processing: application to a Sanskrit tagger.
Journal of Functional Programming 15.4: 573614. URL:
yquem.inria.fr/~huet/PUBLIC/tagger.pdf.
Huet, Grard and Pawan Goyal. 2013. Design of a lean interface
for Sanskrit corpus annotation. Proceedings of ICON 2013,
the 10th International Conference on NLP, pp. 17786.
346
Joseph, George Gheverghese. 2011. The crest of the peacock: nonEuropean roots of mathematics. Princeton: Princeton University Press.
Kedarabhat.t.a. 1942. vrttaratnakara: with the gloss Saubhagyavat
and commentary
Ratnaprabha, ed. and comm. by Nrsimhade astrin. Lahore: Meharacandra Laks.man.adasa.
va S
Mishra, Anand. 2007. Sanskrit metre recognizer. URL: http :
//sanskrit.sai.uniheidelberg.de/Chanda/
HTML/.
Ollett, Andrew. 2013. The gan.acchandas in the Indian metrical
tradition. Contributions to current research in Indology: proceedings of the first International Indology Graduate Research
Symposium, September 2009, Oxford; vol. 1, Pus.pika: tracing
ancient India through texts and traditions, ed. by Nina Mirnig,
Pter-Dniel Sznt, and Michael Williams, ch. 15. Oxford:
Oxbow Books.
Scharf, Peter M., ed. 2011. The Pacakhyanaka of Purn.abhadra:
first XML edition. URL: http : / / sanskritlibrary .
org.
Scharf, Peter M. and Malcolm D. Hyman. 2011. Linguistic issues
in encoding Sanskrit. Delhi: Motilal Banarsidass.
Velankar, H. D., ed. 1949. Jayadaman: a collection of ancient texts
on Sanskrit prosody and a classical list of Sanskrit meters with
an alphabetical index. Haritos.amala 1. Includes the following
texts: Jayadevachandas by Jayadeva, Chandonusa sana by Jayakrti, Vrttaratnakara by Kedarabhat.t.a, and Chandonusa sa
na by Hemacandra.
Bombay: Haritosha Samiti.
On concord and government relations

in Sanskrit
P RASAD P. J OSHI
Abstract: Syntactic relations are chiefly of two types: concord and government. In concord, both items in a construction are marked for the concerned category of case,
number, person, etc., while in government one item is
marked for the category and the other governs or assigns
it. In Sanskrit, where concord and government relations
occupy a prominent place, a noun and an adjective in concord agree in gender, case, and number. A verb and noun
(denoting an agent or direct object) in concord agree in
person and number. A verb governs a karaka-noun by assigning it a specific karaka-role. Speijers Sanskrit Syntax
and V. S. Aptes The Students Guide to Sanskrit Composition provide a sufficient account for concord and government relations in Sanskrit. The present paper attempts
to glimpse into their treatment in the Pan.inian tradition of
Sanskrit grammar. Although rules of concord and government are not explicitly stated by the ancient grammarians,
their concept of samanadhikaran.ya apposition provides
a basis to explain concord relations, and the concept of anabhihita not expressed otherwise provides the ground
to explain government of a noun by a verb-form.
Keywords: concord, government, syntactic relations, thematic roles, karaka, Sanskrit
347
348
P. J OSHI
Introduction
The study of syntax identifies different relations among constituents of a sentence. Agreement is one such relation. We say
that there is grammatical agreement between items when a particular syntactic feature of one item is chosen with reference to that
of another in the sentence. The syntactic features are usually the
grammatical categories like gender, case, number, voice, aspect,
etc. The items in agreement thus show a kind of harmony. Agreement relations are chiefly of two types: concord and government.1
The importance of concord and government
The grammatical categories involved in concord and government

have particular importance in the syntax of Sanskrit and other similar languages which do not have fixed word order. In a language
like English where the word order is fairly fixed, the relation between items is already indicated by position. An adjective normally appears before or after the English noun, e.g. blue lotus
(adj.-noun), mission impossible (noun-adj.). We do not need a
formal marker to indicate its relation with the noun. But in Sanskrit, this being a highly inflectional language and fairly free of
word ordering rules, the concord-government patterns are all important in establishing syntactic structures. The grammatical categories thus involved in concord and government serve the purpose
of indicating syntactic relations. To indicate syntactic relations
is part of their function. Palmer (1971: 2.6) refers to government and concord as being two types of linkages (1) of a word
or class of word requiring a particular form of another word, and
(2) a form of one word requiring a corresponding form of another.
1 Bloomfield (1935: 19394) added cross-reference as a third type of agreement.
O N CONCORD AND GOVERNMENT
349
Speijers Sanskrit Syntax and V. S. Aptes The Students Guide to

Sanskrit Composition provide a sufficient account of the concord
and government relations in Sanskrit. The present paper attempts
to glimpse their treatment in the tradition of Sanskrit grammar as
represented by Pan.ini (c. 4th BCE) and his followers.
Concord
In concord, both the items in the construction are marked for the
concerned category of gender, case, number, or person, etc., e.g. in
English this book, where the pronoun this and noun book
both are marked for singularity, and these books, where both are
marked for plurality. In concord, there is a correspondence of categories between items. Concord is alternatively called agreement.
In Sanskrit, concord is chiefly held between (1) a substantive and
adjective, (2) a substantive and another noun, and (3) a predicate
verb and noun (of agent and object).
3.1
Concord of substantive and adjective
A substantive and adjective in Sanskrit fall in concord and agree in

gender, case, and number, e.g. nlam utpalam (a blue lotus). Here
both the items are marked for gender (ntr.), case (nom.), and number (sg.). Pan.ini does not recognize adjective and substantive as
two different categories as they are perceived by modern linguists.
He has used the term vises.an.a in the As..ta dhyay, but this is not an
equivalent term for what is called adjective in modern grammar.
It only means a qualifying word in general and can include an adjective and also a substantive in apposition (Joshi and Roodbergen
1993: 91). In the Pan.inian system of derivation, both noun and
adjective are equally termed pratipadikas (A. 1.2.45) and liable of
adding case suffixes (termed sup) (A. 4.1.1). Which of the seven
triplets consisting of singular, dual, and plural case suffixes is to be
350
P. J OSHI
employed after a noun is determined by different criteria like the

karaka relation, the adjoining word, expression of other relations,
and the sense of the nominal base. A. 2.3.46 and other rules in A.
2.3 sufficiently provide for this. In the sentence
(1) Ba;a;SmMa k+:fM k+.=+ea;a;ta One makes a terrible mat.
the substantive kat.a directly entering into the karaka relation of
karman with the verb gets an accusative case suffix according to
A. 2.3.2. Now the question is why the adjective bhs.ma should
also get the same accusative suffix as well as the same gender and
number as kat.a. No rule in the As..ta dhyay explains this. Patajali in his Mahabhas.ya on A. 2.3.1 (MBh. 1.442.5) discusses the
point. He says a stem cannot be used without a suffix, and a suffix cannot be used without a stem. Some inflection or other has
to come after the stem. This being so and no other case would
be able to establish the relation of bhs.ma with kat.a here bhs.ma is employed with the accusative. Or alternatively, he further
says, bhs.ma is equally a karman and hence can take an accusative
suffix. Finally changing on this, he says, ;Ta;va;a k+:f O;;va k+:mRa .~ya;a;t,a ta;tsa;a;ma;a;na;a;a;Da;k+.=+ya;a;;
+a;Sma;a;
a;d;Bya;ea ;
a;d;ta;a;ya;a Ba;
a;va;Sya;a;ta Otherwise
let kat.a only be the karman; bhs.ma can take the accusative on account of its being samanadhikaran.a with kat.a. Here it is clear that
Patajali regards samanadhikaran.ya as the basis for agreement of
case between a noun and its adjective. The word samanadhikaran.a means having common reference. The word bhs.ma and
kat.a are two different lexical items, but they refer to one and the
same thing simultaneously and are said in tradition to be samanadhikaran.a. The Kasikavrtti on A. 2.1.49, explains the term as
O;;k+:a;sma;a;Te
follows: ;a;Ba;a;pra;vxa; a:a;a;na;a;ma:a;~ya
/ / / Ra vxa; a:aH .sa;a;ma;a;na;a;a;Da;k+.=+ya;m,a,
which means it is a state of two or more words with different senses
having common reference. In modern terms these are words in ap-
351
position. The agreement of case is thus explained in the tradition

on the basis of samanadhikaran.ya.2
Now, what remains is concord of gender and number. As regards the gender and number of nouns, Pan.ini seems to have very
little to say. Some of his rules deal with them.3 His rule
A. 1.2.53 ta;d; a;Za;SyMa .sMa;a;a;pra;ma;a;a;tva;a;t,a
That should not be taught because the authority for it is
convention.
provides an indication that the prescription of gender and number of nouns lies outside the scope of the As..ta dhyay. Patajali
on A. 2.1.36 (MBh. 1.190.18) says, ;a;l+.*: +.ma; a;Za;SyMa l+.ea;k+:a;(ra;ya;tva;a;
a;+:*: +.~ya. Patajalis discussion on the gender and number of adjectives
provides an explanation. On A. 4.1.3 (MBh. 2.200.1) he says, gua;a;va;.ca;na;a;na;Ma Za;b.d;a;na;a;ma;a;(ra;ya;ta;ea ;a;l+.*: +.va;.ca;na;a;a;na Ba;va;a;nta
/ / ta;d;a;Ta;a Zua;*M : va;~:a;m,a
Zua;* :+:a Za;a;f ;a Zua;* :H k+:}ba;lH Zua;* :+:Ea k+:}ba;l+.Ea Zua;* :+:aH k+:}ba;l+.aH ya;d;sa;Ea d+v.yMa
; a;(ra;ta;ea Ba;va;a;ta gua;a;~ta;~ya ya;
a;+:*:M va;.ca;nMa . ca ta;u+a;~ya;a;
a;pa Ba;va;a;ta , which means
that the words expressive of properties like sukla (white) etc. have
gender and number like that of their a sraya (substantive). He has
repeated this point more than once in the Mahabhas.ya. For other
vises.an.as (qualifying words) which are not gun.avacanas, Patajali
makes a reference in general while commenting on A. 1.2.51. First
he merely observes that they agree with substantives in gender and
2 Panini
has used this term samanadhikaran.a eight times in his As..ta dhyay,
.
but has not defined it. It is supposed to be well known. Patajali in his Mahabhas.ya on A. 2.1.1 discusses the term as follows: ta;tsa;ma;a;na;ma;a;(ra;a;ya;tea ya;tsa;ma;a;nMa Ba;va;a;ta
;Ta;va;a ya;a;va;d U" ;ya;a;tsa;ma;a;na;d+v.yea;ea;a;ta ta;a;va;tsa;ma;a;na;a;a;Da;k+.=+ea;nea;a;ta d+v.yMa ;
a;h l+.ea;ke ;a;Da;k+.=+a;a;ma;tyua;pa;.ca;yRa;tea ta;Ta;a v.ya;a;k+.=+ea ;
a;va;pra;a;ta;
a;Sa:;d M . ca;a;na;a;Da;k+.=+a;va;a;.ca;a;tya;d+v.ya;va;a;.ca;a;a;ta ga;}ya;tea The
term samanadhikaran.ya then means the state of two or more words referring to
the same thing simultaneously, co-referentiality.
3 For example, lu +.
a;pa yua;+:va;d;a;a;+:va;.ca;nea (A. 1.2.51), .=+a:a;a;+:a;h;aH :pMua;a;sa (A. 2.4.29),
etc. Ghatage (1983: 20) writes, The necessity of defining a gender in a grammar
of Sanskrit can only arise if the usage of a word differs from the gender which is
taught for a particular formation, either a compound or a derivative.
352
P. J OSHI
number;4 and provides the ground for it while commenting on the

next sutra. There he reads a line .sa;ma;a;na;a;a;Da;k+.=+tva;a;a;tsa:
/ / ;d ;m,a .sa;ma;a;na;a;a;Da;k+.=+a;a;
a;d;Zea;Sa;a;a;na;Ma yua;+:va;;
+a;va;ea Ba;
a;va;Sya;a;ta, implying that the gender
and number of a visesan.a is like that of its substantive because of
samanadhikaran.atva (apposition).
3.1.1
Adjectives with taddhita-deletion
The adjectives wherein the taddhita termination is deleted in the

process of derivation need a separate treatment. A. 4.2.81 etc. prescribe the deletion of taddhita suffixes. Another rule, lu +.
a;pa yua;+:va;d, v.ya;a;+:va;.ca;nea (A. 1.2.51), says that when a taddhita suffix is lupelided, the word so formed takes the gender and number of the
original word to which the taddhita suffix was added, e.g. 2.
(2) k+:a;l+.*: +.aH .ja;na;pa;dH region which is inhabited by the Kalinga
ks.atriyas
Here the word kalinga loses the taddhita suffix an. prescribed in
the sense of his residence by A. 4.2.81. The suffix an. was added
after kalinga which stands as masculine plural in the analytic paraphrase k+:a;l+.*: +.a;na;Ma ;a;na;va;a;saH .ja;na;pa;dH, so assuming the same gender
and number of the original, it is to be used in the masculine plural. The adjective kalingah. is hence plural despite the coreferential
substantive janapadah. being singular.
s.a trees
(3) g{a;a;maH ; a;Za:=+a;Sa;aH village which is near Sir
Likewise in (3) the adjective thus formed with lup-deletion of a taddhita suffix provides an exception to the agreement of gender and
number with its substantive. They agree only in the single feature
of case.
4 ;nya:a;a;a;Ba;Dea;ya;va;
a;+:
*: +.va;.ca;na;a;a;na Ba;va;a;nta
/ / *: +:a;nya:a lu +.
a;k l+.va;aH .sUa;paH l+.va;a;a ya;va;a;guaH
l+.va;Ma Za;a;k+:m,a
353
Some such adjectives may agree in number and case with their
substantives but not in gender. Katyayana (on A. 1.2.52) gives
words such as hartak as examples as in 4.
(4) h:=+a;ta;k+.aH :P+.l+.a;a;na fruits of the Hartak tree
Here the adjective hartakyah. (with lup-deletion of a taddhita suffix by A. 4.3.167) agrees with its substantive phalani in case and
number but not in gender. It is feminine in accordance with its
original base hartak. In contrast, some adjectives agree in gender
but not in number, as in 5.
(5) Ka;l+.a;ta;kM va;na;a;a;na forests which are near to the mount Khalatika
Here the adjective khalatika (with lup-deletion of a taddhita by
A. 4.2.82) agrees with its neuter substantive in gender but not in
number. It is used in the singular despite its substantive vanani
being in the plural. It agrees in number with the original word
khalatika which is singular in the analytic paraphrase Ka;l+.a;ta;k+:~ya
:pa;vRa;ta;~ya ;dU:=+Ba;va;a;a;na.
On the other hand, the adjective may disagree in both gender
and number with its substantive. Consider the expression in 6.
(6) . ca:*.a ;a;a;Ba:+.paH handsome Caca
The word caca (with lup-deletion of the taddhita suffix kan by A.
5.3.98) means a man who looks similar to a caca (grass model).
Here the adjective abhirupah. (masc.) does not agree with its substantive caca (fem.). Some other words of this kind are vadhrika (fem.) (looking like a eunuch), kharakut. (fem.) (looking like
a barbars bag). The words hartak, khalatika, and abhirupa, although used adjectively, do not agree with their corresponding substantives in gender, number or both.5
5 Cf.
the varttikas h:=+a;ta;k+.a;a;

a;d;Sua v.ya;a;H and Ka;l+.a;ta;k+:a;
a;d;Sua va;.ca;na;m,a on A. 1.2.51
and ma;nua;Sya;lu +.
a;pa :pra;a;ta;Sea;Da;ea va;+:v.yaH on A. 1.2.52.
354
3.1.2
P. J OSHI
Nouns in apposition
Two or more nouns when they refer to the same object simultaneously fall in apposition (samanadhikaran.ya). Nouns in apposition
agree in case but not necessarily in gender and number. For each
noun has its own fixed gender, and its number is determined according to the circumstances. In 7,
(7) ;a;ma;ea vxa:aH mango tree
both nouns are in the nominative as well as agreeing in gender
and number. Yet consider the description of lord Rama in verses
4.15.15c-17d in the Ramayan.a passages in 8.
(8) .=+a;maH :pa:=+ba;l+.a;ma;d ;Ra yua;ga;a;nta;a;a;+;a; a:=+va;ea;a;tTa;taH
/ / 15
;a;na;va;a;sa;vax :aH .sa;a;DUa;na;a;ma;a;pa;a;a;na;Ma :pa:=+a ga;a;taH
;a;ta;Ra;na;Ma .sMa;(ra;ya;(Ea;va ya;Za;sa;(Ea;k+:Ba;a:ja;na;m,a 16
...
...
gua;a;a;na;a;ma;a;k+.=+ea ma;h;a;n,a 17
Here ramah. is the chief noun. It is being qualified by other nouns
viz. nivasavrks.ah. (masc.), para gatih. (fem.), sam
srayah. (masc.)
(ntr.). They all agree in case but disagree in gender.
and bhajanam
Similarly, a verse in the Saduktikarn.a mrta (1.11) reads
(9) ta;~ya;a;sa;a;t,a . . . (ra;a;va;fu;d;a;sa I+.tya;nua;pa;ma;prea;mEa;k+:pa;a:Ma .sa;Ka;a

Vat.udasa, the only resort of matchless love, was a friend to
him.
Another example of this kind is
(10) :vea;d;aH :pra;ma;a;a;m,a The vedas are the authority.
Here vedah. (masc. pl.) occurs in apposition with praman.am
(ntr. sg.). The words agree in case but not in gender and number.6 The words patram, a spadam, sthanam, padam, praman.am,
6 Gadadhara
discusses this point in his Vyutpattivada. I am restricting myself

here to the Pan.inian system.
355
bhajanam, sadhanam, karan.am, rasih., prabhavah. are some of the

common nouns occurring as qualifying words of the chief noun
that agree with it only in the case suffix. The concord between
nouns in apposition can be viewed as partial. For the agreement
of case is the only guarantee. Pan.inis rule ;
a;va;Zea;Sa;a;a;na;Ma . ca;a:ja;a;teaH (A.
1.2.52) has a slender reference to this point. It says that given lupelision of a taddhita suffix, the gender and number of a vises.an.a
(qualifying word) which does not refer to a jati (generic property)
are like those of the taddhita-derived qualified word. Here the
term vises.an.a means a qualifying word in general and can also
stand for a substantive in apposition. A substantive in apposition
not referring to things in a general way but to a particular item does
not have gender and number like that of the qualified word. The
example given by the Kasikavrtti is shown in 11.
(11) :pa:*.a ;a;l+.aH .ja;na;pa;dH the country Pacalah.

Here according to the tradition, janapada has to be a qualifying
word referring to a particular item a single district and hence
has no agreement of gender and number with the plural pacalah.. It is only in the context of taddhita-deleted words, that the
disagreement of gender and number is explained in the As..ta dhyay.
3.2
Concord of a predicate verb and noun (of agent and

object)
A verb of Sanskrit agrees with the noun of agent or object in the

sentence. In 12,
(12) :de;va;d:aH k+:fM k+.=+ea;a;ta Devadatta makes a mat.
the verb karoti and noun devadattah. have shown a concord relation by the agreement of person and number. Both are marked
for third person and singularity. This feature is clearly provided
356
P. J OSHI
for by Pan.ini. In his system the l-suffixes stand for the karakas
(agent and object). For intransitive roots, they refer to the action
instead of an object (A. 3.4.69). For example, in karoti in (12),
the verbal ending -ti (substitute of l-suffix) refers to the agent of
the action denoted by kr-. In kriyate, -te stands for the object. A.
termed madhyama (second person) are
1.4.105 says: the endings
to be employed after a verbal root, when the co-occurring pronoun
yus.mad (you), explicitly used or unused, is samanadhikaran.a (in
apposition) with the l-suffix. The point is that the karaka (agent or
object) referred to by the general l-suffix and the karaka referred to
by the co-occurring pronoun yus.mad has to be identical. When the
agent or object referred to by the l-suffix is same as that referred
to by the pronoun asmad (first person pron.), then those termed uttama (first person) are employed (A. 1.4.107), and in other cases,
the prathama (third person) endings are employed, as in (13).
(13) a. tvMa :pa;.ca;a;sa You cook.
b. ;hM :pa;.ca;a;a;ma I cook.
c. .saH :pa;.ca;a;ta He cooks.
Here the pronouns tvam,
aham and sah. and the respective verbal
endings fall in samanadhikaran.ya (apposition) by referring to the
same karaka of agent as expressed by the l-suffix. They agree with
the verb in person and number. In contrast, in 14
(14) ;hM tva;Ma na;ma;a;a;ma I salute you.
the verbal ending -mi (replacement of an l-suffix) refers to the kartr (agent) and the co-occurring pronoun tvam, accusative singular
of yusmad, refers to the karman (object). Due to the absence of
samanadhikaran.ya, they disagree in features. The word samanadhikaran.e in A. 1.4.105 yua;Sma;du;a;pa;pa;de . . . works as a condition. The
condition is that the upapada and l-suffix denote the same karaka. The upapadas are yus.mad, asmad, or the ses.a (others). The
357
principle of samanadhikaran.ya (apposition) thus provides the basis

for agreement in person and number of a verb and karaka noun.
Government
In the government relation, one of the items in the construction

is marked for the feature in question and the other is the assigner
of that feature to it. Usually prepositions and verbs are governing
elements in any natural language. They govern the case of nouns.
In Sanskrit the government relation is shown by (1) a verb and
noun, (2) indeclinable and noun.
4.1
Government of a verb and a noun
The predicate verb in a Sanskrit sentence governs the case of the

nouns in that sentence. Consider the sentence
(15) :de;va;d:aH k+:fM k+.=+ea;a;ta Devadatta makes a mat.
The predicate verb karoti governs the nominative of the noun devadatta and the accusative of the noun kat.a. This phenomenon of
government is sufficiently explained by Pan.ini. In his system the
verb governs the case of a noun via the assignment of a karaka
role to it. The assignment of different karaka roles to the nouns is
dealt with in the As..ta dhyay under the section headed by the rule
k+:a:=+ke (A. 1.4.23). Another section of the As..ta dhyay headed by the
rule ;na;a;Ba;
a;h;tea (not expressed otherwise) (A. 2.3.1) prescribes the
case suffixes of nouns. The karakas are expressed by (1) a finite
verb ending (tin), (2) a derivational suffix (krt or taddhita), or by
(3) a compound. When the karaka notion of anoun is unexpressed
by any of these means, then to express that notion a case suffix is
added after it by the rules coming under ;na;a;Ba;
a;h;tea (not expressed
otherwise) (A. 2.3.1). If the karaka notion is already expressed by
the mentioned means then the noun is added with a nominative
358
P. J OSHI
suffix to denote the mere sense of its nominal base. In (15), the
verbal ending -ti stands for the karaka notion kartr. So the noun
devadatta whose kartr notion is already expressed,takes the nom
inative suffix (A. 2.3.46).
But the karaka notion karman of kat.a
is unexpressed, and hence kat.a is employed with the accusative
suffix to express the same. While in 16,
(16) k+:fH ;
a;k+:ya;tea A mat is being made.
the notion karman of kat.a is already expressed by the verbal ending
-te (A. 3.4.69), hence the accusative cannot be used. The predicate
verb thus governs the case suffix of a noun via assigning to it a
karaka role. But mere assignment of karaka role is not enough
for governing case. The principle of abhihita/anabhihita works in
between as a condition. The condition is that the karaka notion has
to be unexpressed. If the karaka notion is already expressed then
the noun is required to take the nominative, otherwise it is required
to take the appropriate case denoting the karaka in question. A.
2.3.1 and the following rules deal with the case affixes and the
karaka notions expressed by them.
4.2
Government of an indeclinable and a noun
Several indeclinable words in Sanskrit govern case affixes of

nouns, e.g.
(17) h:=+yea na;maH salutation to Hari.
The indeclinable word namas governs the dative of the noun hari
(A. 2.3.16). The indeclinable antara (between) governs the accusative (A. 2.3.4) as in the following example:
(18) ;nta:=+a tva;Ma ma;Ma . ca k+:ma;q+.lu H The pot is between you and me.
359
This case affixation is traditionally called upapadavibhakti, case

governed by an adjoining word (usually an indeclinable). It includes the items termed karmapravacanya, like anu which governs the accusative (A. 2.3.8) as in
(19) vxa:a;ma;nua towards a tree,
adhi which governs the locative (A. 2.3.9) as in
(20) ;a;Da :pa:*.a ;a;le +.Sua over Pacala,
and apa which governs the ablative (A. 2.3.10) as in
(21) ;pa ;
a:a;ga;teRa;ByaH vxa;;ea :de;vaH It rained except in Trigarta.
Besides an indeclinable, some adjectives also are found to govern the case of nouns. Consider examples (22) and (23).
(22) ma;a;ta;a:= .sa;a;DuaH honest towards his mother
(23) a. k+:mRa; a;a ku+:Za;lH efficient in work
b. k+:maR ;aH ku+:Za;lH efficient in work
The adjective sadhu requires its constituent member to be in the
locative (A. 2.3.43). The adjective kusala requires its constituent
member to be either in the locative or in the genitive (A. 2.3.40).
Similarly the adjective utsuka governs either the locative or instrumental case in accordance with A. 2.3.44, for example,
(24) a. :ke+:Zea;Sua o+.tsua;kH concerned about hair.
b. :ke+:ZEaH o+.tsua;kH concerned with hair.
Conclusions
Concord-government patterns occupy a central place in the syntax

of Sanskrit. Concord relations can be viewed as complete or partial. They are complete for a noun and a gun.avacana (expressive of
360
P. J OSHI
a property) in the sense that the noun and gun.avacana word agree
in all features of gender, case, and number. They are also complete
for a verb and noun agreeing in person and number. The concord
involving taddhita-deleted words and nouns in apposition may be
partial with guarantee only of case. Pan.inian rules of verbal endings provide for the concord between a verb and noun. But there
are no rules in this system which make an obvious statement of
the concord between an adjective and a noun. The reason seems
to be Pan.inis lack of treating adjectives as categorically distinct
from nouns. Patajali explains it taking resort to the principle of
samanadhikaran.ya (apposition). Pan.ini sufficiently provides for
the government relations between a verb and noun, indeclinable
and noun, and adjective and noun. The concept of samanadhikaran.ya (apposition) provides a basis to explain the concord relation
among items. As to the government between a verb and noun, it is
the concept of anabhihita (not expressed otherwise) that provides
the ground. Understanding these concepts would help in analyzing
a Sanskrit sentence.
References
Bloomfield, Leonard. 1935. Language. London: George Allen and
Unwin.
Ghatage, A. M. 1983. Systematics of Pan.inis As..ta dhyay. Proceedings of the International Seminar on Studies in the As.t.a dhyay of Pan.ini, ed. by S. D. Joshi and S. D. Laddu, pp. 17
21. Pune: University of Poona.
Joshi, S. D. and J. A. F. Roodbergen. 1993. The As..ta dhyay of
Pan.ini: with translation and explanatory notes. Vol. 2. New
Delhi: Sahitya Akademi.
Palmer, F. R. 1971. Grammar. Harmondsworth: Penguin.

Abstract: Many idiomatic peculiarities of Sanskrit have been
explicitly noted in the As..ta dhyay. However, a similar
account of others observed in Puran.ic, epic and classical
Sanskrit usage is missing. Charudeva Shastri in his book
Vagvyavaharadarsa has taken a note of numerous such usages. In the second half of this book, he presents a corpus
of six hundred and fifty-five sentences drawn from modern Sanskrit literature which he deems to be erroneous on
various grounds. In around fifty-one of those sentences, he
points out errors in the usage of sup-suffixes. In this paper,
we attempt to comprehend the discussions that explain the
errors in a few of these sentences and to fathom the reasoning involved. Based on the discussions, we wish to come
up with directive rules for avoiding the errors mentioned
and then try to tally them with positive descriptions such as
found in Speijers (1886) Sanskrit syntax. We also attempt
to present karaka trees for these erroneous sentences with
a hope that they prove to be valuable to the machine translation systems being developed by various institutions.
Keywords: erroneous sentences, parse trees, dependency
trees, karaka, modern Sanskrit literature.
Introduction
This work aims to examine the way in which errors in modern

Sanskrit usage are viewed by traditional Sanskrit scholarship. The
361
362
D. K ATIRA AND M. K ULKARNI
work is part of a project undertaken by us to edit, annotate and dis

cuss the three main modern works in this regard, namely, the Sabdapasabdaviveka and Vagvyavaharadarsa by Charudev Shastri
(1955, 1976), and the Suddhikaumud

by Janardan Hegde (2004).
These three works together give us a collection of more than a
thousand sentences that are deemed erroneous by the authors of
these works. While listing erroneous sentences, these authors also
discuss the causes of errors; these discussions are of particular
interest to us. The methodology in analysing these works relies
mainly on comprehending the discussions that explain the errors
and on fathoming the reasoning involved. The idea is to analyse
the reasoning and arguments in as neutral a manner as possible.
For the purpose of the present paper, we focus only on the sentences with errors in the usage of sup-suffixes mentioned in one of
these three texts, namely the Vagvyavaharadarsa.
It is well-known that nominal terminations (sup) in Sanskrit
are deployed in the following situations:
when a certain karaka relation is to be denoted
to denote the same sense as that expressed by the pratipadika
to denote various relations (like servant-served relation etc.)
that are not karaka relations
to denote the connection of a word with a karmapravacanya
when a word cooccurs with another word specifically listed
in the As..ta dhyay.
There cannot be any function-based theoretical explanation for the
assignment of the saja karman to the substrate (adhara) of the
action refered to by adhi-sn , adhi-stha, and adhi-as by
A. 1.4.46 ;a;Da;Za;a;*. :a;a;sa;Ma k+:mRa
and for the consequent second case. The second case is assigned
merely to adhere to the idiom of Sanskrit. Many such idiomatic
PARSE TREES FOR ERRONEOUS SENTENCES
363
peculiarities of Sanskrit have been explicitly noted in the As..ta dhyay. However, a similar account of others observed in Puran.ic,
epic and classical Sanskrit usage is missing. Charudev Shastri has
taken note of numerous such usages, a few of which have been
taken up for discussion in this paper.
Modern Sanskrit
Although the term modern Sanskrit literature is often used by

scholars, the exact period to be referred to by the term modern in
this context is not standardized. However, writers on modern Sanskrit such as Raghunathacarya (2002), Shukla (2002), and Ranganath (2003) consider the literature composed from the 19th century onwards as modern Sanskrit literature. Mishra (2000) in the
preface of his recently compiled work Vim
sasatabdsamskr
taka
vyamrtam remarks,
va;~tua;taH ;
a;va;ga;ta d;ea Za;ta;a;a;b./ d;ya;e<a mea ;a;l+.Kea ga;yea :pua;Sk+:l .sMa;~kx+:ta +;a;ya
k+:a ;a;na;Spa:a .=+a;a;ta .sea ;
a;k+:ya;a ga;ya;a .sa;a;*: +.ea;pa;a;*: ;Dya;ya;na
va;a:*
;a;sa:;d k+.=+ta;a ;hE ;
a;k .sMa;~kx+:ta .=+.ca;na;a;Da;a;mRa;ta;a meMa ;va;Ra;.ca;a;na;ta;a
(;a;Dua;a;na;k+:ta;a) k+:a .sUa:a;pa;a;ta ;a;b.ra;a;f;Za Za;a;sa;na;k+:a;l meMa ;Ga;a;f;ta
:pua;na:ja;Ra;ga:=+a k+:a;l .sea h ;a hu ;a.
Actually, an unbiased and thorough study of the bountiful Sanskrit literature written in the last two centuries
proves that modernness has crept in as a peculiar character of Sanskrit writings since the renaissance that
happened during the British rule.
Error analysis
The word error has different meanings and usages relative to how
it is conceptually applied. The meaning of the Latin word is wandering or straying. An error is hence a deviation from accuracy
364
or correctness. The term error is defined in different ways according to the context and the subject in which it is used. In the
case of language, an individual language users deviations from
standard language norms in grammar, syntax, pronunciation and
punctuation are referred to as errors. The term error is of prime
importance in applied linguistics. Johnson and Johnson (1999)
define this term as breach of a language code resulting in an unacceptable utterance. They also contrast errors with lapse or mistake. According to them, lapses or mistakes are the result of a
failure of performance, while errors are due to an incorrect grasp
of the language.
Errors have always been a subject of interest not only for linguists but also for researchers from other spheres such as psychology, mathematics, etc. Error analysis views the errors positively.
Errors help teachers to find out how far the pupils have progressed
and what remains for them to be learnt. To a researcher in linguistics they provide evidence of how language is learnt and of the
strategies and procedures the learners tend to employ. Errors for
learners are tools for testing their hypotheses about the nature of
the language they are learning. Error analysis is of particular interest in second language learning. Just as the incorrect utterances of
children provide important clues regarding child language acquisition, errors committed during learning a second language provide
evidence of a built-in syllabus (a definite system of language at
every point in the course of its development) of the second language learners. Here an efficient language teaching model may be
prescribed based on clues provided by the errors.
The standard of correct Sanskrit
Generally, to do error analysis in any language, the speech of the

native speakers of that language is considered as standard. Er-
365
rors are determined by contrasting the utterances of those who are

learning that language as their second language with this standard.
By collecting these contrasting utterances, a corpus of deviations
from the target language norms is compiled. The deviations in this
corpus are then classified in various ways. Thus, to determine deviations, we have to determine the standard first. To determine the
standard, the structure underlying the speech of native speakers (at
various levels) is sought to be understood and formally laid down
in the form of rules. The utterences of second language learners
are tested against these rules.
In the present study, all the modern Sanskrit literature that is
printed and published comes under the purview of this work. Our
focus is on the corpus of six hundred and fifty-five (seemingly erroneous) sentences compiled by Charudev Shastri in general, and
on the sentences deemed to be erroneous on the grounds of improper use of sup-suffixes in particular. We critically examine various reasons assigned by Shastri for deeming them erroneous and
attempt to frame directive rules for avoiding such errors and forming sentences agreeable to the idiom of Sanskrit. We wish to see
whether these directive rules throw any light on the positive description of Sanskrit available in works like Speijer (1886). We
also wish to trace the new trends in Sanskrit usage current due to
various factors and try to see which errors (as per traditional grammarians) can be easily accomodated in the structural framework
of Sanskrit. We also take into consideration several usages that
have gained currency post Shastris works and attempt to deliberate upon how they can be accounted for.
As discussed above, determining the standard and contrasting
the usages with the standard forms the core of error analysis. Today it cannot be said with reasonable surety about even a single
speaker of Sanskrit that a person has acquired Sanskrit as his/her
mother tongue through a family tradition that can be traced back
to antiquity. It is therefore preferable to stick to Charudev Shas-
366
tris idea of considering the usages of the learned (sis..ta) as standard. Shastri quotes the famous lines from the Mahabhas.ya on A.
6.3.109 to specify his idea of sis.t.a:
;a;ya;Ra;va;teRa ;a;na;va;a;sea yea b.ra;a;;a;aH ku+:}Ba;a;Da;a;nya;a ;l+.ea;lu +.pa;a ;gxa;h:a;ma;a;a;k+.=+a;aH ;
a;k+:a:*.a ;d;nta:=e+a k+:~ya;a; a;(a;
a;d;d;a;a;ya;aH :pa;a:=+*: +.ta;a;~ta:a;Ba;va;ntaH ; a;Za;;aH
However, one can easily make out from various sources that Shastri cites in his work to determine correct usages that many authors
post Mahabhas.ya also find place in Shastris list of sis.t.as. For example, Shastri (1976: 94) cited the Amarakosa, which is posterior
to the Mahabhas.ya, to evidence a point discussed below in 5.5.
Yet Shastris list of sis.t.as ends with several authentic writers of
Classical Sanskrit.
We accept this idea of sis.t.as from Shastri and further propose
to conceptualize this list as an open-ended one which can include
more authentic writers of modern times as well. Language is a
flow. An attempt to completely curb this flow with a rigid rule set
can lead to its complete ruin. However, language is also a system.
An exhaustive discription of a language is equally desirable at all
times to serve as banks of this flow regularizing it and determining
its domain so that it does not flood and disrupt ordinary behavior (vyavahara). With the advancement of time, new concepts and
situations are created and many old ones become extinct. For a
vibrant language,1 which adapts itself to changing situations, the
entry of new usages with changing times is inevitable. However,
at the same time the inherent system ensures that these new usages
are suitable to the idiom of the language. Slowly these new usages
get assimilated into the system and then govern the entry of newer
1 Details
of 34 periodicals in Sanskrit are available on the following page of

the popular Sanskrit Documents website: http://sanskritdocuments.
org/news/SanskritNewspapersandMagazines.html
367
ones. Holding the usage of sis.t.as as standard is indeed appropriate, especially in the absence of native speakers who can validate
various usages intuitively. However, we propose an open-ended
group (similar to the concept of an a krtigan.a found in Pan.inian
of modern Sanskrit can
grammar) wherein authentic new writers
be assimilated.
Discussion of erroneous sentences
5.1
Sentence 1
(1) I+.h Ba;a:=+tea :pa:*.a ;sa;h;~:a;va;SeRa;ByaH :pra;a;

a;*: +:ma;
a;pa ma;tMa na;a;sa;a;d;nya:a ;vEa;
a;d;k+:a;t,a
(180)2
Here in India, five thousand years ago, there was no religion
other than the Vedic religion.
5.1.1
Shastris comments
:pa:*.a;sa;h;~:a;a;ma;a;ta .sa;ma;a;saH :pa:*.a ;a;a;Da;kM .sa;h;~:a;ma;a;h na tua :pa:*.a .sa;h;~:a;a;a;a;a;ta ;
a;va;
a;d;tMa ;
a;va;du;Sa;a;m,a .tea;na :pa:*.a ;Bya;ea va;SRa;sa;h;~:ea ;Bya I+.a;ta
va;u+:mua;a;.ca;ta;m,a ta;d;
a;pa .sa;d;ea;SMa ;
a;va;va; a:a;ta;a;Ta;Ra;sa;ma;pRa;a;a;t,a na ;
a;h ta;taH :pUa;va ;vEa;
a;d;ke+:ta:=+n}.a;tMa na;a;sa;a;
a;d;a;ta ;
a;va;va:a;a;ta .tea;nea;taH :pa:*.a ;sua
va;SRa;sa;h;~:ea ;a;Sva;
/ a;ta va;+:v.ya;m,a ;sa;kx+:a;a;ya;ma;Ta;eRa ;
a;va;vxa;ta I+.a;ta nea;h
;
a;va;ta;nya;tea
Scholars know well that the compound pacasahasra means one thousand and five and not five thousand. Hence it would be appropriate to say pacabhyo vars.asahasrebhyah. instead of pacasahasravars.ebhyah.. This sentence would also be wrong as it
does not deliver the expected meaning. The author
does not mean that there was no religious belief other
2 The
numbers in parenthesis written after each sentence hereafter stand for

the number of that sentence in Shastri 1976.
368

than belief in Vedic religion before that. Hence, one
should paraphrase the sentence as itah. pacasu vars.asahasres.u. I have already explained this point several times before, so I am not elaborating upon it here
again.
There are two issues that Charudev Shastri discusses regarding

this sentence.
1. The word pacasahasra: Elsewhere in his writing (e.g.
1976: 103), he has already clarified that just as in the case of
words like ekadasa the compound is traditionally analyzed
as ekadhikah. dasa to deliver the precise meaning eleven,
so too here the compound pacasahasra should be analyzed
as pacadhikam
sahasram and thereby mean one thousand
and five and not five thousand as intended by the author
of the sentence under consideration. To mean five thousand the speaker would have to say paca sahasran.i without compounding the two words.
2. Shastri also finds a problem in the phrase pacasahasravars.ebhyah. prak. Interpreted literally, it does not refer to the
situation five thousand years ago; instead it refers to the situation before five thousand years.
Hence Shastri suggests a different and a more apt way of conveying the meaning intended.
(10 ) I+.taH :pa:*.a ;sua va;SRa;sa;h;~:ea;Sua I+.h Ba;a:=+tea ;
a;k+:ma;
a;d;k+:a;t,a
5.1.2
Discussion
According to Charudeva Shastry one should use the uncompounded phrase paca satani to mean five hundred. However, the
369
question then arises, how do we make this whole number an adjective? The phrase paca satani rupyakan.i is not correct because the
word satani, does not denote the enumerated object (samkhyeya).
Hence, it cannot qualify rupyakan.i. Shastri in his comments on

the sentence under consideration provides us with one solution for
this. The compound rupyakasatam should be formed equivalent to
the phrase rupyakan.a m
satam. Now rupyakasatam is a tatpurus.a
compound with the subsequent component sata as the principal element. Hence, its neuter gender and singular number are retained
even in the compounded state. Rupyakasatam can be deemed to be
denoting the number (samkhy
a). The compound as a whole is now
qualifiable by paca. Thus we can say paca rupyakasatani. This

is the basis of using the phrase pacasu vars.asahasres.u suggested
by Charudev Shastri.
Another way of expressing the same can be rupyakan.a m
pacasat in accordance with the usage observed in the Pacatantra,
(1a) :a . ca ma;;a;Ma vxa; a:Ma Bua:*+;a;na;a;na;M
a :pa;a;q+.
/ / ta;a;na;Ma :pa:*.a ;Za;ta;a ;a;ta;+a;ta (Jha
1994: 3).
Here, the compound is derived in the following manner: pacanam
satanam
samaharah. = pacasata + n
p pacasat.3 The
compound when thus derived denotes a samkhy

a. The samkhyeya
has to be connected to it with the sixth case. In the sentence under
consideration, we may alternatively use this method and say
(1b) I+.taH va;Sa;Ra;a;Ma :pa:*.a ;sa;h;~:ya;Ma I+.h Ba;a:=+tea ;
a;k+:ma;
a;pa ma;tMa na;a;sa;a;d;nya:a
;vEa;
a;d;k+:a;t,a
Both methods though grammatically correct are quite complex.
Instead of saying five thousand years, had the author wanted
to say five thousand five hundred and fifty five years, he would
have to say, either (1c) or (1d).
3 A.
2.4.30 varttika: ;k+:a:=+a;nta;ea:a:=+pa;d;ea ;

a;d;guaH ;a;~:
/ a;ya;Ma Ba;a;Sya;tea (MBh. I.480.1) and
A. 4.1.21 ;
a;d;ga;eaH.
370
(1c) I+.taH :pa:*.a ;pa:*.a ;a;Za;du:a:=e+Sua :pa:*.a;sua va;SRa;Za;ta;a;a;Da;ke+:Sua :pa:*.a ;sua va;SRa;sa;h;~:ea;Sua I+.h
Ba;a:=+tea ;
a;k+:ma;
a;d;k+:a;t,a
(1d) I+.taH va;Sa;Ra;a;Ma :pa:*.a ;pa:*.a ;a;Za;du:a:=+a;ya;Ma :pa:*.a ;Za;tya;a;Da;k+:a;ya;Ma :pa:*.a ;sa;h;~:ya;Ma I+.h
Ba;a:=+tea ;
a;k+:ma;
a;d;k+:a;t,a
Either is a verbose expression. Modern languages permit the use
of figures instead. (2007: 127) justifies the expression pacasahasravars.ebhyah. saying the following:
;
a;d;gua; a;a;tMa Za;ta;m,a = ;
a;d;Za;ta;m,a, ;
a:a;gua; a;a;tMa Za;ta;m,a = ;
a:a;Za;ta;m,a I+.tyea;vMa va;yMa ;
a;va;g{a;h;va;a;k+.aM va;d;a;maH ma;Dya;ma;pa;d;l+.ea;pa;a .sa;ma;a;saH :a
va;+:v.yaH .sa . ca ta;tpua:+:Sa;Bea;dH
This is perhaps the only way to justify the use of the expression pacasatam
rupyakan.i found on the Indian currency notes of rupees
500. However, the question as to how more complex numbers may
be represented with convenience still remains unanswered. Just as
one would say pacasatam, should it be also permitted to say pacasahasrapacasatapacapacasat or pacapacasadadhikapacasatadhikapacasahasram? Moreover, for ease of writing, if a
modern writer wishes to write numbers in figures then questions
like the following may arise:
How is he supposed to use it in various cases, e.g. 5555su
vars.es.u?
What would be a uniform method for representing dates like
5/4/2013?4
How to write ordinals, e.g. 555tame vars.e?
4 It should be noted that Shastri (1976: 165) rules out even the use of the word
dinan ka. The following are his comments: . ca;tua;dR ;Za;
a;d;na;a;*: e I+.a;ta k+:a;pa;ta;e
/ / a nUa;ta;na;ea v.ya;va;h;a:=H :pa:=+}.pa:=+a;ea;na v.ya;va;h;a:=e+a ;
a;va;sMa;va;d;ta;a;a;ta na;a;dx ;tyaH ..ca;tua;dR ;Zea ;
a;d;nea I+.tyea;va :pa;ya;Ra;p~ya;a;ta
;*: +.Za;b.de;na na;a;TRaH
371
The construction pacasahasravars.ebhyah. prak is a clear imitation of constructions prevalent in Hindi and other North-Indian
languages. In Hindi a sentence like (1e)
(1e) :pa;<a;.ca h:ja;a:= .sa;a;l :pa;h;le ;vEa;
a;d;k ;Da;mRa :ke ;a;sa;va;a dU;sa:=+a k+:ea;IR ;Da;mRa na;h ;<a
Ta;a
conveys the appropriate sense of five thousand years ago. Although idiomatic in Hindi, such a usage is not idiomatic in Sanskrit. Hence it would be interpreted literally as before five thousand years and the meaning intended by the author would not get
conveyed. This usage is very common in modern Sanskrit. Numerous examples of this usage in modern Sanskrit literature can
be cited. The following, in the preface to the edition of the Suktiratnakara by a revered grammarian Pundit V. B. Bhagvat (1999:
01), is just one of them:
ba;hu;ByaH .sMa;va;tsa:=e+ByaH :pUa;va .sUa;a;+.=+a;a;k+.=H na;a;ma ma;h;a;Ba;a;Sya;f ;a;k+:a h;~ta;a;l+.a;Ka;ta:
+.pa;a ma;ya;a .sa;ma;a;Da;ga;ta;a
Another problem with this usage is that two words, itah. (which
is implied) and pacasahasravars.ebhyah. present themselves to
fulfill the expectancy of prak leading to ambiguity. It would be
difficult to know whether the author means before now or before
five thousand years ago. Hence, a more correct way to construct a
sentence that would deliver the intended meaning accurately would
be (1f).
(1f) I+.taH :pa:*.a ;a;BaH va;SRa;sa;h;~:Ea H :pUa;va I+.h Ba;a:=+tea ;
a;k+:ma;
a;pa ma;tMa na;a;sa;a;d;nya:a
;vEa;
a;d;k+:a;t,a
aH
A. 2.1.31 :pUa;vRa;sa;dx;Za;sa;ma;ea;na;a;TRa;k+:l+.h;a;na;pua;a;a;ma;(ra:(;;E
supports the use of the third case in relation to the word purva.
Also, with such a construction there would be just one word itah.
372
in the ablative case to fulfil the expectancy of purva. Thus, this

kind of a construction would be the most efficient in conveying
the intended meaning. However, Shastri (1976: 40) rules out the
use of such a sentence, though perfect, on the grounds that such a
construction has never been used by the sis.t.as. Shastri remarks:
.sa;vRa;Ta;a ;a;na:=+va;d;a;ea Y;pya;yMa :pra;k+:a:=+ea na ta;a;va;tpra;ma;a;a;k+:ea;a;fM ;a;na;
a;va;Za;tea ya;a;va;a ; a;Za;E H .sa;ma;TRa;na;Ma l+.Ba;tea
Such a construction is also not seen in modern Sanskrit writings.
Hence it does not deserve much attention. Shastri suggests two
other ways of communicating the intended meaning accurately.
One is to make a sentence such as
(1g) ;d;a :pa:*.a va;SRa;sa;h;~:a ;a;a;a;h Ba;a:=+tea ;vEa;
a;d;ke+:ta:=e+Sa;Ma ma;ta;a;na;a;ma;
a;va;d;a;ma;a;na;a;na;a;m,a
Another is the way it has been demonstrated above in (10 ). Both
usages are acceptable to the sis.t.as; however Shastri (1976: 39)
explains the exact difference in the expression and the sensibility
to select the appropriate one of them in different situations in the
following words:
:pra;Ta;maH :pra;k+:a:=+~tua *: +:a;.ca;de ;va .sa;*: +.taH .~ya;a;t,a ta:a ;
a;h k+:a;l+.
a;va;Zea;Sa;~ya;a;a;ta;k+:a;nta;~ya ;
a;va;Zea;Sa;a;a;BUa;ta;a ;
a;k+:ya;a kx+:tpra;tya;ya;a;ntea;na
:Sa;.a;ntea;na;ea;.cya;ta I+.a;ta k+:a;l+.a;pea:a;ya;a :pra;v.ya;+:a ta;~ya;a ga;Ea;a;ta;a ta;sma;a;d;a:Ea;vMa;
a;va;DaH ;
a;k+:ya;a;k+:a;l+.ea Y;a;Ba;prea;ya;tea ta:Ea;vEa;Sa :pra;k+:a:= ;Ea;pa;a;ya;k+:ea nea;ta:=;a ya:a tua ;
a;k+:ya;a :pra;a;Da;a;nyea;na ;
a;va;va:ya;tea ;a;ta;z+.a
. ca;ea;.cya;tea ta:a k+:a;l+.a;na;deR ;ZaH .sa;a;}yEa;va yua; I+.a;ta ;
a;d;ta;a;ya O;;va
:pra;k+:a:=+~ta:a yua;+.+.pa I+.a;ta :pa;Zya;a;maH
The first type (1g) should be employed when the time that has
passed away is intended to be emphasized. The action is expressed
by words ending in a krt suffix. The second type (10 ) is used when
373
the action is to be given more focus by employing verbal terminations (tin)

and the time is indicated in the locative case in a subordinate manner.
Figure 1 summarizes the discussion above.
Figure 1
Sentence 1
Iha bharate pacasahasravars.ebhyah. prak kimapi matam
nasd
anyatra vaidikat.
5.1.3
Directive rules
1. A compound with numerals used for denoting numbers

larger than one hundred should be such that the thing
counted comes as the first component followed by numbers
like sata, sahasra, etc. Additional numbers should be indi
cated by separate words that indicate samkhyeyas
and qualify these compounds.5
5 Editor: Adjectival compounds of the type pacasahasra in the sense of 5,000
are not only common but are justified by Pan.inian grammar. Whitney (1889:
374
2. To indicate a particular period in the past, words like adya

and itah. should be used in the sentence. The words denoting
years should be put in the locative case.6
5.2
Sentence 2
(2) gua:+:a;a ku+:Za;l+.pra:(ea kx+:tea .sa;a;ta ta;a;ma;tTa;mua:a:=+ya;a;ta ;

a;d;l +.a;paH (95)
A question regarding his welfare having been posed by his
teacher, Dilpa answers it in this way.
17981) 477c, d . . . 481 cites several and specifically provides examples of
s.at.satani in the sense of 600 not 106. Bohtlingk and Roth provide examples
with citations, e.g. of dvisata in the sense of 200 as well as 102. The Kasika provides examples of adjectival compounds whose meaning is the product of
two numbers. Under A. 2.2.25, for example, the example dvidasa h. is provided of
a bahuvrhi compound in the sense of a counted object (sankhyeya). On this Jinendrabuddhi specifically describes the sense as multiplicative rather than additive
writing: ;
a;d;d;Za;aH I+.a;ta ;
a;d;dR ;Zea;a;ta ;
a;va;g{a;hH .sua:ja;TeRa Y;yMa .sa;ma;a;saH d;Za;sa;}ba;a;nDa;n
/ / a;a ya;a;vxa;a;ta:=, d;Za;Za;b.de;na l+ya;ma;a;a;a .sa;a ;
a;d;Za;b.de;na;a;K.ya;a;ya;tea Under A. 5.4.73 Jinendrabuddhi gives the examples dvisatah. and trisatah. and comments: ta;Ea;va;
a;pa .sa;* +;ae ;yea va;teRa;ta O;;va, .sa;* +;ae ;ya;ta;yEa;va va;a;tRa;~ya .sua:ja;TRa;~ya . ca;a;a;Ba;Da;a;na;a;t,a Such compounds are therefore
adjectives meaning 2 x 10 = 20, 3 x 10 = 30. Therefore pacasahasra is a valid
bahuvrhi compound in the sense of 5,000 years and the phrase pacasahasres.u vars.es.u is just as good as Shastris recommended pacasu vars.asahasres.u.
Nevertheless, the compounding of such numeric adjectives with nouns denoting
the counted objects is restricted. Although such compounds would be permitted
by the general rule A. 2.1.57 ;
a;va;Zea;Sa;Ma ;
a;va;Zea;Syea;na ba;hu;l+.m,a, A. 2.1.50 (p. 383) serves
to restrict such compounds to sajas such as saptars.ayah.. The compound pacasahasravars.a would be in violation of that restriction. The same issue arrises
in 5.5.
6 Editor: Usages such as navame hani and saptadase hani mean on the ninth
day or seventeenth day from now. They use an ordinal rather than a cardinal
number. To indicate similarly a time in the past 5,000 years ago, one would
say itah. purvam pacame vars.asahasre or pacasahasratame vars.e in the fifth
millenium or in the five-thousandth year before now rather than itah. pacasu vars.asahasres.u in five millenia. However, the usage Sastri proposed but
deemed not used to indicate a time prior to now by 5,000 years (1f) should be
sought further before being dismissed.

5.2.1
375
Shastris comments
gua:+:a;a kx+:tMa ku+:Za;l+.pra:(a;mua:a:=+ya;ta;a;a;ta va;+:v.ya;m,a Ba;a;va;l+a;a;a;ya;aH
.sa;a;}ya;a .nEa;Sa ;
a;va;Sa;ya I+.a;ta :pUa;va;Ra:;d eR .sa;
a;va;~ta:=M ;a;na;ga;
a;d;ta;a;ma;a;ta ta;ta
O;;va;a;va;Da;a;yRa;m,a
One should rather say gurun.a krtam
kusalaprasnam
uttarayati. It has already been clearly

explained in the
first section that the use of the seventh case indicative of action (bhavalaks.an.a saptam) is inappropriate
here. The first section may be referred to for details.
Shastri engages in an elaborate discussion in the first section of

the Vagvyavaharadarsa regarding the locative absolute construction, the interpretation of
A. 2.3.37 ya;~ya . ca Ba;a;vea;na Ba;a;va;l+a;a;m,a (.sa;a;ma;a 36)
The seventh vibhakti occurs after a nominal base the action
of which characterizes another action.
and the opinions of various grammarians regarding it. He (1976:
35) concludes as follows:
ya;~ya k+:tRuaH k+:mRa;a;ea va;a ;
a;k+:ya;a ;
a;k+:ya;a;nta:=+~ya l+a;Ma Ba;va;a;ta
.sa . cea;t,a k+:ta;Ra k+:mRa va;a ;
a;k+:ya;a;nta:=+a;(ra;ya;a;t,a k+:tRuaH k+:mRa;a;ea va;a
;a;Ba;d;ae;ta ta:Ea;va .sa;a;ma;a .~ya;a;t,a, ;Bea;de tua ma;a BUa;
a;d;a;ta.
The kartr or karman bringing about the indicating (laks.aka) action
different from the kartr or karman participating in the
should be
is such a use of the seventh
action indicated (laks.ya). Only then
case appropriate. In (2), the indicating action of posing (karan.a)
and the indicated action of answering (uttaran.a) both belong to the
same object, namely the question (prasna). Hence Shastri deems
this sentence erroneous and prefers (20 ) instead.
(20 ) gua:+:a;a kx+:tMa ku+:Za;l+.pra:(a;mua:a:=+ya;a;ta ;
a;d;l +.a;paH Dilpa answers the
question regarding his welfare posed by his teacher.
376
5.2.2

Discussion
Whether the subjects of the two actions (i.e., the one that indicates
and the one that is indicated) must be different or may be the same
is not clear either from the statement of A. 2.3.37 or from the various discussions available on the rule in the tradition. Shastri quotes
the following words from the Bhas.a vrtti of Purus.ottamadeva:
ya;~ya ;
a;k+:ya;ya;aY;nya;~ya ;
a;k+:ya;a;nta:=M l+ya;tea ta;taH .sa;a;ma;a
The seventh (vibhakti) occurs after that by the action
of which another action is characterized.
Consider the following sentence:
(2a) .=+a;mea va;nMa :pra;a;ta;+ma;a;nea .sa ;
a;pa;tuaH :pra;a;ta;a;a;mea;va ma;na;~ya;k+.=+ea;a;a;nya;a;tk+:
/ a:*.a ;t,a (Shastri 1976: 35)
Rama setting out for the forest, he kept in his mind just his
fathers oath and nothing else.
According to Shastris conclusion, (2a) would not be correct. Here
the kartr of both the indicating action of setting out (prasthana)
and the indicated action of keeping (karan.a) is Rama.
The case of the following sentence is a little different:
(2b) h;tea d;Za;a;~yea ;
a;va;Ba;a;Sa;a;ea l+.*: +.a:=+a:$yea Y;a;Ba;
a;Sa;
a;Sa;.cea .=+a;mea;a (Shastri
1976: 35)
The ten-headed one (Ravan.a) having been slain, Vibhs.an.a
t.as consecrated in the rulership or <r>Lanka by Rama.
Here also the kartr of both the indicating action of slaying (ha action of consecrating (abhisecana) is Rama.
nana) and indicated
.
However, the use of the bhavalaks.an.a saptam would be appropriate here. In the previous sentence, both words indicating action
were in agreement with the kartr. Here although the kartr is the
are in agreement with the kasame, the words indicating action
rman and not with the kartr. Hence, the identity of kartr in this
of the bhavalaksana saptam.
case does not restrict the use
. .
377
The case of (2) is different from that of both (2a) and (2b). In
(2), there are two actions involved, namely posing (karan.a) and answering (uttaran.a). The word indicating the first action (krta) is in
agreement with the word indicating the karman (prasna).Though
the question is also the karman of the second action (uttaran.a),
the word indicating that action is not in agreement with prasna.
Rather, it is in agreement with the word indicating the kartr (dil
pa). Shastri has provided no explanation for deeming the sentence
as erroneous in this situation. Based on the discussions above we
therefore form the directive rule as in 5.2.3. Moreover, if tam be
considered as referring to the teacher (guru) of the previous clause,
then the sentence would also have to be considered as correct.
The tree in Figure 2 summarizes the discussion above.
Figure 2
Sentence 2
Gurun.a kusalaprasne krte sati tam ittham uttarayati dilpah..
378
5.2.3

Directive rule
In the case of a bhavalaks.an.a saptam, the kartr or karman of the

kartr or karman
indicated action should not be same as that of the
with which the indicating action is in agreement.7
5.3
Sentence 3
(3) ya;
a;d na;a;ma ta:a;"+nTa;a;Dya;a;pa;nMa .sMa;~kx+:ta O;;va Ba;vea:a;d;a . cC+.a:a;a ba;hU;pa;kx+:ta;aH
.~yuaH (93)
If the relevant books were taught in Sanskrit only, the students would be benefited a lot.
5.3.1
Shastris comments
.sMa;~kx+:tea;na Ba;vea;
a;d;tyea;va .sa;a;Dua va;a;a;gGa
/ k+.=+Ma d;a:=+ma;Dya;a;pa;na;~yea;a;ta
txa;ta;a;yEa;va yua;+:a ya;d;a;.ca;aY;na;Byua;
a;d;ta;a;ma;tya;a;
a;d;Sua ta;Ta;a d;ZRa;na;a;t,a
Saying samskr
tena bhavet (instead of samskr
te)
since language is an instrumentor a
would be correct,
means of teaching. The same is observed to be so in
the usages such as
7 Editor: In (2), the participles krta and sat refer to and characterize the ques of answering (uttarana) which conditions
tion which is the karman of the action
.
the dvitya vibhakti by A. 2.3.2. In (2b), A. 3.4.21 would require the use of an
absolutive after the root denoting the action that takes place in prior time. Rama
is the agent of both the indicating action of slaying (hanana) and the indicated
action of consecrating (abhis.ecana). Because the slaying occurs previously, A.
3.4.21 provides the affix ktva after the root han. Ravan.a, as the karman of the
action of slaying, would condition the dvitya. The result would be in a sentence
such as (??) .=+a;mea;a d;Za;a;~yMa h;tva;a ;
a;va;Ba;a;Sa;a;ea l+.*: +.a:=+a:$yea Y;a;Ba;
a;Sa;
a;Sa;.cea The sentence
might pass if Ramas agency in the prior act of slaying is not desired to be expressed (avivaks.ita). Similarly by not (2) might pass by not intending the identity
of the prasna as the karman of the action denoted by kr and as the karman of u
ttaran.a.
379
(3a) ya;d;a;.ca;aY;na;Byua;
a;d;ta;m,a Kena Upanis.ad 1.4a
which has not been expressed by speech.
(30 ) ya;
a;d na;a;ma ta:a;"+nTa;a;Dya;a;pa;nMa .sMa;~kx+:tea;nEa;va Ba;vea:a;d;a . cC+.a:a;a ba;hU;pa;kx+:ta;aH
.~yuaH
5.3.2
Discussion
Sentence (3) seems to be a faithful imitation of the case usage in

Hindi, Marathi, etc. in such a context. The name of the language
of instruction or communication is used with the preposition meMa in
Hindi. The above sentence is a direct translation of the following
Hindi sentence: . . . .sMa;~kx+:ta meMa h ;a h;ea . . . . There is no such convention found in Sanskrit. Hence going by the literal meaning, the
language being a medium or instrument of instruction should be
represented in the instrumental case as in (30 ). Shastri substantiates his opinion by providing (3a). Many modern Sanskrit writers
are tempted to use the locative instead. This very error can be
shown with the help of Figure 3.
Figure 3
Sentence 3
Yadi nama tattadgranthadhyapanam
samskr
ta eva bhavet tada
chatra bahupakrtah. syuh..
380
5.3.3

Directive rule
The medium of communication (usually the name of a language)

should be linked to the verb related to communication by the third
case.
5.4
Sentence 4
(4) na;va;ea;Q+eaY;yMa yua;va;aY;a;ta;ma;a:a;ma;nua:=;$ya;tea va;Dva;a;m,a (17)

This youth who has just married loves his wife excessively.
5.4.1
Shastris comments
.~:a;a na;va;ea;Q+a Ba;va;a;ta :pua:+:Sa;~tua va;ea;Q+a va;h;na;
a;k+:ya;a;ya;Ma :pua:+:Sa;~ya
k+:tRxa;tvMa ;a;~:
/ a;ya;a;(a k+:mRa;tvMa :pra;a;sa:;d ;m,a :pa;a:=+vea:a;a;nua:ja;ea Y;nUa;Qe .$yea;e
d;a:=+pa;a:=+g{a;h;a;
a;d;tya:a;a;nUa;Qe I+.tya:a o+a:=+pa;d;l+.ea;pa;ea d+;v.yaH ;nUa;Q+ea Y;nUa;Q+d;a:= I+.tya;TRaH O;;vMa :pa;a:=+a;ya;
a;k+:ya;a;ya;a;ma;
a;pa d+;v.ya;m,a
.tea;na .sa;d;aH kx+:ta;d;a:=H ;a;.ca:=+kx+:ta;
a;va;va;a;hH I+.a;ta va;a va;+:v.ya;m,a
ta:a;a;a;.ca:=M kx+:ta;ea ;
a;va;va;a;h;ea yea;nea;a;ta ;
a;va;g{aH
A woman is called navod.ha newly brought (home).
A man is called vod.ha one who brings her. It is
well-known that a man is the subject and a woman
is an object with reference to the action of bringing
(home), i.e. getting married. In the sentence,
(4a) :pa;a:=+vea:a;a;nua:ja;ea Y;nUa;Qe .$yea;e d;a:=+pa;a:=+g{a;h;a;t,a When an
elder son is unwed, a younger son will marry
after (the elders) taking of a wife.
there is an elision of the subsequent component of
the compound. anud.ha the one who is not brought
(home) means anud.hadara the one who has not
brought his wife (home). The same is the case with
the verb pari + n.. Hence one should use either sadyah. krtadarah. or acirakrtavivahah. (to mean just
381
married). The explanation of acirakrtavivaha is the
one who has married recently.

The adjective navod.ha is usable only for a woman and not for
a man. The reason is that it is the man who performs the function
of bringing along (vahana), and it is the woman who is brought
along by the man.
(40 ) .sa;d;aH kx+:ta;d;a:=H . . .
;a;.ca:=+kx+:ta;
a;va;va;a;h;ea Y;yMa yua;va;a;a;ta;ma;a:a;ma;nua:=;$ya;tea va;Dva;a;m,a
5.4.2
Discussion
Words referring to the ubiquitous institution of marriage differ in

different languages. The word in English is marriage while that
in Sanskrit is vivaha. The primary meaning of the word marriage
is an act of bringing about union. In Sanskrit, one has to look at it
as the act merely of a man. A woman is passive in this ritual. Thus
the man is the kartr of the action referred to by vi + vah, while a
woman is the karman.

The same is the case with other Sanskrit
words denoting the ritual of marriage namely pari + n, upa + yam, pan.igrahan.a etc. The differences in the constructions used to
describe marriage are a classic example of how cultural contexts
affect not only the vocabulary of a language but also syntactic relations, in this case karaka relations.
Figure 4 summarizes the discussion above.
5.4.3
Directive rule
When roots concerning marriage such as vi + vah, pari + n,

etc. are used, the man should be placed in the role of the kartr
and a woman should be placed in the role of the karman.
382

Figure 4
Sentence 4
Navod.ho yam
yuvatimatram anurajyate vadhvam.
5.5
Sentence 5
(5) .sa;h;~:a;
a;d;v.ya;yua;ga;a;na;a;mea;kM b.ra;a;M ;
a;d;nMa Ba;va;a;ta (20)
One day of brahman is equal to a thousand divine yugas.
5.5.1
Shastris comments
.sa;h;~:Ma ;
a;d;v.ya;yua;ga;a;na;a;a;ta va;+:v.ya;m,a ;;a;d;Za ;a;na;mea;Sa;a;~tua k+:a;+a
;
a:Ma;Za:ua ta;aH k+:l+.a I+.tya;a;
a;d;Sua .sMa;a;a;sMa;a;a;na;eaH .sa;a;ma;a;na;a;a;Da;k+.=+yea;na :pra;ya;ea;ga;d;ZRa;na;a;t,a :Sa; +a tua du:+:pa;pa;a;d;a *: +:a;.ca;t,a .sMa;a;a;sMa;a;a;Ba;a;va;ma;
a;va;va; a:a;tva;a k+.=+ea txa;ta;a;ya;a :pra;yua:*+;te
a va;a;a;.ca k+:ea;
a;va;d;aH d;Ea
d;Ea ma;a;ga;Ra;
a;d;ma;a;sa;Ea .~ya;a;dx;tua;~tEa:=+ya;nMa ;
a:a;a;Ba;a:=+tya;ma:=e ya;Ta;a :a
Ba;va;ta;a;a;ta Zea;SaH
One should say sahasram
divyayugani instead, since
the usage of the same case for the name and the thing
named can be seen in the sentence
383
(5a) ;;a;d;Za ;a;na;mea;Sa;a;~tua k+:a;+a ;

a:Ma;Za:ua ta;aH k+:l+.aH (Amarakosa 1.4.11)
Eighteen nimes.as make one kas..tha and thirty
kas.t.has make one kala.
It is difficult to substantiate the use of the sixth case
here. At times, unwilling to express the relationship
of name and the thing named, the experts of speech
use the third case in the sense of instrument as can be
observed in
(5b) d;Ea d;Ea ma;a;ga;Ra;
a;d;ma;a;sa;Ea .~ya;a;dx;tua;~tEa:=+ya;nMa ;
a:a;a;BaH (Amarakosa 1.4.13)
Each of the two months beginning with Marga
is a season (rtu). A course (of the sun) (ayana)
by three of these.
(is constituted)
In this construction the word bhavati becomes is implied.
5.5.2
Discussion
Shastri sees two problems with (5):

1. Compounding is not acceptable between sahasra and divyayuga due to the constraint that such compounds are limited
to conventional nouns by
A. 2.1.50 ;
a;d;#sa;* +;ae .sa;Va;a;ya;a;m,a 8
2. The sixth case in sahasradivyayuganam is inappropriate.
He therefore suggests the following two possible revisions:
(50 ) .sa;h;~:Ma ;
a;d;v.ya;yua;ga;a;a;na . . .
.sa;h;~:ea;a ;
a;d;v.ya;yua;gEaH b.ra;a;M ;
a;d;nMa Ba;va;a;ta
8 Compare
5.1 and note 5.
384
What would have led the author of (5) to use the sixth case?
The author of must have assumed a ses.a relation between the yugas
and a day of brahman and deployed the sixth case in accordance
with A. 2.3.50.9 One may say that yugas are the components (avayava) and a day of brahman is the whole possessed of them (avayav). But then the idea that a thousand divine yugas completely
exhaust a day of brahman is not tapped. One may alternatively
think of yugas as the original material (prakrti) and a day of brahman as its modification (vikrti). But there isno processing of the
yugas in order to make a brahma dina similar to the way one has
to process gold to prepare ornaments. A thousand divine yugas are
the components that make one brahma dina. While neither of these
two relations seems appropriate, the author must have conceived of
some such relation to justify the use of the sixth case.
Is there any rule ordaining the use of a particular case or prohibiting the use of the sixth case in such a usage? There is no
particular rule in Pan.ini ordaining any case or prohibiting the use
of the sixth case in this situation. So the sixth case may be comfortably used without violating any of Pan.inis rules. Mere unavailability of any particular rule in Pan.ini for a particular situation cannot be the basis for deployment of the sixth case according
to Shastri. One has to look for the usages by the sis.t.as in similar situations. Accordingly Shastri advocates two other ways of
expressing this idea drawing support from the usage found in the
Amarakosa.
Shastri conceptualizes two ways of looking at the relationship
between a thousand divine ages (sahasradivyayugas) and a day of
brahman (brahma dina). One way is to look at them as two notions. Further, the notion of brahma dina is based upon the notion
of divya yugas. He thus rightly sees a nameable-name relationship
between the two. The nameable and name have been used in the
9 See
Cardona Extension rules 2.2.2 (p. 61).
385
same case (though their gender and/or number may be different at

times) at numerable places in the classical literature. The following are a couple of examples:
(5c) Za;Ea;.ca;sa;nta;ea;Sa;ta;pa;s~va;a;Dya;a;yea:(;a:=+pra; a;a;Da;a;na;a;a;na
Yogasutra 2.32)
;a;na;ya;ma;aH (Patajali
(5d) +:+:va;tUa ;a;na;+a (A. 1.1.26)

In (5c) the number of both the nameable and the name is the same.
However, while the nameable is in neuter gender, the name is in
masculine gender. In (5d) the nameables are used in the dual number while the name is used in the singular.
One can neglect the nameable-name aspect existing between a
thousand divine ages and a day of brahman and merely say that the
former lead to the formation of the latter. However, as discussed
above, both yuga and a day of brahman are notions. There is no
actual process of formation of a day of brahman with yugas. Hence
the use of the third case as per A. 2.3.18 is not completely justified.
However, such a use is still valid as per Shasrti it is admitted by
the sis.t.as.10
Figure 5 summarises the above discussion.
5.5.3
Directive rule
Several homogenous individual components constituting a whole

are not to be represented in the sixth case assuming that they are
the material causes of the whole. Instead nameable-named relation
between the two should be expressed by representing both of them
10 Editor: The use of the instrumental tribhih in Amarakosa 1.4.13 is accounted
.
for by A. 2.3.23 hetau. If, on the other hand, one supplies the verb as to be, then
the three months are termed karana by A. 1.4.42 and the trtya is provided by A.
2.3.18. In either case, the use is provided for by the a stadhyayi.
386

Figure 5
Sentence 5
Sahasradivyayuganam ekam
brahmam
dinam
bhavati.
in the first case. Alternatively such individual components should

be deemed as instruments triggering the action and represented in
the instrumental case.
5.6
Sentence 6
(6) ma;h;a;tma;a;Ba:=+
a;pa .sa;ma;ma;na;TRa;kM ;
a;d;Sa;a;nta
/ / ma;l +.a;ma;sa;a;Da;yaH (36)
The wicked-hearted nurture hatred even for great men.
5.6.1
Shastris comments
;
a;d;Sa ;pra;ta;a;ta;a;
a;va;a;ta .sa;k+:mRa;kH ta;Ta;a . ca ku+:ma;a:=e :pra;ya;ea;gaH ;
a;d;Sa;a;nta
/ / ma;nd;a;(a;a:=+tMa ma;h;a;tma;na;a;a;ma;a;ta .tea;na ma;h;a;tma;na;ea ;
a;d;Sa;nta;a;a;ta
va;+:v.ya;m,a
The root dvis.a aprattau is transitive. The same is
supported by the following usage in the Kumarasambhava:
387
(6a) ;
a;d;Sa;a;nta
/ / ma;nd;a;(a;a:=+tMa ma;h;a;tma;na;a;m,a
The idle hate the deeds of the great.
Hence one should use mahatmano dvis.anti.
5.6.2
Discussion
It is apt to make a construction like

(6b) .=+a;maH .sua;g{a;a;vea;a (.sa;h) ;a;ma;l+.a;ta
but saying sugrvam
milati* would not be appropriate because the
root mil is intransitive according to Pan.inis dhatupa.tha. Shastri
suspects the same error by the author of (5). The fact is however
that the root dvis. in (5) is transitive; great men (mahatman) are the
karman of dvis.. The correct construction is shown in (60 ). Shastri
in his corpus of 655 sentences has incorporated numerous such examples where transitive roots have been dealt with as if they were
intransitive and vice versa. Figure 6 summarizes the discussion.
(60 ) ma;h;a;tma;na;ea Y;pya;na;TRa;kM ;
a;d;Sa;a;nta
/ / ma;l +.a;ma;sa;a;Da;yaH
5.6.3
Directive rule
The root dvis. hate is transitive in Sanskrit. Hence, the object of

hatred should be expressed in the second case.
5.7
Sentence 7
(7) :ke+:a;.ca;n}.a;nd;a;~tva;a:=+ta;ta:=+ma;Dya;ya;na;a;tpa;a:=+gl+.a;na;a gua:+:ku+:l+.a;t,a .sa;ma;a;va;tRa;ntea (37)

Some lazy ones return from the gurukula as they get bored
with studying very quickly.
388

Figure 6
Sentence 6
Mahatmabhir api samam anarthakam

dvis.anti malmasadhiyah..
5.7.1
Shastris comments
:pa;ya;Ra;d;ya;ea gl+.a;na;a;d;a;TeRa . ca;tua;TyeRa;a;ta va;.ca;na;a;
a;+:*: +.a;d, gl+.a;ya;a;ta;na;a ya;ea;gea . ca;tua;Ta;Ra v.ya;va;h;a:=+a;nua;pa;a;a;ta;na;a Ba;va;a;ta .tea;na;a;Dya;ya;na;a;ya :pa;a:=+gl+.a;na;a I+.a;ta va;+:v.ya;m,a
From the indication of the varttika :pa;ya;Ra;d;ya;ea gl+.a;na;a;d;a;TeRa
. ca;tua;Tya;Ra the fourth case in connection with the root gla
is in accordance with correct usage. Hence one should
use adhyayanaya parigalanah..
5.7.2
Discussion
Though Pan.ini has specified the use of a specific case in relation to

the root glai either under the adhikara karake or under the adhikara anabhinite, Katyayana has indicated the use of the fourth case
for the object towards which glani disinterest is expressed in the
following varttika related to the formation of pradi compounds:
:pa;ya;Ra;d;ya;ea gl+.a;na;a;d;a;TeRa . ca;tua;Tya;Ra (A. 2.2.18 vt. 4, MBh. I.417.1),
389
According to Shastri, the same should be followed while making

sentences using the root glai. Hence the correct formulation of (7)
is shown in (70 ). Figure 7 summarizes the discussion.
(70 ) :ke+:a;.ca;n}.a;nd;a;~tva;a:=+ta;ta:=+ma;Dya;ya;na;a;ya :pa;a:=+gl+.a;na;a gua:+:ku+:l+.a;tsa;ma;a;va;tRa;ntea
Figure 7
Sentence 7
Kecin mandas tvaritataram adhyayanat pariglana gurukulat samavartante.
5.7.3
Directive rule
The tiresome activity is to be represented in the fourth case when

the root glai is used.
390
5.8
Sentence 8
(8) ya;Ta;a ;
a;d;=e +P+.ea Y;nta:=+a:=+a;mea;Sua Bra;a;}ya;npua;Spa;a;a;Ma ma;Dva;va;a;.ca;na;ea;a;ta ta;Ta;a .sa;tyMa
ma;a;a;gRa;ta;a .ja;naH .sa;}.pra;d;a;ya;a;na;Ma .sa;tya;ma;a;d:ea (66)
Just as a honeybee wandering about in the grove sucks the
nectar from the flowers, a person investigating truth imbibes
truth from various sects.
5.8.1
Shastris comments
;a;~ta
/ / k+:a:=+k+:tva;
a;va;va:ea;a;ta :pua;Spa;a; a;a :pua;Spea;Bya I+.a;ta va;a va;+:v.yMa .sa;}.pra;d;a;ya;a;n,a .sa;}.pra;d;a;yea;Bya I+.a;ta . ca .sa;}ba;nDa;a;a;.ca;K.ya;a;sa;a
tua na;a;ta;a;va yua;e+:a;ta :Sa; +a na;ea;pa;pa;d;a;tea gua:+:pUa;va;Ra;nua;k+:mea;a ; a;Za;Sya;pra; a;Za;Syea;Bya;ea d ;a;ya;ma;a;na o+.pa;de ;ZaH .sa;}.pra;d;a;ya;ea Ba;va;a;ta :pra;kx+:tea tua
ma;ta;
a;va;Zea;Sa;a;a;Ba;a;na;
a;va;;a;na;Ma .sa;mua;d;a;ya;ea ;
a;va;va; a:a;taH, .tea;na .sa;}.pra;d;a;ya;Za;b.d;eaY;~Ta;a;nea
Here there is a definite karaka relation to be expressed.
Hence, one may use pus.pan.i or pus.pebhya and sampradayan or sampradayebhyah.. Mere relation is
not desired to be expressed. Hence, the use of sixth
case is inappropriate. The word sampradaya means
the preaching that a disciple gets through the tradition
of gurus. In the present context, the meaning intended
to be expressed is a group of people adhering to a
particular belief system. Hence, the use of the word
sampradaya is inappropriate.
5.8.2
Discussion
Charudev Shastri has two issues with this sentence. Firstly, he

doesnt agree with the use of the sixth case in pus.pa and sampradaya. Indeed, nectar (madhu) is a part of a flower (pus.pa), so the partwhole relationship can be expressed using the sixth case. However,
here, with the use of the root ci, the emphasis is not on the part-
391
whole relationship. Rather it is on the process of separation of the

part, i.e. the nectar from the fixed (dhruva) whole, i.e. the flower.
Hence, the use of fifth case is appropriate. Since, the root ci has
been included among roots that govern two direct objects (dvikarmaka roots) listed in the verse cited under A. 1.4.51 ;k+:a;Ta;tMa . ca
, (first at MBh. I.329.19) the second case may also be optionally
used. These alternatives are shown in (80 ). Figure 8 summarizes
this discussion.
(80 ) ya;Ta;a ;
a;d;=e +P+.ea Y;nta:=+a:=+a;mea;Sua Bra;a;}ya;npua;Spa;a; a;a :pua;Spea;Bya;ea ma;Dva;va;a;.ca;na;ea;a;ta
ta;Ta;a .sa;tyMa ma;a;a;gRa;ta;a .sa;}.pra;a;d;a;ya;a;nsa;}.pra;d;a;yea;ByaH .sa;tya;ma;a;d:ea
Figure 8
Sentence 8
Yatha dvirepho ntararames.u bhramyan pus.pan.a m
madhv avacinoti tatha satyam
margita janah. sampradayanam
satyam a datte.
Secondly, Shastri points out that the word sampradaya is not

used in Sanskrit in the sense in which it is being put to use in modern Indian languages. It doesnt mean group of people adhering
392
to a particular belief system. Numerous words have been borrowed from Sanskrit by modern Indian languages where they have
changed their meanings. They now seem to be re-entering modern Sanskrit with their new connotations. A list of just a few such
words found in Charudev Shastris corpus appears in Table 1.
5.8.3
Directive rule
A part-whole relationship is not to be expressed if such a part is

being separated from the whole. Instead, separation is to be represented by using the fifth case in the word denoting the whole. In
the case of the roots duh, yac, math, mus., ci, and ji the second case
is also permitted there.
5.9
Sentence 9
(9) ;nDMa ta;maH :pra;

a;va;H :pua;ma;a;npa:=e+ZMa :pra;k+:a;Za;a;ya ya;a;.cea;ta (72)
A person who has entered heavy darkness should pray to
God for light.
5.9.1
Shastris comments
:pra;k+:a;ZMa ya;a;.cea;tea;tyea;vMa nya;a;sa;ea ;a;na;d;eRa;SaH .~ya;a;t,a
prakasam
yaceta only is correct.
5.9.2
Discussion
The root yac has also been incorporated in the list of dvikarmaka roots under A. 1.4.51. Hence it can have two karmans. One
candidate for the post of karman here is paresa while the other is
prakasa. So one may either say, paresam
prakasam
yaceta or paresa t prakasam
yaceta. The author of (8) has used the subordinate
karman correctly in the second case, while he has surprisingly used
393
Table 1
Word sense in Sanskrit versus in modern Indian languages
Word
;a;d;ZRa
;a;ma;l,
Ba;a;vua;k
.ja;a;
a;va;ta
.ja;a;va;na
;a;yua;s,a
:pra;kx+:a;ta
:pra;a;ta;Za;ea;Da
:pa:=+ma;a;TRa
:pra;a;nta
;a;.ca;a;nta;ta
/ /
.sMa;k
+:a;Ra;ta;a
.sa:a;a
o+.pa;yua;
k+:a;
a;F+.nya
ma;a;l+.a
.sMa;ya;ea;gea;na
ma;a;Dya;ma
:pa;a;F+.k
Modern meaning
ideal
to get
connoisseur
living/live
life = span of life
age
nature/creation
Sanskrit meaning
mirror
to meet
wellbeing
life
life = sentience
total span of life
unmanifest cause of the
world (mahat)
revenge
use not found in this sense
in classical Sanskrit
offering help to others
an account of what has
happened in the past (bhutartha)
region
border
worried
it was thought (by someone)
narrowness
state of being mixed
power
existence, goodness
useful
used
difficulty (in understanding) toughness, cruelty
necklace
garland
by chance
union, joining
medium
relating to the middle
reader
teacher
394
the candidate for the position of the main karman in the fourth
case.
Perhaps there is some different idea behind this usage. One
may compare the part prakasa ya yaceta in the above sentence with
vanaya mumoca in
(9a) va;na;a;ya :pa;a;ta;pra;a;ta;ba:;d ;va;tsa;Ma ya;Za;ea;Da;na;ea ;Dea;nua;mxa;Sea;mRua;ma;ea;.ca (Raghuvam
sa
2.1).
Mallinatha quotes
A. 2.3.14 ;
a;k+:ya;a;Ta;eRa;pa;pa;d;~ya . ca k+:mRa; a;a .~Ta;a;a;na;naH
The fourth case-affix is employed in denoting the object
(karma) of that verb, which is suppressed (sthanin) in a sentence, and which has in construction (upapada) therewith another verb, denoting an action, performed for the sake of the
future action. (Vasu 1909)
and comments on the word sthaninah. as follows:
;pra;yua:$ya;ma;a;na;~ya ;
a;k+:ya;a;Ta;eRa;pa;pa;d;~ya ;Da;a;ta;eaH ;na;a;Ba;
a;h;tea k+:mRa; a;a k+:a:=+ke . ca;tua;Ta;Ra ;
a;va;Ba;a;+:BRa;va;a;ta (Jijnasu 1964: 212).
Just as vanaya mumoca can be paraphrased as vanam
gantum
mumoca, it is also possible to paraphrase prakasa ya yaceta as prakasam
praptum
yaceta. The only difference between the two examples is that vana in (9a) denotes the karman of gantum whereas it
is not related to the action of the verb mumoca in any way. In (9)
however, the light denoted by prakasa qualifies to be the karman of both praptum
as well as yaceta. Shastris opinion about the
sentence implies that A. 2.3.14 does not apply in such a situation.
(90 ) ;nDMa ta;maH :pra;
a;va;H :pua;ma;a;npa:=e+ZMa (or :pa:=e+Za;a;t,a) prakASaM yAceta.</s>
395
Figure 9
Sentence 9
Andham
tamah. pravis..tah. puman paresam
prakasa ya yaceta.
5.9.3
Directive rule
When the object of the kriyarthakriya and the aprayujyamana dhatu is one and the same, then such an object should be written in the
second case.
5.10
Sentence 10
(10) :pra;a;tya;Ma . ca .sa;h;ya;ea;gea . ca .=+a;";~ya .sMa;(ra;ya;ea Ba;va;a;ta (82)

Betterment of a country lies in affection and cooperation.
5.10.1
Shastris comments
.sMa;(ra;ya;Ma .sMa;(ra;a;ya;ta I+.a;ta va;a .sMa;(ra;yaH o+.Ba;ya;Ta;a;
a;pa :pra;a;tya;a;a;ma;a;ta
.sa;h;ya;ea;ga I+.a;ta . ca .sa;a;}ya;Ea na;ea;pa;pa;d;ae;tea :pra;a;a;taH .sa;h;ya;ea;ga;(a .=+a;";~ya .sMa;(ra;ya I+.tyea;vMa v.ya;va;h:=+a;a;ya;m,a
Sam
sraya may be understood either as the process of
taking refuge (sam
srayan.a) or the one who is resorted
396

to (sam
sryate). In both situations the seventh case of
the words prti and sahayoga would be inappropriate.
One should paraphrase the sentence in the following
way instead: prtih. sahayogasca ras..trasya sam
srayo.
(100 ) :pra;a;a;taH .sa;h;ya;ea;ga;(a .=+a;";~ya .sMa;(ra;ya;ea Ba;va;a;ta

Figure 10
Sentence 10
Prtyam
ca sahayoge ca ras..trasya sam
srayo bhavati.
5.10.2
Directive rule
The thing resorted to should be represented in the first case as a

complement of the subject.
R EFERENCES
397
Conclusion
Various aspects regarding ten sentences from Shastris corpus have

been discussed in detail above. New insights related to the correct
use of karakas have been listed as derivative rules. The tree charts
for erroneous sentences have also been presented. The paper aims
at drawing the attention of scholars towards the need to investigate more aspects of the karaka system that are left undescribed or
undetailed by Pan.ini through the survey of Sanskrit literature and
to formalize them in order to bring more completeness to the description of the karaka system in Sanskrit. It was also an attempt
to trace new trends in modern Sanskrit writings with a view to invoke discussions about the need either to alter them or to accept
and account for them. The parse trees are also expected to be of
some assistance for improving the efficiency of machine translation systems.
References
Bhagvat, V. S., ed. 1999. Sukti-ratnakara (Commentary on Pata . a Narayan.a: fascicule I: a hnikas 1
jalis Mahabhas.ya) by Ses
a.
and 2. Pune: Anand
asrama Samsth
Hegde, Janardan. 2004. Suddhikaumud. Bangalore: Samskrita

Bharati.
. ed. 2007. Bhas.a pakah.. Bangalore: Samskrita Bharati.
vis.n.usarmapran.tam
Jha, R., ed. 1994. Sr
Pacatantram. Varanasi:
Chaukhamba Vidyabhavan.
Jijnasu, B., ed. 1964. As..ta dhyay-bhas.ya-prathamavrtti [pratha
ma-trtyadhyayatmaka prathama bhaga]. Tenth edition,
2008.
Sonipat: Ramlal Kapoor Trust.

Johnson, Keith and Helen Johnson. 1999. Encyclopedic dictionary
of applied linguistics: a handbook for language teaching. Oxford, UK; Cambridge, Mass.: Blackwell Publishers.
398
Mishra, A., ed. 2000. Vim

sasatabdsamskr
takavyamrtam. Delhi:
Delhi Sanskrit Academy.

Raghunathacarya, E. B., ed. 2002. Modern Sanskrit literature, tradition & innovations. Delhi: Sahitya academy.
Ranganath, S., ed. 2003. Studies in Rgveda and Modern Sanskrit
Literature. Delhi: Eastern Book
Linkers.
Shastri, Charudev., ed. 1955. Sabdapasabdaviveka. Delhi: Motilal

Banarasidass. [Vikram 2011.]
. ed. 1976. Vagvyavaharadarsa. Delhi: Motilal Banarasidass.
Shukla, H. L., ed. 2002. Modern Sanskrit Literature. Delhi: New
Bharatiya Book Corp.
Vasu, S. C., ed. 1909. The Siddhanta Kaumud of Bhhat..toji Dks.ita.
Vol. 1. Reprint: Delhi: Motilal Banarasidass, 2003. Allahabad:
The Pan.ini Office.
Whitney, William Dwight. 1889. Sanskrit grammar: including
both the Classical language, and the older dialects, of Veda
and Brahmana. 2nd ed. Cambridge, Mass.: Harvard University Press. [Reprint: 1971.]

The following bibliography of works on Sanskrit syntax consolidates the combined bibliographies of Madhav M. Deshpande and
Hans Henrich Hock, entitled A bibliography of writings on Sanskrit syntax, published in Hock 1991c and the bibliography of
works on Sanskrit syntax published since then compiled by Hans
Henrich Hock for presentation at the Seminar on Sanskrit Syntax
and Discourse Structures held in Paris, 1315 June 2013.
Aalto, Pentti. 1979. On the absolute instrumental in Sanskrit.
Dr. Ludwik Sternbach felicitation volume, ed. by J. P. Sinha,
vol. 1, pp. 4755. Indologica Taurinensia 7. Torino: Edizioni
Jollygrafica; Lukhnow: Akhila Bharatiya Sanskrit Parishad.
Aklujkar, Ashok. 1969. Two textual studies of Bhartr-hari. Journal of the American Oriental Society 89: 54763.
Albino, Marcos. 2000. Finale Dative auf -ya im Rgveda. His
torische Sprachwissenschaft 113: 11731.
Ananthanarayana, H. S. 1970. The karaka theory and case grammar. Indian Linguistics 31: 1427.
Andersen, Paul Kent. 1979. Word order typology and prepositions in Old Indic. Festschrift for Oswald Szemernyi, ed. by
B. Brogyanyi, pp. 2334. Amsterdam: John Benjamins.
. 1980. Typological and functional aspects of word order in Asoka. General Linguistics 20: 122.
. 19821983. Amplified sentences in Asoka. Zeitschrift fr
vergleichende Sprachforschung 96: 1729.
. 1982a. Means of expressing a comparison of inequality in Old
Indic. General Linguistics 22: 17284.
. 1982b. On the word order typology of the Satapathabr

ahman.a. Journal of Indo-European Studies 10: 3742.
399
400
S ANSKRIT SYNTAX
. 1983. Word order typology and comparative constructions.

Amsterdam: John Benjamins.
. 1985a. Die grammatische Kategorie Passiv im Altindischen:
ihre Funktion. Grammatische Kategorien: Funktion und
Geschichte: Akten der VII. Fachtagung der Indogermanischen
Gesellschaft, Berlin, 20.-25. Februar 1983, ed. by Bernfried
Schlerath, in collab. with Veronica Rittner, pp. 4757. Wiesbaden: L. Reichert.
. 1985b. Means of expressing a comparison of inequality in
Old Indic. Proceedings of the Fifth World Sanskrit Conference, ed. by R. N. Dandekar and P. D. Navathe, pp. 1115.
New Delhi: Rashtriya Sanskrit Sansthan.
. 1986. The genitive agent in Rigvedic passive constructions. Collectanea linguistica, pp. 913. Prace Komisji
Jezykoznawstwa 53.

us.an.a Press.
Aralikatti, R. N. 1982. A study of modern spoken Sanskrit with
reference to sentence patterns. Ph.D. dissertation. Tirupati: S.
V. University.
. 1985. A note on the complement structures in sentences in
modern spoken Sanskrit as recorded on tapes. Proceedings of
the Fifth World Sanskrit Conference, ed. by R. N. Dandekar
and P. D. Navathe, pp. 1626. New Delhi: Rashtriya Sanskrit
Sansthan. [Reprinted as Aralikatti 1986.]
. 1986. A note on the complement structures in sentences in
modern spoken Sanskrit as recorded on tapes. Samskr
ta-vi
marsah. 4.2: 18. [Reprint of Aralikatti 1985.]
. 1991. A note on word order in modern spoken Sanskrit and
some positive constraints. Studies in Sanskrit syntax: a vol-
B IBLIOGRAPHY
401
ume in honour of the centennial of Speijers Sanskrit syntax

(18861986), ed. by Hans Henrich Hock, pp. 1318.
Arnold, Edward V. 1897. Outline of the historical grammar of the
Vedas. Journal of the American Oriental Society 18: 20352.
Avery, John. 1880. Contributions to the history of verb inflection in Sanskrit. Journal of the American Oriental Society 10:
219324.
. 1885a. On modes in relative clauses in the Rig-Veda. Journal
of the American Oriental Society 11: cxlviiiclxi. Proceedings
at Boston, May 2nd, 1883.
. 1885b. On relative clauses in the Rig-Veda. Journal of
the American Oriental Society 11: lxivlxvi. Proceedings at
Boston, May 18th, 1881.
Bader, Franoise. 1987. Structure de lnonc en indo-europen.
Papers from the 7th International Conference on Historical
Linguistics, ed. by Anna Giacalone Ramat, Onofrio Carruba,
and Giuliano Bernini, pp. 1334. Amsterdam studies in the
theory and history of linguistic science; series IV, Current issues in linguistic theory 48. Amsterdam: John Benjamins.
Baldi, Philip. 1979. Typology and the Indo-European prepositions. Indogermanische Forschungen 84: 4961.
Balles, Irene. 2000. Die altindische Cvi-Konstruktion: alte Deutungen und neue Wege. Indoarisch, Iranisch und die Indogermanistik, ed. by Bernhard Forssman and Robert Plath, pp. 25
36. Wiesbaden: L. Reichert.
. 2006. Die altindische Cvi-Konstruktion: Form, Funktion, Ursprung. Mnchner Forschungen zur historischen Sprachwissenschaft 4. Bremen: Ute Hempen Verlag.
. 2008. Principles of syntactic reconstruction and morphology
as paleosyntax: the case of some Indo-European secondary
verbal formations. Principles of syntactic reconstruction, ed.
by Gisella Ferraresi and Maria Goldbach, pp. 16186. Amsterdam; Philadelphia: John Benjamins.
402
S ANSKRIT SYNTAX
Bardal, Jhanna and Thrhallur Eythrsson. 2009. The origin of

the oblique subject construction: an Indo-European comparison. Grammatical change in Indo-European languages, ed.
by Vit Bubenik, John Hewson, and Sarah Rose, pp. 17993.
Amsterdam; Philadelphia: John Benjamins.
Barth, A. 1875. Le grondif sanscrit en tva. Mmoires de la Socit de Linguistique de Paris 2: 23840.
Bartholomae, Christian. 1889. Der sogenannte genitivus temporis im Veda. Zeitschrift fr die Kunde der indogermanischen
Sprachen 15: 20021.
. 1892. Altindische Infinitive auf -man und -mani. Indogermanische Forschungen 1: 495500.
. 1907. Zur Gerundivbildung im Arischen. Zeitschrift fr vergleichende Sprachforschung 41: 31935.
Basu, D. N. 1983. A case for the genitive case. Proceedings of
the International Seminar on Studies in the As..ta dhyay of Pan.ini, ed. by S. D. Joshi and S. D. Laddu, pp. 1059. Pune:
Centre for Advanced Study in Sanskrit, University of Poona.
Behaghel, Otto. 1929. Zur Stellung des Verbs im Germanischen
und Indogermanischen. Zeitschrift fr vergleichende Sprachforschung 56: 27681.
Benfey, Theodor. 1872. ber die Entstehung des indogermanischen Vokativs. Abhandlungen der Kniglichen Gesellschaft
der Wissenschaften zu Gttingen 17.4: 392. Vorgelegt in der
Kniglichen Gesellschaft der Wissenschaften den 14. Februar
1872.
Benigny, Julius. 1929. Zum indogermanischen Nominalsatz. Indogermanische Forschungen 47: 12444.
Benveniste, mile. 1935. Les infinitifs avestiques. Paris: AdrienMaisonneuve.
. 1948. Noms dagent et noms daction en indo-europen. Paris:
Adrien-Maisonneuve.
B IBLIOGRAPHY
403
. 1950. La phrase nominale. Bulletin de la Socit de Linguistique de Paris 46.1: 1936.

. 1951. Prtrit et optatif en indo-europen. Bulletin de la Socit de Linguistique de Paris 47.1: 1120.
Bergaigne, Abel. 1879. Essai sur la construction grammaticale
considre dans son dveloppement historique en sanskrit, en
grec, en latin, dans les langues romanes et dans les langues
germaniques. Mmoires de la Socit de Linguistique de Paris
3: 151, 12454, 16986.
. 1887. La syntaxe des comparaisons vdiques. Mlanges Renier, pp. 75101. Paris: Vieweg.
Bhandarkar, Ramkrishna Gopal. 1914. On Sanskrit and the derived
languages: Wilson Philological Lectures, delivered in 1877.
Bombay: Radhabai Atmaram Sagoon.
Bharadwaj, Sudhi Kant. 1982. Linguistic study of dharmasutras.
Rohtak: Manthan Publications.
Bhatnagar, V. 1995. Un-Pan.inian syntax (concord) in the Ramayan.a and the Mahabharata. Modern evaluation of the Mahabharata: Prof. R. K. Sharma felicitation volume, ed. by Satya
Pal Narang, pp. 32844. Delhi: Nag Publishers.
Bhatt, V. M. 2000. Pragmatics of absolute locative in Sanskrit.
Annals of the Bhandarkar Oriental Institute 81: 27982.
Bichlmeier, Harald. 1999. Zur Syntax der ablativischen Adverbien im Rgveda. Mnchener Studien zur Sprachwissenschaft
60: 766.
Biese, Yrj Moses Jalmari. 1931. Der Gebrauch von as- und bhuim Aitareya-Brahman.a. Arctos 2: 4465.
Bloch, Jules. 19061908. La phrase nominale en sanskrit. Mmoires de la Socit de Linguistique de Paris 14: 2796.
. 1934. Lindo-aryen du veda aux temps modernes. Paris:
Adrien-Maisonneuve. [English translation by A. Master 1965.]
404
S ANSKRIT SYNTAX
. 1965. Indo-Aryan from the Vedas to modern times, ed. by

A. Master. Paris: Adrien-Maisonneuve. [Translation of Bloch
1934.]
Bloomfield, Maurice. 1884. On certain irregular Vedic subjunctives or imperatives. American Journal of Philology 5: 115.
. 1912a. On instability in the use of moods in earliest Sanskrit.
American Journal of Philology 33.1: 129.
. 1912b. On the variable position of the finite verb in oldest
Sanskrit. Indogermanische Forschungen 31: 15677.
Bloomfield, Maurice and Franklin Edgerton. 1930. Vedic variants.
3 vols. Philadelphia: Linguistic Society of America.
Bodewitz, H. W. 2001. A Vedic example of no doubt used as
parenthetic clause. Indo-Iranian Journal 44: 1720.
Bhtlingk, Otto. 1888. Ueber den impersonalen Gebrauch der
Participia necess. im Sanskrit. Zeitschrift der Deutschen Morgenlndischen Gesellschaft 42: 36669.
Borooah, Anundoram. 1879. A higher Sanskrit grammar: or, gender and syntax, with copious illustrations from standard Sanskrit authors and references to Latin and Greek grammars.
Calcutta: Mukherjea; London: Trbner.
Brereton, Joel P. 1982. The particle iva in Vedic Prose. Journal
of the American Oriental Society 102: 44350.
. 1986. Tat tvam asi in context. Zeitschrift der Deutschen Morgenlndischen Gesellschaft 136: 98109.
Brereton, Joel P. and Stephanie W. Jamison, eds. 1991. Panels of
the VIIth World Sanskrit Conference, Kern Institute, Leiden,
August 2329, 1987, ed. by Johannes Bronkhorst; vol. 4, Sense
and syntax in Vedic.
Breunis, Andries. 1990. The nominal sentence in Sanskrit and
Middle Indo-Aryan. Leiden: E. J. Brill.
Brockington, J. L. 1968. The syntax and style of the Ramayan.a.
Ph.D. dissertation. Oxford: Oxford University.
B IBLIOGRAPHY
405
Bronkhorst, Johannes, ed. 19901992. Panels of the VIIth World

Sanskrit Conference, Kern Institute, Leiden, August 2329,
1987. 11 vols. Leiden: E. J. Brill.
Brugmann, Karl. 1883. Altind. pura und griech. pros mit
dem Indikativ des Prsens. Abhandlungen der KniglichSchsischen Gesellschaft der Wissenschaften, philosophischhistorische Classe 35: 16973.
. 1904. Die Demonstrativpronomina der indogermanischen
Sprachen: eine bedeutungsgeschichtliche Untersuchung. Abhandlungen der philosophisch-historischen Klasse der Kniglich-Schsischen Gesellschaft der Wissenschaften 22.6.
Leipzig: B. G. Teubner.
. 1925. Die Syntax des einfachen Satzes im Indogermanischen.
Berlin: de Gruyter.
Bubenik, Vit. 1983. Impersonal constructions in functional grammar. Advances in functional grammar, ed. by S. C. Dik,
pp. 183204. Dordrecht: Foris.
. 1987a. Jussive sentences in Sanskrit syntax. Select Papers
from SALA-7, ed. by E. Bashir et al., pp. 3148. Bloomington:
Indiana University Linguistics Club.
. 1987b. Passivized causatives in Sanskrit and Prakrits. Linguistics 25: 687704.
. 1991. Nominal and pronominal objects in Sanskrit and
Prakrit. Studies in Sanskrit syntax: a volume in honour of
the centennial of Speijers Sanskrit syntax (18861986), ed.
by Hans Henrich Hock, pp. 1930.
. 2006a. Cases and prepositions in Indo-Aryan. From case to
adposition: the development of configurational syntax in IndoEuropean languages, ed. by John Hewson and Vit Bubenik,
pp. 10230. Amsterdam; Philadelphia: John Benjamins.
. 2006b. On the evolutionary changes in the Old and Middle
Indo-Aryan systems of case and adpositions. Papers of the
12th World Sanskrit Conference, held in Helsinki, Finland, 13
406
S ANSKRIT SYNTAX
18 July, 2003; vol. 5, Themes and tasks in Old and Middle

Indo-Aryan linguistics, ed. by Bertil Tikkanen and Heinrich
Hettrich, pp. 6588. Delhi.
Burnouf, Eugne. 1824. Sur un usage remarquable de linfinitif
samscrit. Journal asiatique 5: 12024.
Butt, Miriam. 2003. The light verb jungle. Harvard Working Papers in Linguistics 9: 149.
Caland, W. 1891. Zur Syntax der Pronomina im Avesta. Verhandelingen der Koninklijke Akademie van Wetenschappen, Afd.
Letterkunde 20. Amsterdam: Johannes Mller.
. 1897. Eine bereinstimmung zwischen Vedischem und
Avestischem sprachgebrauch. Zeitschrit fr vergleichende
Sprachforschung 34: 4567.
. 19121913. Syntaktisch-exegetische Miszellen. Indogermanische Forschungen 31: 1057.
. 1927. On a paragraph of Vaidic syntax. Acta Orientalia 5:
4951. Leiden.
Canedo, Jos. 1937. Zur Wort- und Satzstellung in der alt- und
mittelindischen Prosa. Gttingen: Vandenhoeck and Ruprecht.
Cardona, George. 1970c. The Indo-Iranian construction mana
(mama) krtam. Language 46: 112.
aninis karakas: agency, animation, and identity.
. 1974c. P
.
. 1976b. Subject in Pan.ini. The notion of subject in South
Asian languages, ed. by M. K. Verma, pp. 138. South Asian
Studies Publications Series 2. Madison: University of Wisconsin.
. 1978a. A note on Indo-Iranian dha (IE *dhe). Indian Linguistics 39; Professor P. B. Pandit memorial volume, pp. 209
24. Pune: Linguistic Society of India.
. 1978b. Relations between causatives and passives in IndoIranian. Studies in the Linguistic Sciences 8.2: 142.
B IBLIOGRAPHY
407
. 1978c. Review of Haudry 1977. Kratylos 23: 7181. [Appeared 1979.]

. 1981. Vedic causatives. Golden Jubilee volume, pp. 1738.
sodhana man.d.ala.
Pune: Vaidika sam
. 1990. A note on dative agents in Sanskrit. Experiencer
Subjects in South Asian Languages: proceedings of the Conference on the Experiencer Subject in South Asian Languages,
November 1988, Madison, Wisconsin, ed. by Manindra K.
Verma and K. P. Mohanan, pp. 14145.
. 1993. On comparatives and superlatives formed to finite verbs
in Sanskrit. Annals of the Bhandarkar Oriental Research Institute 72-73: 40922.
. 1998. kim
tarhi in the Mahabhas.ya. Mr Curad: studies
in honor of Calvert Watkins, ed. by Jay Jasonoff, H. Craig
Melchert, and Lisi Olivier, pp. 6373. Innsbruck: Institut fr
Sprachwissenschaft.
. 2002. The Old Indo-Aryan tense system. Journal of the
. 2012b. On the construction type nat.asya srn.oti. Samskr
ta
Vimarsah. 6: 6584.
Casaretto, Antje. 2010a. Syntax und Wortarten der Lokalpartikeln des Rgveda IX: a nu. Mnchener Studien zur Sprach 65: 764.
wissenschaft
. 2010b. Syntax und Wortarten der Lokalpartikeln des Rgveda
and
VI: prati. International Journal of Diachronic Linguistics
Linguistic Reconstruction 7: 153.
. 2010c. Syntax und Wortarten der Lokalpartikeln des Rgve zur
da XXI: puras, purastat und pura . Mnchener Studien
Sprachwissenschaft 66: 1153.
. 2011. Syntax und Wortarten der Lokalpartikeln des Rgveda
VIII: a bhi. Historische Sprachforschung 123: 97156.
. 2012a. Lokalpartikeln im Rgveda XIV: a ccha. Zeitschrift fr
28: 127.
Indologie und Sdasienstudien
408
S ANSKRIT SYNTAX
. 2012b. Syntax und Wortarten der Lokalpartikeln des Rgveda

XII: v. Historische Sprachforschung 124: 13477.
. 2012c. Syntax und Wortarten der Lokalpartikeln des Rgve
da XVI: a ti und tiras. International Journal of Diachronic
Linguistics and Linguistic Reconstruction 8: 173216.
Channing, Eva. 1887. On negative clauses in the Rigveda. Journal of the American Oriental Society 13: xcixcii.
Charpentier, Jarl. 1912. Die Desiderativbildungen der indoiranischen Sprachen. Uppsala: Appelberg.
Chen, Shu-Fen and Benjamin Slade, eds. 2014. Grammatica et
verba Glamour and verve: studies in South Asian, historical, and Indo-European linguistics in honor of Hans Henrich
Hock on the occasion of his seventy-fifth birthday. Ann Arbor;
New York: Beech Stave Press.
Costello, John R. 1982. The absolute construction in IndoEuropean: a syntagmemic reconstruction. Journal of IndoEuropean Studies 10: 23552.
Cuny, Albert. 19061908. Les prverbes dans le atapathabrahman.a. Mmoires de la Socit de Linguistique de Paris 14:
289331. Paris: Librairie Honore Champion. [Reprinted as
Cuny 1907.]
. 1907. Les prverbes dans le atapathabrahman.a. Thse pour
le doctorat, prsente la Facult des lettres de lUniversit de
Paris. Paris: Impr. nationale. [Reprint of Cuny 19061908.]
Dahl, Eystein. 2008. Performative sentences in Vedic. Journal of
South Asian Linguistics 1.1: 525.
. 2009a. Semantische und pragmatisch-kontextuelle Faktoren
in der Entwicklung des altindoarischen Perfekts. Pragmatische Kategorien: Form, Funktion und Diachronie, ed. by E.
Rieken and P. Widmer, pp. 3550. Wiesbaden: L. Reichert.
. 2009b. Some semantic and pragmatic aspects of object alternation in early Vedic. The role of semantics and pragmatics
B IBLIOGRAPHY
409
in the development of case, ed. by J. Bardal and Sh. Chelliah,

. 2009c. Time, tense and aspect in early Vedic grammar. Ph.D.
dissertation. Oslo: University of Oslo.
. 2010. Time, tense and aspect in Early Vedic grammar. Leiden:
E. J. Brill.
. 2012. Evidence for evidentiality in Late Vedic. Indic across
the millennia: from the Rigveda to Modern Indo-Aryan: 14th
World Sanskrit Conference, Kyoto, Japan, September 1st5th,
2009, proceedings of the Linguistic Section, ed. by Jared S.
Klein and Kazuhiko Yoshida, pp. 922.
Dahl, Eystein and Chiara Fedriani. 2012. The argument structure of experience: experiential constructions in Early Vedic,
Homeric Greek and Early Latin. Transactions of the Philological Society 110.3: 34262.
Dash, Siniruddha and Brendan Gillon. 1995. Adjectives in Sanskrit. Adyar Library Bulletin 59: 28594.
Davison, Alice. 2009. Adjunction, features, and locality in
Sanskrit and Hindi-Urdu correlatives. Correlatives crosslinguistically, ed. by Anik Liptk, pp. 22362. Amsterdam;
Philadelphia: John Benjamins.
Debrunner, Albert. 1928. Zeitadverbia als Nominative. Indogermanische Forschungen 46: 182.
. 1948. Indirekte Rede im Altindischen. Acta Orientalia 20:
12032.
Delbrck, Bertold. 1867. Ablativ, localis, instrumentalis im altindischen, lateinischen, griechischen und deutschen. Berlin:
Dmmler.
. 1869. Ueber den indogermanischen, speciell den vedischen
Dativ. Zeitschrift fr vergleichende Sprachforschung 18: 81
106.
410
S ANSKRIT SYNTAX
. 1871. Syntaktische Forschungen; vol. 1, Der Gebrauch des

Conjunctivs und Optativs im Sanskrit und Griechischen. Halle:
Waisenhaus.
. 1876. Syntaktische Forschungen; vol. 2, Altindische Tempuslehre. Halle: Waisenhaus.
. 1878. Syntaktische Forschungen; vol. 3, Die altindische Wort
folge aus dem Satapathabr
ahman.a dargestellt. Halle: Waisenhaus.
. 1888. Syntaktische Forschungen; vol. 5, Altindische Syntax.
Halle: Waisenhaus. [Reprint: Darmstadt: Wissenschaftliche
Buchgesellschaft, 1968.]
. 1900. Grundri der vergleichenden Grammatik der indogermanischen Sprachen, by K. Brugmann and B. Delbrck;
vol. 5, Vergleichende Syntax der indogermanischen Sprachen.
Straburg: Trbner.
Deshpande, Madhav Murlidhar. 1980. Evolution of syntactic theory in Sanskrit grammar: syntax of the Sanskrit infinitive tumUN. Ann Arbor: Karoma.
. 1981a. Pan.ini and the Vedic evidence: a peep into the past.
sodhana
Golden Jubilee volume, pp. 5265. Pune: Vaidika Sam
Man.d.ala.
. 1981b. Sanskrit gerund constructions: syntactic disputations.
Indo-Iranian Journal 23: 16785.
. 1983. Pan.ini as a frontier grammarian. Papers from the 19th
Regional Meeting of the Chicago Linguistic Society, pp. 110
16. Chicago: Chicago Linguistic Society.
. 1985. Ellipsis and syntactic overlapping: current issues in Pan.inian syntactic theory. Pune: Bhandarkar Oriental Research
Institute.
. 1987a. From uttarapadalopa to madhyamapadalopa: implications for theoretical change. Annals of the Bhandarkar Oriental Research Institute 67: 25157.
B IBLIOGRAPHY
411
. 1987b. Pan.inian reflections on Vedic infinitives: On infinitives with -tosUN and -KasUN. Paper presented at the VIIth
World Sanskrit Conference, Kern Institute, Leiden, 2329 August 1987. Unpublished.
. 1987c. Pan.inian syntax and the changing notion of sentence.
R. G. Bhandarkar 150th birth-anniversary volume, pp. 5598.
Annals of the Bhandarkar Oriental Research Institute 68. Pune:
Bhandarkar Oriental Research Institute.
. 1987d. Pan.inian syntax of gerund constructions: an alternative view. Adyar Library Bulletin; Supplement, Ludo Rocher
felicitation volume, pp. 24266. Madras: Adyar Library.
. 1988a. Ellipsis in Pan.inian grammar: a justifiable gap? Paper presented at the Third Sanskrit Syntax Symposium, South
Asian Languages Analysis Roundtable, Seattle, WA, 1988.
Unpublished.
. 1988b. Pan.ini and the northwestern dialect: some suggestions
on sutra 3.3.10. Languages and cultures: studies in honor of
Edgar C. Polom, ed. by Mohammad Ali Jazayery and Werner
Winter, pp. 11122. Berlin: Mouton de Gruyter.
. 1988c. Prototypes in Pan.inian syntax. Paper presented at
the 198th annual meeting of the American Oriental Society,
Chicago, 1988. Unpublished.
. 1989. Vedic syntax: a select bibliography. New horizons of
research in Indology: silver jubilee volume, ed. by V. N. Jha,
pp. 2948. Publication of the Centre of Advanced Study in
Sanskrit, class E 10. Pune: Centre for Advanced Study of Sanskrit, University of Poona.
. 1990. Some features of the sampradana karaka in Pan.ini. Experiencer Subjects in South Asian Languages: proceedings of
the Conference on the Experiencer Subject in South Asian Languages, November 1988, Madison, Wisconsin, ed. by Manindra K. Verma and K. P. Mohanan, pp. 14760.
412
S ANSKRIT SYNTAX
. 1991a. Ditransitive passive in Pan.ini. Indo-Iranian Journal

34.1: 1935.
. ed. 1991b. Panels of the VIIth World Sanskrit Conference,
Kern Institute, Leiden, August 2329, 1987, ed. by Johannes
Bronkhorst; vol. 5, Pan.ini and the Veda.
. 1991c. Pan.inian reflections on Vedic infinitives: on the meaning of tumartha. Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986),
ed. by Hans Henrich Hock, pp. 4554.
. 1991d. Pan.inian syntax and changing notion of sentence.
Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986), ed. by Hans
Henrich Hock, pp. 3144.
Deshpande, Madhav Murlidhar and Hans Henrich Hock. 1991. A
bibliography of writings on Sanskrit syntax. Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986), ed. by Hans Henrich Hock,
pp. 21941.
Devasthali, G. V. 1974. Vakya according to the munitraya of
Sanskrit grammar. Charudeva Shastri felicitation volume,
pp. 20615. Delhi: Charudeva Shastri Felicitation Committee.
Disterheft (Haas), Dorothy. 1976. The voice of the infinitive in
the RigVeda. Current progress in historical linguistics: proceedings of the Second International Conference on Historical Linguistics, Tucson, Arizona, 1216 January 1976, ed.
by William M. Christie, pp. 10727. Amsterdam; New York:
North-Holland.
Disterheft, Dorothy. 1980. The syntactic development of the infinitive in Indo-European. Columbus, Ohio: Slavica.
. 1981. The Indo-Iranian predicate infinitive. Zeitschrift fr
vergleichende Sprachforschung 95: 11021.
Dressler, Wolfgang. 1968. Ved. dive-dive und die idg. Iterativkomposita. Studien zur Sprachwissenschaft und Kul-
B IBLIOGRAPHY
413
turkunde: Gedenkschrift fr Wilhelm Brandenstein, pp. 3947.

Innsbruck: Institut fr Vergleichende Sprachwissenschaft der
Leopold-Franzens-Universitt.
. 1969. Eine textsyntaktische Regel der idg. Wortstellung.
Zeitschrift fr vergleichende Sprachforschung 83: 125.
. 1971. ber die Rekonstruktion der idg. Syntax. Zeitschrift
fr vergleichende Sprachforschung 85: 522.
Dunkel, George E. 1976. Repetition and deletion of preverbs and
verbs in early Indic and Greek. Ph.D. dissertation. Philadelphia: Department of Linguistics, University of Pennsylvania.
. 1979. Preverb repetition. Mnchener Studien zur Sprachwissenschaft 38: 4182.
. 1981. Amred
. ita and iteration of preverbs in Vedic and Hittite. Zeitschrift fr vergleichende Sprachforschung 95: 214
26.
. 1982a. Autour de a rg-vdique. Indo-Iranian Journal 24:
89102.
. 1982b. The original syntax of conjunctive *-kw e. Die
Sprache 28: 12943.
. 1985. IE hortatory *ey, *eyte: Ved. e ta . . . stavama, Hitt.
ehu=wa it, Hom. ei dge. Mnchener Studien zur Sprach
wissenschaft 46: 4779.
. 1990. J. Wackernagel und die idg. Partikeln *so, *ke, *kem
und *an. Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute: Kolloquium der Indogermanischen Gesellschaft . . . 1988, ed. by H. Eichner and H. Rix,
pp. 10030. Wiesbaden: L. Reichert.
. 1992. Die Grammatik der Partikeln. Grammatische Kategorien: Akten der 8. Fachtagung der Indogermanischen
Gesellschaft, Leiden, ed. by R. Beekes et al., pp. 15377. Innsbruck: Institut fr Sprachwissenschaft der Universitt.
. 1997. Conjunctive u and invariable sa in the Rgveda. In
dogermanische Forschungen 102: 15678.
414
S ANSKRIT SYNTAX
Durkin, Desmond. 1991. Konditionalstze im Satapathabr

ahman.a. Freiburger Beitrge zur Indologie 26. Wiesbaden: Harrassowitz.
Eaton, A. J. 1884. The a tmanepada in the Rigveda. Ph.D. dissertation. Leipzig: Universitt Leipzig.
Elizarenkova, Tatyana J. 1960a. Aorist v Rigvede. [The aorist in
the Rig-Veda]. Moscow: Izdatelstvo Vostokhnoj Literatury.
. 1960b. On the problem of the development of tenses in OldIndo-Aryan (the R.gveda). Vol. 4. Moscow: Oriental Literature
Publishing House.
. 1982. Review of Haudry 1977. Indo-Iranian Journal 23: 43
53.
. 1995. Language and style of the Vedic rs.is. New York: State
University of New York Press.

Elizarenkova, Tatyana J. and V. N. Toporov. 1976. The Pali language. Moscow: Nauka.
Emeneau, Murray B. 1969. Sanskrit syntactic particles: kila, khalu, nunam. Indo-Iranian Journal 11: 24168. [Reprinted in
Nooten 1988: 12947.]
. 1971. Dravidian and Indo-Aryan: the Indian linguistic area.
Symposium on Dravidian civilization, ed. by Andre F.
Sjoberg, pp. 3368. Austin; New York: Jenkins. [Reprinted in
Emeneau 1980: 16796.]
. 1980. Language and linguistic area. Essays collected, ed. by
A. S. Dil. Stanford, CA: Stanford University Press.
Etter, Annemarie. 1985. Die Fragestze im Rgveda. Berlin: de
Gruyter.
Fahs, A. 1963. Beitrge zur Syntax der Palisprache. Ph.D. dissertation. Rostock: Universitt Rostock.
Fahs, Achim. 1989. Grammatik des Pali. Leipzig: Verlag Enzyklopdie.
Filliozat, Pierre. 1983. The Pan.inian conception of karmakartr.
Proceedings of the International Seminar on Studies in the A-
B IBLIOGRAPHY
415
s..ta dhyay of Pan.ini, ed. by S. D. Joshi and S. D. Laddu, pp. 11

16. Pune: Centre for Advanced Study in Sanskrit, University of
Poona.
Forssman, Bernhard. 1981. Zur altpersischen Syntax: DB I 50.
Mnchener Studien zur Sprachwissenschaft 40: 3547.
. 1985. Fortfhrung von Relativstzen im Avestischen.
Franke, Otto. 1890. Die casuslehre des Pan.ini verglichen mit dem
gebrauch der casus im Pali und in den Asoka-inschriften.
Beitrge zur Kunde der indogermanischen Sprachen 16: 64
120.
Friedrich, Paul. 1975. Proto-Indo-European syntax: the order
of meaningful elements. Journal of Indo-European Studies;
monograph 1.
Frst, Alfons. 1916. Der Sprachgebrauch der lteren Upanis.ads verglichen mit dem der frheren vedischen Perioden und
dem des klassischen Sanskrit. Zeitschrift fr vergleichende
Gaedicke, Carl. 1880. Der Accusativ im Veda. Breslau: Koebner.
Garca-Ramn and Jos Luis. 1995. Zum Akkusativ der Richtung
im Vedischen und im Indogermanischen. Verba et structurae:
Festschrift fr Klaus Strunk zum 65. Geburtstag, ed. by Heinrich Hettrich et al., pp. 3352. Innsbruck: Institut fr Sprachwissenschaft.
. 1997. Infinitivos y abstractos verbales en indoiranio. Syntaxe des langues indo-iraniennes anciennes, ed. by E. Pirart,
pp. 2949. Barcelona: Editorial Ausa.
Ghosal, S. N. 19661967. The genitive in the role of the nongenitive. Journal of the Oriental Institute 16: 296314.
Gillon, Brendan S. 1987a. A GB approach to the syntax of classical Sanskrit. Paper read at the 9th South Asian Languages
Analysis Roundtable, Cornell and Syracuse Universities, June
1987. Unpublished.
416
S ANSKRIT SYNTAX
. 1987b. Two forms of negations in Sanskrit: Prasajyapratis.edha and paryudasapratis.edha. Lokapraja 1.1: 8189. Puri:
Lokabhas.a Pracara Samitih., Bada Odiya Matha.
. 1993. Bhartrharis solution to the problem of asamartha com hari: philosopher and grammarian: proceedpounds. Bhartr
International Conference on Bhartrhari, ed.
ings of the First
Bern:
by Saroja Bhate and Johannes Bronkhorst, pp. 11733.
Peter Lang. [Reprint: Delhi: Motilal Banarsidass, 1994.]
. 1995. Autonomy of word formation: evidence from Classical
Sanskrit. Indian Linguistics 56.14: 1552.
. 1996. Word order in Classical Sanskrit. Indian Linguistics
57.14: 136.
. 2004. Subject predicate order in Classical Sanskrit. Language and grammar: studies in mathematical linguistics and
natural language, ed. by Philip Scott, Claudia Casadio, and
Robert Seely, pp. 21125. Stanford: Center for the Study of
Language and Information, Stanford University.
. 2007a. Exocentric (bahuvrhi) compounds in Classical Sanskrit. Proceedings of the First International Symposium on
Sanskrit Computational Linguistics, ed. by Grard Huet and
Amba Kulkarni, pp. 112. Rocquencourt: Institut National
de Recherche en Informatique et en Automatique. URL:
http : / / sanskrit . inria . fr / Symposium /
Proceedings.pdf.
. 2010a. Linguistic investigations into ellipsis in Classical Sanskrit. Sanskrit computational linguistics: 4th International
Symposium, New Delhi, India, December 2010, Proceedings,
ed. by Girish Nath Jha, pp. 21830. Lecture Notes in Artificial
Intelligence 6465.
B IBLIOGRAPHY
417
Gippert, Jost. 1978. Zur Syntax der infinitivischen Bildungen

in den indogermanischen Sprachen. Europische Hochschulschriften, Reihe 21, Linguistik 3. Frankfurt: Lang. [Originally
presented as the authors thesis, Freie Universitt Berlin.]
. 1984. Zum prdikativen Infinitiv. Zeitschrift fr vergleichende Sprachforschung 97: 20520.
. 1985. Verbum dicendi + Infinitiv im Indoiranischen. Mnchener Studien zur Sprachwissenschaft 44: 2957.
. 1995. Zur Syntax des Infinitivs auf -tum im Altindischen.
Kuryowicz memorial volume, ed. by Wojciech Smoczynski,
vol. 1, pp. 25577. Analecta Indoevropaea Cracoviensia 2.
Krakw: Universitas.
. 2004. Indo-European word order in main and subordinate
clauses in a diachronic perspective. Analecta homini universali dicata: Arbeiten zur Indogermanistik, Linguistik, Philologie, Politik, Musik und Dichtung: Festschrift fr Oswald
Panagl zum 65. Geburtstag, ed. by Thomas Krisch, Thomas
Lindner, and Ulrich Mller, vol. 1, pp. 4868. Stuttgarter Arbeiten zur Germanistik 421. Stuttgart: Heinz.
Gonda, Jan. 1942. Bemerkungen zum Gebrauch der Pronomina
der 1. und 2. Person als Subjekt im Altindischen. Acta Orientalia 19: 21179. [Reprinted in Gonda 1975b: 11179.]
. 1948. Quelques observations sur lemploi du verb simple au
lieu dun compos etc. dans la langue sanskrite. Acta Orientalia 20: 167205.
. 1951a. La place de la particule ngative na dans la phrase en
vieil indien. Leiden: E. J. Brill.
. 1951b. Remarks on the Sanskrit passive. Leiden: E. J. Brill.
. 1952. Remarques sur la place du verbe dans la phrase active
et moyenne en langue sanscrite. Utrecht: Oosthoek.
. 1954. The original character of the Indo-European relative
pronoun io-. Lingua 4: 141. [Reprinted in Gonda 1975a:
164204.]
418
S ANSKRIT SYNTAX
. 1955. Two notes on Sanskrit syntax. Journal of Oriental Research, Madras 24: 18.
. 1956a. Defining the nominative. Lingua 5: 28897.
. 1956b. The character of the Indo-European moods. Wiesbaden: Harrassowitz.
. 1957a. A critical survey of the publications on the periphrastic future in Sanskrit. Lingua 6: 15879.
. 1957b. A note on the functions of the accusative. S. K. Belvalkar felicitation volume, pp. 7280. Banaras: Motilal Banarsidass.
. 1957c. Attraction and coordination in the Veda. Bulletin
of the School of African and Oriental Studies 20: 27989.
[Reprinted in Gonda 1975b: 22737.]
. 1957d. The character of the Sanskrit accusative. Miscelnea
homenaje a Andr Martinet, ed. by D. Cataln, pp. 4765. Canarias. [Reprinted in Gonda 1975a: 4462.]
. 1957e. The use of the particle ca. Vak 5: 173. [Reprinted in
Gonda 1975a: 32092.]
. 1958. Syntax and verse structure in the Veda. Indian Linguistics 20: 3543. [Reprinted in Gonda 1975b: 25967.]
. 1959a. Four studies in the language of the Veda. The Hague:
Mouton.
. 1959b. On amplifed sentences and similar structures in the
Veda. Four studies in the language of the Veda, pp. 770. The
Hague: Mouton.
. 1959c. Stylistic repetition in the Veda. Verhandelingen der
Koninklijke Nederlandse Akademie van Wetenschappen,
Afdeeling Letterkunde, Verhandelingen, nieuwe reeks 65.3.
Amsterdam: Noord-Hollandsche Uitgevers Maatschappij.
. 1960a. Ellipsis, brachylogy, and other forms of brevity in
speech in the Rgveda. Verhandelingen der Koninklijke Neder van Wetenschappen, Afdeeling Letterkunde,
landse Akademie
B IBLIOGRAPHY
419
Verhandelingen, nieuwe reeks 67.4. Amsterdam: NoordHollandsche Uitgevers Maatschappij.

. 1960b. Reflections on the Indo-European medium. Lingua
9: 3067, 17593.
. 1960c. Some notes on the position of the attributive adjective
in early Indian prose. Bulletin of the Deccan College Research
Institute 20: 30318. [Reprinted in Gonda 1975b: 95110.]
. 1962a. The aspectual functions of the R.gvedic present and
aorist. The Hague: Mouton.
. 1962b. The unity of the Vedic dative. Lingua 11: 14150.
[Reprinted in Gonda 1975a: 3443.]
. 1967. On the use of the absolutive in Sanskrit. Kaviraj Abhinandana Grantha, ed. by B. Saksena, pp. 26265. Lucknow:
Akhil Bharatiy Sanskrit Parishad. [Reprinted in Gonda 1975b:
9194.]
. 1968. The Sanskrit particle api. Lingua.
. 1971. Old Indian. Handbuch der Orientalistik 2.1.1. Leiden: E.
J. Brill.
. 1975a. Selected studies / J. Gonda; vol. 1, Indo-European linguistics.
. 1975b. Selected studies / J. Gonda; vol. 3, Sanskrit, grammatical and philological studies.
. 1975c. Selected studies / J. Gonda: presented to the author by
the staff of the Oriental Institute, Utrecht University, on the
occasion of his 70th birthday. 5 vols. Leiden: E. J. Brill.
. 1979. The medium in the Rgveda. Leiden: E. J. Brill.
Goto, Toshifumi. 2002. Funktionen

des Akkusativs und Rektionsarten des Verbums anhand des Altindoarischen. Indogermanische Syntax: Fragen und Perspektiven, ed. by Heinrich Hettrich and Jeong-Soo Kim, pp. 2142. Wiesbaden: L. Reichert.
. 2012. Grammatical irregularities in the R.gveda, book IV.
Indic across the millennia: from the Rigveda to Modern
Indo-Aryan: 14th World Sanskrit Conference, Kyoto, Japan,
420
S ANSKRIT SYNTAX
September 1st5th, 2009, proceedings of the Linguistic Section, ed. by Jared S. Klein and Kazuhiko Yoshida, pp. 2336.
Grace E. C., Jr. 1971. The order of constituents in IndoEuropean. Ph.D. dissertation. Austin: University of Texas.
Green, Alexander. 1966. The dative of agency: a chapter of IndoEuropean case syntax. New York: AMS Press.
Grgoire, A. 1899. Les infinitifs de lAvesta. Zeitschrift fr vergleichende Sprachforschung 35: 79140.
Gren-Eklund, Gunilla. 1978. A study of nominal sentences in the
oldest Upanis.ads. Ph.D. dissertation. Uppsala: Uppsala University.
Gune, Pandurang D. 1913. Die altindischen Absolutiva, besonders
im Rgveda, Aitareya- und Satapatha-br

ahman.a: ein Beitrag
zur altindischen Syntax. Leipzig: Kreysing.

Hackstein, Olav. 2003. Apposition and word-order typology in
Indo-European. Festschrift for Werner Winter, ed. by Brigitte
Bauer and Georges-Jean Pinault, pp. 13152. Berlin; New
York: Mouton de Gruyter.
. 2010. Apposition and nominal classification in Indo-European
and beyond. sterreichische Akademie der Wissenschaften,
Philophisch-Historische Klasse, Sitzungsberichte 798. Wien:
sterreichischen Akademie der Wissenschaften.
Hahn, E. Adelaide. 1950. The infinitive with subject accusative.
Transactions of the American Philological Association 81:
11729.
. 1953a. Some Hittite-Sanskrit parallels. Language 29: 242
54.
. 1953b. Subjunctive and optative: their origin as futures. Philological Monographs 16. New York: American Philological Association.
. 1964. Relative and antecedent. Transactions of the American
Philological Association 95: 11141.
B IBLIOGRAPHY
421
Hale, Mark Robert. 1987a. Deictic fronting in Vedic Prose. Paper

presented at the annual meeting of the American Linguistic
Society, San Francisco, December 1987. Unpublished.
. 1987b. Studies in the comparative syntax of the oldest IndoIranian languages. Ph.D. dissertation. Cambridge, Mass.:
Harvard University.
. 1987c. Wackernagels law and the language of the Rig Veda.
Studies in memory of Warren Cowgill, ed. by C. Watkins,
pp. 3850. Berlin: de Gruyter.
. 1988. The phonology-syntax interface in the language of the
Rig Veda. Paper presented at the Third Sanskrit Syntax Symposium, South Asian Languages Analysis Roundtable, Seattle,
WA, 1988. Unpublished.
. 1990. Preliminaries to the study of the relationship between
syntax and sandhi in Rigvedic Sanskrit. Mnchener Studien
zur Sprachwissenschaft 51: 7796.
. 1991. Some observations on intersentential pronominaliza a. Panels of the
tion in the language of the Taittirya Samhit
VIIth World Sanskrit Conference, Kern Institute, Leiden, August 2329, 1987; vol. 4, Sense and syntax in Vedic, ed. by
Joel P. Brereton and Stephanie W. Jamison, pp. 221.
. 1996. Deriving Wackernagels Law: prosodic and syntactic
factors determining clitic placement in the language of the
Rigveda. Approaching second: second position clitics and related phenomena, ed. by A. Halpern and A. Zwicky, pp. 165
97. Stanford: Center for the Study of Language and Information, Stanford University.
Hamp, Eric P. 1976. Why syntax needs phonology. Papers from
the Parasession on Diachronic Syntax, pp. 34864. Chicago:
Linguistic Society.
. 1984. The reconstruction of particles and syntax. Historical
syntax, ed. by J. Fisiak, pp. 17382. Berlin: Mouton.
422
S ANSKRIT SYNTAX
. 1985. Transitive and causative in Indo-European. Papers

from the Parasession on Causatives and Agentivity, ed. by
W. E. Eilfort et al., pp. 6466. Chicago: Linguistic Society.
Hartman, C. G. 1966. Emphasizing and connecting particles in the
thirteen principal upanishads. Suomalaisen Tiedeakatemian
Toimituksia. Annales Academi Scientiarum Fennic, SarjaSer. B 143.2. Ph.D. dissertation, University of Helsinki, 1966.
Helsinki: Suomalainen Tiedeakatemia.
Hartmann, P. 1955. Nominale Ausdrucksformen im wissenschaftlichen Sanskrit. Heidelberg: Winter.
Harweg, Roland. 1964. Kompositum und Katalysationstext
vornehmlich im spten Sanskrit. The Hague: Mouton.
Haudry, Jean. 1973. Parataxe, hypotaxe et corrlation dans la
phrase latine. Bulletin de la Socit de Linguistique de Paris
68.1: 14786.
. 1977. Lemploi des cas en vdique: introduction ltude des
cas en indo-europen. Les Hommes et les lettres 5. Lyon:
LHermes.
. 1980. Review of Klein 1978. Kratylos 25: 1225.
Havers, Wilhelm. 1911. Untersuchungen zur Kasussyntax der indogermanischen Sprachen. Untersuchungen zur indogermanischen sprach-und kulturwissenschaft 3. Straburg: Trbner.
. 1926. Der sog. Nominativus pendens. Indogermanische
Forschungen 43: 20757.
Hejib, Alaka and Arvind Sharma. 1976. The prohibitive use of
na in lieu of ma with the augmentless aorist in the RgVeda.
Adyar Library Bulletin 40: 10814.

. 1977. The negative particle na when used with the aorist system in the Rg-Veda. Folia Linguistica 11: 27780.
1944. Syntax of the infinite verb-forms of Pali.
Hendriksen, Hans.
Thesis of the faculty of philosophy of the University of Copenhagen. Copenhagen: Munksgaard.
B IBLIOGRAPHY
423
. 1947. A syntactic rule in Pali and Ardhamagadh. Acta Orientalia 20: 81106.
. 1948. On the meaning of Sanskrit bhavati. Acta Orientalia
20: 20615.
Hermann, Eduard. 1895. Gab es im Indogermanischen Nebenstze? Zeitschrift fr vergleichende Sprachforschung 33:
481535.
Herzog, E. 1873. Die Syntax des Infinitivs. Jahrbcher fr klassische Philologie: 133.
Hettrich, Heinrich. 1984. Zur historischen Syntax der nomina actionis im R.gveda: der doppelte Dativ. Mnchener Studien
zur Sprachwissenschaft 43: 55106.
. 1985. Review of Ch. Lehmann 1984. Kratylos 30: 4252.
. 1987. Zur Entwicklung der Finalstze altindogermanischer
Sprachen. Zeitschrift fr vergleichende Sprachforschung 100:
22137.
. 1988. Untersuchungen zur Hypotaxe im Vedischen. Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft 4. Berlin; New York: de Gruyter.
. 1990. Der Agens in passivischen Stzen altindogermanischer Sprachen. Nachrichten der Akademie der Wissenschaften
in Gttingen, Philologisch-historische Klasse 2. Gttingen:
Akademie der Wissenschaften.
. 1991. Syntax und Wortarten der Lokalpartikeln des R.gveda,
I: a dhi. Mnchener Studien zur Sprachwissenschaft 52: 27
76.
. 1993. Syntax und Wortarten der Lokalpartikeln des R.gveda. II: antar. Mnchener Studien zur Sprachwissenschaft 54:
14776. Appeared 1994.
. 1994. Semantische und syntaktische Betrachtungen zum doppelten Akkusativ. Frh-, Mittel-, Sptindogermanisch: Akten
der IX. Fachtagung der Indogermanischen Gesellschaft, ed. by
George E. Dunkel et al., pp. 11134. Wiesbaden: L. Reichert.
424
S ANSKRIT SYNTAX
. 1995. Zur funktionalen Variationsbreite altindogermanischer

Kasus. Verba et structurae: Festschrift fr Klaus Strunk zum
65. Geburtstag, ed. by Heinrich Hettrich et al., pp. 5372. Innsbruck: Institut fr Sprachwissenschaft.
. 2002a. Das Projekt einer Kasussyntax des Rgveda. Der In und Perspekstrumental. Indogermanische Syntax: Fragen
tiven, ed. by Heinrich Hettrich and Jeong-Soo Kim, pp. 43
. 2002b. Syntax und Wortarten der Lokalpartikeln des R.gveda.
III: pari. Novalis Indogermanica: Festschrift fr Gnter Neumann zum 80. Geburtstag, ed. by Matthias Fritz and Susanne
Zeilfelder, pp. 21542. Grazer Vergleichende Arbeiten 17.
Graz: Leykam.
. 2007. Materialien zu einer Kasussyntax des Rgveda.
Wrzburg: Institut fr Altertumswissenschaften.

. 2011. Konkurrierender Gebrauch obliquer Kasus im Rgveda. Indogermanistik und Linguistik im Dialog: Akten derXIII.
Fachtagung der Indogermanischen Gesellschaft, ed. by Th.
Krisch and Th. Lindner, pp. 22331. Wiesbaden: L. Reichert.
Hettrich, Heinrich, Antje Casaretto, and Carolin Schneider. 2010.
Syntax und Wortarten der Lokalpartikeln des Rgveda IV: All zur Sprachgemeines, II: u pa, III: a va. Mnchener Studien
wissenschaft 64: 17130.
Hiersche, Rolf. 1976. Zu ai. e d (aid) + Akk. siehe (da) in den
Brahman.as. Die Sprache 22: 13136.
Hinber, Oskar von. 1968. Studien zur Kasussyntax des Pali,
besonders des Vinaya-Pit.aka. Mnchener Studien zur Sprachwissenschaft Beihefte 2. Issued also as thesis, Mainz, 1966.
Mnchen.
Hirt, Hermann. 1937. Indogermanische Grammatik; vol. 7, Syntax. Indogermanische bibliothek. 1. Abt. Sammlung indogermanischer Lehr- und Handbcher. 1. Reihe: Grammatiken 13.
Heidelberg: Winter.
B IBLIOGRAPHY
425
Hock, Hans Henrich. 1975. Substratum influence on (Rig-Vedic)

Sanskrit? Studies in the Linguistic Sciences 5.2: 76125.
. 1981. Sanskrit causative syntax: a diachronic study. Studies
in the Linguistic Sciences 11.2: 933.
. 1982a. Clitic verbs in PIE or discourse-based verb fronting?:
Sanskrit sa hovaca ga rgyah. and congeners in Avestan and
Homeric Greek. Studies in the Linguistic Sciences 12.2: 1
38.
. 1982b. The Sanskrit passive: synchronic behavior and diachronic development. Studies in South Asian languages and
linguistics, ed. by P. J. Mistry, pp. 12737. South Asian Review
6.
. 1982c. The Sanskrit quotative: a historical and comparative
study. Studies in the Linguistic Sciences 12.2: 3985.
. 1983a. Language-death phenomena in Sanskrit: grammatical
evidence for attrition in contemporary spoken Sanskrit. Studies in the Linguistic Sciences 13.2: 2135.
. 1983b. Pronominal reflexivization in Sanskrit. Paper presented at the South Asian Languages Analysis Roundtable,
1983, Syracuse, NY.
. 1984. (Pre-)Rig-Vedic convergence of Indo-Aryan with Dravidian? Another look at the evidence. Studies in the Linguistic
Sciences 14.1: 89108.
. 1985a. Pronoun fronting and the notion verb-second position in Beowulf. Germanic linguistics: papers from a symposium at the University of Chicago, April 24, 1985, ed. by
J. T. Faarlund, pp. 7086. Bloomington: Indiana University
Linguistics Club.
. 1985b. Sanskrit double-object constructions: Will the real object please stand up? Prac-Bhas.a -Vijana: Indian Journal of
Linguistics 12: 5070. Calcutta: West Bengal Institute of Linguistics.
426
S ANSKRIT SYNTAX
. 1985c. Transitivity as a gradient feature? Evidence from

Indo-Aryan, especially Sanskrit and Hindi. Proceedings of the
Conference on Participant Roles: South Asia and adjacent areas, ed. by A. R. K. Zide et al., pp. 24763. Bloomington:
Indiana University Linguistics Club.
. 1986a. P-oriented constructions in Sanskrit. South Asian
languages: Structure, convergence, and diglossia, ed. by
Bh. Krishnamurti et al., pp. 1526. MLBD series in linguistics 3. Delhi: Motilal Banarsidass.
. 1986b. Principles of historical linguistics. Berlin; New York:
Mouton de Gruyter.
. 1986c. Review of Meenakshi 1983. Journal of the American
Oriental Society 106.2: pp. 345346.
. 1987. Reduced-clause and clause-union absolutives and participles in Vedic Prose. Select papers from SALA-7, ed. by
Elena Bashir et al., pp. 18298. Bloomington: Indiana University Linguistics Club.
. 1988. Voice, mood, and the gerundive (krtya) in Sanskrit. Indologica Taurinensia 13: 81102. Torino:Istituto di indologia.
. 1989a. Conjoined we stand: Theoretical implications of Sanskrit relative clauses. Studies in the Linguistic Sciences 19.1:
93126.
. 1989b. Research on Sanskrit syntax: a status report. New
horizons of research in Indology: silver jubilee volume, ed. by
Vashishtha Narayan Jha, pp. 90107. 1st ed. Publications of
the Centre of Advanced Study in Sanskrit, Class E 10. Pune:
Centre for Advanced Study in Sanskrit, University of Pune.
. 1990. Oblique subjects in Sanskrit? Experiencer Subjects
in South Asian Languages: proceedings of the Conference on
the Experiencer Subject in South Asian Languages, November 1988, Madison, Wisconsin, ed. by Manindra K. Verma and
K. P. Mohanan, pp. 11939.
B IBLIOGRAPHY
427
. 1991a. Causees, passive agents, or instruments? Instrumental

NPs with causatives in early and later Vedic Prose. Studies
in Sanskrit syntax: a volume in honour of the centennial of
Speijers Sanskrit syntax (18861986), ed. by Hans Henrich
Hock, pp. 7194.
. 1991b. Possessive agents in Sanskrit? Studies in Sanskrit
syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986), ed. by Hans Henrich Hock, pp. 55
70.
. ed. 1991c. Studies in Sanskrit syntax: a volume in honour of
the centennial of Speijers Sanskrit syntax (18861986). Delhi:
. 1993. Some peculiarities of Vedic-prose relative clauses.
Wiener Zeitschrift fr die Kunde Sdasiens 35; Supplement
1993, pp. 929.
. 1994a. Discourse linkage in Sanskrit narratives with special
emphasis on the story of Nala. Papers from the Fifteenth
South Asian Language Analysis Roundtable, 1993, ed. by Alice Davison and Frederick M. Smith, pp. 11739. Iowa City,
IA: South Asian Studies Program.
. 1994b. Narrative linkage in the Mahabharata. Modern evaluation of the Mahabharata: Prof. R. K. Sharma felicitation volume, ed. by S. P. Narang, pp. 295313. Delhi: Nag Publishers.
. 1996. Whos On First: toward a prosodic account of P2 clitics. Approaching second: second position clitics and related
phenomena, ed. by A. Halpern and A. Zwicky, pp. 199270.
Stanford: Center for the Study of Language and Information,
Stanford University.
. 1997a. Chronology or genre? Problems in Vedic syntax. Inside the texts beyond the texts: new approaches to the study
of the Vedas, ed. by Michael Witzel, pp. 10326. Harvard Oriental Series, Opera Minora 2. Cambridge, Mass.: Harvard University.
428
S ANSKRIT SYNTAX
. 1997b. Nexus and extraclausality in Vedic, or sa-fig

all over again: a historical (re)examination. Historical,
Indo-European, and lexicographical studies: a festschrift for
Ladislav Zgusta on the occasion of his 70th birthday, ed. by
Hans Henrich Hock, pp. 4978. Berlin; New York: Mouton de
Gruyter.
. 2000. Genre, discourse, and syntax in Sanskrit. Textual parameters in older languages, ed. by S. Herring, P. van Reenen,
and L. Schsler, pp. 16395. Amsterdam; Philadelphia: John
Benjamins.
. 2002. Vedic e ta . . . stavama: subordinate, coordinate, or
what? Indo-European perspectives, ed. by Mark R. V. Southern, vol. 43, pp. 89102. Journal of Indo-European Studies;
monograph. Washington, DC: Institute for the Study of Man.
. 2005. How strict is Strict OV?: a family of typological constraints with focus on South Asia. Yearbook of South Asian
Languages and Linguistics 2005, ed. by Rajendra Singh and
Tanmoy Bhattacharya, pp. 14563. New Delhi; Thousand
Oaks: Sage Publications.
. 2006a. Reflexivization in the Rig-Veda (and beyond). Papers of the 12th World Sanskrit Conference, held in Helsinki,
Finland, 1318 July, 2003; vol. 5, Themes and tasks in Old
and Middle Indo-Aryan linguistics, ed. by Bertil Tikkanen and
Heinrich Hettrich, pp. 1944.
. 2006b. Ta-participles with genitive agents in Vedic Sanskrit:
Was Pan.ini right? Sanskrit Studies Centre Journal 2: 2449.
Bangkok: Silpakorn University.
. 2007. A short history of Vedic prefix-verb compound accentuation. Paper presented at the Texas Vedic Workshop 2007.
Unpublished.
. 2009a. Some aspects of Sanskrit agreement. Paper presented
at the 28th South Asian Languages Analysis Roundtable, 911
October 2009, University of North Texas. Unpublished.
B IBLIOGRAPHY
429
. 2009b. Stmme oder Wurzeln im Sanskrit? Primre vs.

sekundre Verbalstammbildung und das Kausativ. International Conference on Morphology and Digitisation, ed. by Jost
Gippert, pp. 6380. stav srovnvac jazykovedy Chatrear
2009. Prague.
. 2012a. Issues in Sanskrit agreement. Indic across the millennia: from the Rigveda to Modern Indo-Aryan: 14th World
Sanskrit Conference, Kyoto, Japan, September 1st5th, 2009,
proceedings of the Linguistic Section, ed. by Jared S. Klein and
Kazuhiko Yoshida, pp. 4958.
. 2012b. Sanskrit and Pan.ini Core and periphery. Samskr
ta
Vimarsa, new series 6; World Sanskrit Conference Special,
pp.
85102. New Delhi: Rashtriya Sanskrit Sansthan.
. 2012c. Some notes on Indo-European double direct-object
constructions. Historische Sprachforschung 125; Von Fall zu
Fall: Studien zur indogermanischen Syntax, ed. by Alfred
Bammesberger, Olav Hackstein, and Sabine Ziegler, pp. 151
64. Appeared 2014.
. 2013. Proto-Indo-European verb-finality: reconstruction, typology, validation. Journal of Historical Linguistics 3; ProtoIndo-European syntax and its development: proceedings of the
Workshop on Proto-Indo-European Syntax and its Development, ed. by Leonid. I. Kulikov and Nikolas Lavidas, pp. 49
76. Amsterdam; Philadelphia: John Benjamins.
. 2014a. Predicate order in Vedic Prose. Proceedings of the
15th World Sanskrit Conference; vol. 1, Vedic studies: language, texts, culture, and philosophy: proceedings of the Veda
Section, 15th World Sanskrit Conference, New Delhi, 2012, ed.
by Hans Henrich Hock, pp. 517.
. ed. 2014b. Proceedings of the 15th World Sanskrit Conference; vol. 1, Vedic studies: language, texts, culture, and philosophy: proceedings of the Veda Section, 15th World Sanskrit
430
S ANSKRIT SYNTAX
Conference, New Delhi, 2012. New Delhi: Rashriya Sanskrit

Sansthan and D. K. Printworld.
. 2014c. Vedic verb accent revisited. Papers of the 13th World
Sanskrit Conference, Edinburgh, 1014 July 2006; vol. 3,
Vedic and Sanskrit historical linguistics, ed. by Jared S. Klein
and Elizabeth Tucker, pp. 15378.
. Forthcoming(a). Narrative linkage in Sanskrit. To be submitted to a festschrift (2014). In preparation.
. Forthcoming(b). The relation of the Indian grammatical tradition to modern linguistics. Bulletin dtudes Indiennes; Proceedings of the Panels on the History of Indian Grammatical
Traditions, Potsdam, 2008, ed. by Emilie Aussant and JeanLuc Chevillard, pp. Forthcoming.
Hoenigswald, Henry M. 1980. Notes on reconstruction, word order, and stress. Linguistic reconstruction and Indo-European
syntax: proceedings of the colloquium of the Indogermanische
Gesellschaft, University of Pavia, ed. by Paolo Ramat, Onofrio
Carruba, Anna Giacalone Ramat, and Giorgio Graffi, pp. 609
87. Amsterdam; Philadelphia: John Benjamins.
Hfer, K. 1840. Vom Infinitiv, besonders im Sanskrit: eine
etymologisch-syntactische abhandlung als probe einer sanskritsyntax. Berlin: Vof.
Hoffmann, Karl. 19521956. Zum prdikativen Adverb. Mnchener Studien zur Sprachwissenschaft 1: 5460. [Reprinted in
Hoffmann 19751976.]
. 1956. Zur Parenthese im Altpersischen. Mnchener Studien
zur Sprachwissenschaft 9: 7986. [Reprinted in Hoffmann
19751976.]
. 1961. Die Ortsnamen-Parenthese im Altpersischen und
Vedischen. Zeitschrift der Deutschen Morgenlndischen
Gesellschaft 110: 6473. Wiesbaden: Kommissionsverlag F.
Steiner. [Reprinted in Hoffmann 19751976.]
B IBLIOGRAPHY
431
. 1967. Der Injunktiv im Veda: eine synchronische Funktionsuntersuchung. Indogermanische Bibliothek, Dritte Reihe, Untersuchungen. Heidelberg: Winter.
. 19751976. Aufstze zur Indoiranistik. 3 vols. Vol. 3 published
as a festschrift for Karl Hoffmanns 75th birthday, and edited
by Sonja Glauch, Robert Plath, and Sabine Ziegler. Wiesbaden: L. Reichert.
Holland, Gary B. 1984. Subordination and relativization in Early
Indo-European. Proceedings of the tenth annual meeting of
the Berkeley Linguistic Society, February 1720, 1984, ed. by
Claudia Brugman, pp. 60922. Berkeley: Berkeley Linguistics
Society.
. 1986a. A constraint on the position of the relative in Vedic
Sanskrit. Paper presented at the First Sanskrit Syntax Symposium, South Asian Languages Analysis Roundtable, Urbana,
IL.
. 1986b. Nominal sentences and the origin of absolute constructions in Indo-European. Zeitschrift fr vergleichende
Sprachforschung 99.2: 16393. Gttingen: Vandenhoeck and
Ruprecht.
. 1991. Definiteness and relativization in the Rigveda. Panels
of the VIIth World Sanskrit Conference, Kern Institute, Leiden,
August 2329, 1987; vol. 4, Sense and syntax in Vedic, ed. by
Joel P. Brereton and Stephanie W. Jamison, pp. 2232.
. 1996. Relativization and word order in Old Irish and Vedic
Sanskrit. Journal of Indo-European Studies 24: 32329.
Holland, Gary B. and Nancy Ickler. 1978. Some observations on
relatives and demonstratives in Greek and Sanskrit. Proceedings of the fourth annual meeting of the Berkeley Linguistic
Society, ed. by Jeri J. Jaeger, pp. 43245. Berkeley: Berkeley
Linguistics Society.
Holtzmann, Adolf. 1884. Grammatisches aus dem Mahabharata.
Leipzig: Breitkopf and Hrtel.
432
S ANSKRIT SYNTAX
Hook, Peter Edwin. 1976. As.t.a dhyay 3.4.21 and the role of semantics in Paninian linguistics. Papers from the twelfth regional meeting of the Chicago Linguistic Society, ed. by Salikoko S. Mufwene, Carol A. Walker, and Sanford B. Steever,
pp. 30212. Chicago: Chicago Linguistic Society.
. 1980. As.t.a dhyay 3.3.158 and the notion of subject in Pan.ini.
Revue Roumaine de Linguistique 25: 7987.
. 1984a. Kasmrasabdamrta 8.3.3: an account of the ergative in
the Paninian linguistic tradition.

Indian Linguistics 44: 3942.
. 1984b. Paninis as.t.a dhyay: a two-storey house for a threestorey language? Paper presented at the VIth World Sanskrit
Conference, 1320 October 1984, Philadelphia, Pennsylvania.
. 1985. Coexistent analyses and participant roles in IndoAryan. Proceedings of the Conference on Participant Roles:
South Asia and adjacent areas, ed. by A. R. K. Zide et
al., pp. 26483. Bloomington: Indiana University Linguistics
Club.
. 1991. On identifying the conceptual restructuring of passive
as ergative in Indo-Aryan. Pan.inian studies: S. D. Joshi felicitation volume, ed. by Madhav Murlidhar Deshpande and
Saroja Bhate, pp. 17799. Ann Arbor: Center for South and
Southeast Asian Studies, University of Michigan.
. 1995. Covert absolutivity in Vedic. Sthapakasraddha: a
commemorative volume for Prof. Georgij A. Zograph, ed. by
N. V. Gurov and Y. V. Vasilkov, pp. 8293. St. Petersburg:
Russian Academy of Sciences, Institute of Oriental Studies.
Hopkins, E. W. 1892. The Aryan future. American Journal
of Philology 13: 150. Baltimore: Johns Hopkins University
Press.
. 1906. The Vedic dative reconsidered. Transactions and Proceedings of the American Philological Association 37: 87120.
. 1907. Aspects of the Vedic dative. Journal of the American
Oriental Society 28: 360406.
B IBLIOGRAPHY
433
. 1918. The origin of the ablative case. Journal of the American Oriental Society 38: 4759.
Humbach, Helmut. 1982. Vedic indrasca vayo Old Avestan mazdsca ahurNho. Mnchener Studien zur Sprachwissenschaft 41: 95102.
Ickler, Ingeborg. 1973. Untersuchungen zur Wortstellung und Syntax der Chandogyopanis.ad. Gppinger Akademische Beitrge
75. Gppingen: Kmmerle.
. 1976. Die vedische Partikel kila. Zeitschrift fr vergleichende Sprachforschung 90: 5086.
Insler, Stanley. The origin of the Sanskrit passive aorist. Indogermanische Forschungen 73: 31246.
. 1998. Mitra varun.a or mitra varun.a ? Mr Curad: Festschrift
for Calvert Watkins, ed. by J. Jasonoff, C. Melchert, and L.
Oliver, pp. 28590. Innsbruck: Institut fr Sprachwissenschaft.
Jacobi, Hermann. 1895. Die Inversion von Subjekt und Prdikat
im Indischen. Indogermanische Forschungen 5: 3358.
. 1896. Zur Entwicklung des indischen Satzbaus. Indogermanischen Forschungen 6: 15256. Abstract in A. Thumbs report on Die indogermanische Sektion auf der Klner Philologenversammlung. Anzeiger fr indogermanische Sprach- und
Altertumskunde, Beiblatt zu den Indogermanischen Forschungen.
. 1897. Compositum und Nebensatz. Bonn: Cohen.
. 1899. ber das periphrastische perfect im Sanskrit.
. 1903. ber den nominalen Stil des wissenschaftlichen Sanskrit. Indogermanische Forschungen 14: 23651.
Jamison, Stephanie W. 1976. Functional ambiguity and syntactic
change: the Sanskrit accusative. Papers from the Parasession
on Diachronic Syntax, pp. 12635. Chicago: Linguistic Society.
434
S ANSKRIT SYNTAX
. 1977. Function and form in the -aya-formations of the

Rigveda and Atharvaveda. Ph.D. dissertation. New Haven:
Yale University. [Revised as Jamison 1983.]
. 1979a. Remarks on the expression of agency with the passive in Vedic and Indo-European. Zeitschrift fr vergleichende
. 1979b. The case of the agent in Indo-European. Die Sprache
25: 12943.
. 1982. Case disharmony in R.gvedic similes. Indo-Iranian
Journal 24: 25171.
. 1983. Function and form in the -aya-formations of the Rigveda
and Atharvaveda. Ergnzungshefte zur Zeitschrift fr vergleichende Sprachforschung 13. Gttingen: Vandenhoeck and
Ruprecht. [Revision of Jamison 1977.]
. 1984. The Vedic passive optative and its functional equivalents: a study of the syntax of the gerundive. Journal of the
. 1988. Some intersections of semantics and syntax in Vedic.
Paper presented at the Third Sanskrit Syntax Symposium,
South Asian Languages Analysis Roundtable, Seattle, WA,
1988. Unpublished.
. 1991a. The syntax of direct speech in Vedic. Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986), ed. by Hans Henrich Hock,
pp. 95112. [Also published as Jamison 1991b.]
. 1991b. The syntax of direct speech in Vedic. Panels of the
VIIth World Sanskrit Conference, Kern Institute, Leiden, August 2329, 1987; vol. 4, Sense and syntax in Vedic, ed. by
Joel P. Brereton and Stephanie W. Jamison, pp. 4061. [Also
published as Jamison 1991a.]
. 1997. Conjunctive u and invariable sa in the Rgveda: questions of method (with a note on Greek dialectal
and ).
Indogermanische Forschungen 102: 15678.
B IBLIOGRAPHY
435
Jeffers, R. 1972. The infinitives of the Indo-European languages.

Ph.D. dissertation. Ithaca, NY: Department of Linguistics,
Cornell University.
. 1975. Remarks on Indo-European infinitives. Language 51:
13348.
. 1987. On methodology in syntactic reconstruction: reconstructing inter-clause syntax in prehistoric Indo-European.
Papers from the 7th International Conference on Historical
Linguistics, ed. by A. G. Ramat et al., pp. 30523. Amsterdam: John Benjamins.
Jeffers, R. and R. Kantor. 1984. A history of the Sanskrit gerund.
Indogermanische Forschungen 89: 89103.
Jeffers, R. and W. J. Pepicello. 1979. The expression of purpose
in Indo-European. Indogermanische Forschungen 84: 118.
Job, Michael. 1994. Zur Funktion des Perfekts im Rgveda. In Schmidt
dogermanica et caucasica: Festschrift fr Karl Horst
zum 65. Geburtstag, ed. by Roland Bielmeier and Reinhard
Stempel, pp. 4162. Berlin; New York: de Gruyter.
Johnson, David E. 1974. Towards a relationally based grammar.
Ph.D. dissertation. Urbana: University of Illinois.
Jolly, Julius. 1872. Ein Kapitel vergleichender Syntax: der Conjunctiv und Optativ und die Nebenstze im Zend und Altpersischen in Vergleich mit dem Sanskrit und Griechischen. Mnchen: T. Ackermann.
. 1873. Geschichte des Infinitivs im Indogermanischen. Mnchen: Ackermann.
Joseph, Brian D. 1981. Hittite iwar, wa(r) and Sanskrit iva.
. 1991. A diachronic phonological solution to the syntax of
Vedic negative particles. Studies in Sanskrit syntax: a volume
in honour of the centennial of Speijers Sanskrit syntax (1886
1986), ed. by Hans Henrich Hock, pp. 11322.
436
S ANSKRIT SYNTAX
Joshi, S. D. 1969. Sentence structure according to Pan.ini. Indian

Antiquary, 3rd ser. 3: 1326.
Kale, Moreshwar Ramchandra. 1894. A higher Sanskrit grammar:
for the use of schools and colleges. 1st edition. Bombay: Gopal
Narayen & Co. [Revised as Kale 1898.]
. 1898. A higher Sanskrit grammar. Bombay: Gopal Narayen &
Co. [Revision of Kale 1894. Reprinted with different pagination: Delhi: Motilal Banarsidass, 1961.]
Keith, A. Berriedale. 1906. The use of the gerund as passive in
Sanskrit. Journal of the Royal Asiatic Society 38.3: 69395.
. 1907. The use of the passive gerund in Sanskrit. Journal of
the Royal Asiatic Society 39.1: 16466.
. 1908. An unusual use of the nominative. Journal of the Royal
Asiatic Society 40.4: 112427.
. 1909a. Notes on syntax. Journal of the Royal Asiatic Society
41.2: 42832.
. 1909b. Participles as finite verbs. Zeitschrift der Deutschen
Morgenlndischen Gesellschaft 63: 34649.
a. Journal
. 1909c. Tenses and moods in the Kat.haka Samhit
of the Royal Asiatic Society 1909: 14954.
. 1910a. Grammatical notes. Journal of the Royal Asiatic Society 42.1: 15159.
. 1910b. Peculiarities in the use of iti. Journal of the Royal
Asiatic Society 42.4: 131721.
. 1912. Notes on Vedic syntax. Journal of the Royal Asiatic
Society 44.3: 72126.
Keydana, Gtz. 1997. Absolute Konstruktionen in altindogermanischen Sprachen. Gttingen: Vandenhoeck and Ruprecht.
. 2000. Prdikativa im Altindischen. Vividharatnakaran.d.aka:
Festgabe fr Adelheid Mette, ed. by Ch. Chojnacki et al.,
pp. 36978. Swisttal-Odendorf: Indica et Tibetica Verlag.
. 2009. Latente Objekte und altindische Diskursgrammatik.
Pragmatische Kategorien: Form, Funktion und Diachronie,
B IBLIOGRAPHY
437
ed. by E. Rieken and P. Widmer, pp. 12542. Wiesbaden: L.

Reichert.
. 2011. Wackernagel in the language of the Rigveda: a reassessment. Historische Sprachforschung 124: 80107.
Kieckers, Ernst. 19151916a. Zur oratio recta in den indogermanischen Sprachen 1. Indogermanische Forschungen 35: 193.
. 19151916b. Zur oratio recta in den indogermanischen Sprachen 2. Indogermanische Forschungen 36: 170.
Kiparsky, Paul. 1968. Tense and mood in Indo-European syntax.
Foundations of Language 4: 3057.
. 1982. Some theoretical problems in Pan.inis grammar. Postgraduate and Research Department series 16. Pune: Bhandarkar Oriental Research Institute.
. 1995. Indo-European origins of Germanic syntax. Clause
structure and language change, ed. by A. Battye and I.
Roberts, pp. 14069. Oxford; New York: Oxford University
Press.
. 1998. Aspect and event structure in Vedic. Yearbook of South
Asian Languages and Linguistics 1: 2961. URL: http :
/ / www . stanford . edu / ~kiparsky / Papers /
semanticsfest.pdf.
. 2005. The Vedic injunctive: historical and synchronic implications. The Yearbook of South Asian Languages and Linguistics 2005, ed. by Tanmoy Bhattacharya, pp. 21936. Berlin;
New York: Mouton de Gruyter.
. 2009. On the architecture of Pan.inis grammar. Sanskrit
computational linguistics: first and second international symposia, Rocquencourt, France, October 2007; Providence, RI,
USA, May 2008; Revised selected and invited papers, ed. by
Grard Huet, Amba Kulkarni, and Peter M. Scharf, pp. 3394.
Lecture Notes in Artificial Intelligence 5402.
. 2010. Dvandvas, blocking, and the associative: the bumpy
ride from phrase to word. Language 86: 30231.
438
S ANSKRIT SYNTAX
Kiparsky, Paul and J. F. Staal. 1969. Syntactic and semantic relations in Panini. Foundations of Language 5: 83117.
Klaiman, M. H. 1976. Correlative clauses and IE syntactic reconstruction. Papers from the Parasession on Diachronic Syntax,
pp. 15968. Chicago: Linguistic Society.
. 1978. Arguments against a passive origin of the IA ergative.
Papers from the 4th Regional Meeting of the Chicago Linguistic Society, pp. 20416.
. 1981. The diachronic relationship of the IA ergative and passive constructions. Syntactic change, ed. by B. B. Johns and
D. R. Strong, pp. 13557. Ann Arbor: Department of Linguistics, University of Michigan.
. 1987. The Sanskrit passive in lexical-functional grammar.
Papers from the 23rd Regional Meeting of the Chicago Linguistic Society, pp. 196214.
Klein, Jared S. 1978a. The diachronic syntax of the particle u
in the Rigveda. Journal of the American Oriental Society 98:
26676.
. 1978b. The particle u in the Rigveda: a synchronic and diachronic study. Ergnzungshefte zur Zeitschrift fr vergleichende Sprachforschung 27. Gttingen: Vandenhoeck and
Ruprecht.
. 1978c. The system of coordinate conjunctions in the
Rigveda. Indo-Iranian Journal 20: 123.
a, a dha, and a typology of Rigvedic conjunction.
. 1980. Ath
Indo-Iranian Journal 22: 195219.
. 1981. The origin of the Rigvedic va yav ndras ca construction. Mnchener Studien zur Sprachwissenschaft 40: 7391.
. 1982a. Rigvedic tu and su. Die Sprache 28: 126.
. 1982b. Sanskrit ca, Indo-European *kw e, and the semantics
of coordinate conjoined structures in the Rigveda. Studies in
South Asian languages and linguistics, ed. by P. J. Mistry,
pp. 6577. South Asian Review 6.
B IBLIOGRAPHY
439
. 1985a. Toward a discourse grammar of the Rigveda; vol. 1,

Coordinate conjunction. 2 parts. Indogermanische Bibliothek
1. Heidelberg: Winter.
. 1985b. The origin and syntax of the Rigvedic construction ya(. . . ) ka//u- ca. Mnchener Studien zur Sprachwissenschaft
44: 10521.
. 1987. The two senses of the term anaphora and their functional unity: evidence from the Rigveda. Festschrift for Henry
Hoenigswald on the occasion of his seventieth birthday, ed. by
George Cardona and N. H. Zide, pp. 19299. Tbingen: Narr.
. 1991a. Syntactic and discourse correlates of verb-initial sentences in the Rigveda. Studies in Sanskrit syntax: a volume
in honour of the centennial of Speijers Sanskrit syntax (1886
1986), ed. by Hans Henrich Hock, pp. 12343. [Also published
as Klein 1991b.]
. 1991b. Syntactic and discourse correlates of verb-initial sentences in the Rigveda. Panels of the VIIth World Sanskrit
Conference, Kern Institute, Leiden, August 2329, 1987; vol.
4, Sense and syntax in Vedic, ed. by Joel P. Brereton and
Stephanie W. Jamison, pp. 6680. [Also published as Klein
1991a.]
. 1992. On verbal accentuation in the Rigveda. American Oriental Society Essay 11. New Haven, CT: American Oriental
Society.
. 1996. Sa fig and Indo-European deixis. Historische Sprachforschung 109: 2139.
. 1999. A triadic structural feature of nominal Anaphora in the
Rigveda. Gering und doch von Herzen: 25 Indogermanistische Beitrge, Bernhard Forssman zum 65. Geburtstag, ed. by
J. Habisreitinger, R. Plath, and S. Ziegler, pp. 11934. Wiesbaden: L. Reichert.
440
S ANSKRIT SYNTAX
. 2000. Polyptoton and paronomasia in the Rigveda. Anusantatyai: Festschrift fr Johanna Narten zum 70. Geburtstag, ed.
by Almut Hintze and Eva Tichy, pp. 13555. Dettelbach: Rll.
. 2002a. Homoioteleuton in the Rigveda. Indo-European perspectives, ed. by Mark R. V. Southern, vol. 43, pp. 199222.
Journal of Indo-European Studies; monograph. Washington,
DC: Institute for the Study of Man.
. 2002b. Responsion in the Rigveda. Journal of the American
Oriental Society 122: 31117.
. 2002c. The syntax of style: a typology of word and
morpheme-level repetition in the Rigveda. Indogermanische
Syntax: Fragen und Perspektiven, ed. by Heinrich Hettrich and
Jeong-Soo Kim, pp. 23148. Wiesbaden: L. Reichert.
. 2004a. Amreditas and related constellations in the Rigveda.
Journal of the American Oriental Society 123: 773802.
. 2004b. Nominal and adverbial amreditas and the etymology
of Rgvedic nana. The Vedas: text, language & ritual: proceedings of the Third International Vedic Workshop, Leiden
2002, ed. by Arlo Griffiths and Jan E. M. Houben, pp. 25160.
Groningen: Egbert Fortsen.
. 2004c. Patterns of verb repetition in the Rigveda. General
Linguistics 41: 22256.
. 2004d. Phrasal repetition in the Rigveda. Bulletin dtudes
indiennes 21: 745.
. 2006. Aspects of the rhetorical poetics of the Rigveda. La
langue potique indo-europenne: Actes du Colloque de travail de la Socit des tudes Indo-Europennes, Paris, 2224
octobre 2003, ed. by Georges-Jean Pinault and Daniel Petit,
pp. 195211. Leuven-Paris: Peeters.
. 2007a. On the nature and function of preverb repetition in the
Rigveda. Studien zur Indologie und Iranistik 24: 91103.
B IBLIOGRAPHY
441
. 2007b. Relative pronoun sequences in the Rigveda: a syntactic, rhetorical-poetic, and discourse-level study. Bulletin
dtudes indiennes 22-23: 495536.
. 2007c. Sequential negation in the Rigveda. Verba docenti:
studies in historical and Indo-European linguistics presented
to Jay H. Jasanoff by students, colleagues, and friends, ed. by
Alan Nussbaum, pp. 21329. Ann Arbor: Beech Stave Press.
. 2008a. Adverbial repetition in the Rigveda. Indologica: T.
Ya. Elizarenkova memorial volume, ed. by L. Kulikov and M.
Rusanov, vol. 1, pp. 16782. Orientalia et Classica: Papers of
the Institute of Oriental and Classical Studies 20. Moscow:
Russian State University for the Humanities.
. 2008b. Interrogative sequences in the Rigveda. Evidence
and counter-evidence: essays in honour of F. Kortlandt, ed.
by Alexander Lubotsky et al., vol. 1, pp. 297308. Studies
in Slavic and general linguistics 32. Amsterdam; New York:
Rodopi.
. 2008c. Numeral repetition in the Rigveda. Proceedings of
the Nineteenth Annual UCLA Indo-European Conference, ed.
by Karlene Jones-Bley et al., vol. 54, pp. 17191. Journal of
Indo-European Studies; monograph. Washington: Institute for
the Study of Man.
. 2010. Personal pronoun sequences in the Rigveda. Ex Anatolia lux: Anatolian and Indo-European studies in honor of
H. Craig Melchert on the occasion of his sixty-fifth birthday,
ed. by Ronald Kim et al., pp. 20416. Ann Arbor; New York:
Beech Stave Press.
. 2012. Deictic pronoun repetition in the Rigveda. Indic across
2009, proceedings of the Linguistic Section, ed. by Jared S.
Klein and Kazuhiko Yoshida, pp. 5976.
442
S ANSKRIT SYNTAX
. 2014a. Conjunctive sequences in the Rigveda. Papers of the

13th World Sanskrit Conference, Edinburgh, 1014 July 2006;
vol. 3, Vedic and Sanskrit historical linguistics, ed. by Jared S.
Klein and Elizabeth Tucker, pp. 26592.
. 2014b. Interstanzaic repetition in the Rigveda. Proceedings
of the 15th World Sanskrit Conference; vol. 1, Vedic studies: language, texts, culture, and philosophy: proceedings of
the Veda Section, 15th World Sanskrit Conference, New Delhi,
2012, ed. by Hans Henrich Hock, pp. 1838.
. 2014c. Some rhetorical aspects of adjacent interstanzaic
phrasal repetition in the Rigveda. Grammatica et verba
Glamour and verve: studies in South Asian, historical, and
Indo-European linguistics in honor of Hans Henrich Hock on
the occasion of his seventy-fifth birthday, ed. by Shu-Fen Chen
and Benjamin Slade, pp. 12844.
Klein, Jared S. and Elizabeth Tucker, eds. 2014. Papers of the 13th
World Sanskrit Conference, Edinburgh, 1014 July 2006; vol.
3, Vedic and Sanskrit historical linguistics. Delhi: Motilal Banarsidass.
Klein, Jared S. and Kazuhiko Yoshida, eds. 2012. Indic across
2009, proceedings of the Linguistic Section. Bremen: Ute
Hempen Verlag.
Knobl, Werner. 2006. Studies on the present participle, 1: the
present participle expressive of intentionality. Journal of Indological Studies 1617: 65108.
Kobayashi, Masato. 2012. Information structure and the particles
va and eva in Vedic Prose. Indic across the millennia: from
the Rigveda to Modern Indo-Aryan: 14th World Sanskrit Conference, Kyoto, Japan, September 1st5th, 2009, proceedings
of the Linguistic Section, ed. by Jared S. Klein and Kazuhiko
Yoshida, pp. 7792.
B IBLIOGRAPHY
443
Krisch, Thomas. 1990. Das Wackernagelsche Gesetz aus heutiger

Sicht. Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute: Kolloquium der Indogermanischen Gesellschaft, Basel, 1988, ed. by Heiner Eichner
and Helmut Rix, pp. 6481. Wiesbaden: L. Reichert.
. 1994. Betrachtungen zur Wortstellung im Altindischen.
Frh-, Mittel-, Sptindogermanisch: Akten der IX. Fachtagung
der Indogermanischen Gesellschaft, ed. by George E. Dunkel
et al., pp. 16983. Wiesbaden: L. Reichert.
. 1996. Zur Genese und Funktion der altindischen Perfekta
mit langem Reduplikationsvokal: mit kommentierter Materialsammlung. Innsbruck: Institut fr Sprachwissenschaft.
. 1997. B. Delbrcks Arbeiten zur Wortstellung aus heutiger
Sicht. Berthold Delbrck y la sintaxis indoeuropea hoy: actas del coloquio de la Indogermanische Gesellschaft 1994,
Madrid, ed. by E. Crespo and J.-L. Garca-Ramn, pp. 283
. 2002. Indogermanische Wortstellung. Indogermanische Syntax: Fragen und Perspektiven, ed. by Heinrich Hettrich and
Jeong-Soo Kim, pp. 24961. Wiesbaden: L. Reichert.
. 2009. On the syntax of silence in Proto-Indo-European.
Information theory and language change: new approaches
to word order variation, ed. by R. Hinterhlzl and Svetlana
Petrova, pp. 191220. Berlin; New York: Mouton de Gruyter.
. 2011. Some remarks on the position of adverbials in Greek
and Vedic sentences. Indogermanistik und Linguistik im
Dialog: Akten der XIII. Fachtagung der Indogermanischen
Gesellschaft, ed. by Th. Krisch and Th. Lindner, pp. 3009.
Wiesbaden: L. Reichert.
Kuiper, F. B. J. 1974. The genesis of a linguistic area. International Journal of Dravidian Linguistics 3: 13553.
Kulikov, Leonid. 1997. Vedic mriyate and other pseudo-passives:
note on an accent shift. Indo-European, Nostratic, and be-
444
.
.
.
.
S ANSKRIT SYNTAX
yond: Festschrift for V. V. Shevoroshkin, ed. by I. Hegedus
et al., pp. 198205. Journal of Indo-European Studies; monograph 22. Washington: Institute for the Study of Man.
1998. Passive, anticausative, and classification of verbs: the
case of Vedic. Typology of verbal categories: papers presented to Vladimir Nedjalkov on the occasion of his 70th birthday, ed. by L. Kulikov and H. Vater, pp. 13953. Tbingen:
Niemeyer.
2001. The Vedic -ya-presents. Proefschrift. Leiden: Universiteit Leiden.
2006a. Passive and middle in Indo-European: reconstructing
the early Vedic passive paradigm. Passivization and typology:
Form and function, ed. by Werner Abraham and Larisa Leisl,
2006b. The Sanskrit -yet-optative: a formation not yet
recorded in Sanskrit grammars. Wiener Zeitschrift fr die
Kunde Sdasiens 50: 2768.
2006c. The Vedic medio-passive aorists, statives and their
participles: reconsidering the paradigm. Papers of the 12th
World Sanskrit Conference, held in Helsinki, Finland, 1318
July, 2003; vol. 5, Themes and tasks in Old and Middle IndoAryan linguistics, ed. by Bertil Tikkanen and Heinrich Hettrich, pp. 4564.
2007a. Reciprocal constructions in Vedic. Reciprocal constructions, ed. by V. P. Nedjalkov et al., vol. 2, pp. 70938.
Amsterdam; Philadelphia: John Benjamins.
2007b. The reflexive pronouns in Vedic: a diachronic and typological perspective. Lingua 117118: 141233.
2008. The Vedic causative samkhy
apayati / samk
sa payati reconsidered. Indologica: T. Ya. Elizarenkova memorial volume, ed. by L. Kulikov and M. Rusanov, vol. 1, pp. 24561.
Orientalia et Classica: Papers of the Institute of Oriental and
B IBLIOGRAPHY
445
Classical Studies 20. Moscow: Russian State University for the

Humanities.
. 2009. Valency-changing categories in Indo-Aryan and IndoEuropean: reconstructing the early Vedic paradigm. Multilingualism: proceedings of the 23rd Scandinavian Conference of
Linguistics, ed. by Anju Saxena and ke Viberg, pp. 7592.
Uppsala: Uppsala Universitet.
. 2010. Nominal composition, noun incorporation and nonfinite formations in Sanskrit: delimiting the boundaries of the
astram: Indological and linguisverbal paradigm. Anantam S
tic studies in honour of Bertil Tikkanen, ed. by Klaus Karttunen, pp. 11131. Studia Orientalia 108. Helsinki: Finnish
Oriental Society.
. 2012a. The Vedic -ya-presents: passives and intransitivity in
Old Indo-Aryan. Amsterdam: Rodopi.
. 2012b. Vedic preverbs as markers of valency-changing
derivations: transitivity and objecthood in Indo-European (evidence from Old Indo-Aryan). Studies in Language 36.4: 721
46.
. 2013. Language vs. grammatical tradition in Ancient India:
How real was Pan.inian Sanskrit? Evidence from the history of
late Sanskrit passives and pseudo-passives. Folia Linguistica
Historica 34: 5992.
Kumar, B. S. 1976. Sanskrit syntax and the grammar of case.
Delhi: Motilal Banarsidass.
Kmmel, Martin Joachim. 1996. Stativ und Passivaorist im Indoiranischen. Gttingen: Vandenhoeck and Ruprecht.
. 2000. Das Perfekt im Indoiranischen. Wiesbaden: L. Reichert.
. 2014. Pan.ini 5.3.5 and the function of Sanskrit etad. Proceedings of the 15th World Sanskrit Conference; vol. 1, Vedic
studies: language, texts, culture, and philosophy: proceedings
of the Veda Section, 15th World Sanskrit Conference, New
Delhi, 2012, ed. by Hans Henrich Hock, pp. 3956.
446
S ANSKRIT SYNTAX
Kupfer, Katharina. 2000. Die Demonstrativpronomina im Rigveda.

Frankfurt: Peter Lang.
Kurzov, Helena. 1981. Der Relativsatz in den indoeuropischen
Sprachen. Hamburg: H. Buske; Praha: Academia, Verlag der
Tschechoslowakischen Akademie der Wissenschaften.
Lahiri, P. C. 1933. Studies in the word order of Sanskrit prose.
Ph.D. dissertation. London: University of London.
Lanman, C. R. 1880. Noun inflection in the Veda. Journal of the
Lazzeroni, Romano. 2004. Inaccusativit indoeuropea e alternanza causative vedica. Archivio Glottologico Italiano 89.2:
13964.
Lehmann, Christian. 1984. Der Relativsatz: Typologie seiner
Strukturen, Theorie seiner Funktionen, Kompendium seiner
Grammatik. Language Universals Series 3. Tbingen: Narr.
Lehmann, Winfred P. 1969. Proto-Indo-European compounds in
relation to other Proto-Indo-European syntactic patterns. Acta
Linguistica Hafniensia 12: 120.
. 1973. Explanations for some syntactic phenomena of ProtoIndo-European. Glossa 7: 8190.
. 1974. Proto-Indo-European syntax. Austin: University of
Texas Press.
. 1976. From topic to subject in Indo-European. Subject and
topic, ed. by C. N. Li, pp. 44656. New York: Academic Press.
. 1980. The reconstruction of non-simple sentences in ProtoIndo-European. Linguistic reconstruction and Indo-European
syntax, ed. by P. Ramat et al., pp. 11344. Amsterdam: John
Benjamins.
. 1985. Development of conditional clauses in early Sanskrit.
Sprachwissenschaftliche Forschungen: Festschrift fr Johann
Knobloch, ed. by H. M. lberg et al., pp. 2316. Innsbruck.
B IBLIOGRAPHY
447
Lehmann, Winfred P. and Undirapola Ratanajoti. 1975. Typo

logical syntactical characteristics of the Satapathabr
ahman.a.
Journal of Indo-European Studies 3: 14759.
Liebert, Gsta. 1951. ber die Partikeln su und tu im Vedischen.
Studia Linguistica 5: 5388.
. 1954. Zum Gebrauch der w-Demonstrativa im ltesten Indoiranischen. Lunds Universitets rsskrift, Afd. 1, n.F. 50.9.
Lund: Gleerup.
Liebich, Bruno. 1886. Die casuslehre der indischen grammatiker verglichen mit dem gebrauch der casus im AitareyaBrhman.a. Beitrge zur Kunde der indogermanischen Sprachen 10: 20534. Continued in 11: 273-315.
Lindner, Bruno. 1878. Altindische Nominalbildung nach den Sam
hitas. Jena: Costenoble.
Longobardi, Giuseppe. 1980. Les relatives nominales indoeuropnnes. Linguistic reconstruction and Indo-European syntax: proceedings of the colloquium of the Indogermanische
Gesellschaft, University of Pavia, 6-7 September 1979, ed. by
Paolo Ramat, pp. 17182. Amsterdam studies in the theory
and history of linguistic science, series IV, Current issues in
linguistic theory 19. Amsterdam: John Benjamins.
Lowe, John J. 2011a. Negated participles in R.gvedic Sanskrit and
Proto-Indo-European. Indo-Iranian Journal 54: 1938.
. 2011b. Rgvedic clitics and prosodic movement. Proceedings of
the LFG11 conference, ed. by Miriam Butt and
Tracy Holloway King, pp. 36080. Stanford: CSLI Publications. URL: http : / / web . stanford . edu / group /
cslipublications/cslipublications/LFG/16/
lfg11.html.
. 2013. Indo-European Caland adjectives in *-nt- and participles in Sanskrit. Historische Sprachforschung 126.
Ludwig, Alfred. 1871. Der Infinitiv im Veda. Prague: Calve.
448
S ANSKRIT SYNTAX
. t und
. 1898. ber die absoluten verbalformen des Samskr
den praedicatsauszdruck im allgemeinen. Sitzungsberichte
der Kniglichen Bhmischen Gesellschaft der Wissenschaften,
Classe fr Philosophie, Geschichte und Philologie 1897.7: 1
92. Prague.
Lhr, Rosemarie. 1997. Zur Kasusattraktion in altindischen dativischen Infinitivkonstruktionen: ein Fall von syntaktischer
Analogie? Sound law and analogy: papers in honor of Robert
S. P. Beekes on the occasion of his 60th birthday, ed. by
Alexander Lubotsky, pp. 15570. Leiden Studies in IndoEuropean 9. Leiden: Rodopi.
. 2000a. Der Nebensatz und seine Konkurrenten in der Indogermania: der altindische Relativsatz. Historische Sprachforschung 113: 7187.
. 2000b. Zum Modalfeld im Altindischen. Indoarisch,
Iranisch und die Indogermanistik: Arbeitstagung der Indogermanischen Gesellschaft vom 2. bis 5. Oktober 1997 in Erlangen, ed. by Bernhard Forssman and Robert Plath, pp. 32738.
. 2004a. Nominalkomposition im Altindischen und Altgriechischen. Komplexe Wortstrukturen: Komposition, Inkorporation, Polysynthese, ed. by E. Nowak, pp. 107214. Berlin:
Technische Universitt Berlin, Institut fr Sprache und Kommunikation.
. 2004b. Thematische Rollen und Kasus: zu Agens und Patiens
beim Passiv im Altindischen. Linguistische Arbeitsberichte
81: 99126.
. 2008a. Competitive Indo-European syntax. Principles of
syntactic reconstruction, ed. by Gisella Ferraresi and Maria
Goldbach, pp. 12160. Amsterdam; Philadelphia: John Benjamins.
. 2008b. Old Indic clauses between subordination and coordination. Subordination versus coordination in sentence
B IBLIOGRAPHY
449
and text, ed. by Cathrine Fabricius-Hansen and Wiebke Ramm,

Macdonell, Arthur Anthony. 1916. A Vedic grammar for students.
London: Oxford University Press. [Reprint: Bombay; Calcutta; Madras: Oxford University Press, 1955.]
. 1927. A Sanskrit grammar for students. 3rd edition. Oxford:
Oxford University Press.
Matilal, Bimal Krishna. 1966. Indian theorists on the nature of
the sentence. Foundations of Language 2: 37793.
. 1985. Logic, language, and reality. Delhi: Motilal Banarsidass.
Mayr, Aurel. 1871. Beitrge aus dem Rgveda zur Accentuirung des
Verbum finitum. Wien: Gerolds Sohn.

Meenakshi, K. 1983. Epic syntax. Delhi: Meharchand Lacchmandas. [Revision of Ph.D. dissertation, University of Poona,
1963.]
. 1986. The quotative in Indo-Aryan. South Asian languages:
structure, convergence, and diglossia, ed. by Bh. Krishnamurti
et al., pp. 20918. Delhi: Motilal Banarsidass.
. 1991a. The genitive in Pan.ini and in Epic Sanskrit. Studies
in Sanskrit syntax: a volume in honour of the centennial of
Speijers Sanskrit syntax (18861986), ed. by Hans Henrich
Hock, pp. 14552.
. 1991b. Vedic infinitive and Pan.ini. Panels of the VIIth World
Sanskrit Conference, Kern Institute, Leiden, August 2329,
1987; vol. 5, Pan.ini and the Veda, ed. by Madhav Murlidhar
Deshpande, pp. 4654.
Meillet, Antoine. 19061908. La phrase nominale en indoeuropen. Mmoires de la Socit de Linguistique de Paris
14: 126.
. 1931. Le cas employs linfinitif en indo-europen. Bulletin de la Socit de Linguistique de Paris 32.2: 18893.
Melazzo, Roberta. 2010. I bahuvrhi del Rig Veda. Innsbruck: Institut fr Sprachwissenschaft.
450
S ANSKRIT SYNTAX
Michelson, Truman. 1904. Linguistic archaisms of the Ramayan.a. Journal of the American Oriental Society 25: 89145.
Miehle, Helen Louise. 1974. Relative constructions in the RigVeda, book five. Journal of Indo-European Studies 2: 407
34.
Migron, Saul. 1985. The Rgvedic stanza as a syntactical unit:
a study of selected trimeter passages. Ph.D. dissertation.
Jerusalem: The Hebrew University.
. 1993. Catena and climax in Vedic prose. Die Sprache 35:
7180.
. 1994. The cleft sentence in Vedic Prose: a sketch. Mnchener Studien zur Sprachwissenschaft 55: 99122.
Miller, D. Gary. 1975. Indo-European: VSO, SOV, SVO, or all
three? Lingua 37: 3152.
Minard, Armand. 1936. La subordination dans la prose vdique:
tudes sur le Satapatha-Br

ahman.a, 1. Annales de lUniversit
de Lyon, nouvelle srie 3 3. Thse, Paris. Paris: Socit
ddition Les Belles Lettres.
. 1949. Trois nigmes sur les cent chemins: recherches sur le Satapatha-Brahman.a. Annales de lUniversit de Lyon, nouvelle
srie 3 17. Paris: Les Belles Lettres.
Molina Muoz, Adriana. 2009. A reexamination of the perfect
and the imperfect in Vedic Sanskrit. Qualifying examination
paper. Urbana: Department of Linguistics, University of Illinois, Urbana-Champaign.
. 2014. Sanskrit compounds and the architecture of the grammar. Grammatica et verba Glamour and verve: studies
in South Asian, historical, and Indo-European linguistics in
honor of Hans Henrich Hock on the occasion of his seventyfifth birthday, ed. by Shu-Fen Chen and Benjamin Slade,
pp. 181201.
Mumm, Peter Arnold. 1995. Verbale Definitheit und der vedische
Injunktiv. Verba et structurae: Festschrift fr Klaus Strunk
B IBLIOGRAPHY
451
zum 65. Geburtstag, ed. by Heinrich Hettrich et al., pp. 169

93. Innsbruck: Institut fr Sprachwissenschaft.
. 2002. Retrospektivitt im Rigveda: Aorist und Perfekt. Indogermanische Syntax: Fragen und Perspektiven, ed. by Heinrich Hettrich and Jeong-Soo Kim, pp. 15788. Wiesbaden: L.
Reichert.
. 2004. Altindisch sma, Teil 1: Rig und Atharvaveda. International Journal of Diachronic Linguistics and Linguistic Reconstruction 1: 1968.
Negelein, Julius von. 1897. Das Verbalsystem des Atharvaveda.
Norden: Soltau.
Nooten, B. A. van, ed. 1988. Sanskrit studies of M. B. Emeneau.
Occasional papers 13. Berkeley: Center for South and Southeast Asian Studies, University of California at Berkeley.
Oberlies, Thomas. 1999. ca if, when, ce(d) and wor(l)ds
upside-down? Vidyopasana: studies in honour of Harivallabh
C. Bhayani, ed. by P. J. Mistry and Bharati Modi, pp. 16972.
Mumbai; Ahmedabad: Image Publications.
. 2003. A grammar of Epic Sanskrit. Indian Philology and South
Asian Studies. Berlin: Walter de Gruyter.
Oertel, Hanns. 1923. Zum disjunkten Gebrauch des Nominativs
in der Brahman.aprosa. Antidoron: Festschrift Jakob Wackernagel, pp. 4550. Gttingen: Vandenhoeck and Ruprecht.
. 1926. The syntax of cases in the narrative and descriptive
prose of the Brahman.as: the disjunct use of cases. Indogermanische Bibliothek; Abt. 1, Sammlung indogermanischer
Lehr- und Handbcher; Reihe 1, Grammatiken 18. Heidelberg:
Winter.
. 1935. Syntaktische quivalenz des Genitivs und Ablativs bei
Verben der Trennung in der altindischen Prosa. Sitzungsberichte der Bayerischen Akademie der Wissenschaften,
philosophisch-historische Abteilung 1935.12. Mnchen: Bayerische Akademie der Wissenschafter.
452
S ANSKRIT SYNTAX
. 1936. Ai. Gen.-Abl. der Feminina auf -as in dativischer Funktion. Zeitschrift fr vergleichende Sprachforschung 63: 206.
. 1937a. Zu den Kasusvariationen in der vedischen Prosa,
1. Sitzungsberichte der Bayerischen Akademie der Wissenschaften, philosophisch-historische Abteilung 1937.8. Mnchen: Bayerische Akademie der Wissenschafter.
. 1937b. Zum altindischen Ausdrucksverstrkungstypus satyasya satyam das Wahre des Wahren = die Quintessenz
des Wahren. Sitzungsberichte der Bayerischen Akademie der
Wissenschaften, philosophisch-historische Abteilung 1937.3.
Mnchen: Bayerische Akademie der Wissenschafter.
. 1938a. Die Konstruktion von svara in der vedischen Prosa.
. 1938b. Zu den Kasusvariationen in der vedischen Prosa,
. 1939a. The Genitivus personae with verbs of eating . . . and
accepting . . . in Vedic Prose. A volume of Eastern and Indian studies presented to Professor F. W. Thomas, C.I.E., on
his 72nd birth-day, 21st March 1939, ed. by S. M. Katre and
P. K. Gode, pp. 16065. New Indian Antiquary extra series 1.
Bombay: Karnatak Publishing.
. 1939b. Zu den Kasusvariationen in der vedischen Prosa,
. 1940. Zu den Wortstellungsvarianten der Mantras des A
tharvaveda in der Saunakaund Paipalada-Rezension und
des Samaveda in der Kauthuma- und Jaiminya-Rezension.
Sitzungsberichte der Bayerischen Akademie der Wissenschaften, philosophisch-historische Abteilung 1940.7. Mnchen: Bayerische Akademie der Wissenschafter.
B IBLIOGRAPHY
453
. 1941. Die Dativi finales abstrakter Nomina und andere

Beispiele nominaler Satzfgung in der vedischen Prosa.
Sitzungsberichte der Bayerischen Akademie der Wissenschaften, philosophisch-historische Abteilung 1941.2, 9. Mnchen: Bayerische Akademie der Wissenschafter.
. 1942. Zu den ai. Ellipsen. Zeitschrift fr vergleichende
. 1944. Zu den ai. Ellipsen. Zeitschrift fr vergleichende
Oettinger, Norbert. 1983. Altavestisch yas . . . cica jeder, der.
. 1986. Syntax des Relativsatzes und pluralischer Instrumental
im Avestischen. Indo-Iranian Journal 29: 4548.
Oguibnine, Boris. 2006. Notes on the instrumental case of the
subject/agent vs. other cases in Buddhist Sanskrit. Papers of
the 12th World Sanskrit Conference, held in Helsinki, Finland,
1318 July, 2003; vol. 5, Themes and tasks in Old and Middle
Indo-Aryan linguistics, ed. by Bertil Tikkanen and Heinrich
Hettrich, pp. 89120.
Oldenberg, Hermann. 1888. Die Hymnen des Rigveda, 1:
metrische und textgeschichtliche Prolegomena. Berlin: Wilhelm Hertz.
. 1906. Vedische Untersuchungen, 16: die Verbalenklisis
im R.gveda. Zeitschrift der Deutschen Morgenlndischen
Gesellschaft 60: 70740.
. 1907a. Vedische Untersuchungen, 18: zu den Verbalprfixen. Zeitschrift der Deutschen Morgenlndischen Gesellschaft
61: 80315.
. 1907b. Vedische Untersuchungen, 20: die enklitischen Formen des Pronominalstammes a-. Zeitschrift der Deutschen
Morgenlndischen Gesellschaft 61: 82530.
. 19091912. Rgveda: textkritische und exegetische Noten. Abhandlungen
der Kniglichen Gesellschaft der Wissenschaften
454
S ANSKRIT SYNTAX
zu Gttingen, Philologisch-Historische Klasse. n.F. 11.5; 13.3.

Bd. 1: Erstes bis sechstes Buch. Bd. 2: Siebentes bis zehntes
Buch. Zu Grunde gelegt ist [T.] Aufrechts Text, 2. Auflage
[Bonn, 1877], v. 1, p. v. Berlin: Weidmannsche Buchhandlung.
. 1917. Zur Geschichte der altindischen Prosa: mit besonderer Bercksichtigung der prosaisch-poetischen Erzhlung.
Abhandlungen der Kniglichen Gesellschaft der Wissenschaften zu Gttingen, Philologisch-Historische Klasse. n.F.
16.6. Berlin: Weidmannsche Buchhandlung.
Oranskaya, T. 1981. On the realisation of the category of
definiteness-indefiniteness in the Vedic Sanskrit. Summaries
of papers presented by Soviet scholars to the 5th World Sanskrit Conference, p. 218. Moscow.
rterer, Georg. 1873. Beitrge zur vergleichenden Casuslehre des
Zend und Sanskrit. Ph.D. dissertation, Universitt Mnchen,
1873. Mnchen: Straub.
Ostler, Nicholas D. M. 1979. Case-linking: a theory of case and
verb diathesis applied to Classical Sanskrit. Ph.D. dissertation. Cambridge, Mass.: MIT.
Papke, Julia Kay Porter. 2010. Classical Sanskrit preverb ordering: a diachronic study. Ph.D. dissertation. Columbus: Ohio
State University.
Patyal, Hukam Chand. 1994. Linguistic, syntactic, and stylistic
peculiarities of the Brahman.as. Sruti-cint

aman.ih.: Prof. C. G.
Kashikar felicitation volume, ed. by S. S. Bahulkar and S.
Paranjpe, pp. 11221. Pune: Tilak Maharashtra Vidyapeeth.
Pinault, Georges-Jean. 1985. Emploi et analyse des adverbes
comparatifs sanskrits en -vat. Grammatische Kategorien:
Funktion und Geschichte: Akten der 7. Fachtagung der Indogermanischen Gesellschaft, Berlin, 10.-25. Februar 1983,
ed. by Bernfried Schlerath, pp. 34069. Wiesbaden: L. Reichert.
B IBLIOGRAPHY
455
. 19951996. Distribution des particules comparatives dans la

a. Bulletin dtudes indiennes 1314:
langue de la Rk-Samhit
30767.
. 1997. Distribution de la particule negative na dans la Rk
a. Bulletin dtudes indiennes 15: 21346.
Samhit
. 2001. Vdique tanu et la notion de personne en indo-iranien.
Bulletin de la Socit de Linguistique de Paris 96.1: 181206.
Pisani, Vittore. 1954. ber die Inifinite auf sanskr. -adhyai.
Porzig, W. 1923. Die Hypotaxe im Rigveda, 1. Indogermanische
Forschungen 41: 210303.
. 1942. Die Namen fr Satzinhalte im Griechischen und im Indogermanischen. Berlin: de Gruyter.
Pray, Bruce. 1976. From passive to ergative in Indo-Aryan. The
notion of subject in South Asian languages, ed. by M. K.
Verma, pp. 195211. South Asian Studies Publications Series
2. Madison: University of Wisconsin.
Proferes, Theodore. 2003. Kuru kings, Tura Kavaseya, and the
-tvaya gerund. Bulletin of the School of Oriental and African
Studies 66.2: 21019.
Puhvel, Jaan. 1973. Nature and means of comparison in ProtoIndo-European. Journal of Indo-European Studies 1: 14554.
Raja, Kunjunni. 1958. The elliptical sentence: Indian theories.
Adyar Library Bulletin 22: 2531.
. 1976. Pan.inis concept of a sentence. Adyar Library Bulletin
40: 16570.
Raster, Peter. 1993. Die indische Grammatiktradition. Syntax:
ein internationales Handbuch zeitgenssischer Forschung /
An international handbook of contemporary research, ed. by
Joachim Jacobs et al., pp. 199207. Berlin; New York: de
Gruyter Mouton.
456
S ANSKRIT SYNTAX
Reinhl, Uta. 2013. Grammaticization and configurationality: the

emergence of postpositional phrases in Indo-Aryan. Doctoral
dissertation. Kln: Universitt Kln.
Renou, Louis. 1925. La valeur du parfait dans les hymns vdiques.
Collection linguistique 18. Paris: Champion.
. 1928a. Les formes dites dinjonctif dans le rgveda. Etrennes
de linguistique: offertes par quelques amis
mile Benveniste:
avant-propos de Antoine Meillet, pp. 6380. Paris: Paul Geuthner.
. 1928b. Les matres de la philologie vdique. Paris.
. 1932. A propos du subjonctif vdique. Bulletin de la Socit
Linguistique de Paris 33.1.
. 1933. La sparation du prverbe et du verbe en vdique. Bulletin de la Socit Linguistique de Paris 34.1: 4996.
. 1935. Labsolutif sanskrit en -am. Mmoires de la Socit
Linguistique de Paris 23: 35992.
. 1937a. Infinitifs et drivs nominaux dans le r.gveda. Bulletin
de la Socit Linguistique de Paris 38.1: 6987.
. 1937b. La dcadence et la disparition du subjonctif. Monographies sanskrites 1. Paris: Adrien-Maisonneuve.
. 1937c. Le suffixe -tu et la constitution des infinitifs. Monographies sanskrites 2. Paris: Adrien-Maisonneuve.
. 1940. Notes de grammaire vdique, 1. Bulletin de la Socit
Linguistique de Paris 41.1: 20814.
. 1947. Sur la phrase ngative dans le rgveda. Bulletin de la
Socit Linguistique de Paris 43.1: 4349.

. 1952. Grammaire de la langue vdique. Collection Les
langues du monde, Srie Grammaire, philologie, littrature 9.
Lyon: IAC.
. 1954. Le passage du nom daction linfinitif dans le
R.gveda. Sprachgeschichte und Wortbedeutung: Festschrift
Albert Debrunner, pp. 3837. Bern: Francke.
B IBLIOGRAPHY
457
. 19551969. tudes vdiques et pan.inennes. 17 vols. Paris:

Boccard.
. 1956. Histoire de la langue sanskrite. Lyon: IAC.
. 1961. Grammaire sanscrite. 2e dition revue, corrige et augmente. Paris: Adrien-Maisonneuve.
. 1965. Remarques gnrales sur la phrase vdique. Symbolae
linguisticae in honorem Georgii Kuryowicz. Polska Akademia
Nauk, Komisja Jezykoznawstwa, Prace 5. Wrocaw: Zakad
Narodowy im. Ossolinskich.
Richter, Oswald. 1897. Die unechten Nominalkomposita des Altindischen und Altiranischen. Straburg: Trbner.
Rocher, Ludo. 1979. Review of Verpoorten 1977. Journal of the
American Oriental Society 99.2: pp. 348349.
. 1980. A note on the Sanskrit gerund. Recherches de linguistique: hommages Maurice Leroy, ed. by J. Bingen, Andr
Coupez, and Francine Mawet, pp. 18188. [Travaux] Universit Libre de Bruxelles, Faculte de Philosphie et Lettres 73.
Brussels: ditions de lUniversit de Bruxelles.
Rouse, W. H. D. 1906. The use of the passive gerund in Sanskrit.
Journal of the Royal Asiatic Society 1906: 992.
Ruppel, Antonia. 2012a. Absolute constructions in early IndoEuropean. Cambridge: University Press.
. 2012b. The development of the genitive absolute in Sanskrit. Indic across the millennia: from the Rigveda to Modern
Indo-Aryan: 14th World Sanskrit Conference, Kyoto, Japan,
September 1st5th, 2009, proceedings of the Linguistic Section, ed. by Jared S. Klein and Kazuhiko Yoshida, pp. 13544.
Sadovski, Velizar. 2020. Untersuchungen zu Sprache und Stil des
ltesten Indo-Iranischen (Veda und Avesta). Wien: sterreichische Akademie der Wissenschaften.
Salomon, Richard. 1981. A linguistic analysis of the Mun.d.aka
Upanis.ad. Wiener Zeitschrift fr die Kunde Sdasiens 30: 91
105.
458
S ANSKRIT SYNTAX
. 1982. The ukti-vyakti-prakaran.a as a manual of spoken Sanskrit. Indo-Iranian Journal 24: 1325.
. 1988. The preterite optative in Indic. Paper presented at the
Third Sanskrit Syntax Symposium, South Asian Languages
Analysis Roundtable, Seattle, WA, 1988. Unpublished.
Satya Vrat Shastri. 1964. The Ramayan.a: a linguistic study. Delhi:
Munshi Ram Manohar Lal.
. 1977. Sanskrit usage. Indologica Taurinensia 3/4: 44954.
Saussure, Ferdinand de. 1881. De lemploi du gnitif absolu
en sanscrit. Ph.D. dissertation, Universitt Leipzig, 1880.
Geneva: Fick.
Scharf, Peter M. 1997. Clause-initial dvayam: one less case of an
extraposed adverb. 40: 32738.
. 2013b. Linguistics in India. The Oxford handbook of the history of linguistics, ed. by Keith Allen, pp. 230264. Oxford:
Oxford University Press.
Scharfe, Hartmut. 2006. Indo-Aryan and Dravidian convergence:
gerunds and noun composition. Papers of the 12th World
Sanskrit Conference, held in Helsinki, Finland, 1318 July,
2003; vol. 5, Themes and tasks in Old and Middle IndoAryan linguistics, ed. by Bertil Tikkanen and Heinrich Hettrich, pp. 197254.
Schufele, Steven. 1988a. Adpositions authentic and inauthentic:
some questions in the syntax of Vedic Sanskrit. Paper presented at the Third Sanskrit Syntax Symposium, South Asian
Languages Analysis Roundtable, Seattle, WA, 1988. Unpublished.
. 1988b. Wheres my NP?: non-transformational analyses of
Vedic pronominal fronting. Studies in the Linguistic Sciences
18.2: 12962.
. 1990. Free word-order syntax: the challenge from Vedic Sanskrit to contemporary formal syntactic theory. Ph.D. dissertation. Urbana: University of Illinois.
B IBLIOGRAPHY
459
. 1991a. Single-word topicalization in Vedic Prose: a challenge

to Government & Binding? Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax
(18861986), ed. by Hans Henrich Hock, pp. 15173.
. 1991b. Verb-medial clauses in Vedic: some theoretical implications. Studies in Sanskrit syntax: a volume in honour of
the centennial of Speijers Sanskrit syntax (18861986), ed. by
Hans Henrich Hock, pp. 17796.
. 1993. The Vedic clause-initial string and Universal Grammar. Studies in the Linguistic Sciences 23.1: 13161.
Scherer, Anton. 1973. Die ursprngliche Funktion des Konjunktivs. Indogermanische und allgemeine Sprachwissenschaft:
Akten der IV Fachtagung der Indogermanischen Gesellschaft,
Bern, 28. Juli1. August 1969, ed. by Georges Redard, pp. 99
Schmidt, Gernot. 1977. ber indogermanische nominale Relativkonstruktionen. Indogermanische Forschungen 82: 6174.
Schmitt-Brandt, Robert. 1973. Vergleich der indogermanischen
Nebensatzkonstruktionen. Indogermanische und allgemeine
Sprachwissenschaft: Akten der IV Fachtagung der Indogermanischen Gesellschaft, Bern, 28. Juli1. August 1969, ed. by
Georges Redard, pp. 12541. Wiesbaden: L. Reichert.
Schneider, Carolin. 2009. Syntax und Wortarten der Lokalpartikeln des R.gveda V: n. Historische Sprachforschung 122:
11869.
. 2010a. Lokalpartikel im R.gveda (Folge 13): a pa VII: ns.
Studien zur Indologie und Iranistik 27: 135.
. 2010b. Syntax und Wortarten der Lokalpartikeln des R.gveda
VII: ns. International Journal of Diachronic Linguistics and
Linguistic Reconstruction 7: 14993.
. 2011. Syntax und Wortarten der Lokalpartikeln des R.gveda
X: u d. Mnchener Studien zur Sprachwissenschaft 65: 185
241.
460
S ANSKRIT SYNTAX
. 2012a. Local particles in the R.gveda, part XVII: para.

. 2012b. Syntax und Wortarten der Lokalpartikeln des R.gveda.
XV.2: sakam, satra , s(u)mad und sataso. International Journal of Diachronic Linguistics and Linguistic Reconstruction 9:
10136.
. 2013. Syntax und Wortarten der Lokalpartikeln des R.gveda
XIX: pra. Indogermanische Forschungen 117: 1574.
Seiler, Hansjakob. 1960. Relativsatz, Attribut und Apposition.
Wiesbaden: Harrassowitz.
. 1962. Zum Verhltnis von Wort und Satz in indogermanischen Sprachen. II. Fachtagung fr indogermanische und allgemeine Sprachwissenschaft. Innsbrucker Beitrge zur Kulturwissenschaft, Sonderheft 15. Innsbruck.
Sen, N. M. 1952a. On the syntax of cases in the Ramayan.a. Journal of the Oriental Institute 2: 31126. Baroda.
. 1952b. Syntax of tenses in the Ramayan.a. Journal of the Oriental Institute 1: 3017. Baroda.
Sen, Sukumar. 1925. Notes on the use of cases in the Kat.haka
a. Journal and Proceedings of the Asiatic Society of
Samhit
Bengal 21: 18.
. 19261930. The syntax of cases in Vedic prose. Annals of the
Bhandarkar Oriental Research Institute 8: 34778. Continued
in 9:33-48, 91-170; 10:45-76, 219-234.
. 1928. An outline of syntax of Buddhistic Sanskrit. Journal
of the Department of Letters 17: 165. Calcutta University.
. 1953. Historical syntax of Middle Indo-Aryan. Indian Linguistics 13: 352473.
Sgall, Petr. 1958. Die Infinitive im Rgveda. Acta Universitatis
Prague.
Carolinae, Philologica 2: 135268.
Siecke, Ernst Ludwig. 1869. De genetivi in lingua sanscritica, imprimis vedica usu. Ph.D. dissertation. Berlin: Vogt.
B IBLIOGRAPHY
461
Sinha, Anjani C. 1973. Generative semantics and Pan.inis karakas. Journal of the Oriental Institute 23: 2739. Baroda.
Suszkiewicz, E. 1972. Some remarks on Sanskrit and Greek absolutive genitives. India Maior: congratulatory volume presented to J. Gonda, ed. by J. Ensink and Peter Gaeffke,
pp. 19699. Leiden: E. J. Brill.
Shnen, Renate. 1979. Untersuchungen zur Komposition von Reden und Gesprchen im Ramayan.a. Studien zur Indologie und
Iranistik 6. Reinbeck: Wezler.
. 1985. Die Konstruktionen des Absolutivs im Sanskrit.
Grammatische Kategorien: Funktion und Geschichte: Akten
der VII. Fachtagung der Indogermanischen Gesellschaft, ed.
by Bernfried Schlerath, pp. 47889. Wiesbaden: L. Reichert.
Sommer, Ferdinand. 1977. estai. Ferdinand Sommer: Schriften
aus dem Nachla, ed. by Bernhard Forssman, pp. 15477.
Mnchener Studien zur Sprachwissenschaft Beihefte, n. F. 1.
Mnchen: Kitzinger.
. 1896. Vedische und Sanskrit-Syntax. Grundriss der IndoArischen Philologie und Altertumskunde 1.6. Strassburg:
Trbner. [Reprint: Graz, Austria: Akademische Druck- u. Verlagsanstalt, 1974.]
Stchoupak, Nadine. 1920. Le complment du nom dans le Satapathabrahman.a. Mmoires de la Socit de Linguistique de
Paris 21: 138.
Storck, Frid. Guil. Paul. 1862. Casuum in lingua palica formatio comparata cum sancscritae linguae ratione. Monasterii:
Brunn.
462
S ANSKRIT SYNTAX
Strunk, Klaus. 1969. Wortstellung und Univerbierung altpersischer Korrelativverbindungen. Zeitschrift fr vergleichende
. 1980. Zum indogermanischen Medium und konkurrierenden Kategorien. Wege zur Universalienforschung: sprachwissenschaftliche Beitrge zum 60. Geburtstag von H. Seiler, ed.
by G. Brettschneider and Ch. Lehmann, pp. 32137. Tbingen:
Narr.
. 1983. Typische Merkmale von Fragestzen und die altindische
Pluti. Bayerische Akademie der Wissenschaften, Phil.-Hist.
Kl., Sitzungsberichte 1983.8. Mnchen: Beck.
Stump, Gregory T. 1979. The loss of the subjunctive mood in
yadi-conditionals from Vedic to Epic Sanskrit. Papers of the
1978 Mid-America Linguistics Conference at Oklahoma, ed.
by Ralph E. Cooley, Mervin R. Barnes, and John A. Dunn,
pp. 24251. University of Oklahoma.
Sturtevant, Edgar H. 1930. Relatives in Indo-European and Hittite. Language, Language Monograph 6.7; Curme volume of
linguistic studies, pp. 14149. Washington, DC: Linguistic Society of America. URL: http : / / www . jstor . org /
stable/521994.
Subramaniam, P. S. 1973. Deep structure and surface structure in
Pan.ini. Proceedings of the third All-India Conference of Linguists, Hyderabad, December 29-30, 1972, pp. 2021. Pune:
Linguistic Society of India.
Szemernyi, Oswald. 1985. Syntax, meaning, and origin of
the Indo-European particle kw e. Collectanea philologica:
Festschrift fr Helmut Gipper zum 65. Geburtstag, ed. by Gnther Heintz and Peter Schmitter, vol. 2, pp. 74775. Saecula
spiritalia 15. Baden-Baden: Koerner.
Taraporewala, Irach J. S. 1967. Sanskrit syntax. Delhi: Munshiram
Manoharlal.
B IBLIOGRAPHY
463
Thieme, Paul. 1929. Das Plusquamperfektum im Veda. Ergnzungshefte zur Zeitschrift fr vergleichende Sprachforschung
auf dem Gebiete der indogermanischen Sprachen 7. Gttingen:
Vandenhoeck and Ruprecht.
. 1944. Zum parenthetischen Nominalsatz im Indischen.
. 1949. Untersuchungen zur Wortkunde und Auslegung des
Rigveda. Hallische Monographien 7. Halle: Niemeyer.
Thomas, E. J. 19211922. Some linguistic superstitions. Philologica: Journal of Comparative Philology 1: 96101.
Thommen, Eduard. 1903. Die Wortstellung im nachvedischen Altindischen und im Mittelindischen. Ph.D. dissertation, GeorgAugust-Universitt zu Gttingen. Gtersloh: Bertelsmann.
Thornton, Randolph. 1987. The verbal focus system of Epic Sanskrit: South Asian Languages Analysis Roundtable Conference held in Ann Arbor, Mich., May 1719, 1985. Select
Papers from SALA-7, ed. by Elena Bashir, Madhav M. Deshpande, and Peter Edwin Hook, pp. 35463. Bloomington: Indiana University Linguistics Club.
Thumb, Albert. 1959. Handbuch des Sanskrit: mit Texten und
Glossar: eine Einfhrung in das sprachwissenschaftliche
Studium des Altindischen, rev. by Richard Hauschild; vol. 2,
Texte und Glossar. 3rd edition, revised. Indogermanische Bibliothek. Erste Reihe, Lehr- und Handbcher 1.2. Heidelberg:
Carl Winter.
Thurneysen, Rudolph. 1908. Altindisch e tava. Mlanges de linguistique: offerts M. Ferdinand de Saussure, pp. 2257. Collection linguistique 2. Paris: Librarie de la Socit de Linguistique de Paris. [Reprint: Genve: Slatkine Reprints, 1975.]
Tichy, Eva. 1980. Zum Kasusgebrauch bei Kausativa transitiver
Verben. Die Sprache 26: 118.
. 1993. Transponierte Rollen und Ergnzungen beim vedischen Kausativ. Indogermanica et Italica: Festschrift fr Hel-
464
S ANSKRIT SYNTAX
mut Rix zum 65. Geburtstag, ed. by Gerhard Meiser, in collab. with Jn Axel Hararson und Christiane Schaefer Jadwiga
Bendahman, pp. 43660. Innsbrucker Beitrge zur Sprachwissenschaft der Universitt Innsbruck 72. Innsbruck: Institut fr
Sprachwissenschaft.
. 1995a. Die Nomina agentis auf -tar- im Vedischen. Heidelberg:
Winter.
. 1995b. Vedisch e d: Funktion, Herkunft und subordinierte
Transpositionen. Verba et structurae: Festschrift fr Klaus
Strunk zum 65. Geburtstag, ed. by Heinrich Hettrich et al.,
pp. 31943. Innsbrucker Beitrge zur Sprachwissenschaft der
Universitt Innsbruck 83. Innsbruck: Institut fr Sprachwissenschaft.
. 1997. Vom indogermanischen Tempus/Aspekt-System zum
vedischen Zeitstufensystem. Berthold Delbrck y la sintaxis
indoeuropea hoy: actas del coloquio de la Indogermanische
Gesellschaft 1994, Madrid, ed. by Emilio Crespo and Jos L.
Garca-Ramn, pp. 589609. Madrid: Ediciones de la UAM;
. 2006. Der Konjunktiv und seine Nachbarkategorien: studien
zum indogermanischen Verbum, ausgehend von der lteren
vedischen Prosa. Bremen: Ute Hempen Verlag.
Tikkanen, Bertil. 1984. On the tense value, functions, and
forms of the Sanskrit gerund. Studia Orientalia 55: 47197.
Helsinki.
. 1987. The Sanskrit gerund: a synchronic, diachronic, and typological analysis. Studia Orientalia 62. Ph.D. dissertation.
Helsinki: Societas Orientalis Fennica.
. 1991. On the syntax of Sanskrit gerund constructions: a functional approach. Studies in Sanskrit syntax: a volume in honour of the centennial of Speijers Sanskrit syntax (18861986),
ed. by Hans Henrich Hock, pp. 197208.
B IBLIOGRAPHY
465
Tikkanen, Bertil and Heinrich Hettrich, eds. 2006. Papers of the

12th World Sanskrit Conference, held in Helsinki, Finland, 13
18 July, 2003, general editors Petteri Koskikallio and Asko
Parpola; vol. 5, Themes and tasks in Old and Middle IndoAryan linguistics. Delhi: Motilal Banarsidass.
Tsiang, Sarah. 1988a. The discourse function of the absolutive in
the Pacatantra. Studies in the Linguistic Sciences 18.2: 163
81.
. 1988b. The syntactic features of speech and context in the
Pacatantra. Paper presented at the Third Sanskrit Syntax
Symposium, South Asian Languages Analysis Roundtable,
Seattle, WA, 1988. Unpublished.
Tsiang, Sarah and Albert Watanabe. 1987. The Pacatantra and
Aesops Fables: a comparison of rhetorical structure in classical Indian and western literature. Studies in the Linguistic
Sciences 17.1: 13746.
Tsiang-Starcevic, Sarah. 1997. The discourse functions of subordinate constructions in Classical Sanskrit narrative texts.
Ph.D. dissertation. Urbana: University of Illinois.
Tubb, Gary A. and Emery R. Boose. 2007. Scholastic Sanskrit: a
manual for students. New York: American Institute for Buddhist Studies.
Vale, R. N. 1948. Verbal composition in Indo-Aryan. Deccan College dissertation series 6. Pune: Deccan College.
Van de Walle, Lieve. 1987. Pragmatic motivation of the passive
voice in Classical Sanskrit. 7th World Sanskrit Conference.
Leiden.
. 1988. Semantic functions eligible for subject position: a case
study in Classical Sanskrit. Paper presented at the Third Sanskrit Syntax Symposium, South Asian Languages Analysis
Roundtable, Seattle, WA, 1988. Unpublished.
466
S ANSKRIT SYNTAX
. 1991. Pragmatics and Classical Sanskrit: a pilot study in linguistic politeness. Ph.D. dissertation. Antwerp: University of
Antwerp. [Revised as Van de Walle 1993.]
. 1992. On explaining double object constructions. Meaning
and grammar: cross-linguistic perspectives, ed. by Michel Kefer and Johan van der Auwera, pp. 20322. Berlin; New York:
Mouton de Gruyter.
. 1993. Pragmatics and Classical Sanskrit: a pilot study in linguistic politeness. Amsterdam; Philadelphia: John Benjamins.
[Revision of Van de Walle 1991.]
Varma, Siddheshwar. 1937. Syntax of the dative case in the Rg
veda. Jha commemoration volume: essays on oriental subjects presented to Vidyasagara Mahamahopadyaya Pan.d.ita
Ganganatha Jha on his completing the 60th year on 25th
september, 1932 by his pupils, friends and admirers, ed.
by Surendra Nath Dasgupta and Shripad Krishna Belvalkar,
pp. 43556. Poona oriental series 39. Pune: Oriental Book
Agency.
Vekerdi, J. 1955. On past tense and verbal aspect in the Rgveda.
Acta Orientalia 5: 7599.

Vendryes, J. 1914. Sur lemploi de linfinitf au genitif dans
quelques langues indoeuropennes. Mmoires de la Socit
Linguistique de Paris 16: 24760.
Verma, Manindra K. and K. P. Mohanan, eds. 1990. Experiencer
Subjects in South Asian Languages: proceedings of the Conference on the Experiencer Subject in South Asian Languages,
November 1988, Madison, Wisconsin. Palo Alto: Center for the
Study of Language and Information, Stanford University.
Verpoorten, J.-M. 1977. Lordre des mots dans lAitareya-Brahman.a. Bibliothque de la Facult de Philosophie et Lettres
de lUniversit de Lige 216. Ph.D. dissertation, Universit de
Lige. Paris: Les Belles Lettres.
B IBLIOGRAPHY
467
abara-Bha. 1991. Iti dans le Jaiminya-Brahman.a et dans le S

s.ya. Panels of the VIIth World Sanskrit Conference, Kern Institute, Leiden, August 2329, 1987; vol. 4, Sense and syntax
in Vedic, ed. by Joel P. Brereton and Stephanie W. Jamison,
pp. 8693.
Vine, Brent. 1997. On the expression of reflexive possession in
the Rig-Veda: RV sva. Syntaxe des langues indo-iraniennes
anciennes: colloque international Sitges (Barcelona) 4-5
mai 1993 / organis par lInstitut du Proche-Orient Ancien
(Universit de Barcelone); actes, ed. by Eric Pirart, pp. 203
14. Aula orientalis; supplementa 6. Barcelona: Editorial Ausa.
Viti, Carlotta. 2004. Funzioni semantiche e pragmatiche nelle
strategie di possesso dellantico indiano. Archivio Glottologico Italiano 89: 4183.
. 2007. Strategies of subordination in Vedic. Milano: Francoangeli.
. 2010. The information structure of OVS in Vedic. Diachronic studies on information structure: language acquisition and change, ed. by Gisella Ferraresi and Rosemarie Lhr,
pp. 3762. Language, context, and cognition 10. Berlin: de
Gruyter.
Voyles, Joseph B. 1970. The infinitive and participle in IndoEuropean: a syntactic reconstruction. Linguistics 58: 6891.
Wackernagel, Jacob. 1892. ber ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen 1: 333
436. [Reprint: Kleine Schriften. Gttingen: Vandenhoeck and
Ruprecht, 19551979.]
. 1926. Vorlesungen ber Syntax. Basel: Birkhuser.
. 1940. Indogermanisch -qu e als alte nebensatzeinleitende
Konjunktion. Zeitschrift fr vergleichende Sprachforschung
67: 15.
Wackernagel, Jacob and Albert Debrunner. 18961957. Altindische Grammatik. 3 volumes in 4. I. Lautlehre. II. 1. Einleitung
468
S ANSKRIT SYNTAX
zur Wortlehre. Nominalkomposition. 2. Die Nominalsuffixe,

von A. Debrunner. III. Nominalflexion, Zahlwort, Pronomen,
von A. Debrunner und J. Wackernagel. Vol. 1 reissued in 1957
with additions by (Renou and) Debrunner. Gttingen: Vandenhoeck and Ruprecht.
Wallace, William D. 1984. The interaction of word order and
pragmatics in a Sanskrit text. Studies in the Linguistic Sciences 14.1: 16788.
Watkins, Calvert. 1976. Towards Proto-Indo-European syntax:
problems and pseudo-problems. Papers from the Parasession on Diachronic Syntax, April 22, 1976, ed. by Sanford B.
Steever, Carol A. Walker, and Salikoko S. Mufwene, pp. 305
26. Chicago: Linguistic Society.
. 1982. Aspects of Indo-European poetics. The IndoEuropeans in the fourth and third millenia, ed. by Edgar C.
Polom, pp. 10420. Ann Arbor: Karoma.
Wecker, Otto. 1906. Der gebrauch der kasus in der lteren Upanis.ad-literatur. Beitrge zur Kunde der indogermanischen Sprachen 30: 161.
Wenzel, Heinrich. 1879. ber den Instrumentalis im Rigveda.
Tbingen: Laupp.
Whitney, William Dwight. 1883. Remarks upon M. Bloomfields
paper on certain irregular Vedic subjunctives or imperatives.
Journal of the American Oriental Society 11: clxiiclxiv.
. 1889. Sanskrit grammar: including both the Classical language, and the older dialects, of Veda and Brahmana. 2nd ed.
Cambridge, Mass.: Harvard University Press. [Reprint: 1971.]
. 1892a. On Delbrcks Vedic syntax. American Journal of
Philology 13: 271306.
. 1892b. On the narrative use of imperfect and perfect in the
Brahman.as. Transactions of the American Philological Association 23: 534.
B IBLIOGRAPHY
469
Wilhelm, Eugen. 1873. De infinitivi linguarum sanscritae, bactricae, persicae, graecae, oscae, umbricae, latinae, goticae
forma et usu. Eisenach: Bachmeister.
Willman-Grabowska, Helena. 1928. Le locatif dans le Rig-Veda.
Thse complmentaire. Paris: Varsovie.
Windisch, Ernst. 1869. Untersuchungen ber den Ursprung des
Relativpronomens in den indogermanischen Sprachen. Studien zur griechischen und lateinischen Grammatik, ed. by
Georg Curtius, vol. 2.2, pp. 201419. Leipzig: S. Hirzel.
Witzel, Michael. 1989. Tracing the Vedic dialects. Dialectes
dans les litteratures indo-aryennes, ed. by Colette Caillat,
pp. 97265. Paris: Institut de Civilisation Indienne.
Wolff, Fritz. 1905. Die Infinitive des Indischen und Iranischen, 1:
die ablativisch-genetivischen und die accusativischen Infinitive. Gtersloh: Bertelsmann.
. 1906. Zur Frage des Akkusativs mit dem Infinitiv. Zeitschrift
fr vergleichende Sprachforschung 39: 490500.
Wst, Walther. 1923. Der Schaltsatz im Rgveda. Ph.D. Disserta
tion. Mnchen: Ludwig-Maximilians-Universitt.
. 1928. Stilgeschichte und Chronologie des Rgveda. Abhandlun Leipzig.
gen fr die Kunde des Morgenlandes 17.4.
Zakharyin, B. A. 1998. Pa:ninis ktv:-type formations in Sanskrit and Hindi. Vagbharat: proceedings of the International
Conference on South Asian Languages (July 14, 1997), ed. by
L. V. Khokhlova and Atul Sawani, pp. 22134. Moscow: Institute of Asian and African Studies, Moscow State University.
Zehnder, Thomas. 2011. Zur Funktion der Infinitive im Veda. Indogermanistik und Linguistik im Dialog: Akten der XIII. Fachtagung der Indogermanischen Gesellschaft, ed. by Thomas
Krisch and Thomas Lindner, pp. 63238. Wiesbaden: L. Reichert.
Ziegler, Sabine. 2002. Zur Entstehung des Locativus Absolutus im Altindischen. Indogermanische Syntax: Fragen und
470
S ANSKRIT SYNTAX
Perspektiven, ed. by Heinrich Hettrich and Jeong-Soo Kim,

pp. 7886. Wiesbaden: L. Reichert.
Zubat, Joseph. 1907. Die man-Stze. Zeitschrift fr vergleichende Sprachforschung 40: 748519.
Zwolanek, Rene. 1970. Va yav ndrasca: Studien zu Anrufungsformen im Vedischen, Avestischen und Griechischen. Mnchener Studien zur Sprachwissenschaft, Beihefte 5. Mnchen:
Kitzinger.
As..ta dhyay sutra index

A. 1.2.64, 13, 94
A. 1.2.66, 122, 143
A. 1.2.73, 13
A. 1.3.1, 79
A. 1.3.12, 130, 159, 161,
185188
A. 1.3.13, 161, 182
A. 1.3.14, 162, 185
A. 1.3.15, 120, 185, 186
A. 1.3.16, 163, 182, 185
A. 1.3.17, 162, 164
A. 1.3.18, 100
A. 1.3.19, 175, 176
A. 1.3.2, 161, 181
A. 1.3.20, 184
A. 1.3.23, 162, 179
A. 1.3.4, 178
A. 1.3.43, 163
A. 1.3.62, 119, 120, 135, 136
A. 1.3.63, 135, 136, 176
A. 1.3.72, 119, 187, 188
A. 1.3.77, 185, 188
A. 1.3.78, 162, 185, 187, 188
A. 1.3.80, 187
A. 1.3.83, 186, 187
A. 1.3.84, 187, 189
A. 1.3.85, 163, 165, 188, 189
A. 1.3.9, 161, 181
A. 1.3.93, 159, 161
A. 1.1.2, 131
A. 1.1.21, 121
A. 1.1.23, 121
A. 1.1.26, 385
A. 1.1.27, 113
A. 1.1.3, 97
A. 1.1.37, 89, 131, 230
A. 1.1.46, 93
A. 1.1.49, 89, 9196, 166
A. 1.1.5, 123
A. 1.1.51, 130
A. 1.1.52, 97
A. 1.1.56, 111, 113, 124,
128, 134, 137, 144
A. 1.1.62, 59
A. 1.1.64, 68, 130
A. 1.1.66, 72, 9496, 99, 166
A. 1.1.67, 9497, 99, 100,
166
A. 1.1.72, 97
A. 1.2.1, 123, 124
A. 1.2.11, 98
A. 1.2.19, 123, 124, 126
A. 1.2.4, 123, 124, 126, 127
A. 1.2.45, 227, 349
A. 1.2.46, 59, 130, 228
A. 1.2.51, 351353
A. 1.2.52, 353, 355
A. 1.2.53, 351
471
472
A. 1.4.1, 77
A. 1.4.10, 329
A. 1.4.100, 130, 158
A. 1.4.103, 130
A. 1.4.104, 8
A. 1.4.105, 99, 356
A. 1.4.107, 8, 99, 356
A. 1.4.108, 99
A. 1.4.11, 329
A. 1.4.12, 329
A. 1.4.13, 111
A. 1.4.14, 60, 228
A. 1.4.2, 77
A. 1.4.21, 58
A. 1.4.22, 58, 129, 130
A. 1.4.23, 357
A. 1.4.24, 78
A. 1.4.42, 66, 77, 385
A. 1.4.46, 362
A. 1.4.49, 27, 129
A. 1.4.50, 27
A. 1.4.51, 26, 27, 391, 392
A. 1.4.52, 27, 30, 31
A. 1.4.53, 27, 30
A. 1.4.54, 130
A. 1.4.56, 230
A. 1.4.57, 129, 230
A. 1.4.59, 230
A. 1.4.80, 100
A. 1.4.99, 158
A. 2.1.1, 32, 57, 59, 351
A. 2.1.2, 118
A. 2.1.3, 59
S UTRA
INDEX
A. 2.1.31, 371
A. 2.1.36, 351
A. 2.1.4, 33
A. 2.1.49, 350
A. 2.1.50, 374, 383
A. 2.1.57, 374
A. 2.2.18, 388
A. 2.2.25, 374
A. 2.2.8, 59
A. 2.3.1, 66, 89, 350,
358
A. 2.3.10, 359
A. 2.3.14, 394
A. 2.3.16, 358
A. 2.3.18, 66, 385
A. 2.3.2, 350, 378
A. 2.3.23, 385
A. 2.3.29, 95
A. 2.3.36, 70
A. 2.3.37, 70, 98, 99,
376
A. 2.3.4, 358
A. 2.3.40, 359
A. 2.3.41, 94
A. 2.3.43, 359
A. 2.3.44, 359
A. 2.3.46, 62, 63, 129,
350, 358
A. 2.3.50, 6163, 91,
384
A. 2.3.72, 130
A. 2.3.8, 359
A. 2.3.9, 359
357,
375,
130,
149,
S ANSKRIT SYNTAX
A. 2.4.26, 146
A. 2.4.29, 351
A. 2.4.30, 369
A. 2.4.64, 122, 143
A. 2.4.71, 130
A. 2.4.82, 131
A. 3.1.1, 72, 99, 114, 270
A. 3.1.16, 228
A. 3.1.2, 99, 270
A. 3.1.26, 6971
A. 3.1.3, 121
A. 3.1.35, 135
A. 3.1.40, 176
A. 3.1.67, 125
A. 3.1.68, 130
A. 3.1.87, 124, 126, 134
A. 3.1.91, 72, 100, 171
A. 3.1.97, 99, 100
A. 3.2.123, 130, 158, 171
A. 3.2.124, 25
A. 3.2.16, 60
A. 3.2.37, 18
A. 3.2.97, 60
A. 3.3.1, 171
A. 3.3.132, 147
A. 3.3.139, 10, 71, 72
A. 3.3.140, 10, 71
A. 3.3.156, 10, 71, 73
A. 3.3.18, 133
A. 3.4.111, 83
A. 3.4.112, 83
A. 3.4.18, 23
A. 3.4.19, 23
473
A. 3.4.20, 21, 23
A. 3.4.21, 20, 23, 378
A. 3.4.69, 125, 126, 130,
158, 356, 358
A. 3.4.70, 25
A. 3.4.71, 25
A. 3.4.77, 130, 134
A. 3.4.78, 130, 158, 185
A. 3.4.79, 83, 130
A. 3.4.85, 134
A. 4.1.1, 100, 349
A. 4.1.16, 143
A. 4.1.21, 369
A. 4.1.3, 70, 71, 143, 351
A. 4.1.4, 68, 99, 100
A. 4.1.76, 114, 129
A. 4.1.82, 114, 129
A. 4.1.83, 114
A. 4.2.124, 144
A. 4.2.24, 115, 116
A. 4.2.30, 115
A. 4.2.34, 115117, 144
A. 4.2.46, 144, 147
A. 4.2.81, 352
A. 4.2.82, 353
A. 4.2.92, 113115
A. 4.3.11, 114, 116
A. 4.3.156, 144
A. 4.3.167, 353
A. 4.3.17, 114
A. 4.3.53, 114117
A. 4.3.54, 114116
A. 4.3.80, 144
474
A. 4.4.2, 114
A. 5.1.115, 129, 130, 133
135, 138, 140, 149
A. 5.1.116, 132, 134, 135,
137, 139, 149
A. 5.1.96, 144
A. 5.3.98, 353
A. 5.4.22, 144
A. 5.4.73, 374
A. 6.1.103, 143
A. 6.1.107, 130
A. 6.1.125, 148
A. 6.1.129, 148
A. 6.1.162, 173, 181
A. 6.1.198, 118
A. 6.1.72, 96
A. 6.1.77, 95, 96
A. 6.1.87, 112, 131
A. 6.1.88, 112
A. 6.1.97, 83, 130
A. 6.2.148, 118
A. 6.2.172, 149
A. 6.2.175, 149
A. 6.3.109, 366
A. 6.3.14, 60
A. 6.3.34, 117, 145
A. 6.3.42, 145
A. 6.4.1, 111
A. 6.4.22, 145, 148
A. 6.4.34, 90, 91
A. 6.4.66, 123
A. 6.4.71, 93
A. 6.4.72, 93
S UTRA
INDEX
A. 7.1.12, 112, 130

A. 7.1.13, 112
A. 7.1.14, 112
A. 7.1.2, 229, 230
A. 7.1.33, 67
A. 7.1.41, 83
A. 7.1.46, 82
A. 7.1.5, 83
A. 7.1.52, 67
A. 7.1.8, 83
A. 7.1.85, 92
A. 7.1.86, 92
A. 7.1.87, 92
A. 7.1.88, 92
A. 7.1.9, 112
A. 7.1.91, 124
A. 7.1.95, 137140
A. 7.1.97, 138, 140
A. 7.2.1, 97, 98
A. 7.2.102, 130
A. 7.2.103, 112
A. 7.2.114, 98
A. 7.2.115, 124
A. 7.2.90, 67
A. 7.3.102, 112
A. 7.3.50, 114
A. 7.3.51, 114
A. 7.3.86, 98, 130
A. 8.1.1, 146, 148
A. 8.1.11, 148
A. 8.1.28, 95, 96
A. 8.1.8, 228
A. 8.1.9, 146
S ANSKRIT SYNTAX
A. 8.2.39, 131
A. 8.2.7, 60
A. 8.2.82, 228
A. 8.3.18, 93
A. 8.3.19, 93
A. 8.3.22, 93
A. 8.3.23, 131
A. 8.3.59, 175, 176
A. 8.4.1, 97
A. 8.4.2, 98
A. 8.4.58, 131
A. 8.4.59, 131
475
476
S UTRA
INDEX
Author index
Aalto, Pentti, 399
Abhyankar, Vasudev Kashinath,
150, 151
Ajotikar, Anuja, 168, 195
Ajotikar, Tanuja, 168, 195
Aklujkar, Ashok, 195, 399
Albino, Marcos, 399
Amarendramohan, 264, 266
Ananthanarayana, H. S., 399
Andersen, Paul Kent, 399,
400
Andrews Avery Delano, III,
10, 50
Apte, Vaman Shivaram, 7,
264, 288, 301, 302,
325, 328, 329, 332,
342, 345, 400
Aralikatti, R. N., 400
Arnold, Doug, 15, 50
Arnold, Edward V., 401
Aryavaraguru,
Jagannathaswamy,
150, 152
Aufrecht, Theodor, 151, 152
Auroux, Sylvain, 223, 224,
233
Avery, John, 401
Bader, Franoise, 401
Baldi, Philip, 401
477
Balles, Irene, 401

Barth, A., 402
Bartholomae, Christian, 402
Bardal, Jhanna, 402
Basu, D. N., 402
Behaghel, Otto, 402
Behera, Laxmidhar, 170, 196
Benfey, Theodor, 402
Benigny, Julius, 402
Benveniste, mile, 402, 403
Bergaigne, Abel, 403
Bhagvat, V. S., 371, 397
Bhandarkar, Ramkrishna Gopal,
403
Bharadwaj, Sudhi Kant, 403
Bharati, Akshar, 279, 302
arya,
Bhat.t.anathaswamy, Ach
150, 152
Bhate, Saroja, 225, 226, 228,
233, 234
Bhatia, Medhavi, 279, 302
Bhatnagar, V., 403
Bhatt, V. M., 403
Bichlmeier, Harald, 403
Biese, Yrj Moses Jalmari,
403
Bloch, Jules, 40, 50, 403,
404
478
Bloomfield, Leonard, 348,
360
Bloomfield, Maurice, 404
Boas, Franz, 233, 234
Bodewitz, H. W., 404
Bodirsky, M., 282, 302
Bhtlingk, Otto von, 157,
160, 195, 404
Bolkestein, A. Machteld, 5,
50
Boose, Emery R., 465
Borooah, Anundoram, 14,
404
Brereton, Joel P., 2, 404
Breunis, Andries, 404
Brockington, J. L., 404
Bronkhorst, Johannes, 102,
103, 405
Brown, La, 223, 234
Brugmann, Karl, 405
Bubenik, Vit, 405
Bunker, Ralph, 322
Burnouf, Eugne, 2, 406
Butt, Miriam, 406
Caland, W., 406
Canedo, Jos, 406
Cappeller, Carl, 160, 198
Cardona, George, 8, 19, 24,
50, 54, 61, 70, 76
78, 89, 91, 94, 98,
102104, 151, 152,
AUTHOR INDEX
166, 168170, 195,
406, 407
Casaretto, Antje, 407, 408,
424
Chaitanya, Vineet, 279, 302
Channing, Eva, 408
Charpentier, Jarl, 408
Chen, Shu-Fen, 408
Cherry, C., 272, 304
Chung, Sandra, 205, 206,
216
Colombat, Bernard, 223, 234
Comrie, Bernard, 205, 215,
216
Costello, John R., 408
Culotta, A., 272, 302
Cuny, Albert, 408
Dahl, Eystein, 408, 409
Dangarikar, Chaitali, 197,
323
Dasgupta, Probal, 10, 50
Dash, Siniruddha, 409
Dave, J. H., 150, 152
Davison, Alice, 12, 409
Debrunner, Albert, 409, 467
Delbrck, Bertold, 4, 10, 24,
38, 44, 207, 216,
409, 410
Deshpande, Khanderao, 150,
154
Deshpande, Madhav Murlidhar, v, 14, 8, 24,
S ANSKRIT SYNTAX
26, 31, 87, 104,
232, 234, 410412
Devasthali, G. V., 412
Dharurkar, Chinmay, 141,
152
Disterheft (Haas), Dorothy,
412
Disterheft, Dorothy, 412
Dressler, Wolfgang, 412, 413
Dryer, Matthew S., 223, 234
Dunkel, George E., 413
Durkin, Desmond, 414
Dwarika Das Shastri, 102,
104, 151, 152
Eaton, A. J., 414
Edgerton, Franklin, 404
Elizarenkova, Tatyana J., 414
Emeneau, Murray B., 40,
414
Etter, Annemarie, 414
Eythrsson, Thrhallur, 402
Fahs, A., 414
Fahs, Achim, 414
Farmer, Anne, 5, 51
Fedriani, Chiara, 409
Filliozat, Pierre, 414
Forssman, Bernhard, 415
Franke, Otto, 415
Friedrich, Paul, 415
Frst, Alfons, 415
Gaedicke, Carl, 415
479
Gang, Sh, 259, 265
Garca-Ramn, 415
Ghatage, A. M., 351, 360
Ghosal, S. N., 415
Ghosh, Manomohan, 102,
104
Gillon, Brendan S., 4, 5, 7,
32, 33, 167, 196,
240, 256, 258, 261,
265, 271, 286, 288,
289, 302, 307, 322,
409, 415, 416
Gippert, Jost, 417
Giridharasarma, 151, 152
Gnoli, Raniero, 264, 265
Godabole, Narayan.a Balakr.shn.a,
264, 265
Gode, Parshuram Krishna,
325, 328, 329, 332,
342, 345
Gonda, Jan, 35, 43, 417419
Goto, Toshifumi, 419
Govindacharya, 302
Goyal, Pawan, 164, 170,
196, 310, 311, 322,
344, 345
Grace E. C., Jr., 420
Green, Alexander, 420
Greenbaum, Sidney, 266
Greenberg, Joseph H., 203,
204, 216
Grgoire, A., 420
Gren-Eklund, Gunilla, 420
480
Grice, H. P, 47, 51
Gune, Pandurang D., 420
Hackstein, Olav, 420
Haghighi, A. D., 272, 302
Hahn, E. Adelaide, 420
Haiman, John, 40, 51
Hajicov, E., 279, 304
Hale, Kenneth, 5, 10, 51
Hale, Mark Robert, 35, 37,
421
Halle, Morris, 17, 33, 51
Hamp, Eric P., 421, 422
Hartman, C. G., 422
Hartmann, P., 422
Harweg, Roland, 422
Haspelmath, Martin, 223,
233, 234
Haudry, Jean, 422
Havelka, J., 282, 303
Havers, Wilhelm, 422
Hegde, Janardan, 362, 370,
397
Hejib, Alaka, 422
Hellwig, Oliver, 308, 322
Hendriksen, Hans, 422, 423
Hermann, Eduard, 423
Hertel, Johannes, 310, 322,
342, 345
Herzog, E., 423
Hettrich, Heinrich, 3, 28, 36,
423, 424, 465
Hiersche, Rolf, 424
AUTHOR INDEX
Hinber, Oskar von, 424
Hirt, Hermann, 424
Hock, Hans Henrich, v, 1
4, 810, 12, 14, 16,
19, 20, 24, 25, 27,
30, 31, 3538, 40,
41, 43, 44, 47, 203,
212214, 216, 399,
412, 425430
Hfer, K., 430
Hoenigswald, Henry M., 430
Hoffmann, Karl, 430, 431
Holland, Gary B., 431
Holtzmann, Adolf, 431
Hook, Peter Edwin, 24, 432
Hopkins, E. W., 432, 433
Houben, Jan E. M., 7477,
8082, 86, 105
Huddleston, Rodney, 253,
266
Hudson, R., 279, 283, 303
Hueckstedt, Robert, 215, 216
Huet, Grard, 105, 164, 196,
311, 322, 323, 344,
345
Humbach, Helmut, 433
Hyman, Malcolm D., 166,
167, 174, 196, 200,
327, 346
Ickler, Ingeborg, 433
Ickler, Nancy, 431
Ingalls, David, 305, 309, 323
S ANSKRIT SYNTAX
Insler, Stanley, 433
Jackendoff, Ray, 205, 206,
217
Jacobi, Hermann, 433
Jamison, Stephanie W., 2,
35, 404, 433, 434
Jeffers, R., 435
Jere, Atmaram Narayan, 276,
303
Jha, Girish Nath, 166, 197
Jha, R., 369, 397
Jijnasu, B., 394, 397
Job, Michael, 435
Johnson, Cynthia, 15, 51
Johnson, David E., 435
Johnson, Helen, 364, 397
Johnson, Keith, 364, 397
Jolly, Julius, 435
Joseph, Brian D., 435
Joseph, George Gheverghese, 327, 346
Joshi, Ganeshashastri Ambadas, 150, 151
Joshi, K. R., 277, 278, 303
Joshi, S. D., 19, 51, 88, 105,
170, 197, 349, 360,
436
Joshi, Shivram Dattatray, 264,
265
Kale, Moreshwar Ramchandra, 436
481
Kalika Prasad Shukla, 102,
104, 151, 152
Kane, Pandurang Vaman,
264, 266
Kantor, R., 435
Karve, Cintamana Ganesa,
325, 328, 329, 332,
342, 345
Kashikar, C. G., 102, 107,
233, 235
Katira, Dipesh, 141, 152
Katuri, Shivakumari, 166,
198
Kedarabhat.t.a, 327, 329, 346
Keenan, Edward L, 10, 51
Kehler, Andrew, 247, 266
Keith, A. Berriedale, 436
Keydana, Gtz, 37, 436, 437
Kieckers, Ernst, 437
Kielhorn, Lorenz Franz, 102,
105, 150, 153
Kiparsky, Paul, 11, 33, 167,
170, 197, 437, 438
Klaiman, M. H., 438
Klein, Jared S., 438442
Knobl, Werner, 442
Kobayashi, Masato, 442
Krisch, Thomas, 443
Kmmel, Martin Joachim,
445
Kuhlmann, M., 282, 302
Kuiper, F. B. J., 443
482
Kulikov, Leonid, 20, 443
445
Kulkarni, Amba, 105, 166,
170, 196, 198, 303,
308, 312, 322, 323
Kulkarni, Malhar, 141, 152,
166, 197, 323
Kumar, B. S., 445
Kupfer, Katharina, 446
Kurzov, Helena, 446
Lagarde, Jean-Pierre, 223,
234
Lahiri, P. C., 4, 38, 446
Lakshmi Narasimham, S.,
141, 153
Lallot, Jean, 222, 223, 234
Lanman, C. R., 446
Lazcano, Elisabeth, 234
Lazzeroni, Romano, 446
Leech, Geoffrey, 266
Lehmann, Christian, 10, 446
Lehmann, Winfred P., 446,
447
Leumann, Ernst, 160, 198
Liebert, Gsta, 447
Liebich, Bruno, 447
Lin, Dekang, 271, 303
Lindner, Bruno, 447
Ling, Vivian, 259, 266
Longacre, Robert E., 40, 52
Longobardi, Giuseppe, 447
Lowe, John J., 447
AUTHOR INDEX
Ludwig, Alfred, 447, 448
Lhr, Rosemarie, 448
Luis, Jos, 415
MacCartney, Bill, 272, 303
Macdonell, Arthur Anthony,
449
Mangal Deva Shastri, 102,
106
Manning, C. D., 272, 302
Manning, Christopher D.,
272, 303
Marantz, Alec, 17, 33, 51
Marcus, S., 282, 303
Marneffe, Marie-Catherine
de, 272, 303
Matilal, Bimal Krishna, 449
Mayr, Aurel, 449
McCloskey, James, 205, 206,
216
Meenakshi, K., 426, 449,
509
Meillet, Antoine, 449
Melazzo, Roberta, 449
Melcuk, I., 279, 304
Menezes, A., 272, 304
Michelson, Truman, 450
Miehle, Helen Louise, 450
Migron, Saul, 40, 450
Miller, D. Gary, 450
Minard, Armand, 10, 450
Mishra, A., 363, 398
S ANSKRIT SYNTAX
Mishra, Anand, 167, 197,
198, 326, 346
Mishra, Sudhir K., 166, 197
Misra, Bankelala, 150, 153
Misra, Muraldhara, 151,
153
Mhl, M., 282, 302
Mohanan, K. P., 19, 466
Molina Muoz, Adriana, 33,
450
Monier-Williams, Monier, 160,
198
Mumm, Peter Arnold, 450,
451
Munro, Pamela, 40, 51
Narasimhacharya, M. S., 102,
106
Negelein, Julius von, 451
Ng, A. Y., 272, 302
Nivre, J., 282, 304
Nooten, B. A. van, 414, 451
Oberlies, Thomas, 157, 160,
198, 451
Oertel, Hanns, 451453
rterer, Georg, 454
Oettinger, Norbert, 453
Ogawa, Hideyo, 195
Oguibnine, Boris, 453
Oldenberg, Hermann, 453,
454
Ollett, Andrew, 327, 346
483
Oranskaya, T., 454
Ostler, Nicholas D. M., 27,
454
Padhye, D. G., 150, 154
Palmer, F. R., 348, 360
Pandit, Shankar Pandurang,
264, 266
Panevov, J., 279, 304
Papke, Julia Kay Porter, 454
Paramesvaranandasarma, 151,
152
Patel, Dhaval, 166, 198
Patyal, Hukam Chand, 454
Payne, John, 253, 266
Pepicello, W. J., 435
Petersen, Wiebke, 167, 198
Peterson, Peter, 264, 266
Pinault, Georges-Jean, 454,
455
Pisani, Vittore, 455
Pokar, Sheetal, 303, 308, 323
Pontillo, Tiziana, 94, 106
Porzig, W., 455
Pray, Bruce, 455
Proferes, Theodore, 455
Puhvel, Jaan, 455
Pullum, Geoffrey K, 5, 51
Quirk, C., 272, 304
Quirk, Randolph, 250, 266
Radford, Andrew, 206, 217
Raghunatha Sarm
a, 102, 106
484
Raghunathacarya, E. B., 363,
398
Raja, K. Kunjunni, 276, 304
Raja, Kunjunni, 455
Ramakrishnamacharyulu, K.
V., 312, 323
Rangacharya, Rao Bahadur
M., 151, 153
Ranganath, S., 363, 398
Raster, Peter, 455
Ratanajoti, Undirapola, 447
Rau, Wilhelm, 102, 106
Rgnier, Adolphe, 233, 235
Reinhl, Uta, 456
Renou, Louis, 456, 457
Richter, Oswald, 457
Rocher, Ludo, 457
Roodbergen, J. A. F., 19, 51,
264, 265, 349, 360
Ross, John Robert, 5, 51
Roth, Rudolf von, 157, 160,
195
Rouse, W. H. D., 457
Ruppel, Antonia, 457
Sadler, Louisa, 15, 50
Sadovski, Velizar, 457
Salomon, Richard, 457, 458
Sam
. kr.tyayana, Rahula, 264,
266
Sangal, Rajeev, 279, 302
Sarup, Lakshman, 151, 153,
224, 225, 233, 235
AUTHOR INDEX
Sastri, Gangadhara, 274, 304
Satuluri, Pavan Kumar, 166,
198
Satya Vrat Shastri, 458
Saussure, Ferdinand de, 458
Schufele, Steven, 4, 5, 458,
459
Scharf, Peter, 322
Scharf, Peter M., 76, 88, 107,
166171, 174, 184,
189, 195, 196, 199,
200, 310, 323, 327,
342, 346, 458
Scharfe, Hartmut, 458
Scherer, Anton, 459
Schmidt, Gernot, 459
Schmitt-Brandt, Robert, 459
Schneider, Carolin, 424, 459,
460
Segall, P., 279, 304
Seiler, Hansjakob, 460
Sellmer, Sven, 305, 309, 324
Sen, N. M., 460
Sen, Sukumar, 460
Sgall, Petr, 460
Shaer, Benjamin, 5, 7, 256,
265, 307, 322, 416
Sharma, Arvind, 422
Sharma, Aryendra, 150, 154
Shastri, Charudev., 362, 366
368, 370, 372, 375,
376, 398
S ANSKRIT SYNTAX
Shukl, Devanand, 303, 308,
323
Shukla, H. L., 363, 398
Siecke, Ernst Ludwig, 460
Sinha, Anjani C., 461
Sitaram Shastri, 150, 151,
154
Slade, Benjamin, 408
Sleator, D., 279, 304
Snijders, Liselotte, 5, 52
Shnen, Renate, 461
Sommer, Ferdinand, 461
Sonatakke, N. S., 102, 107,
233, 235
Sorensen, J., 272, 302
Speijer, J. S., 4, 10, 1416,
1921, 24, 25, 30,
38, 270, 304, 361,
365, 398, 461
Srivastav, Veneeta, 10, 52
Staal, J. F., 5, 7, 207, 214,
217, 270, 304, 307,
308, 324, 438, 461
Staal, J. Frits, 166, 200
Stchoupak, Nadine, 461
Storck, Frid. Guil. Paul, 461
Strunk, Klaus, 462
Stump, Gregory T., 462
Sturtevant, Edgar H., 462
Subbanna, Sridhar, 170, 200
Subramania Iyer, K. A., 102,
107, 129, 151, 154
Subramaniam, P. S., 462
485
Svartik, Jan, 266
Swamiji, Shivamurthy, 166,
201
Szemernyi, Oswald, 462
Suszkiewicz, E., 461
Taranatha, 264, 266
Taraporewala, Irach J. S.,
462
Telang, Kashinath Trimbak,
264, 267
Temperley, D., 279, 304
Tesnire, L., 279, 304
Thieme, Paul, 463
Thomas, E. J., 463
Thommen, Eduard, 463
Thompson, Sandra A., 40, 52
Thornton, Randolph, 463
Thumb, Albert, 463
Thurneysen, Rudolph, 463
Tichy, Eva, 463, 464
Tikkanen, Bertil, 464, 465
Toporov, V. N., 414
Trivedi, Kamlashankar Pranashankar,
151, 154
Tsiang, Sarah, 36, 43, 465
Tsiang-Starcevic, Sarah, 36,
465
Tubb, Gary A., 465
Tucker, Elizabeth, 442
Vale, R. N., 465
486
Van de Walle, Lieve, 27, 35,
45, 465, 466
Varakhedi, Srinivas, 170,
200
Varma, Siddheshwar, 466
Vasu, S. C., 394, 398
Vedavrata, 102, 107, 150,
151, 154
Vekerdi, J., 466
Velankar, H. D., 325, 327,
328, 346
Vendryes, J., 466
Vennemann, Theo, 204, 217
Verma, Manindra K., 19, 466
Verpoorten, J.-M., 457, 466,
467, 509
Vijayapala, 102, 107
Villavicencio, Aline, 15, 50
Vine, Brent, 20, 25, 467
Viti, Carlotta, 467
Voyles, Joseph B., 467
Wackernagel, Jacob, 467
Wallace, William D., 45,
215, 217, 468
Ward, Gregory, 247, 266
Watanabe, Albert, 36, 43,
465
Watkins, Calvert, 468
Wecker, Otto, 468
Wenzel, Heinrich, 468
AUTHOR INDEX
Whitney, William Dwight,
32, 157, 201, 373,
398, 468
Wilhelm, Eugen, 469
Willman-Grabowska, Helena,
469
Windisch, Ernst, 469
Witzel, Michael, 469
Wolff, Fritz, 469
Wst, Walther, 469
Yoon, James Hye Suk, 17, 52
Yoshida, Kazuhiko, 442
Zakharyin, B. A., 20, 23, 24,
469
Zehnder, Thomas, 469
Ziegler, Sabine, 469
Zubat, Joseph, 470
Zwolanek, Rene, 470
Title index
A propos du subjonctif vdique, 456
Ablativ, localis, instrumentalis im altindischen, lateinischen,
griechischen und deutschen, 409
The absolute construction in Indo-European, 408
Absolute constructions in early Indo-European, 457
Absolute Konstruktionen in altindogermanischen Sprachen, 436
Labsolutif sanskrit en -am, 456
arya-Dharmakrteh. Praman.avarttikam (svarthanumana-paricchedah.)
Ac
Svopajavr.ttya Karn.akagomi-viracitaya tat..tkaya ca sahitam,
264, 266
Der Accusativ im Veda, 415
Adhyaya 2 Pada 14, 102, 106
Adjectives in Sanskrit, 409
Adjunction, features, and locality in Sanskrit and Hindi-Urdu correlatives, 12, 409
Adpositions authentic and inauthentic, 458
Adverbial clauses, 40, 52
Adverbial repetition in the Rigveda, 441
Der Agens in passivischen Stzen altindogermanischer Sprachen,
423
Ai. Gen.-Abl. der Feminina auf -as in dativischer Funktion, 452
Algebraic linguistics, 282, 303
Altavestisch yas . . . cica jeder, der, 453
Altind. pura und griech. pros mit dem Indikativ des Prsens, 405
Altindisch tava, 463
Altindisch sma, Teil 1, 451
Die altindische Cvi-Konstruktion, 401
Altindische Grammatik, 467
Altindische Infinitive auf -man und -mani, 402
487
488
T ITLE INDEX
Altindische Nominalbildung nach den Samhit

as, 447
Altindische Syntax, 4, 10, 24, 44, 410
Altindische Tempuslehre, 410
Die altindische Wortfolge aus dem Satapathabr

ahman.a dargestellt,
4, 38, 207, 216, 410
Die altindischen Absolutiva, besonders im Rgveda, Aitareya- und
Satapatha-br
ahman.a, 420
Amplified sentences in Asoka, 399
Amred
. ita and iteration of preverbs in Vedic and Hittite, 413
Amreditas and related constellations in the Rigveda, 440
An analytic database of the As.t.a dhyay, 167, 200
An outline of syntax of Buddhistic Sanskrit, 460
An unusual use of the nominative, 436
An XML formalization of the As.t.a dhyay, 189, 200
Analogy in generative grammar, 204, 217
Annotating Sanskrit texts based on sa bdabodha systems, 312, 323
Aorist v Rigvede, 414
Apposition and nominal classification in Indo-European and beyond, 420
Apposition and word-order typology in Indo-European, 420
Are taddhita affixes provided after pratipadikas or padas?, 168,
200
The argument structure of experience, 409
Arguments against a passive origin of the IA ergative, 438
The Aryan future, 432
Aspect and event structure in Vedic, 437
Aspects of Indo-European poetics, 468
Aspects of the rhetorical poetics of the Rigveda, 440
Aspects of the Vedic dative, 432
The aspectual functions of the R.gvedic present and aorist, 419
As..ta dhyay 3.3.158 and the notion of subject in Pan.ini, 432
As..ta dhyay 3.4.21 and the role of semantics in Paninian linguistics,
24, 432
S ANSKRIT SYNTAX
489
The As..ta dhyay of Pan.ini, 349, 360

As..ta dhyay-bhas.ya-prathamavrtti [prathama-trtyadhyayatmaka
prathama bhaga], 394, 397

tha, dha, and a typology of Rigvedic conjunction, 438
The a tmanepada in the Rigveda, 414
Attraction and coordination in the Veda, 418
Aufstze zur Indoiranistik, 430, 431
Autonomy of word formation, 33, 240, 265, 416
Autour de a rg-vdique, 413
B. Delbrcks Arbeiten zur Wortstellung aus heutiger Sicht, 443

Background and Introduction, 102, 103, 151, 152, 169, 195
Background of the As..ta dhyay, 88, 105
Beitrge aus dem Rgveda zur Accentuirung des Verbum finitum,
449
Beitrge zur Syntax der Palisprache, 414
Beitrge zur vergleichenden Casuslehre des Zend und Sanskrit,
454
Bemerkungen zum Gebrauch der Pronomina der 1. und 2. Person
als Subjekt im Altindischen, 35, 417
Betrachtungen zur Wortstellung im Altindischen, 443
Bhartr.hari as a cognitive linguist, 74, 105
Bhartr.hari on taddhita formations involving comparison, 129, 154
Bhartrharis rule for unexpressed karakas, 261, 265
Bhartrharis solution to the problem of asamartha compounds, 32,
416
33,
Bhartr.haris Vakyapadya, 102, 106
Bhas.a pakah., 370, 397
Bhat..tikavi, 150, 153
Bhat..tikavyam of Sr
A bibliography of writings on Sanskrit syntax, v, 13, 412
Brhat-Sabdendu
sekhara by Nagesa Bhat..ta, 150, 154
ca if, when, ce(d) and wor(l)ds upside-down?, 451
490
T ITLE INDEX
Le cas employs linfinitif en indo-europen, 449

Case disharmony in R.gvedic similes, 434
A case for the genitive case, 402
The case of the agent in Indo-European, 434
Case-linking, 27, 454
Cases and prepositions in Indo-Aryan, 405
Die casuslehre der indischen grammatiker verglichen mit dem gebrauch der casus im Aitareya-Brhman.a, 447
Die casuslehre des Pan.ini verglichen mit dem gebrauch der casus
im Pali und in den Asoka-inschriften, 415
Casuum in lingua palica formatio comparata cum sancscritae linguae ratione, 461
Catena and climax in Vedic prose, 40, 450
Causees, passive agents, or instruments? Instrumental NPs with
causatives in early and later Vedic Prose, 427
The character of the Indo-European moods, 418
The character of the Sanskrit accusative, 418
Chronology or genre? Problems in Vedic syntax, 9, 35, 37, 427
Classical Sanskrit, wild trees and the properties of free word order languages, 5, 7, 256, 265, 307, 322, 416
Classical Sanskrit preverb ordering, 454
Classification of terms, 225, 226, 228, 234
Classifications of terms, 233
Clause-initial dvaym, 458
The cleft sentence in Vedic Prose, 450
Clitic verbs in PIE or discourse-based verb fronting?, 35, 37, 425
Coexistent analyses and participant roles in Indo-Aryan, 432
Comparative concepts and descriptive categories in crosslinguistic studies, 223, 233, 234
Competitive Indo-European syntax, 448
Le complment du nom dans le Satapathabr

ahman.a, 461
Completeness analysis of a Sanskrit reader, 164, 196
A comprehensive grammar of the English language, 250, 266
S ANSKRIT SYNTAX
491
Compositum und Nebensatz, 433

Computational structure of the As.t.a dhyay and conflict resolution
techniques, 170, 200
Computer simulation of As..ta dhyay, 170, 196
Conjoined we stand, 8, 10, 12, 203, 212214, 216, 426
Conjunctive sequences in the Rigveda, 442
Conjunctive u and invariable s in the Rgveda, 413, 434
in Vedic Sanskrit, 431
A constraint on the position of the relative
Constraints on ellipsis and event reference, 247, 266
Constraints on non-projective dependency parsing, 282, 304
Constraints on variables in syntax, 5, 51
Context-sensitive rules in Pan.ini, 166, 200
Contributions to the history of verb inflection in Sanskrit, 401
Coordinate conjunction, 439
Corpus reprsentatif des grammaires et des traditions linguistiques, 234
Correlative clauses and IE syntactic reconstruction, 438
Covert absolutivity in Vedic, 432
The crest of the peacock, 327, 346
A critical survey of the publications on the periphrastic future in
Sanskrit, 418
The dative of agency, 420
Die Dativi finales abstrakter Nomina und andere Beispiele nominaler Satzfgung in der vedischen Prosa, 453
De genetivi in lingua sanscritica, imprimis vedica usu, 460
De infinitivi linguarum sanscritae, bactricae, persicae, graecae,
oscae, umbricae, latinae, goticae forma et usu, 469
De lemploi du gnitif absolu en sanscrit, 458
La dcadence et la disparition du subjonctif, 456
Deep structure and surface structure in Pan.ini, 462
Defining the nominative, 418
Definiteness and relativization in the Rigveda, 431
492
T ITLE INDEX
Deictic fronting in Vedic Prose, 421

Deictic pronoun repetition in the Rigveda, 441
Die Demonstrativpronomina der indogermanischen Sprachen, 405
Die Demonstrativpronomina im Rigveda, 446
Denys le thrace, 223, 234
Dependency syntax, 279, 304
Dependency tree kernels for relation extraction, 272, 302
Dependency treelet translation, 272, 304
Dependency-based evaluation of MINIPAR, 271, 303
Deriving Wackernagels Law, 37, 421
Die Desiderativbildungen der indoiranischen Sprachen, 408
Design of a lean interface for Sanskrit corpus annotation, 311,
322, 344, 345
Designing a constraint based parser for Sanskrit, 303, 308, 323
A deterministic dependency parser with dynamic programming for
Sanskrit, 303
Development of conditional clauses in early Sanskrit, 446
The development of the genitive absolute in Sanskrit, 457
A diachronic phonological solution to the syntax of Vedic negative
particles, 435
The diachronic relationship of the IA ergative and passive constructions, 438
The diachronic syntax of the particle u in the Rigveda, 438
The discourse function of the absolutive in the Pacatantra, 465
The discourse functions of subordinate constructions in Classical
Sanskrit narrative texts, 36, 465
Discourse linkage in Sanskrit narratives with special emphasis on
the story of Nala, 41, 427
Distributed Morphology and the pieces of inflection, 17, 33, 51
A distributed platform for Sanskrit processing, 310, 322
Distribution de la particule negative n dans la Rk-Samhit
a, 455
langue de la RkDistribution des particules comparatives dans la
Samhit
a, 455
S ANSKRIT SYNTAX
493
Ditransitive passive in Pan.ini, 412

Dravidian and Indo-Aryan, 40, 414
Dvandvas, blocking, and the associative, 437
lments de syntaxe structurale, 279, 304
Ellipsis and syntactic overlapping, 410
Ellipsis, brachylogy, and other forms of brevity in speech in the
Rgveda, 418
in Paninian grammar, 411
Ellipsis
.
The elliptical sentence, 455
Emphasizing and connecting particles in the thirteen principal upanishads, 422
Lemploi des cas en vdique, 422
Emploi et analyse des adverbes comparatifs sanskrits en -vt, 454
Encyclopedic dictionary of applied linguistics, 364, 397
Epic syntax, 426, 449, 509
Essai sur la construction grammaticale considre dans son
dveloppement historique en sanskrit, en grec, en latin, dans
les langues romanes et dans les langues germaniques, 403
estai, 461
tudes sur la grammaire vdique, 233, 235
tudes vdiques et pan.inennes, 457
Evidence for evidentiality in Late Vedic, 409
Evolution of syntactic theory in Sanskrit grammar, 410
Exocentric (bahuvrihi) compounds in Classical Sanskrit, 258, 265
Exocentric (bahuvrhi) compounds in Classical Sanskrit, 416
Experiencer Subjects in South Asian Languages, 19, 466
Explanations for some syntactic phenomena of Proto-IndoEuropean, 446
The expression of purpose in Indo-European, 435
The extended siddha-principle, 170, 197
Extracting dependency trees from Sanskrit texts, 308, 322
494
T ITLE INDEX
Finale Dative auf -ya im Rgveda, 399

A formal re-presentation
of the As..ta dhyay, 167, 198
Les formes dites dinjonctif dans le rgveda, 456
Fortfhrung von Relativstzen im Avestischen,

415
Four studies in the language of the Veda, 418
Die Fragestze im Rgveda, 414
phrase structure rules, 5, 51
Free word order and
Free word-order syntax, 4, 5, 458
From Pan.inian sandhi to finite state calculus, 166, 196
From passive to ergative in Indo-Aryan, 455
From topic to subject in Indo-European, 446
From uttarapadalopa to madhyamapadalopa, 410
Function and form in the -aya-formations of the Rigveda and
Atharvaveda, 434
Functional ambiguity and syntactic change, 433
A functional toolkit for morphological and phonological processing, 344, 345
Funktionen des Akkusativs und Rektionsarten des Verbums anhand
des Altindoarischen, 419
Funzioni semantiche e pragmatiche nelle strategie di possesso
dellantico indiano, 467
Gab es im Indogermanischen Nebenstze?, 423
The gan.acchandas in the Indian metrical tradition, 327, 346
Gan.akas..ta dhyay, 166, 201
Gaps in grammar and culture, 5, 10, 51
A GB approach to the syntax of classical Sanskrit, 415
Der gebrauch der kasus in der lteren Upanis.ad-literatur, 468
Der Gebrauch des Conjunctivs und Optativs im Sanskrit und
Griechischen, 410
Der Gebrauch von as- und bhuim Aitareya-Brahman.a, 403
Generating typed dependency parses from phrase structure parses,
272, 303
S ANSKRIT SYNTAX
495
Generation of Sanskrit compounds, 166, 198

Generative semantics and Pan.inis karakas, 461
The genesis of a linguistic area, 443
The genitive agent in Rigvedic passive constructions, 400
The genitive in Pan.ini and in Epic Sanskrit, 449
The genitive in the role of the non-genitive, 415
The Genitivus personae with verbs of eating . . . and accepting
. . . in Vedic Prose, 452
Genre, discourse, and syntax in Sanskrit, 35, 428
Le grondif sanscrit en tva, 402
Geschichte des Infinitivs im Indogermanischen, 435
Government, barriers, and small clauses in modern Irish, 205,
206, 216
Grammaire de la langue vdique, 456
La grammaire gnrale et les fondements philosophiques des
classements de mots, 223, 224, 233
Grammaire sanscrite, 457
Grammar, 348, 360
A grammar of Epic Sanskrit, 157, 160, 198, 451
Grammatica et verba Glamour and verve, 408
Grammatical irregularities in the R.gveda, book IV, 419
Grammatical notes, 436
Grammaticization and configurationality, 456
Die Grammatik der Partikeln, 413
Grammatik des Pali, 414
Die grammatische Kategorie Passiv im Altindischen, 400
Grammatisches aus dem Mahabharata, 431
Handbook of American Indian languages, 233, 234
The Harshacarita of Ban.abhat..ta, 264, 266
A higher Sanskrit grammar, 14, 404, 436
Histoire de la langue sanskrite, 457
Historical syntax of Middle Indo-Aryan, 460
496
T ITLE INDEX
A history of the Sanskrit gerund, 435

Hittite iwar, wa(r) and Sanskrit iva, 435
Homoioteleuton in the Rigveda, 440
How strict is Strict OV?, 8, 428
Die Hymnen des Rigveda, 1, 453
Die Hypotaxe im Rigveda, 1, 455
I bahuvrhi del Rig Veda, 449
IE hortatory *ey, *eyte, 413
Impersonal constructions in functional grammar, 405
Inaccusativit indoeuropea e alternanza causative vedica, 446
The independent reader, 259, 266
Indian theories of meaning, 276, 304
Indian theorists on the nature of the sentence, 449
Indic across the millennia, 442
Indirekte Rede im Altindischen, 409
Die indische Grammatiktradition, 455
Indo-Aryan and Dravidian convergence, 458
Indo-Aryan from the Vedas to modern times, 404
Lindo-aryen du vda aux temps modernes, 403
Indo-European, 450
Indo-European Caland adjectives in *-nt- and participles in Sanskrit, 447
Indo-European linguistics, 417419
Indo-European origins of Germanic syntax, 11, 437
Indo-European word order in main and subordinate clauses in a
diachronic perspective, 417
Indogermanisch -qu e als alte nebensatzeinleitende Konjunktion,
467
Indogermanische Wortstellung, 443
The Indo-Iranian construction mana (mama) krtam, 406
The Indo-Iranian predicate infinitive, 412

Infinite syntax., 5, 51
S ANSKRIT SYNTAX
497
Les infinitifs avestiques, 402

Les infinitifs de lAvesta, 420
Infinitifs et drivs nominaux dans le r.gveda, 456
Der Infinitiv im Veda, 447
The infinitive and participle in Indo-European, 467
Die Infinitive des Indischen und Iranischen, 1, 469
Die Infinitive im Rgveda, 460
subject accusative, 420
The infinitive with
The infinitives of the Indo-European languages, 435
Infinitivos y abstractos verbales en indoiranio, 415
Information structure and the particles va and ev in Vedic Prose,
442
The information structure of OVS in Vedic, 467
Der Injunktiv im Veda, 431
The interaction of word order and pragmatics in a Sanskrit text,
45, 215, 217, 468
Interrogative sequences in the Rigveda, 441
Interstanzaic repetition in the Rigveda, 442
Die Inversion von Subjekt und Prdikat im Indischen, 433
Issues concerning constraints on discontinuous NPs in Latin, 5, 52
Issues in Sanskrit agreement, 14, 429
abara-Bhas.ya, 467
Iti dans le Jaiminya-Brahman.a et dans le S
J. Wackernagel und die idg. Partikeln *s, *ke, *kem und *an, 413
Jayadaman, 325, 327, 328, 346
Jussive sentences in Sanskrit syntax, 405
Kadambar by Bana and his son, 264, 266
Ein Kapitel vergleichender Syntax, 435
The karaka theory and case grammar, 399
Karikaval, 276, 303
Kasika, 150, 154
Kasmrasabdamrta 8.3.3, 432
498
T ITLE INDEX
tarhi in the Mahabhas.ya, 407

kim
Kompositum und Katalysationstext vornehmlich im spten Sanskrit, 422
Konditionalstze im Satapathabr
ahman.a, 414
Der Konjunktiv und seine Nachbarkategorien, 464
Konkurrierender Gebrauch obliquer Kasus im Rgveda, 424
Die Konstruktion von svara in der vedischen Prosa,

452
Die Konstruktionen des Absolutivs im Sanskrit, 461
Kuru kings, Tura Kavaseya, and the -tvaya gerund, 455
Language, 348, 360
Language and linguistic area, 414
Language and style of the Vedic rs.is, 414
in Ancient India, 445
Language vs. grammatical tradition
Language-death phenomena in Sanskrit, 425
Latente Objekte und altindische Diskursgrammatik, 436
Levels in Pan.inis As..ta dhyay, 76, 107, 167, 199
The light verb jungle, 406
Linguistic analysis and some Indian traditions, 70, 103
A linguistic analysis of the Mun.d.aka Upanis.ad, 457
Linguistic archaisms of the Ramayan.a, 450
Linguistic investigations into ellipsis in Classical Sanskrit, 416
Linguistic issues and intelligent technological solutions in encoding Sanskrit, 167, 200
Linguistic issues in encoding Sanskrit, 167, 174, 200, 327, 346
Linguistic study of dharmasutras, 403
Linguistic, syntactic, and stylistic peculiarities of the Brahman.as,
454
Linguistic typology, 205, 216
Linguistics in India, 458
Local particles in the R.gveda, part XVII, 460
Le locatif dans le Rig-Veda, 469
Logic and conversation, 47, 51
S ANSKRIT SYNTAX
499
Logic, language, and reality, 449

Lokalpartikel im R.gveda (Folge 13): pa VII: ns, 459
Lokalpartikeln im Rgveda XIV, 407
The loss of the subjunctive

mood in yadi-conditionals from Vedic
to Epic Sanskrit, 462
Madhavya Dhatuvrtti canonical index, 167, 171, 199
The Mahabharata,305, 309, 323
Mahabhas.ya-Dpika of Bhartr.hari, 102, 103
Les matres de la philologie vdique, 456
Die man-Stze, 470
Manu-Smriti with Nine Commentaries by Medhatithi, Sarvajanarayan.a, Kulluka, Raghavananda, Nandana, Ramacandra, Man.irama, Govindaraja, and Bharuci, 150, 152
Materialien zu einer Kasussyntax des Rgveda, 3, 424
The meaning of nouns, 87, 104, 232, 234

The meaning of the sentence in its semantic and pragmatic aspects,
279, 304
Meaning statements in Pan.inis grammar: on the purpose and
context of the As..ta dhyay, 74, 76, 77, 8082, 86, 105
Means of expressing a comparison of inequality in Old Indic, 399,
400
The medium in the Rgveda, 419
a vruna ?, 433
Mitra vrun.a or mitr
.
Mixed-gender antecedent agreement in Latin, 15, 51
Modeling Pan.inian grammar, 166, 167, 171, 184, 199
Modelling As..ta dhyay, 167, 198
Modelling the grammatical circle of the Pan.inian system of Sanskrit grammar, 167, 197
Modern Sanskrit Literature, 363, 398
Modern Sanskrit literature, tradition & innovations, 363, 398
500
T ITLE INDEX
Mudrarakshasa written by Visa khadatta, with Commentary of

D
. hund.hiraja (written in 1713 AD) edited with critical and explanatory notes, 264, 267
Die Namen fr Satzinhalte im Griechischen und im Indogermanischen, 455
Narrative linkage in Sanskrit, 40, 43, 430
Narrative linkage in the Mahabharata, 41, 427
Natural language processing, 279, 302
Nature and means of comparison in Proto-Indo-European, 455
Der Nebensatz und seine Konkurrenten in der Indogermania, 448
Negated participles in R.gvedic Sanskrit and Proto-Indo-European,
447
The negative particle n when used with the aorist system in the
Rg-Veda, 422
and extraclausality in Vedic, or sa-fig all over again,
Nexus
428
The Nighan..tu and the Nirukta, 151, 153, 224, 225, 233, 235
Die Nomina agentis auf -tar- im Vedischen, 464
Nominal and adverbial amreditas and the etymology of Rgvedic
nana, 440
Nominal and pronominal objects in Sanskrit and Prakrit, 405
Nominal composition, noun incorporation and non-finite formations in Sanskrit, 445
The nominal sentence in Sanskrit and Middle Indo-Aryan, 404
Nominal sentences and the origin of absolute constructions in
Indo-European, 431
Nominale Ausdrucksformen im wissenschaftlichen Sanskrit, 422
Nominalkomposition im Altindischen und Altgriechischen, 448
Noms dagent et noms daction en indo-europen, 402
A note on dative agents in Sanskrit, 24, 407
A note on Indo-Iranian dha (IE *dhe), 406
S ANSKRIT SYNTAX
501
A note on the complement structures in sentences in modern spoken Sanskrit as recorded on tapes, 400
A note on the functions of the accusative, 418
A note on the Sanskrit gerund, 457
A note on Vakyapadya 1.45/46: a tmabhedas tayoh. kecid . . . , 54,
104
A note on word order in modern spoken Sanskrit and some positive
constraints, 400
Notes de grammaire vdique, 1, 456
Notes on reconstruction, word order, and stress, 430
Notes on syntax, 436
Notes on the instrumental case of the subject/agent vs. other cases
in Buddhist Sanskrit, 453
Notes on the use of cases in the Ka.thaka Samhit
a, 460
Notes on Vedic syntax, 436
Noun inflection in the Veda, 446
Nouns and noun phrases, 253, 266
Numeral repetition in the Rigveda, 441
arya Jinendrabuddhipada and PadamaNyasa or Pacika of Ac
jar of Haradatta Misra on the Kasikavr.tti [Commentary on
the As..ta dhyay of Pan.ini] of Vamana-Jayaditya, 102, 104, 151,
152
Nyaya-darsanam with Vatsyayanas Bhas.ya, Uddyotakaras
Varttika, Vacaspati Misras Tatparyat.ka and Visvanathas
Vr.tti, 264, 266
Nyayasiddhantamuktaval, 277, 278, 303
Oblique subjects in Sanskrit?, 19, 20, 24, 426
Old Indian, 419
Old Indic clauses between subordination and coordination, 448
The Old Indo-Aryan tense system, 407
On a paragraph of Vaidic syntax, 406
On amplifed sentences and similar structures in the Veda, 43, 418
502
T ITLE INDEX
On certain irregular Vedic subjunctives or imperatives, 404

On comparatives and superlatives formed to finite verbs in Sanskrit, 407
On Delbrcks Vedic syntax, 468
On explaining double object constructions, 27, 466
On identifying the conceptual restructuring of passive as ergative
in Indo-Aryan, 432
On instability in the use of moods in earliest Sanskrit, 404
On methodology in syntactic reconstruction, 435
On modes in relative clauses in the Rig-Veda, 401
On negative clauses in the Rigveda, 408
On Pan.inis metalinguistic use of cases, 89, 91, 94, 103, 166, 195
On past tense and verbal aspect in the Rgveda, 466
On relative clauses in the Rig-Veda, 401

On Sanskrit and the derived languages, 403
On the syntax of silence in Proto-Indo-European, 443
On the absolute instrumental in Sanskrit, 399
On the architecture of Pan.inis grammar, 33, 167, 197, 437
On the construction type nat.asya srn.oti, 407
Old and Middle Indo-Aryan
On the evolutionary changes in the
systems of case and adpositions, 405
On the expression of reflexive possession in the Rig-Veda, 20, 25,
467
On the interaction of morphology and syntax, 5, 51
On the meaning of Sanskrit bhavati, 423
On the narrative use of imperfect and perfect in the Brahman.as,
468
On the nature and function of preverb repetition in the Rigveda,
440
On the problem of the development of tenses in Old-Indo-Aryan
(the R.gveda), 414
On the realisation of the category of definiteness-indefiniteness in
the Vedic Sanskrit, 454
S ANSKRIT SYNTAX
503
On the semantic foundation of Pan.inian derivational procedure,

167, 199
On the semantic foundation of Pan.inian derivational procedure:
the derivation of kumbhakara, 88, 107
On the structure of Pan.inis system, 76, 103
On the syntax of cases in the Ramayan.a, 460
On the syntax of Sanskrit gerund constructions, 464
On the tense value, functions, and forms of the Sanskrit gerund,
464
On the use of the absolutive in Sanskrit, 419
On the variable position of the finite verb in oldest Sanskrit, 404
On the word order typology of the Satapathabr

ahman.a, 399
On translating and formalizing Pan.inian rules, 166, 195
On verbal accentuation in the Rigveda, 439
The order of constituents in Indo-European, 420
Lordre des mots dans lAitareya-Brahman.a, 457, 466, 509
The origin and syntax of the Rigvedic construction y- (. . . )
k/ / - ca, 439
The origin of the ablative case, 433
The origin of the oblique subject construction, 402
The origin of the Rigvedic va yav ndras ca construction, 438
The origin of the Sanskrit passive aorist, 433
The original character of the Indo-European relative pronoun io-,
417
w
The original syntax of conjunctive *-k e, 413
Origines et dveloppement de la thorie des parties du discours en
grce, 222, 234
Die Ortsnamen-Parenthese im Altpersischen und Vedischen, 430
Outline of the historical grammar of the Vedas, 401
The Pali language, 414
The Pacakhyanaka of Purn.abhadra, 310, 323, 342, 346
The Pacatantra and Aesops Fables, 36, 43, 465
504
T ITLE INDEX
The Panchatantra, 310, 322, 342, 345

Panels of the VIIth World Sanskrit Conference, Kern Institute, Leiden, August 2329, 1987, 405
Pan.ini 5.3.5 and the function of Sanskrit etd, 445
Pan.ini: A survey of research, 8, 19, 50
Pan.ini and Pan.inyas on ses.a relations, 61, 103, 104
Pan.ini and the northwestern dialect, 411
Pan.ini and the Veda, 412
Pan.ini and the Vedic evidence, 410
Pan.ini as a frontier grammarian, 4, 31, 410
Pan.ini, vivaks.a , and karaka-rule-ordering, 170, 199
Pan.inian accounts of the class eight presents, 168, 199
Pan.inian accounts of the Vedic subjunctive, 168, 199
The Pan.inian conception of karmakartr, 414
Paninian grammar framework appliedto English, 279, 302
Pan.inian reflections on Vedic infinitives, 412
Pan.inian reflections on Vedic infinitives: On infinitives with -tosUN and -KasUN, 411
Pan.inian sutras of the type anyebhyo pi drsyate, 77, 103
Pan.inian syntax and changing notion of sentence,

8, 412
Pan.inian syntax and the changing notion of sentence, 26, 411
Pan.inian syntax of gerund constructions, 411
. a or the Siks
. a Vedan ga ascribed to Pan.ini (being the
Pan.inya Siks
most ancient work on Indo-Aryan phonetics), 102, 104
Pan.inye atidesa nuslanam, 141, 153
Paninis as..ta dhyay, 24, 432
Pan.inis As.t.a dhyay and linguistic theory, 167, 196
Pan.inis concept of a sentence, 455
Pan.inis grammar and its computerization, 74, 75, 105
Pan.inis karakas, 78, 103
Pan.inis karakas, 406
Pa:ninis ktv:-type formations in Sanskrit and Hindi, 20, 23, 24,
469
S ANSKRIT SYNTAX
505
Parataxe, hypotaxe et corrlation dans la phrase latine, 422

Parsing English with a link grammar, 279, 304
Participles as finite verbs, 436
The particle iva in Vedic Prose, 404
The particle u in the Rigveda, 438
Les parties du discours dans la linguistique moderne et contemporaine, 223, 234
Les parties du discours (partes orationis) et la reconstruction
dune syntaxe latine au XVIe sicle, 223, 234
Le passage du nom daction linfinitif dans le R.gveda, 456
Passive and middle in Indo-European, 444
Passive, anticausative, and classification of verbs, 444
Passivized causatives in Sanskrit and Prakrits, 405
Patajalis Vyakaran.a-Mahabhas.ya Paspasa hnika, introduction,
text, translation and notes, 264, 265
Patajalis Vyakaran.a-Mahabhas.ya: Vibhaktyahnika (P. 2.3.18
2.3.45), 19, 51
Patterns of verb repetition in the Rigveda, 440
Peculiarities in the use of iti, 436
Das Perfekt im Indoiranischen, 445
Performative sentences in Vedic, 408
Personal pronoun sequences in the Rigveda, 441
Phonological overgeneration in Pan.inian system, 166, 197
The phonology-syntax interface in the language of the Rig Veda,
421
Phrasal repetition in the Rigveda, 440
La phrase nominale, 403
La phrase nominale en indo-europen, 449
La phrase nominale en sanskrit, 403
La place de la particule ngative na dans la phrase en vieil indien,
417
Das Plusquamperfektum im Veda, 463
Polyptoton and paronomasia in the Rigveda, 440
506
T ITLE INDEX
P-oriented constructions in Sanskrit, 19, 20, 25, 426

Portuguese: Corpora, coordination and agreement, 15, 50
Possessive agents in Sanskrit?, 19, 20, 24, 427
Prdikativa im Altindischen, 436
Pragmatic motivation of the passive voice in Classical Sanskrit,
465
Pragmatics and Classical Sanskrit, 35, 45, 466
Pragmatics of absolute locative in Sanskrit, 403
Prakriyakaumud by Ramacandracarya, 151, 153
The Prakriykaumud of Rmachandra, 151, 154
Prakriyapradars.in., 166, 198
Praman.avarttikam, 264, 265
Bhat..toji Dkshita, 151, 154
Praud.hamanorama of Sr
Predicate order in Vedic Prose, 8, 44, 429
Preliminaries to the study of the relationship between syntax and
sandhi in Rigvedic Sanskrit, 421
(Pre-)Rig-Vedic convergence of Indo-Aryan with Dravidian? Another look at the evidence, 8, 425
Prtrit et optatif en indo-europen, 403
The preterite optative in Indic, 458
Preverb repetition, 413
Les prverbes dans le atapathabrahman.a, 408
The prohibitive use of na in lieu of ma with the augmentless aorist
in the RgVeda, 422
of historical linguistics, 426
Principles
Principles of syntactic reconstruction and morphology as paleosyntax, 401
Proceedings of the Fifth International Sanskrit Computational
Linguistics Symposium, 197, 323
Projectivity in totally ordered rooted trees, 282, 303
Das Projekt einer Kasussyntax des Rgveda. Der Instrumental, 424
425
Pronominal reflexivization in Sanskrit,
S ANSKRIT SYNTAX
507
Pronoun fronting and the notion verb-second position in Beowulf, 425

Proto-Indo-European compounds in relation to other Proto-IndoEuropean syntactic patterns, 446
Proto-Indo-European syntax, 415, 446
Proto-Indo-European verb-finality, 38, 429
Prototypes in Pan.inian syntax, 411
Purvatrasiddham and a srayat siddham, 170, 195
Quelques observations sur lemploi du verb simple au lieu dun
compos etc. dans la langue sanskrite, 417
Questions and relative and complement clauses in a Bangla grammar, 10, 50
The quotative in Indo-Aryan, 449
The Ramayan.a, 458
Ramopakhyana the story of Rama in the Mahabharata, 310, 323
Ramus, Petrus, 223, 234
Random scrambling? Constraints on discontinuity in Latin noun
phrases, 5, 50
rng wo-mn li d Lao Z (Let us read Laozi), 259, 265
Reciprocal constructions in Vedic, 444
The reconstruction of non-simple sentences in Proto-IndoEuropean, 446
The reconstruction of particles and syntax, 421
Reduced-clause and clause-union absolutives and participles in
Vedic Prose, 19, 20, 24, 25, 426
A reexamination of the perfect and the imperfect in Vedic Sanskrit,
450
Reflections on the Indo-European medium, 419
The reflexive pronouns in Vedic, 20, 444
Reflexivization in the Rig-Veda (and beyond), 20, 428
508
T ITLE INDEX
The relation of the Indian grammatical tradition to modern linguistics, 3, 8, 430

Relations between causatives and passives in Indo-Iranian, 406
Relative and antecedent, 420
Relative clauses, 10, 51
Relative clauses in Hindi and learnability, 10, 52
Relative constructions in the Rig-Veda, book five, 450
Relative pronoun sequences in the Rigveda, 441
Relatives in Indo-European and Hittite, 462
Les relatives nominales indoeuropnnes, 447
Der Relativsatz, 10, 446
Der Relativsatz in den indoeuropischen Sprachen, 446
Relativization and word order in Old Irish and Vedic Sanskrit, 431
Relativsatz, Attribut und Apposition, 460
Remarks on Indo-European infinitives, 435
Remarks on the expression of agency with the passive in Vedic and
Indo-European, 434
Remarks on the Sanskrit passive, 417
Remarks upon M. Bloomfields paper on certain irregular Vedic
subjunctives or imperatives, 468
Remarques gnrales sur la phrase vdique, 457
Remarques sur la place du verbe dans la phrase active et moyenne
en langue sanscrite, 417
Repetition and deletion of preverbs and verbs in early Indic and
Greek, 413
Research on Sanskrit syntax, 3, 426
Responsion in the Rigveda, 440
Retrospektivitt im Rigveda, 451
Review of Ch. Lehmann 1984, 423
Review of Haudry 1977, 407, 414
Review of Meenakshi 1983, 426
Review of Verpoorten 1977, 457
Review of Klein 1978, 422
S ANSKRIT SYNTAX
509
Revised and enlarged edition of Prin. V. S. Aptes The practical

Sanskrit-English dictionary, 325, 328, 329, 332, 342, 345
Rgveda, 453
R.gveda-Samhit
a, 102, 107, 233, 235
Rgvedic clitics and prosodic movement, 447
The Rgvedic stanza as a syntactical unit, 450

Rigvedic t and s, 438
Robust textual inference via graph matching, 272, 302
The roots, verb-forms, and primary derivatives of the Sanskrit language, 157, 201
Rule selection in the As.t.a dhyay or Is Pan.inis grammar mechanistic?, 170, 199
Rule-blocking and forward-looking conditions in the computational modeling of Pan.inian derivation, 166, 169, 199
The Rupavatara of Dharmakrti, 151, 153
Russian, 215, 216
Sa fig and Indo-European deixis, 439
Sabd
apasabdaviveka, 362, 398
Sakuntala, 264, 265
Sanskrit and Pan.ini Core and periphery, 4, 31, 429
Sanskrit ca, Indo-European *kw e, and the semantics of coordinate
conjoined structures in the Rigveda, 438
Sanskrit causative syntax, 4, 31, 425
Sanskrit compounds and the architecture of the grammar, 33, 450
Sanskrit computational linguistics, 105, 196, 197, 323
Sanskrit double-object constructions, 27, 30, 425
The Sanskrit gerund, 464
Sanskrit gerund constructions, 410
Sanskrit grammar, 32, 373, 398, 468
A Sanskrit grammar for students, 449
Sanskrit, grammatical and philological studies, 417419
Sanskrit metre recognizer, 326, 346
510
T ITLE INDEX
The Sanskrit particle api, 419

The Sanskrit passive, 19, 20, 425
The Sanskrit passive in lexical-functional grammar, 438
The Sanskrit quotative, 425
Sanskrit studies of M. B. Emeneau, 414, 451
Sanskrit syntactic particles, 414
Sanskrit syntax, 4, 10, 1416, 1921, 24, 25, 30, 38, 270, 304, 361,
365, 398, 461, 462
Sanskrit syntax and the grammar of case, 445
Sanskrit usage, 458
The Sanskrit -yet-optative, 444
A Sanskrit-English dictionary, 160, 198
Sanskrit-Wrterbuch, 157, 160, 195
Der Schaltsatz im Rgveda, 469
465
Scholastic Sanskrit,
Selected studies / J. Gonda, 419
Semantic functions eligible for subject position, 465
Semantic processing in Pan.inis karaka system, 166, 197
Semantische und pragmatisch-kontextuelle Faktoren in der Entwicklung des altindoarischen Perfekts, 408
Semantische und syntaktische Betrachtungen zum doppelten
Akkusativ, 28, 423
Sense and syntax in Vedic, 2, 404
Sentence structure according to Pan.ini, 436
La sparation du prverbe et du verbe en vdique, 456
Sequential negation in the Rigveda, 441
A short history of Vedic prefix-verb compound accentuation, 428
The Siddhanta Kaumud of Bhhat..toji Dks.ita, 394, 398
Simulating the Pan.inian system of Sanskrit grammar, 167, 198
Single-word topicalization in Vedic Prose, 4, 459
Der sog. Nominativus pendens, 422
Der sogenannte genitivus temporis im Veda, 402
Some aspects of Sanskrit agreement, 16, 428
S ANSKRIT SYNTAX
511
Some features of the sampradana karaka in Pan.ini, 24, 411

Some Hittite-Sanskrit parallels, 420
Some intersections of semantics and syntax in Vedic, 434
Some issues in the computational implementation of the As..ta dhyay, 168, 195
Some linguistic superstitions, 463
Some notes on Indo-European double direct-object constructions,
27, 30, 31, 429
Some notes on the position of the attributive adjective in early Indian prose, 419
Some observations on intersentential pronominalization in the language of the Taittirya Samhit
a, 35, 421
Some observations on relatives and demonstratives in Greek and
Sanskrit, 431
Some peculiarities of Vedic-prose relative clauses, 36, 47, 427
Some principles of Pan.inis grammar, 98, 103, 168, 169, 195
Some problems of Indo-Aryan philology, 40, 50
Some relation-specific issues, 312, 323
Some remarks on Sanskrit and Greek absolutive genitives, 461
Some remarks on the position of adverbials in Greek and Vedic
sentences, 443
Some rhetorical aspects of adjacent interstanzaic phrasal repetition in the Rigveda, 442
Some semantic and pragmatic aspects of object alternation in
early Vedic, 408
Some theoretical problems in Pan.inis grammar, 437
Some universals of grammar with particular reference to the order
of meaningful elements, 203, 204, 216
Der Sprachgebrauch der lteren Upanis.ads verglichen mit dem
der frheren vedischen Perioden und dem des klassischen Sanskrit, 415
Sr
Vyakaran.a-mahabhas.yam, 102,
107, 150, 151, 154
512
T ITLE INDEX
madbhat..tojidks.itaviracita Vaiyakaran.asiddhantakaumud . . . Sr
Sr
madvasudevadks.itapran.taya Balamanoramakhyavyakhyaya
Srimajj
anendrasarasvatviracitaya Tattvabodhinyakhyavyakhyaya ca sanathita, 151, 152
majjaiminipran.te Mmams
Sr
adarsane . . . , 150, 151
vamanajayadityaviracita Pan.inyas..ta dhyaysutravrttih. Kasika,
Sr
102, 107
vis.n.usarmapran.tam
Sr
Pacatantram, 369, 397
Stmme oder Wurzeln im Sanskrit? Primre vs. sekundre Verbalstammbildung und das Kausativ, 3, 429
Stativ und Passivaorist im Indoiranischen, 445
Stilgeschichte und Chronologie des Rgveda, 469
Strategies of subordination in Vedic,467
Structure and implementation of a digital edition of the As..ta dhyay, 167, 198
Structure de lnonc en indo-europen, 401
The students guide to Sanskrit composition, 7, 264, 288, 301, 302,
400
Studien zur Kasussyntax des Pali, besonders des Vinaya-Pit.aka,
424
Studies in Indias Vedic grammarians, 1, 82, 105
Studies in Indian grammarians I, 166, 195
Studies in Rgveda and Modern Sanskrit Literature, 363, 398
Studies in
Sanskrit grammars, 195
Studies in Sanskrit syntax, v, 2, 399, 427
Studies in the comparative syntax of the oldest Indo-Iranian languages, 421
Studies in the syntax of relative and comparative clauses, 10, 50
Studies in the word order of Sanskrit prose, 4, 38, 446
Studies on the present participle, 1, 442
A study of modern spoken Sanskrit with reference to sentence patterns, 400
A study of nominal sentences in the oldest Upanis.ads, 420
S ANSKRIT SYNTAX
513
The style of Ban.a, 215, 216

Stylistic repetition in the Veda, 418
Subject in Pan.ini, 406
Subject predicate order in Classical Sanskrit, 416
Subjunctive and optative, 420
Subordination and relativization in Early Indo-European, 431
La subordination dans la prose vdique, 10, 450
Substratum influence on (Rig-Vedic) Sanskrit?, 425
Suddhikaumud
, 362, 397
Le suffixe -tu et la constitution des infinitifs, 456
Sukti-ratnakara (Commentary on Patajalis Mahabhas.ya) by Ses.a Narayan.a, 371, 397

Sur lemploi de linfinitf au genitif dans quelques langues indoeuropennes, 466
Sur la phrase ngative dans le rgveda, 456
Sur un usage remarquable de linfinitif

samscrit, 2, 406
Switch reference and universal grammar, 40, 51
Syntactic and discourse correlates of verb-initial sentences in the
Rigveda, 439
Syntactic and semantic relations in Panini, 438
A syntactic approach to category-changing phrasal morphology:
Nominalizations in Korean and English, 17, 52
The syntactic development of the infinitive in Indo-European, 412
The syntactic features of speech and context in the Pacatantra,
465
A syntactic rule in Pali and Ardhamagadh, 423
Syntaktische quivalenz des Genitivs und Ablativs bei Verben der
Trennung in der altindischen Prosa, 451
Syntaktisch-exegetische Miszellen, 406
Syntax, 424
Syntax: a minimalist introduction, 206, 217
The syntax and style of the Ramayan.a, 404
Syntax and verse structure in the Veda, 418
514
T ITLE INDEX
Die Syntax des einfachen Satzes im Indogermanischen, 405

Die Syntax des Infinitivs, 423
Syntax des Relativsatzes und pluralischer Instrumental im Avestischen, 453
Syntax, meaning, and origin of the Indo-European particle kw e,
462
The syntax of cases in the narrative and descriptive prose of the
Brahman.as, 451
The syntax of cases in Vedic prose, 460
The syntax of direct speech in Vedic, 35, 434
The syntax of style, 440
Syntax of tenses in the Ramayan.a, 460
Syntax of the dative case in the Rgveda, 466
Syntax of the infinite verb-formsof Pali, 422
Syntax und Wortarten der Lokalpartikeln des R.gveda, I, 423
Syntax und Wortarten der Lokalpartikeln des R.gveda. II, 423
Syntax und Wortarten der Lokalpartikeln des R.gveda. III, 424
Syntax und Wortarten der Lokalpartikeln des Rgveda IV: Allge
meines, II: pa, III: va, 424
Syntax und Wortarten der Lokalpartikeln des Rgveda IX, 407
Syntax und Wortarten der Lokalpartikeln des
R.gveda V, 459
Syntax und Wortarten der Lokalpartikeln des Rgveda VI, 407
R.gveda VII, 459
Syntax und Wortarten der Lokalpartikeln des Rgveda VIII, 407
R.gveda X, 459
Syntax und Wortarten der Lokalpartikeln des Rgveda XII, 408
R.gveda XIX, 460
Syntax und Wortarten der Lokalpartikeln des R.gveda. XV.2, 460
Syntax und Wortarten der Lokalpartikeln des Rgveda XVI, 408
Rgveda XXI, 407
La syntaxe des comparaisons vdiques, 403

The system of coordinate conjunctions in the Rigveda, 438
Systematics of Pan.inis As.t.a dhyay, 351, 360
S ANSKRIT SYNTAX
515
Tantravartikam, 274, 304

Ta-participles with genitive agents in Vedic Sanskrit, 428
Tat tvam asi in context, 404
Teleology and the simplification of accentuation in Pan.inian
derivation, 169, 200
Tense and mood in Indo-European syntax, 437
Tenses and moods in the Ka.thaka Samhit
a, 436
Text in Sutra-form, 102, 106
Texte und Glossar, 463
Eine textsyntaktische Regel der idg. Wortstellung, 413
Thematische Rollen und Kasus, 448
Themes and tasks in Old and Middle Indo-Aryan linguistics, 465
Three myths in modern Pan.inian studies, 74, 75, 82, 105
Time, tense and aspect in Early Vedic grammar, 409
Time, tense and aspect in early Vedic grammar, 409
Towards a relationally based grammar, 435
Towards Proto-Indo-European syntax, 468
Tracing the Vedic dialects, 469
Transitive and causative in Indo-European, 422
Transitivity as a gradient feature? Evidence from Indo-Aryan, especially Sanskrit and Hindi, 426
Transponierte Rollen und Ergnzungen beim vedischen Kausativ,
463
A triadic structural feature of nominal Anaphora in the Rigveda,
439
Trois nigmes sur les cent chemins, 450
Two forms of negations in Sanskrit, 240, 265, 416
Two notes on Sanskrit syntax, 418
The two senses of the term anaphora and their functional unity,
439
Two textual studies of Bhartr-hari, 399
Typische Merkmale von Fragestzen

und die altindische Pluti,
462
516
T ITLE INDEX
Typological and functional aspects of word order in Asoka, 399
Typological syntactical characteristics of the Satapathabr

ahman.a,
447
Typology and the Indo-European prepositions, 401
ber das periphrastische perfect im Sanskrit, 433
Ueber den impersonalen Gebrauch der Participia necess. im Sanskrit, 404
Ueber den indogermanischen, speciell den vedischen Dativ, 409
ber den Instrumentalis im Rigveda, 468
ber den nominalen Stil des wissenschaftlichen Sanskrit, 433
ber die Entstehung des indogermanischen Vokativs, 402
ber die Inifinite auf sanskr. -adhyai, 455
ber die Partikeln s und t im Vedischen, 447
ber die Rekonstruktion der idg. Syntax, 413
ber die absoluten verbalformen des Samskr
. t und den praedicatsauszdruck im allgemeinen, 448
ber ein Gesetz der indogermanischen Wortstellung, 467
ber indogermanische nominale Relativkonstruktionen, 459
Eine bereinstimmung zwischen Vedischem und Avestischem
sprachgebrauch, 406
Ujjvaladattas Commentary on the Un.a disutras, 151, 152
The ukti-vyakti-prakaran.a as a manual of spoken Sanskrit, 458
Die unechten Nominalkomposita des Altindischen und Altiranischen, 457
The unity of the Vedic dative, 419
Un-Pan.inian syntax (concord) in the Ramayan.a and the Mahabharata, 403
Untersuchungen ber den Ursprung des Relativpronomens in den
indogermanischen Sprachen, 469
Untersuchungen zu Sprache und Stil des ltesten Indo-Iranischen
(Veda und Avesta), 457
Untersuchungen zur Hypotaxe im Vedischen, 36, 423
S ANSKRIT SYNTAX
517
Untersuchungen zur Kasussyntax der indogermanischen Sprachen, 422

Untersuchungen zur Komposition von Reden und Gesprchen im
Ramayan.a, 461
Untersuchungen zur Wortkunde und Auslegung des Rigveda, 463
Untersuchungen zur Wortstellung und Syntax der Chandogyopanis.ad, 433
Die ursprngliche Funktion des Konjunktivs, 459
The use of the gerund as passive in Sanskrit, 436
The use of the particle ca, 418
The use of the passive gerund in Sanskrit, 436, 457
Vagvyavaharadarsa, 362, 366368, 370, 372, 375, 376, 398
Vaiyakaran.asiddhantakaumud, 302
Vakya according to the munitraya of Sanskrit grammar, 412
Vakyapadya of Bhartr.hari (An ancient Treatise on the Philosophy
of Sanskrit Grammar) Containing the T.ka of Pun.yaraja and
the Ancient Vr.tti, 102, 107, 151, 154
Vakyapadya of Bhartr.hari with the Commentary of Helaraja, 102,
107
Vakyapadya of Bhartr.hari with the Vr.tti and the Paddhati of
Vr.s.abhadeva, 102, 107
Vakyapadyam [Part III, Vol. II] [Bhuyodravya-Gun.a-Dik-Sadhana-Kriya-Kala-Purus.a-Sankhya-Upagraha and Lingasamuddesa], 102, 106
Valency-changing categories in Indo-Aryan and Indo-European,
445
La valeur du parfait dans les hymns vdiques, 456
Vat, 141, 152
Va yav ndrasca, 470
Ved. dive-dive und die idg. Iterativkomposita, 412
Vedic and Sanskrit historical linguistics, 442
518
T ITLE INDEX
sa payati reconsidered,
The Vedic causative samkhy
apayati / samk
444
Vedic causatives, 407
The Vedic clause-initial string and Universal Grammar, 459
The Vedic dative reconsidered, 432
Vedic ta . . . stvama, 428
A Vedic example of no doubt used as parenthetic clause, 404
A Vedic grammar for students, 449
Vedic indrasca vayo Old Avestan mazdsca ahurNho, 433
Vedic infinitive and Pan.ini, 449
The Vedic injunctive, 437
The Vedic medio-passive aorists, statives and their participles, 444
Vedic mriyte and other pseudo-passives, 443
The Vedic passive optative and its functional equivalents, 434
Vedic preverbs as markers of valency-changing derivations, 445
Vedic studies, 429
Vedic syntax, 411
Vedic variants, 404
Vedic verb accent revisited, 430
The Vedic -ya-presents, 444
The Vedic -ya-presents, 445
Vdique tanu et la notion de personne en indo-iranien, 455
Vedisch d, 464
Die vedische Partikel kila, 433
Vedische und Sanskrit-Syntax, 4, 10, 38, 461
Vedische Untersuchungen, 20, 453
Verbal composition in Indo-Aryan, 465
The verbal focus system of Epic Sanskrit, 463
Verbale Definitheit und der vedische Injunktiv, 450
Das Verbalsystem des Atharvaveda, 451
Verb-medial clauses in Vedic, 4, 459
S ANSKRIT SYNTAX
519
The verbs for and in Walman, a Torricelli language of Papua

New Guinea, 223, 234
Verbum dicendi + Infinitiv im Indoiranischen, 417
Vergleich der indogermanischen Nebensatzkonstruktionen, 459
Vergleichende Syntax der indogermanischen Sprachen, 410
Vikramorvasyam, a drama in five acts by Kalidasa, edited with
English notes, 264, 266
Vim
sasatabdsamskr
takavyamrtam, 363, 398
gerundive
(krtya) in Sanskrit, 426
Voice, mood, and the
The voice of the infinitive in the RigVeda,

412
Vom indogermanischen Tempus/Aspekt-System zum vedischen
Zeitstufensystem, 464
Vom Infinitiv, besonders im Sanskrit, 430
Vorlesungen ber Syntax, 467
vrttaratnakara, 327, 329, 346
Vyakarana = Mahabhasya of Patajali, 102, 105, 150, 153
The
.
.
Vyakaran.amitakshara, a Gloss on Pan.inis Grammatical Apho Annambhat..ta, 150, 152
risms by Sr
Wackernagel in the language of the Rigveda, 37, 437
Wackernagels law and the language of the Rig Veda, 37, 421
Das Wackernagelsche Gesetz aus heutiger Sicht, 443
Warlpiri and the grammar of non-configurational languages, 5, 51
Well-nested drawings as models of syntactic structure, 282, 302
Where the sense is intended although the corresponding speech
unit is not employed, 94, 106
Wheres my NP?, 458
Whos On First, 37, 427
Why syntax needs phonology, 421
Word grammar, 279, 283, 303
Word order in Classical Sanskrit, 4, 7, 256, 265, 271, 286, 288,
289, 302, 307, 322, 416
520
T ITLE INDEX
Word order in Sanskrit and universal grammar, 5, 7, 207, 214,

217, 270, 304, 307, 308, 324, 461
Word order typology and comparative constructions, 400
Word order typology and prepositions in Old Indic, 399
Words in metrical milieux, 305, 309, 324
Die Wortstellung im nachvedischen Altindischen und im Mittelindischen, 463
Wortstellung und Univerbierung altpersischer Korrelativverbindungen, 462
X syntax: a study of phrase structure, 205, 206, 217
Zeitadverbia als Nominative, 409
Zu ai. d (aid) + Akk. siehe (da) in den Brahman.as, 424
Zu den ai. Ellipsen, 453
Zu den Kasusvariationen in der vedischen Prosa, 1, 452
Zu den Wortstellungsvarianten der Mantras des Atharvaveda in
der Saunakaund Paipalada-Rezension und des Samaveda in

der Kauthuma- und Jaiminya-Rezension, 452
Zum prdikativen Infinitiv, 417
Zum Akkusativ der Richtung im Vedischen und im Indogermanischen, 415
Zum altindischen Ausdrucksverstrkungstypus satyasya satyam
das Wahre des Wahren = die Quintessenz des Wahren, 452
Zum disjunkten Gebrauch des Nominativs in der Brahman.aprosa,
451
Zum Gebrauch der w-Demonstrativa im ltesten Indoiranischen,
447
Zum indogermanischen Medium und konkurrierenden Kategorien,
462
Zum indogermanischen Nominalsatz, 402
S ANSKRIT SYNTAX
521
Zum Kasusgebrauch bei Kausativa transitiver Verben, 463

Zum Modalfeld im Altindischen, 448
Zum parenthetischen Nominalsatz im Indischen, 463
Zum prdikativen Adverb, 430
Zum Verhltnis von Wort und Satz in indogermanischen Sprachen.
II., 460
Zur Kasusattraktion in altindischen dativischen Infinitivkonstruktionen, 448
Zur altpersischen Syntax, 415
Zur Entstehung des Locativus Absolutus im Altindischen, 469
Zur Entwicklung der Finalstze altindogermanischer Sprachen,
423
Zur Entwicklung des indischen Satzbaus, 433
Zur Frage des Akkusativs mit dem Infinitiv, 469
Zur Funktion der Infinitive im Veda, 469
Zur Funktion des Perfekts im Rgveda, 435
altindogermanischer Kasus, 424
Zur funktionalen Variationsbreite
Zur Genese und Funktion der altindischen Perfekta mit langem
Reduplikationsvokal, 443
Zur Gerundivbildung im Arischen, 402
Zur Geschichte der altindischen Prosa, 454
Zur historischen Syntax der nomina actionis im R.gveda, 423
Zur oratio recta in den indogermanischen Sprachen 1, 437
Zur oratio recta in den indogermanischen Sprachen 2, 437
Zur Parenthese im Altpersischen, 430
Zur Stellung des Verbs im Germanischen und Indogermanischen,
402
Zur Syntax der ablativischen Adverbien im Rgveda, 403
Zur Syntax der infinitivischen Bildungen in
den indogermanischen
Sprachen, 417
Zur Syntax der Pronomina im Avesta, 406
Zur Syntax des Infinitivs auf -tum im Altindischen, 417
522
T ITLE INDEX
Zur Wort- und Satzstellung in der alt- und mittelindischen Prosa,

406

Sanskrit Syntax

Uploaded by

Copyright:

Available Formats

Sanskrit Syntax

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sanskrit Syntax

Uploaded by

Copyright:

Available Formats

Sanskrit syntax

Selected papers presented at the seminar

Sanskrit, and some worked directly on Sanskrit grammatical texts

in the organization of the Seminar on Sanskrit syntax and discourse

derivational procedure. In the second, he deals with the syntax

the Romanization of Satapathabr

H ANS H ENRICH H OCK

Some issues in Sanskrit syntax

Derivation and interpretation in Pan.inis system

Extension rules and the syntax of As..ta dhyay sutras

Voice, preverb, and transitivity restrictions

Interrogatives and word-order in Sanskrit

To classify words: European and Indian

Constituency and cotextual dependence

How free is free word order in Sanskrit?

Parse trees for erroneous sentences

A bibliography of Sanskrit syntax

As..ta dhyay sutra

Detailed table of contents

Some issues in Sanskrit syntax

Derivation and interpretation in Pan.inis system

Extension rules and the syntax of As..ta dhyay sutras

P ETER S CHARF, PAWAN G OYAL,

Interrogatives and word-order in Sanskrit

To classify words: European and Indian

Appended coordination (stripping) . . . . . . . . . . . . . .

How free is free word order in Sanskrit?

Sannidhi violation involving two

Meter identification of Sanskrit verse

A bibliography of Sanskrit syntax

As..ta dhyay sutra

Derivation and interpretation in Pan.inis system

Voice, preverb, and transitivity restrictions

Constituency and cotextual dependence

To classify words: European and Indian . . . approaches 219

Distinctive features of poetic syntax

Relative position of a direct object with respect to

Parse trees for erroneous sentences

Some issues in Sanskrit syntax

A brief survey of recent publications

The Sanskrit syntax bibliography provided at the end of this book

S OME ISSUES IN S ANSKRIT SYNTAX

pects of Sanskrit linguistics. The following statistics are based on

ent grammatical structures and categories in the language of the

In the following section of my paper I address formal issues that

Free Word Order and related issues

S OME ISSUES IN S ANSKRIT SYNTAX

examples come from my collections.

mate (SB. 1.1.2.13)

S OME ISSUES IN S ANSKRIT SYNTAX

Schufeles approach contrasts with that of Staal and Gillon.

S OME ISSUES IN S ANSKRIT SYNTAX

As in the case of word order, Pan.ini has remarkably little to say

S OME ISSUES IN S ANSKRIT SYNTAX

example similar to (6a) was brought to my attention in 1989 by

gus have praised you, powerful Indra, nevertheless hear

tasyasmarayato vyaktam adharmo n

(7) a. Z!a;ya;Ra;ta;ea h va;!a IR +a;M!a;.ca;ke ya;a;tk+:

dam a padi]CC + iti (SB.

ata thought, Because I have done what?, therefore I

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

(29) a. k+:du b.ra;vaH . . . naX ;n,a

c. ;a;.ca:a-:pra;ma;a;a;Ta;na;a ba;a;l+.a :de;va;a;na;a;ma;