(Cambridge Books Online) Subbārāo, Kārumūri V - South Asian Languages - A Syntactic Typology-Cambridge University Press (2012)

South Asian Languages
South Asian languages are rich in linguistic diversity and number. This book
explores the similarities and differences of sixty languages from the four
different language families (Austro-Asiatic, Dravidian, Indo-Aryan [Indo-
European], and Tibeto-Burman [Sino-Tibetan]). It focuses on the syntactic
typology of these languages and the high degree of syntactic convergence,
with special reference to the notion of “India as a linguistic area.” Several
areas of current theoretical interest such as anaphora, control theory, case and
agreement, relative clauses, and the significance of thematic roles in grammar
are discussed. The analysis presented has significant implications for current
theories of syntax, verbal semantics, first and second language acquisition,
structural language typology, and historical linguistics. The book will be of
interest to linguists working on the description of South Asian languages, as
well as syntacticians wishing to discover more about the common structure
of languages within this region.
k ā r u m ū r i v . s u b b ā r ā o is Radhakrishnan Chair Professor in Humani-

ties at the University of Hyderabad, India. He has been an elected member of
the Linguistic Society of America since 2003 and the Linguistic Society of
Nepal since 2005. He has taught at several universities in India and abroad.
South Asian Languages:
A Syntactic Typology
Kārumūri V. Subbārāo
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, São Paulo, Delhi, Tokyo, Mexico City
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521861489

C K. V. Subbārāo 2012
This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2012
Printed in the United Kingdom at the University Press, Cambridge
A catalogue record for this publication is available from the British Library
Library of Congress Cataloguing in Publication data

Subbārāo, K. V. (Kārumūri V.), 1941–
South Asian languages : a syntactic typology / Kārumūri V. Subbārāo.
p. cm.
Includes index.
ISBN 978-0-521-86148-9
1. South Asia – Languages. 2. South Asia – Languages – Grammar. 3. Grammar,
Comparative and general – Syntax. I. Title.
P381.S58.S83 2012
409.54 – dc23 2011040839
ISBN 978-0-521-86148-9 Hardback
Additional resources for this publication at www.cambridge.org/subbarao
Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.
Dedicated
to my parents
Karumuri Suseela and Karumuri V. Rathnam
&
to my most respected teacher
D. V. Rama Krishna Rao
Contents
Preface page xi
Acknowledgements xii
List of abbreviations xvi
1 Introduction 1
1.1 Languages of South Asia 1
1.2 Aim of the book 2
1.3 Linguistic theory, language universals and language typology 3
1.4 Inductive and deductive approaches to language analysis 4
1.5 Relevance of linguistic theory: some illustrations 5
1.6 The framework 8
1.7 The data 8
1.8 Relevance of the present work 11
1.9 The richness and complexity of the data and their relevance 11
1.10 Contributions of cross-linguistic studies to our understanding of the nature
of language 12
1.11 The binding link for all chapters 16
2 South Asian languages: a preview 18

2.1 Introduction 18
2.2 Typology of South Asian languages 18
2.3 Typological characteristics of SALs 20
2.4 Evidence in support of “South Asia as a linguistic area” 22
2.5 Parametric variation and SALs 27
2.6 The effects of language contact 31
2.7 Unique features of each language family 33
2.8 Conclusion 41
3 Lexical anaphors and pronouns in South Asian languages 43

3.1 Introduction 43
3.2 Form of the anaphor in South Asian languages 44
3.3 The occurrence of the anaphor in different language families 49
3.4 Multi-functionality of the verbal reflexive/reciprocal 58
3.5 Non-nominative subjects as antecedents to a lexical anaphor 68
3.6 The dative subject construction and the verbal clitic 74
3.7 Long-distance binding and Principle A 75
3.8 Small Clauses and long-distance binding 78
vii
viii Contents
3.9 The reflexive in a locative PP 82

3.10 Binding principles A, B and C 83
3.11 Anaphors and scrambling 85
3.12 Blocking Effects 87
3.13 Sub-commanding NP as an antecedent 88
3.14 Reduplication and Case Copying in complex anaphors 89
3.15 Conclusion 91
4 Case and agreement 93

4.1 Introduction 93
4.2 Agreement typology of SALs 95
4.3 Adposition incorporation 109
4.4 Incorporation in polysynthetic languages 111
4.5 SALs and long-distance agreement 114
4.6 Pronominal Strength Hierarchy in agreement 122
4.7 ECM cases (Small Clauses) and long-distance agreement 123
4.8 Language contact and agreement 126
4.9 Case and agreement in SALs 128
4.10 Conclusion 132
5 Non-nominative subjects 134

5.1 Introduction 134
5.2 NNS constructions in SALs 135
5.3 Domains of occurrence of NNSs in SALs, and their nature 146
5.4 Some subject and non-subject properties of the NNS construction 154
5.5 Nature of the predicate in DSCs 170
5.6 Inherent case assignment in DSCs 181
5.7 Double/multiple Dative Subject constructions 189
5.8 Conclusion 192
Appendix – formal representation of NNSs 192
6 Complementation 193
6.2 The position of occurrence of the COMP and its functions 196
6.3 Structural differences between an IC and an FC 204
6.4 Some crucial issues concerning FCs and ICs 212
6.5 Position of occurrence of an IC- and an FC-clause and the Case Resistance
Principle (CRP) 218
6.6 Narrow Scope and Wide Scope 223
6.7 Arguments against Rightward Extraposition 239
6.8 Conclusion 240
Appendix – the Final-over-Final Constraint (FOFC) and diachronic change 240
7 Backward Control 246

7.2 Forward and Backward Control 247
7.3 Evidence for Backward Control 249
7.4 Backward Control in Malagasy 256
7.5 Conclusion 257
7.6 Postscript 258
Appendix – the Case-marked nature of PRO 259
Contents ix
8 Noun modification: relative clauses 263

8.2 Types of relative clauses 265
8.3 Grammatical functions accessible to relativization in sentential relative clauses 271
8.4 The Externally Headed Relative Clause (non-finite) (the EHRC) 277
8.5 The Internally Headed Relative Clause (the IHRC) 295
8.6 Comitative PP as head 299
8.7 Implications of Postposition Incorporation 308
8.8 Conclusion 311
Notes 313
Bibliography 336
Author index 356
Language index 359
General index 362
(Some relevant material concerning the syntactic typology of SALs is placed

on the web at the URL www.cambridge.org/subbarao – hereafter referred to as
“the Website”)
Preface
This book takes forward the work I have been doing for the last thirty years
on the syntactic typology and convergence in South Asian languages. South
Asian languages, many of which are not even well studied and documented,
belong to different genetic groups. Though they share common structural traits,
they exhibit their own unique properties, providing a great opportunity for
investigation. In this book several new phenomena that I found in South Asian
languages have been analyzed in the Principles and Parameters approach.
The fact that there is no work available that compares and contrasts the rich
variety of data of these languages in a suitable theoretical framework was a
major motivating factor to write a book of this nature. The bulk of the empirical
evidence that I gathered, and the insights gained, further encouraged me to
pursue this work.
Writing this book – which required an enormous effort in understanding the
intricacies of the data and the structure of many South Asian languages – the
issues involved, and attempting to present them collectively within the limita-
tions of space and time, entailed an enormous amount of work and organization.
The findings presented in here will be of great utility to scholars working on
individual languages, to typologists as well as to theoretical linguists interested
in the study of language universals and parametric variation.
The range of topics chosen, the number of issues discussed and the amount
of data provided to substantiate our arguments will enable the reader to gain a
comprehensive view of the intricacies of the syntax of South Asian languages.
This book will demonstrate the relevance and significance of structural typol-
ogy that derives its insights from a formal theoretical framework, and employs
it fruitfully for typological work.
This work, I hope, will provide an incentive for further research on the
syntactic typology of each language family and on individual languages of the
subcontinent.
There is a detailed discussion of a number of issues that are relevant to the
study of the syntactic typology of South Asian languages available at the URL
www.cambridge.org/subbarao.
xi
Acknowledgements
It is my great pleasure and privilege first to thank all those who have been work-
ing on grammars of individual languages which contribute to the understanding
of human language.
The late Professor P. B. Pandit encouraged me to initiate work on syntactic
typology in 1974, at a time when the field was less known. The late Professor
James (Jim) McCawley also persuaded me to write a book on the syntactic
typology of South Asian languages. Jim would have been delighted to see the
present volume. The late Professor W. F. Lehman, too, evinced a keen interest
in my work. I am grateful to Professors Pandit, Lehman and McCawley for
their encouragement. Professor A. R. Kelkar, too, has been very enthusiastic
about the work on the typology of South Asian languages.
In 2005 Professor Peri Bhaskararao of the Institute for the Languages and
Cultures of Asia and Africa, Tokyo University of Foreign Studies (ILCAA),
invited me there as Visiting Professor for a year and provided me an intellectual
environment with unlimited infrastructural facilities in which to pursue my
research goals. It was at the Institute in Tokyo that the first draft of this book
was written. But for his generous help, active support and sincere commitment
towards an academic cause, the book would never have been completed. I
greatly benefited from the discussions on Telugu syntax with him. I owe a lot
to him and to his Institute, which cannot be expressed in words. I am deeply
indebted to Bhaskararao.
My greatest admiration and deep gratitude goes to Alice Davison, who has
been a friend, philosopher and guide throughout the writing of this book.
She went through all the chapters meticulously, making insightful comments,
offering helpful criticism and concrete suggestions, resulting in a balanced
approach between empirical data and the theoretical issues involved. Though
her professional commitments were too numerous, she took time off to read
the manuscript at various stages of its preparation, which I greatly appreciate
and admire. Her perseverance, commitment to promote an academic pursuit,
encouragement and intellectual contribution have enhanced the quality of this
work.
xii
Acknowledgements xiii
I am grateful to Rajesh Bhatt for going through the entire manuscript very
carefully during the pre-final stages and for his constructive criticisms, his help-
ful, incisive and insightful suggestions and comments, and for raising thought-
provoking questions, which have bettered the present work. His suggestions
helped me in radically reorganizing chapters 2 and 8, which has enhanced
conceptual clarity. His unassuming style of functioning during discussions has
really impressed me. It has been fun discussing syntactic issues with Rajesh. I
am deeply indebted to him for all his help.
I am grateful to Probal Dasgupta for having patiently read the chapters and
for commenting on them in terms of data and the theoretical and empirical
issues involved. His insights into Bangla syntax in particular and syntax in
general, with solid arguments, enabled me to proceed in the right direction. I
am deeply indebted to him too.
Both Alice and Probal have been a constant source of encouragement by
repeatedly telling me that this kind of typological work has an immense value.
Thanks are due to Peter Hook, who commented on the data and content
of several chapters and helped me to think more on the issues in Indo-Aryan
languages. His insights into the functioning of these languages helped me a
great deal. His sound comments proved to be very helpful.
Colin (Nick) Masica has also been a great source of encouragement. His vast
and in-depth knowledge of the structure of Indo-Aryan and Dravidian languages
and his valuable insights have been of immense value for this book. He read
some chapters and commented on them. Whenever I needed some help, Nick
was always there with advice and suggestions. It was his monumental work on
the typology of South Asian languages and Indo-Aryan that encouraged me to
follow it up further.
I am grateful to Martin Everaert for his keen interest, support and help at
various stages of the preparation of the manuscript and also for his valuable
inputs and comments on some chapters of the book.
I am thankful to John Peterson, who provided me with his published and
unpublished work on Kharia and who went through many chapters carefully
and commented on them meticulously. I greatly benefited from his knowledge
of Kharia syntax. He was always very enthusiastic to discuss issues of Kharia
syntax, which helped me enormously.
Josef Bayer, Frans Plank, Hans H. Hock and Halldor A. Sigurdsson read
some or most of the chapters and provided helpful insights and significant
inputs to the work. I am grateful to all of them for their valuable suggestions
and support in many ways.
There are many scholars who helped me in different ways by sending rel-
evant material and answering questions on data and analysis. They include:
P. J. Mistry, Bh. Krishnamurti, Annie Montaut, Kashi Wali, Hans H. Hock,
Barbara Lust, James W. Gair, Miriam Butt, K. A. Jayaseelan, R. Amritavalli,
xiv Acknowledgements
Balthasar Bickel, Gisbert Fanselow, Yogendra Yadava, Udaya Narayana Singh,

Swapan Banerjee, Matt Shibatani, Tatiana Oranskaia, Ludmilla Khokhlova,
Boris Zakharyin, Juanita War, O. N. Koul, Prashant Pardeshi, Pritha Chandra
and Rajesh Kumar.
I’d like to express my deepest sense of appreciation and gratitude to Sri
D. V. Rama Krishna Rao, my teacher of Hindi at the undergraduate level, who
selflessly worked for the welfare of all his students.
I’d like to thank my teachers Professors Rama Nath Sahai, M. L. Uptreiti,
Braj B. Kachru, “Yamuna ji” (Yamuna Kachru), Herbert Stalke, Jerry Morgan,
Charles Kisseberth and Mike Geis, who have had a tremendous intellectual
impact on my academic career. It was Yamuna ji who taught me the intricacies
of Hindi syntax at the graduate school, which provided a solid base for my
research on Hindi syntax in particular and language in general. I owe a great
academic debt to “Yamuna ji.”
Shukla Basu deserves special mention. Not only did she provide the Bangla
data enthusiastically, she also put in an enormous amount of hard work with
patience and care, going through all the chapters painstakingly making editorial
corrections. My discussions on Bangla syntax with her have been very fruitful
and I benefited a great deal from such discussions. I am deeply indebted to her.
Unless otherwise mentioned, the entire Bangla data in this work were pro-
vided by Shukla Basu, the Manipuri data by Sarju Devi, the Khasi data by
Grace Temsen and the Tenyidie data by Mimi Kevichüsa. I’m grateful to them
for providing data with enthusiasm and for their keen interest in the analysis of
their languages.
This work owes a great deal, for the data and help in terms of discussions
patiently provided, to a considerable number of native speakers of the various
languages. I wish to record my deep sense of gratitude to all the speakers.
They include Probal Dasgupta and Paromita Bagchi for Bangla; N. Pramodini,
Surmangol Sharma and Nandaraj Amom for Manipuri; Curiously Bareh for
Pnar, Jowai and Rymbai; D. Kuolie for Tenyidie; Harbir Arora for Dakkhini;
N. Varija for Bhalavali Bhasha and Kannada; Hemananda Bisebli for Kannada;
Revd. Van Lal Bapui and Vanlal Englien for Hmar; O. Rosanga, Revd V. L.
Zaithanga, Lalremzami Chinzah and several others for Mizo; Pauthang Haokip
for Thadou; Peri Bhaskararao, M. Subrahmanyam, M. Girija and Usha Rani
for Telugu; P. Sreekumar, Sobha Nair and Hany Babu for Malayalam; Arul
Mozhi (Bangalore) and Parameshwari Krishnamurthy for Tamil; Prakash Pat-
naik for Oriya; Upendra Rabha for Rabha; Modhumita Bora, Gautam Bora,
Manashi Dutta-Gogoi, Atreyee Sharma, Suranjana Barua, Chandana Bhuyan
and Mridusmita for Assamese; Aadil Kak, Sadaf Munshi and O. N. Koul for
Kashmiri; Yogendra Yadava for Maithili; Diwakar Sashtri, Lekhnath Pathak
and Pawan Upadhyaya for Nepali; Prashant Pardeshi and Ashok Aklujkar for
Marathi; Lalita Dhareshwar for Mangalore Konkani and Marathi; P. J. Mistry
and Trupti Nissar for Gujarati; Om Prakash Arora, Harbir Arora, Sandeep
Acknowledgements xv
Gupta and Rama Kant Agnihotri for Punjabi; Teresa Sundi, Bir Singh Sinku
and C. B. Deogam for Ho; Revd. Naphurai Jamatia, Binoy Deb Burman, San-
jeeb Deb Burman and Atul Deb Burman for Kokborok; Bhupen Narzaree,
Birlang Narzaree, Chandan Borgoyari and Ganga Brahma for Bodo; Parichita
Langthasa and Ron Kemprai for Dimasa; Jugen Pegu, Shikha Dele and Darika
Pegu for Mising; Rosemary Achumi, P. Toiho Sema, Hekani Jakhula, Ruth
Murru, Vikali Arkha and Toli Yeptho for Sema; Pangersenla for Ao; and Matt
Shibatani for Japanese and Korean. Their help is acknowledged with great
appreciation.
I’m grateful to G. Uma Maheshwar Rao for the enlightening discussions
we have had on several issues of Telugu syntax and for providing me a good
academic environment during my tenure as Radhakrishnan Chair Professor at
the University of Hyderabad. J. Prabhakarao helped me in many ways. I am
deeply indebted to both of them. I am thankful to my fellow faculty members
and students at the Centre for Applied Linguistics who made my stay of one
year enjoyable and memorable. My special thanks are due to Mayuri Dilip,
Kenei Mepthii-o, Parameshwari Krishnamurthy and Mimi Kevichüsa for their
help during the final stages of this work.
My friends Om Prakash Arora, Harbir Arora, Amit Mitra, R. C. Cowsik,
Yugal Swaroop, Alka Goel, and Satpal and Chhaya Gupta have always provided
their unstinted support in many ways. Rajesh Kumar’s and Lalita Dhareshwar’s
help in a variety of areas is fondly acknowledged.
I’d like to thank U. Sreehari too, for his untiring effort in patiently editing
the manuscript and for helping me in many ways.
Thanks to Sunita Topo for her caring service.
I greatly benefited from descriptive grammars written on South Asian lan-
guages, and the Routledge series of grammatical descriptions of South Asian
languages.
Many thanks to Andrew Winnard of Cambridge University Press, with whom
I first discussed the project and who got it initiated. I appreciate his patience
in bearing with the delays that occurred due to a variety of factors. Thanks are
also due to Sarah Green for her keen interest and patience at various stages of
preparation of the manuscript, and to Tom O’Reilly for all his help. I wish to
express my deep sense of gratitude to Leigh Mueller, the book’s copy-editor,
for having patiently read the manuscript and for making stylistic changes with
meticulous care, paying attention to the minutest possible detail. Her insightful
editorial style is appreciable. It was fun working with Leigh.
My wife Sarala, our daughters Sudhita, Varnita and Vidita, our sons-in-law
Srini, Sandeep and Sanyam, and our grandsons Srijay, Ishaan and Sahaj have
always been a great source of inspiration. Sarala’s support throughout my
academic career deserves special mention. I record my deep appreciation and
sense of gratitude to her in particular and to our family in general. I am thankful
to my extended family for their support and encouragement.
Abbreviations
AA Austro-Asiatic
abl ablative
abs absolutive case
acc accusative
adjr adjectivalizer
adv adverb
AGR/agr agreement
cem collaborative effort marker
caus causative
CFC contrastive focus clitic
cl classifier
com comitative
COMP/comp complementizer
cond conditional
conj conjunction
corr correlative
CP Complement Phrase
cp conjunctive participle
cpm conjunctive participial marker
CRP Case Resistance Principle
dat dative
DD1 definite determiner 1
DD2 definite determiner 2
decl declarative
def definite marker
def agr default agreement
det determiner
dis disjunction
DM/dm deictic marker
DO direct object
Dom differential object marking
DP Determiner Phrase
xvi
List of abbreviations xvii
DR Dravidian
DSC dative subject construction
dub dubitative
ECM Exceptional Case Marking
EHRC Externally Headed Relative Clause
emph emphatic marker
epen epenthetic
epm emotive predicate morpheme
EPP Extended Prejection Principle
erg ergative
ev evidentiality
excl exclusive
f feminine
FC final (right peripheral) complementizer
fin finite
foc focus
FOFC Final-over-Final Constraint
fut future
gen genitive
GF Grammatical Function
gp mkr group marker
gpm generic possession marker
hon honorific
IA Indo-Aryan
IC initial (left peripheral) complementizer
IHRC Internally Headed Relative Clause
imp imperative
imperf imperfect
inch inchoative
incl inclusive
ind indicative
INFL inflection
instr instrumental
intr intransitive
IO indirect object
IP Inflection Phrase
LF Logical Form
loc locative
m masculine
midhon mid-honorific
mkr marker
neg negative
xviii List of abbreviations
neg pple negative participle

neut neuter
nh non-human
nm non-masculine
NNS non-nominative subject
nom nominative
nonfut non-future
non hon non-honorific
NP Noun Phrase
nozr nominalizer
NPAH Noun Phrase Accessibility Hierarchy
NPI Negative Polarity Item
O Object
oam object agreement marker
o ben other-benefactive
obl oblique
OGEN Object of the Genitive
OO oblique object
OV Object Verb
p plural
P and P Principles and Parameters approach
approach
pass passive
per person
perf perfect
PIC Phase Impenetrability Constraint
pm procrastination marker
pn person
pol q mkr / polarity (yes/no) question marker
pol
poss possessive
PP postpositional phrase
pple participle
pres present
PrIC Propositional Island Constraint
PRO Big PRO (uncase-marked, un-governed)
progr progressive
pron pronominal
prox proximate
pst past
q question
quot quotative
List of abbreviations xix
rel relative
S Subject
S-O-V Subject-Object-Verb
S-V-O Subject-Verb-Object
s singular
SA South Asian / South Asia
sam subject agreement marker
self aff self-affective
self ben self-benefactive
Spec Specifier
sub subject
sup mkr superlative marker
TB Tibeto-Burman
TEC Thematic Eligibility Condition
thematic S thematic Sentence
TP Tense Phrase
tr transitive
V Verb
VP/vP Verb Phrase
VR verbal reflexive
VREC verbal reciprocal
V-S-O Verb-Subject-Object
1 Introduction
1.1 Languages of South Asia

South Asian languages (hereafter, SALs), comprising four major language
families (Austro-Asiatic with two sub-branches – Mon-Khmer and Munda –
Dravidian, Indo-Aryan and Tibeto-Burman), are rich in linguistic diversity,
providing a great opportunity for investigation.1 Though estimates differ as to
how many languages are spoken in the subcontinent, it is generally agreed that
the number runs to more than 100.
This book focuses on the syntactic typology of SALs in general, and the
high degree of syntactic convergence in particular, with special reference to
the notion of “India as a linguistic area” (Emeneau 1956; Masica 1976; see
chapter 2 for details).
The study of syntactic typology has gained enormous importance in the
last twenty-five years with emphasis on data and analysis of unrelated and
genetically distinct languages. It enables us to understand the intricate nature
of the Universal Grammar (UG) and parametric variation. Universal Grammar
is a set of principles that are commonly shared by all languages of the world.
According to Chomsky (1975: 118), a general principle found in a language
“belongs to universal grammar, as part of ‘pre-existent’ knowledge that makes
learning possible.” Languages do not differ from each other in innumerable or
random ways; rather, they differ from each other in terms of a limited set of
parameters relating to the subsystems of a formal grammar. Based on data from
the four major South Asian language families mentioned, we wish to examine
the principles that SALs share with other languages in terms of the UG, and
the parameters of syntax and morphology according to which they differ. We
wish to demonstrate how a formal theory of language such as the Government
and Binding framework allows us to capture the commonalities and differences
amongst languages in a principled manner.
Our work also focuses on South Asia as a “linguistic area.” Languages in
the subcontinent share a number of features at the phonological, syntactic and
semantic levels. However, they do differ from each other in their own unique
1
2 Introduction
ways, demonstrating that each language retains its identity in spite of intense
contact with other languages either due to multilingualism, or areal contact,
or both. The similarities found in SALs are due to the fact that all the lan-
guages, except for a few such as Khasi, Rymbai and Pnar (Mon-Khmer), are
verb-final. As one would expect, verb-final languages have several features
in common. Our work further demonstrates how languages that are contigu-
ously located in an area can affect each other in a radical and in a principled
manner.
1.2 Aim of the book

The aim of this book is to provide a typology of major constructions selected
from the principal components of grammar (such as relative clauses, case,
agreement, anaphora, complementation, conjunctive participles, etc.) with a
view to arriving at an understanding of the intricate nature of such construc-
tions and to see how such analysis can enrich our understanding of Universal
Grammar and parametric variation. This will be accomplished first by present-
ing, in a theory-neutral fashion, the basic facts of each of the topics chosen, and
then by providing an explanation of the phenomena in the modular approach
of the Government and Binding framework. Phrase structure, the role of Case,
binding, thematic relations, control theory and movement as the various mod-
ules of the Government and Binding framework, and significant notions such
as government and c-command, have come into focus in the last three decades.
Never before has the study of language universals and parametric variation
received as much attention as it has since the eighties. The topics chosen and
the issues discussed are intended to enable the reader to get a perspective on the
nature of language variation, and the universal principles involved, using the
concepts of a formal theory of grammar, and also to define exactly the features
involved in convergence.
While providing an overall typology of the aspects of SALs, care was taken
in each chapter to focus on the issues that concern individual languages and the
language family to which they belong. For example, there has been a consid-
erable debate over the issue of one finite verb per sentence and the Strict OV
Constraint in Dravidian. This issue relates to the occurrence of the embedded
complement clauses with the marker -ō, and the relative-correlative construc-
tions with the same marker -ō. The nature of the finiteness phenomenon has
implications for the nature of contact and convergence between Dravidian and
Indo-Aryan (see Hock 2005 for the most recent discussion). In the discussion
of conjunctive participles, complementation and relative clauses, this issue has
been focused upon and discussed.
1.3 Linguistic theory 3
1.3 Linguistic theory, language universals and language typology

Language universals and parametric variation have been the focus of study for
the last thirty years, and significant insights have been gained in our understand-
ing of the nature of human language. An explanatorily adequate theory needs
to be supported and substantiated by empirical facts, while empirical facts in
isolation do not have much relevance unless they are properly explained in an
appropriate theoretical framework. The last three decades have witnessed an
enormous emphasis on presenting evidence based on empirical data so that
the foundations prepared for theoretical claims are well laid (Bhaskararao and
Subbarao 2004: ix) – for instance, as in the case of lexical anaphors (Lust,
Wali, Gair and Subbarao 2000) or non-nominative subjects (Bhaskararao and
Subbarao 2004). Thus, to build an explanatorily adequate theory, it is essential
to have a sound database. A fruitful combination of theory and data alone will
yield the desired results. Language typology attempts to study linguistic pat-
terns and language variation, and arrive at generalizations from cross-linguistic
comparison (Croft 1990). One of the goals of typological studies should be to
advance our understanding of central theoretical questions of language. Hence,
typology must focus on various phenomena of current theoretical and empirical
relevance, and must attempt to provide solutions supported by rich empirical
data and valid arguments. We feel that a symbiotic relationship between lin-
guistic theory and language typology which primarily emphasizes choosing
data from a wide variety of languages and analyzing them in a theoretical
framework is thus a desideratum (Subbarao 1999). A similar view is articu-
lated in Cinque (2007: 93), who does not consider “(‘formal’) linguistic theory
and linguistic typology as two separate approaches.” Our work in this volume
attempts to combine, as Baker and McCloskey (2007: 291) succinctly put it, “a
motivated degree of abstractness in the analysis of particular languages, as is
typical in formal linguistics, with an interest in sampling widely from a range of
languages, as is typical in typological studies [emphasis added].” Baker (2001)
labels such an approach as the “the middle way.”
In a sense, this book is a manifestation of what Baker (2008) calls Formal
Generative Typology, aiming to find out “what properties of natural human lan-
guages are genuinely universal,” “what properties of natural human languages
vary from one human language to another” and “which aspects of variation
are patterned, systematic, and grammatical in nature, and which aspects of
variation are random, idiosyncratic, and lexical in nature.”
While it is the linguistic theory that enables linguists to formulate a hypoth-
esis in a given framework, helps them to look for the relevant data, offers
them a direction in which to look, it is the data that not only substantiate the
claims made in the theory, but also provide a window to look for inadequacies,
4 Introduction
counterexamples to the theoretical claims made, enabling linguists to revise and

enrich the theoretical base, if necessary. Hence, an active interaction between
data and theory is a must. Any theory that is proposed without a sound theo-
retical base and any theoretical claims made regarding the nature of human
language without a strong database will not be explanatorily adequate.
We attempt to demonstrate with our findings that it is the linguistic theory
with a sound theoretical base that works as a powerful tool in explicating
commonness and variation in languages. This book is based on a rich set of
data from a larger number of languages and these data have been analyzed
uniformly through the perspective of one theory.
1.4 Inductive and deductive approaches to language analysis

Most of the studies in linguistic typology have been done either purely in
rationalist (deductive) or in empiricist (inductive) approaches. It was in some
studies in relational grammar, a study by Keenan and Comrie (1977) and some
studies in the Government and Binding framework that attempts were made to
strike a reasonable balance between theory and data in analysis.
Though not overtly spelt out, there have been two distinct approaches to
the study of language universals: inductive and deductive. While inductivists
believe in drawing their generalizations based on data collected from a wide
range of languages, deductivists may arrive at these constraints based on a
metalinguistic theory, drawing their evidence from a single language or a limited
number of languages. Scholars who are opposed to a totally deductive approach
argue that “it is necessary to have data from a wide range of languages” (Comrie
1981: 4) to carry out research on language universals. Thus, there is a great
emphasis on arriving at language universals on the basis of “concrete rather
than abstract analyses.”
Deductivists, in contrast, argue that the theory of grammar has the capacity
to predict language universals. It was generally agreed that the study of even
a single language is sufficient to predict language universals. Chomsky (1975:
118) points out: “the principles that appear to have explanatory adequacy for
English are the principles of universal grammar.” Thus, “[a] great deal can be
learned about UG from the study of a single language” (Chomsky 1981: 6).
Coopmans (1983: 458), too, articulates a similar view: “The universals of
generative grammar have a different basis. If we assume that these principles
are universally available, in the sense that they are genetically encoded, then
by a detailed study of one particular language, we will be able to discover
some properties of UG and thus, deepen our insights on the acquisition process
[emphasis added].”
The deductivist favors innateness as the explanation for language univer-
sals. In view of the methodology followed and the choice of data collected for
1.5 Relevance of linguistic theory 5
linguistic inquiry, we argued elsewhere that the deductivist is a microinduc-

tivist and the inductivist is a microdeductivist and, therefore, a strict dichotomy
between the inductive (empiricist) and deductive (rationalist) approaches can-
not be maintained (Subbarao and Saxena 1987; Subbarao 1999). We proposed
an integrated approach that combines these two. This can be summed up as:
Limited Induction Base → Deduction → Induction
(See Subbarao and Saxena 1987 for details.)

Thus, to build an explanatorily adequate theory, it is essential to have a
sound database. A fruitful conflation of theory and data alone will yield the
desired results. According to Lust et al. (2000: 1): “Without basic theory,
the fundamental questions of the nature of human competence for a language
cannot be addressed. However, in the absence of real language data, proposed
answers may not be relevant to the real questions.” Lust et al. (2000) label their
approach as “principled typology.” Consequently, it should be the concern of
the theoretical linguist as well as the typologist to combine theory and empirical
language data.
With this goal in mind, each chapter in this book strikes a balance between
theoretical assumptions and the issues that arise from the data presented from
sixty languages belonging to the four language families of the South Asian
subcontinent.
In each chapter, we provide a detailed analysis of the data in theory-neutral
terms first. It is envisaged that any researcher, irrespective of their theoretical
background, can understand such analysis. Once this is done, points in the
theoretical framework of the Principles and Parameters approach are analyzed
(Chomsky and Lasnik 1995).
1.5 Relevance of linguistic theory: some illustrations

We attempt to demonstrate that it is the linguistic theory with a sound theoret-
ical base that works as a powerful tool to explicate differences in languages.
We discuss below some specific cases from SALs where theory provides an
explanation to account for the facts.
(i) The first example comes from the occurrence of nominal anaphors and
the verbal anaphor. SALs have two forms of the nominal anaphor – simplex
and complex – and two forms of the verbal anaphor – verbal reflexive and
verbal reciprocal. The verbal anaphor is universally monomorphemic (a simplex
form). The complex nominal anaphor is either a partially or fully reduplicated
form of the simplex form.
Except in Marathi (Indo-Aryan [IA]), long-distance binding is blocked when
the complex form of the nominal anaphor occurs in the embedded clause. We
6 Introduction
demonstrate that this fact can be explained by invoking Head to Head movement
(see chapter 3 for details).
When the verbal anaphor occurs in the embedded clause, long-distance
binding is blocked in all the languages.
Thus, the occurrence of either a nominal complex anaphor or a verbal anaphor
or both blocks long-distance binding.
Leaving the case of Marathi (IA) aside for a moment, this can be neatly
formulated in terms of the following parameter.
Parameter 1:
[−/+ complex nominal anaphor] / [−/+ verbal anaphor] directly correlates
with [+/− long-distance binding].
That is, the occurrence of the complex nominal anaphor and verbal anaphor are
negatively specified for the phenomenon of long-distance binding.
(ii) The second example concerns the occurrence of the verbal reflexive
and the locative PP. A locative PP may either be a subcategorized or a non-
subcategorized argument of a predicate. When an anaphor occurs in a sub-
categorized locative PP position, the occurrence of the verbal reflexive is
obligatory – otherwise, it is not (Lust et al. 2000). Thus, the notion of
subcategorization2 helps us in providing an explanation of the occurrence
of the verbal reflexive.
This can be formulated in terms of the following parameter.
Parameter 2:
[+/− the occurrence of a verbal anaphor] directly correlates with [+/− the
occurrence of a nominal anaphor in a subcategorized locative PP position]
(see chapter 3 for details).
That is, the occurrence of a nominal anaphor in a subcategorized locative PP

position is positively specified for the occurrence of a verbal anaphor.
(iii) The third example concerns the case of the occurrence of initial comple-
mentizer (IC) in complement clauses in verb-final languages and wide-scope
reading of wh-expressions.
Word order typology and the Head Direction parameter help us predict wide-
and narrow-scope interpretations of wh-expressions in complement clauses.
Bayer (2001) argues that in verb-final Indo-Aryan languages, wide-scope read-
ing of wh-expressions in IC clauses is blocked. This is due to the fact that the IC
clause falls outside the pattern of argument licensing in head-final languages.
Our study substantiates the claim made in Bayer (2001).
In contrast, verb-final languages with a final complementizer (FC) which
is a quotative permit wide-scope reading of the wh-expressions. The various
parameters that affect the scope of a question word in complement clauses are
discussed and ten parameters are suggested (see chapter 6 for details).
1.5 Relevance of linguistic theory 7
This can be formulated in terms of the following parameter.3

Parameter 3:
[+/− initial complementizer] in verb-final languages directly correlates
with [−/+ wide-scope reading of the wh-expression].
That is, wide-scope reading of the wh-expression is negatively specified for the
feature of initial complementizer.
(iv) The fourth example comes from relative clauses in SALs. Some Tibeto-
Burman languages (Angami [a.k.a. Tenyidie], Sema, Mizo, Manipuri, Sangtam
and Konyak) have both externally headed and internally headed relative clauses.
In externally headed relatives, using a gap rather than movement in many SALs,
a comitative PP cannot head a gap (infinitival) relative clause. In those languages
in which it can, we have demonstrated that a thematic relation needs to be
established between the predicate of the embedded clause and the comitative
head of the externally headed and internally headed relative clause, either in
terms of an overt case marker or some verbal marker/clitic in the embedded
verb to indicate accompaniment. A similar phenomenon is found in Khasi and
Pnar (Mon-Khmer) too. Though the notion of thematic relations has run into
rough weather in recent years (see Newmeyer 2007: 140–141) and it is difficult
to agree on which “thematic roles exist” and how “to independently justify the
assignment of noun phrases to thematic roles in particular sentences” (Dowty
1989: 70, as quoted in Newmeyer 2007: 140–141) in an objective manner,
our analysis demonstrates that the phenomenon of incorporation of either an
adverb such as together or a postposition or a reciprocal marker as a group
marker in the embedded verb enables the comitative PP to meet the Thematic
Eligibility Condition (TEC) proposed in this study, and a comitative PP can
consequently head a non-finite relative clause. Thus, the TEC needs to be
included in the grammar in order to explicate the occurrence or non-occurrence
of a comitative PP as head in the gap (infinitival) relative (see section 8.7 for a
detailed discussion).4
This can be formulated in terms of the following parameter.
Parameter 4:
[+/− the occurrence of a comitative adposition with the head NP / an
adverb denoting together along with a verbal clitic / an incorporated
comitative adposition in the embedded verb] directly correlates with
[+/− the occurrence of a comitative PP as head] in the gap relative.
That is, the occurrence of a comitative PP is underspecified for the occurrence

of a comitative PP as head.
(v) The fifth example concerns the accusative case-marking in nominative–
accusative sentences and the non-nominative (dative/genitive subject) construc-
tions. The “objects” are differentially marked depending upon features such
8 Introduction
as specificity and animacy. Thus, with regard to differential object marking

(DOM) in SALs, when the noun phrase denoting specificity/animacy occurs,
the accusative marker may occur even though the predicate is [−transitive]
in the dative/genitive subject construction (DSC) in Bangla (IA), Malayalam
and Tamil (DR). When the predicate is [+transitive] in the nominative subject
construction, the object is accusative case-marked when it denotes specificity/
animacy. Based on these facts, we propose the following parameter to account
for this variation.
The Differential Object Marking (DOM) parameter: when the noun phrase is
accusative case-marked, the object marker is either associated with
animacy/specificity depending on transitivity in the [−NNS construction],5
or purely animacy/specificity independent of transitivity in the [+NNS
construction], as in parameter 5.
Parameter 5:
The occurrence of the accusative case marker denoting animacy/specificity
directly correlates with [−/+transitivity] in the [+/−NNS construction]
(see chapter 5 for a detailed discussion).
We present further arguments in this work to demonstrate that typology without

theory, and theory without typology, are incomplete, and this, we hope, will
further strengthen our claim that there is a symbiotic relation between the two.
1.6 The framework

In each chapter, after the presentation of data, there is an analysis of the
significant points in the theoretical framework of the Principles and Parameters
approach.
1.6.1 The Principles and Parameters approach

Languages appear to differ from each other in an innumerable number of ways.
According to the Principles and Parameters approach (Chomsky and Lasnik
1995), this variation is not random, and is due to a specific set of parameters,
which are limited in number. It is the interaction of the parameters with the
universal set of principles that are common to all languages that constitutes
human language.
1.7 The data

Data for the volume are drawn from the following SALs. All the languages
are verb-final in sentence structure, except for Khasi (Mon-Khmer), in which
1.7 The data 9
the verb occurs in the medial position, and Kashmiri (Indo-Aryan), in which
the verb occurs in the V2 position, though it is underlyingly verb-final (see
A list of the languages from the four South Asian language families from
which our work draws its data is given below:
ia Indo-Aryan
Assamese
Bangla
Gujarati
Hindi-Urdu
Kashmiri
Konkani
Maithili
Marathi
Nepali
Oriya
Punjabi
Eastern Shina
Shina of Gultari
Shina of Skardu
Sinhala
Swat-Dir Kohistani
Torwali
ib Contact languages
Bhalavali Bhasha (a.k.a. Bhalawali Marathi)
Dakkhini (Hindi-Urdu)
Eastern Bangla (Sylheti)
Mangalore Konkani
Subzapuri
ii Dravidian
Kannada
Malayalam
Manda
Tamil
Telugu
Toda
iii Austro-Asiatic
iiia Mon-Khmer
Khasi
Rymbai
Pnar
10 Introduction
iiib Munda
Ho
Juang
Kharia
Mundari
Santali
Sora/Savara
iv Tibeto-Burman
Ao (Mongsen)
Bodo
Dimasa
Dumi
Garo
Hmar
Kham
Kokborok
Konyak
Ladakhi
Manipuri
Mao
Mising
Mizo
Paite
Rabha
Sangtam
Sema
Tenyidie
Thadou
Tiwa
Most of the data on Tibeto-Burman languages and on some Munda languages
were collected in field trips in the northeastern parts of India and in the states
of West Bengal and Jharkhand.
It should also be mentioned that there are very few syntactic descriptions
available for Tibeto-Burman, Munda and Mon-Khmer languages, especially
at the level of detail and comprehensiveness which are necessary for useful
analysis. Traditional grammars that are available do not address the conceptual
issues that the linguist of today is interested in, in terms of the Universal
Grammar and parametric variation.
The data collected and analyzed by the author with a team of researchers
in the University Grants Commission’s projects on syntactic typology, over a
ten-year period, were also used in this work.
1.9 Richness of data 11
1.8 Relevance of the present work

Most of the discussions in current linguistic theory have been on the basis of
data from English, European languages, and Asian languages such as Korean
and Japanese. However, in recent years, significant work with rich insights on
issues related to individual languages or specific language families of the South
Asian subcontinent (e.g., Indo-Aryan, Dravidian, Austro-Asiatic and Tibeto-
Burman) has been done by many scholars. There is no work available on the
general typology of South Asian languages except for Masica (1976), which
was based mainly on collation of existing published materials, and therefore did
not include much data and analysis from Tibeto-Burman and Austro-Asiatic
languages – on which less work has been done – as the present volume does.
This fact, along with the general neglect of these two families by linguists, and
the issue that some of these languages deserve special attention because they
are on the verge of extinction, were the motivating factors for presenting the
findings of our research in a book form. There is no work available that draws
so much of its data from the Tibeto-Burman and Austro-Asiatic (Munda and
Mon-Khmer) languages.
Moreover, the present work draws its data from primary sources, that is,
from fieldwork and from language consultants. It had the advantage of being
able to check and recheck, or elicit further data, in support of any specific
issue discussed.
1.9 The richness and complexity of the data and their relevance
Languages belonging to genetically different language families have been
in intense contact for thousands of years in the subcontinent, as a result of
which convergence at various levels (phonological, morphological, syntactic
and semantic) has been in process between them. Syntactic convergence results
in the reanalysis of existing functional categories as new ones, and we shall
show how a contact language optimally utilizes the limited set of functional cat-
egories that it has. Convergence may also result in change in the position of the
verb from verb-medial to verb-final in a sentence, as in the case of the Munda
languages and Kamti Tai, whose source language is Thai, which is SVO.
Our work shows that, though the position of the head changes from VO to OV,
the language may still retain some of its VO patterns with regard to phenomena
such as noun incorporation, or the order of the matrix verb (higher predicate)
following or preceding the embedded verb, as in desiderative constructions
with an object complement in Sora (see chapter 4).
Out of the four different language families of the subcontinent, Austro-
Asiatic (Mon-Khmer and Munda) languages are “perhaps the most divergent
12 Introduction
in the world. They are opposite in structure at every level” (Donegan and Stampe
2004: 3). While Khasi retained its SVO structure in spite of being surrounded by
verb-final languages, the Munda languages not only changed to SOV from the
ancestral SVO,6 but also have been influenced by the neighboring Indo-Aryan
and, possibly, Dravidian languages.
The number of languages spoken, with the richness of diversity in structural
patterns, and the interaction between the languages of the different language
families for a prolonged period provides a great opportunity for investigation,
and enables us to avoid “many spurious generalizations resulting from their
very small sample size, and miss[ing] many true generalizations by not looking
at enough languages” (Baker and McLoskey 2007: 290).
Just to cite an example: the subject in these languages may be nominative
case-marked or non-nominative case-marked. The non-nominative subject con-
struction includes dative, accusative, ergative, genitive, instrumental and loca-
tive case-marked subjects. The agreement patterns in these languages interact
with the case-marking of the subject.
It is generally assumed that agreement is only possible with nominative case-
marked subjects. While one set of South Asian languages permits agreement
with nominative case-marked subjects, the other set permits agreement with
lexically case-marked subjects and even oblique PPs, which are not subjects
(see chapters 4 and 5 for further elaboration and discussion).
1.10 Contributions of cross-linguistic studies to our understanding

of the nature of language
Universal Grammar (UG) consists of “various subsystems – X-bar theory, bind-
ing theory, Case theory, theta theory, bounding theory, each containing certain
principles with a limited degree of parametric variation” (Chomsky 1986: 102).
In this book, we deal with issues related to phrase structure, Case (structural
and inherent), thematic roles, binding, control and movement. Hence, the Head
Direction parameter, binding principles, the status of PRO with regard to case-
marking, problems involved in dispensing with PRO, Forward and Backward
Control, Head to Head Movement, the thematic structure of a predicate, struc-
tural Case assignment/checking, assignment of inherent Case, etc., play an
important role in our study. We shall now present a brief discussion of our
findings, keeping Universal Grammar and parametric variation in view.
1. We shall show that the order in which the constituents occur in a sen-
tence depends on the position of the head (verb) in a sentence. We pro-
vide evidence from verb-final languages, a non-verb-final language and a
V2 (verb-second) language. The occurrence of initial complementizers in
some verb-final languages (see chapter 6), the nature of incorporation and
the order of verbs in the desiderative construction in Sora (Munda) (see
1.10 Contributions of cross-linguistic studies 13
chapter 4) and the non-verb-final structures in Kokborok (Tibeto-Burman

[TB]) (see chapter 2) are ‘inconsistent’ and, thus, do not conform to word
order universals.
In Sora, incorporated elements – nouns, quantifiers – occur to the right
of the verb stem as in verb-final languages, contrary to Kayne’s hypothe-
sis. However, in desiderative constructions with an infinitival object com-
plement, the matrix verb (higher predicate) precedes the embedded verb
just as in non-verb-final languages. We point out that only a diachronic
explanation can provide a solution to account for the apparent synchronic
anomalies (see chapter 4 for a detailed discussion).
2. All the languages of the South Asian subcontinent that were studied strictly
obey Principles A, B and C of the Binding Theory, except Marathi (IA).
Marathi permits an anaphor to occur in a finite complement clause with
an antecedent in a higher clause, which is a clear violation of Principle A,
which states that an anaphor is bound in its governing category.
3. Control Theory stipulates that PRO, the notional subject in an embedded
clause, must be ungoverned and uncase-marked (this formulation stands in
also for notational variants introduced in later versions of the standard theo-
ry for theory-internal reasons, variants that appeal to a new notion of ‘null
Case’ and drop the apparatus of government [see R. Martin 2001]). Our
study provides evidence to show that PRO is case-marked and governed,
just as in Icelandic (Sigurdsson 1989), contrary to the standard assump-
tion that it is uncase-marked and ungoverned (see chapter 7 for details).
We show that case-marking of PRO has implications for long-distance
agreement in some SALs such as Hindi-Urdu and Punjabi.
Long-distance anaphoric binding and agreement have two features in
common: (i) a subcommanding NP (possessor) can be an antecedent
to an anaphor, and can also trigger agreement in some languages; and
(ii) anaphoric binding and long-distance agreement are permitted only
when the embedded clause is non-finite. This, in our opinion, shows that
both are anaphoric processes, which require an antecedent (see chapters 3
and 4 for a detailed discussion).
4. Our work on agreement brings out several issues that would be of inter-
est to theoretical linguists and typologists: (i) SALs conform to the gen-
erally observed Pronominal Strength Hierarchy in agreement, where, I
person > II person > III person, where > indicates ‘is stronger than’;
(ii) evidence from SALs demonstrates that (a) the verb may agree with
(subcategorized) arguments as well as non-subcategorized arguments,
(b) there is possessor agreement too in some languages; and (iii) dif-
ferent categories such as nouns, postpositions, quantifiers, emphasizers,
intensifiers, Negative Polarity Items and adverbs are incorporated in the
verb in SALs.
14 Introduction
We demonstrate that when an adposition is incorporated into the verb,

the underived form of the host verb is transitivized and, consequently, the
underived subject in the nominative case gets ergative case-marked after
incorporation (see chapter 4 for details).
We shall also demonstrate that it is the incorporation of the comitative
postposition ‘with’ (either with, or without, the verbal reciprocal) that
permits the comitative PP to head an externally headed non-finite relative
clause in some Tibeto-Burman languages (see section 8.7 for a detailed
discussion).
5. We show that, with regard to clitics, the occurrence of particles or verbal
clitics, the process of copying the head and the presence vs. absence of redu-
plicated forms are some of the processes that block a specific interpretation
and facilitate the other intended interpretation. We demonstrate that in sen-
tences involving clitics, the notion of syntactic dependency domain (Lohse,
Hawkins and Wasow 2004) helps in sentence processing and enables us
to explain the two different interpretations of specific sentences. Further,
it also enables us to explain why the occurrence of some specific particles
facilitates one interpretation, while the occurrence of some others does not
(see chapter 10 on the Website for a detailed discussion).
6. Our analysis demonstrates that reduplication is not just a phenomenon
restricted to the area of morphology alone, and that it has syntactic impli-
cations to the extent that it can help disambiguate a sentence. We demon-
strate that the reduplicated form of the complex anaphor does not permit
long-distance binding in Telugu (DR) while a non-reduplicated simplex
form does (see chapter 3, and also chapter 10 on the Website for a detailed
discussion).
7. In areal language contact situations, a functional category such as a pre-
sentential complementizer may syntactically be reanalyzed as a post-
sentential element, and, consequently, may be reassigned an entirely dif-
ferent set of functions due to convergence with the features of the donor
language (see chapter 6, and also appendix 6.5 on the Website for a detailed
discussion).
8. Backward Control, a phenomenon that has passed unnoticed in the gram-
mars of SALs thus far, is widespread in SALs, and Forward Control is
not even a permitted option in a couple of instances, demonstrating that
Backward Control is not a marked option (see chapter 7). We demon-
strate that it could be used as a heuristic tool to determine whether a lan-
guage is a dialect of a specific language, or an independent language (see
appendix 7.4 on the Website for a detailed discussion).
9. For the first time, our study brings into focus the fact that the reduplication
of an aspect marker and the partial reduplication of the verb stem are
the two strategies used in the formation of polarity (yes/no) questions in
1.10 Contributions of cross-linguistic studies 15
Kokborok (TB), a phenomenon unparalleled in any South Asian language,

and possibly in any language to the best of our knowledge (see the appendix
to chapter 2 on the Website for a detailed discussion). (Also see Subbarao,
Malhotra and Barua 2010.)
10. Our work also demonstrates that Keenan and Comrie’s (1977) Noun Phrase
Accessibility Hierarchy (NPAH) is quite robust and accounts for data on
relative clauses from three language families – Dravidian, Tibeto-Burman
and Austro-Asiatic (Mon-Khmer and Munda). There are, however, a few
counterexamples to the NPAH in one language family, namely Indo-Aryan
(see chapter 8 for a detailed discussion).
11. We illustrate that the NPAH can be used as a powerful tool to predict
whether a DP can be a potential candidate to qualify for ambiguous inter-
pretation in externally headed and internally headed relative clauses. Thus,
while the Noun Phrase Accessibility Hierarchy can be a potent tool to
account for data from a wide range of SALs, we show that it has predictive
power too. Using the NPAH as a tool, our analysis answers the question of
why it is only in Dravidian and Tibeto-Burman languages that an ambigu-
ity arises in gap relatives with a specific set of predicates, while in most of
the Indo-Aryan languages (except in Oriya) it does not (see appendix 8.6
on the Website for a detailed discussion; see also chapter 8 for a discussion
of related issues).
12. We argue that: (i) the predicate in the dative subject construction (DSC) is
[−transitive] and unaccusative; (ii) all non-nominative subjects (NNSs) are
inherently Case-marked; (iii) such inherent Case-marking cannot be done
by an intransitive verb alone, but by the whole predicate compositionally
consisting of a theme or an adjective along with the [−transitive] verb; and
(iv) information concerning agreement should be available vP-internally
(in the lower thematic S) for proper assignment of inherent Case to the
non-nominative subject (see chapter 5 for a detailed discussion).
13. Subjects in general have primacy over objects. The primacy of subject
in verb agreement is violated in Mizo (TB) and it is the direct object
marker alone that occurs in this language, a language that has subject and
non-subject agreement markers co-occurring in the verb.
Further, in possessor agreement in Munda and Tibeto-Burman lan-
guages, it is the possessor of the direct object and indirect object that
triggers agreement, and not the possessor of the subject, thus exhibiting
subject–object asymmetry (see chapter 4 on case and agreement).
14. The next issue concerns the case of complement clauses in many Germanic
verb-second (V2 ) languages and in Kashmiri, an Indo-Aryan language with
an Indo-European ancestor. There is an asymmetry in word order in root and
subordinate clauses in German, while in Yiddish, Icelandic and Kashmiri
there is no such asymmetry. In German, the V2 order is found in the root
16 Introduction
clause, while in complement constructions there is verb-final order in the

embedded complement. In contrast, in Yiddish, Icelandic and Kashmiri,
the order is V2 in the root as well as the embedded clause. Bhatt (1999)
provides an explanation for this asymmetry by positing a Mood phrase in
the latter type of languages that permits the movement of the embedded
V to the empty mood node, and argues that the complementizer in such
languages functions merely as a pure subordinator. In contrast, in German,
the COMP conflates both the Mood and the complementizer, hence the
absence of the Mood node, as a result of which the embedded verb cannot
move and, subsequently, there is verb-final order in German complement
clauses. While a typologist can pinpoint the differences in word orders
in German, Yiddish, Icelandic and Kashmiri, we feel that it is only the
theory that can provide an explanatorily adequate solution to account for
it (for further details on complementation in SALs, see chapter 6, and also
appendix 6.1 on the Website).
15. Our work demonstrates that the verbal reflexive (VR) / verbal reciprocal
(VREC) detransitivizes a predicate and, consequently, the ergative marker
is not permitted with the subject when a nominal anaphor occurs in direct
object position in Hmar, Mizo and Thadou (TB). Further, verbal reflexive
(VR) / verbal reciprocal (VREC) and the detransitivizer/anti-causative suf-
fix in Dravidian and many Tibeto-Burman languages are homophonous. We
show that the verbal anaphor shares several features cross-linguistically.
We hypothesize – without concrete evidence – that this might plausibly
reflect the cognitive capability of the human mind to assign several seem-
ingly unrelated functions to one and only one syntactic entity, namely the
verbal anaphor (see chapter 3 for a detailed discussion).
16. Our work provides empirical evidence from Mizo, Hmar and Thadou
(TB) to demonstrate that case and agreement are intrinsically linked. In
relative clauses, passives and clefts, the ergative case marker and subject
agreement marker are either overtly present together or absent together (see
chapter 4 for a full elaboration).
1.11 The binding link for all chapters

All the chapters together will provide a comprehensive view of the sentence
structure of SALs. Each chapter will provide a general typological picture
of the specific feature under consideration in the languages of the Indian
subcontinent – in each language family – and of any special feature of a specific
language. Due to limitations of space, it was not feasible to provide a detailed
descriptive account of all the structures in all the languages under considera-
tion, with each structure analyzed and the implications of the data considered
1.11 Binding link for all chapters 17
for the structure of each language and language family; nor was it possible to
provide examples from all languages for each of the issues considered.
The binding link for all chapters will be to show how languages of the
subcontinent, which are verb-final (except Khasi, Rymbai and Pnar which are
SVO, and Kashmiri which is a V2 language), differ from each other while
sharing a number of common features.
We present data and analysis of specific constructions from SALs. We assume
that the nature of the evidence presented and the conclusions drawn on the basis
of data from the SALs chosen will be compatible with comparable data from
other SALs.
Finally, we wish to conclude by reiterating our contention that typology
is not a theory. For typologists, theory is a guiding principle that facilitates
arriving at linguistically significant generalizations about the intricate nature of
language. Thus, for them theory is a tool and a means to achieve their goals of
research, not an end itself. In contrast, for a theoretical linguist, theory building
is the primary concern, the ultimate end, to which data may be a means. Thus,
we believe that typological studies should neither be too descriptive to become
taxonomic in nature, nor should they be too abstract for linguistically significant
generalizations to be obliterated in theoretical controversies.
2 South Asian languages: a preview
2.1 Introduction
In this chapter, we present a brief survey of a few syntactic features of South
Asian languages to familiarize the reader with some specific structures found
in SALs. We shall present evidence from the verb-final languages of the four
major language families of South Asia.1 The four major language families in the
subcontinent are: (i) Austro-Asiatic (Mon-Khmer and Munda), (ii) Dravidian,
(iii) Indo-Aryan and (iv) Tibeto-Burman.2
2.2 Typology of South Asian languages

The typological features of SALs have been discussed earlier in several studies
(Chatterji 1926; Emeneau 1956; Ramanujan and Masica 1969; Masica 1976;
Schiffman and Shapiro 1981; Subbarao 2007). Emeneau (1956) was seminal in
its conceptualization of the subcontinent as a linguistic area. Emeneau (1956: 16
n.28) defines a linguistic area as “an area which includes languages belonging
to more than one family but showing traits in common which are found not
to belong to the other members of (at least) one of the families.” We provide
evidence in support of the notion of “South Asia as a linguistic area” from
major and “minor” languages.3
In the South Asian subcontinent, languages belonging to different genetic
groups exhibit common structural traits largely due to two main reasons:
(i) thousands of years of prolonged language contact amongst languages due to
intense bi- and multilingualism; and (ii) the fact that all these languages (except
Khasi,4 which is verb-medial, and Kashmiri, which is a V2 language) are verb-
final, and the position of the verb in terms of verb-finality or non-verb-finality
plays an instrumental role in the manifestation of many identical structural
traits. Some South Asian languages from the Indo-Aryan language family do
This chapter is a radically revised and enlarged version of Subbarao (2007). I am grateful to Rajesh
Bhatt for his very insightful comments which enabled me to reorganize and restructure the chapter.
Thanks to Alice Davison too for her constructive and helpful suggestions.
18
2.2 Typology of South Asian languages 19
exhibit certain syntactic traits which are “disharmonic” (Greenberg 1966) and
are not in conformity with word universals in terms of “consistency.” We shall
show that such structures obey the FOFC (Final-over-Final Constraint) pro-
posed in Holmberg (2000) and discussed in detail in Biberauer, Newton and
Sheehan (2009).5
In this chapter, the focus is on the following aspects of the typology of
SALs.6
Section 2.3 discusses some word order universals found in SALs; section 2.4
discusses some significant syntactic features that provide evidence in support
of the notion of “South Asia as a linguistic area.”7 Section 2.5 focuses on the
parametric variation found in SALs. Section 2.6 briefly mentions the effects of
language contact in SALs. Section 2.7 focuses on the unique syntactic features
that each language family possesses, distinguishing one family from another.
Section 2.8 is the conclusion.
All SALs except Khasi, which is a verb-medial language, share common
structural characteristics at the level of the sentence, and this could mainly be
due to either or both of the following two alternative reasons.
(i) Since all SALs except Khasi (Austro-Asiatic) are SOV (Subject-Object-
Verb) languages,8 they share a number of word order universals proposed
in Greenberg (1966) for verb-final languages.
(ii) All SALs have been in intense language contact with each other for a long
period of time, thus giving rise to the creation of a “linguistic area” or
Sprachbund which literally means ‘language league’ (Hock 1991: 494).
We shall first examine the former alternative. It is generally agreed in the studies
on word order typology as well as in recent theoretical frameworks that it is
the position of the verb (head, as it is generally labeled) in a sentence that
plays a crucial role in the order of occurrence of elements in a sentence. In
the Government and Binding framework, it is referred to as “Head (Direction)
Parameter.” The complement (object) in SOV languages occurs to the left of
the head, the verb or a postposition. Hence, the direction of case assignment in
verb-final languages in a theory such as Government and Binding framework
is from right to left.
In the unmarked order in all SALs, which are SOV (except Khasi), the
complement (direct object, for example) invariably precedes the verb (head)
just as the object of postposition precedes the postposition (head). This is in
contrast to English or French or Khasi, where the complement follows the
head. Thus, there is a mirror image relationship between the complement and
the head in verb-final and non-verb-final languages. Note that a majority of
languages in the world are of the following three types: verb-final, verb-medial
and verb-initial.
We shall now examine the second alternative. In the South Asian subconti-
nent, multilingualism is not an exception but a norm (Pandit 1972; Agnihotri
1992, 2001), and contact between different languages is not an obstacle but
a facilitator in communication (Pandit 1972). Thus, multilingualism implies
intense language contact as a result of which characteristic features of one
language family or language may be “transferred” to another language or lan-
guage family. Such a phenomenon is labeled “convergence” (see Gumperz and
Wilson 1971; Hock 1991).
We shall discuss some of the syntactic features shared by verb-final SALs.9
2.3 Typological characteristics of SALs

The word order in SALs is relatively free. An example of a sentence from
Hindi-Urdu reflecting Subject (S) – Object (O) – Verb (V) word order is given
in (1).
The perfect aspect marker, -ā, and the past tense marker, thā, together con-
stitute the auxiliary. We label them as Aux1 and Aux2 for convenience.
Hindi-Urdu (IA)
(1) āp ne mujhe dekh- ā thā
subject object verb aux1 aux2
you erg me see- perf pst
‘You had seen me.’
In most SALs, the verb occurs in the final position of a sentence as in (2).
However, in Khasi, the verb occurs in the medial position, and in Kashmiri
(IA) (B. Kachru 1973; Bhatt 1999; Raina 2002; Wali and Koul 1997), the finite
form of the verb – be it the main verb or the auxiliary – occurs in the second
position in a sentence. This is generally referred to as the V2 position.
As in all ‘well-behaved’ verb-final languages, in SALs the auxiliary verb
follows the main verb; the indirect object precedes the direct object except in
Sema (Achumi 2000) and Tenyidie (TB) (Kevichüsa 2007) where the direct
object precedes the indirect object in the unmarked word order, if the indirect
object is not overtly dative case-marked. If the dative case marker overtly occurs
with the indirect object, the indirect object precedes the direct object just as in
the other SALs. The set of ditransitive verbs in Tenyidie (TB) includes ts ‘to
give,’ petha ‘to teach,’ pesi ‘to inform’ and rəchə ‘to talk to’ (Subbarao and
Kevichüsa 1998).
do–io order
Tenyidie (TB)
(2) abunɔ- e leda puo- pie a ts
Abuno- nom book one- acc me gave
do io
‘Abuno gave a book to me.’
2.3 Typological characteristics of SALs 21
When the indirect object is dative case-marked, the order is io–do just as in
the other verb-final languages.
io–do order
(3) abunɔ- e a- ki lea puo ketse
Abuno- nom I- dat letter one send o ben
io do
‘Abuno sent a letter to me.’
All SALs, except Khasi, Rymbai and Pnar which are verb-medial, have post-
positions. Since SOV languages have postpositions, the genitive precedes the
head noun as predicted by the implicational universal for verb-final languages.
Comparative and superlative constructions in all SALs (except Khasi) use a
postposition, comparable to than in English, to mark the standard of compari-
son. They all strictly adhere to Greenberg’s word order universal 22.10
Most SALs (except some Tibeto-Burman languages) lack bound comparative
and superlative morphemes, comparable to the -er and -est of English. As
expected, in a verb-final language such as Hindi, the marker of comparison se
‘than’ follows the standard of comparison rādhā ‘Radha’ in (4).
Hindi-Urdu (IA)
(4) raghu rādhā se lambā hai
Raghu Radha than tall is
‘Raghu is taller than Radha.’
Some Tibeto-Burman languages, such as Tenyidie, Bodo, Hmar, Mising and

Sema, and the Mon-Khmer Khasi do have bound markers for comparison that
correspond to the bound forms -er and -est in English. For example, in Hmar
(TB) in-sāŋ is ‘tall,’ in-sāŋ-lem is ‘tall-er’ and in-sāŋ-tak is ‘tall-est,’ as in
sentence (5) below.
Hmar (TB)
(5) lali lala- nekin a- in-11 sāŋ- lem
Lali Lala- than 3s- VR- taller
‘Lali is taller than Lala.’
Thus, while some Tibeto-Burman languages and Khasi share this feature of
bound comparative and superlative markers, Dravidian and Indo-Aryan lan-
guages do not. It may be noted that Sanskrit has the markers -tar (comparative)
and -tam (superlative) occurring to the right of the adjective, as in ucca-tar
‘high-er’ and ucca-tam ‘high-est.’ Many Modern Indo-Aryan languages do
have these markers, used only in formal contexts and not productively, unlike
in Tibeto-Burman languages and Khasi.
Time adverbials (T) precede place adverbials (P) in SOV languages. Thus,
the order of their occurrence is TP.
Time and place adverbials occur in descending order in verb-final languages
(Subbarao 1984a). By descending order, we mean the superordinate chunk
of place or time occurs first, then a subordinate chunk, and then a chunk
subordinate to that follows.12 Examples of time adverbials in Tenyidie (TB)
and place adverbials in Hindi-Urdu (IA) are given in sentences (6) and (7),
respectively.
time adverbials
Tenyidie (TB)
(6) ticie 2000 do khr tarik kerepeŋu theva keba thetha ki
year January month date fifteen night time eight time
‘At eight o’clock at night on the fifteenth of January in the year 2000’
(Kevichüsa 1996; D. Kuolie, p.c.)
place adverbials
Hindi-Urdu (IA)
(7) gāndhı̄ nagar mẽ paccı̄s nambar kı̄ kot.hı̄ kı̄ chat par
Gandhi Nagar in twenty five number of house of roof on
‘On the roof of house number twenty-five in Gandhi Nagar’
(Subbarao 1984a: 18)
The order in which the constituents in time and place adverbials occur in (6)
and (7) demonstrates that all SALs except Khasi are left-branching.
We have discussed some word order universals in the verb-final SALs. We
shall now present some features that are commonly shared by SALs, and
these are the features that provide support to the notion of “South Asia as a
linguistic area.” These include compound verbs, conjunct verbs, reduplication,
echo words, conjunctive participles and the quotative.
2.4 Evidence in support of “South Asia as a linguistic area”

2.4.1 Compound verbs
A characteristic feature of SALs is the use of compound verbs (Chatterji 1926;
Masica 1976). A compound verb is “a sequence of two verbs AB (‘polar A’
plus ‘vector B’) that alternates with A (‘the polar’) with little or no difference
in meaning” (Hook 2001: 101). In simple terms, the polar verb is the main verb
either in its stem form or a conjunctive participial form depending upon the
specific language, and it is the vector verb (that is, semantically bleached) that
takes the auxiliary verb in all SALs. The vector verb, as Masica (1991: 326) puts
2.4 “South Asia as a linguistic area” 23
it, “modifies the meaning of the main verb in terms of ‘manner-specification’

(including directionality, completeness, suddenness, violence, deliberateness,
stubbornness, benefaction, affectivity, etc.) [emphasis in the original].”13
The vector verb contributes to the meaning in terms of perfectivity (Hook
1991), directionality towards/away from the speaker, suddenness (Masica
1976: 143–144), inception/completion (Butt 1995) or the attitude/feeling of
the speaker towards the event.
Hindi-Urdu (IA)
(8) vah subah-subah hamāre ghar ā- bait.hā
he early in the morning our home come- sat
Literally: ‘He came-sat our home early in the morning.’
‘He came to our home early in the morning’ where the speaker expresses
his unhappiness with the person’s coming early in the morning.
The verb ā ‘come’ may occur alone in the sentence, but alone it does not convey
the intended meaning of the speaker’s unhappiness about the event of coming.
In all Dravidian languages, and some Tibeto-Burman (Bodo and Rabha)
and Indo-Aryan (Assamese, Bangla, Oriya, Marathi, Konkani and Mangalore
Konkani) languages, the main verb is in its conjunctive participial form.
Telugu (DR)
(9) (vād.u)i appud.(u)- appud.u madhyāhnam kūd.ā vacc- (i) vēs-
(he) then- then afternoon also come- cpm DROP-
tā- d.ui
imperf- 3m,s
‘Sometimes he comes in the afternoon, too.’
In Tamil and Telugu (DR), two vector verbs can occur in a row (see sentence
(43) in chapter 9 on the Conjunctive Participle on the Website).
2.4.2 Conjunct verbs

Another characteristic feature of SALs is the productive use of noun/adjective
plus a light verb as predicate. It is the light verb that carries the tense, aspect
and agreement markers: “When the non-verbal element is a noun, it may
be treated syntactically as the verb’s object, never taking Dat–Acc marking,
but governing verbal agreement in perfective and infinitival constructions in
Central and Western languages. The Patient of the Conjunct Verbal expression
is often expressed syntactically as a Genitive adjunct of such a noun [in a few
Indo-Aryan languages]” (Masica 1991: 368).
Hindi-Urdu (IA)
(10) mãı̃ ne rām kı̄ madad kı̄
I erg Ram gen help (f.s) did.f.s.
‘I helped Ram.’
(Masica 1991: 368)
In all the Dravidian languages and in many Indo-Aryan languages (Assamese,

Bangla and Oriya) the patient of the conjunct verb takes a dative case marker.
In Marathi and Mangalore Konkani, it may either take a dative or an accusative
case marker. Having the dative case marker, however, is the preferred option
in Marathi, according to Lalita Dhareshwar (p.c.). Thus, in Telugu the patient
rādha ‘Radha’ is dative case-marked in (11) and, unlike the patient in Hindi-
Urdu, it cannot be genitive case-marked. In (11), sahāyam ‘help (noun)’ and
ceyy- ‘do’ together constitute the conjunct verb.
Telugu (DR)
(11) āvid.a rādha ki/ nā.ku/ *nā sahāyam cēsin-di
she Radha dat/ I.obl.dat/ I.gen help did-3s,nm
‘She helped Radha/me.’
2.4.3 Reduplication
Reduplication and echo formation are two characteristic features associated
with SALs. Adjectives, question words, verbs and adverbs all have reduplicated
forms (see Abbi 1992). Reduplication normally provides emphasis or imparts
distributive meaning. We shall also demonstrate that reduplication has syntactic
implications.
Repetition of the verb stem

Marathi and Konkani (IA) are the only Indo-Aryan languages that we came
across where “the repetition of the verb stem, which intensifies the meaning of
the verb” (Pandharipande 1997: 534), is used as a reduplicating device.
Marathi (IA)
(12) tı̄ bol- bol bol-lı̄
she talk- talk talk-pst.3 f,s
(stem form) (stem form)
‘She talked a lot.’
(Pandharipande 1997: 534)
2.4 “South Asia as a linguistic area” 25
Konkani (IA)
(13) ti rao- rao rao-l.i
she stand- stand stand-perf.3 f,s
(stem form) (stem form)
‘She stood (waited).’ / ‘She stood (waited) for a long time.’
(Shobha Pedhnekar and Snehal Naik, p.c.)
In Kharia (Munda), it is only the monosyllabic stem that is reduplicated and the
second stem carries the tense marker (Peterson, p.c.).
Kharia (Munda)
(14) lam- o?
search- pst
‘She/he looked for it.’
(15) lam- lam- ki
search- search- pst
‘She/he looked for [it] for a long time.’
In Ho (Munda), in verb stem reduplication, either (i) the first syllable is redu-
plicated, or (ii) only the first vowel of the stem is lengthened. In the former
case, the first syllable may also be lengthened (Deeney 1979: 58).
E.g., Case 1:
jom - ‘eat’ jō-jom - ‘repeatedly eat’
sen - ‘walk’ se-sen - ‘repeatedly walk’
Case 2:
id. - take ı̄d. - ‘habitually take’
em - ‘give’ ēm - ‘habitually give’
The lengthening of the first vowel of the verb stem seems to be unique to Ho
(Munda), as it is not found elsewhere in the subcontinent in the languages that
we studied.14
In Sema (Tibeto-Burman) too, the verb stem is reduplicated.
Sema (TB)
(16) pa- ye cu- cu- ce- ni
he- nom eat- eat- imperf- pres
‘He eats and eats.’
(Achumi 2000)
2.4.4 Echo words

Formation of echo words
Echo words result from a partial reduplication of words where an initial con-
sonant or syllable is replaced in the reduplicated word. Each language has its
preferred consonant or syllable for such formations; in Hindi-Urdu, it is w-; in
Kashmiri, -; in Bangla, .ta-; and in Telugu, -.
Hindi-Urdu (IA)
(17) cāy ‘tea’ cāy-wāy ‘tea and the like’
khānā ‘food’ khānā-wānā ‘food and the like’
Bangla (IA)
(18) hɔr ‘house’ ghɔr-t.ɔr ‘house and the like’
bhat ‘cooked rice’ bhat-t.at ‘cooked rice and the like’
Telugu (DR)
(19) puli ‘tiger’ puli-ili ‘tiger and the like’
ceruku ‘sugarcane’ ceruku-iruku ‘sugarcane and the like’15
2.4.5 Conjunctive participles

SALs have a non-finite construction generally labeled “the conjunctive par-
ticipial construction.” In this construction, the sentence carries only one finite
verb, and all the subordinate clauses carry a participial form of the verb
which is non-finite. This non-finite form occurs in languages like Hindi-Urdu,
Punjabi and Kashmiri. In contrast, the conjunctive participle has a marker
which is homophonous with the past tense marker in some Indo-Aryan lan-
guages. According to Masica (1991: 271), in Indo-Aryan languages such as
Bangla, for example, the conjunctive participle marker is a “Perfect marking.”
According to Krishnamurti (2003: 291), “(I)n South Dravidian I and South
Dravidian II, the past stem is the same in finite verb inflection (past tense) and
in the formation of past verbal adjective (relative participle) and the gerund
(perfect participle) [conjunctive participle in our terminology here].”
Some of the functions of the conjunctive participle include imparting the
meaning of a coordinating conjunction indicating a sequential action; a con-
cessive interpretation if used with an inclusive particle with the conjunctive
participle and a negative having the matrix verb in its scope; and the interpre-
tation of an alternative action in the sense of ‘instead of’ when the conjunctive
participle is in the scope of a negative (Davison 1981; Y. Kachru 1981) (see
chapter 9 on the Website for more comprehensive discussion).
2.4.6 The quotative

Dravidian, Tibeto-Burman, Munda and some Indo-Aryan languages, such as
Nepali, Assamese, Bangla, Oriya, Marathi and Konkani have a post-sentential
complementizer that is normally a form derived from the verb ‘to say,’ and
it is labeled as the ‘quotative’ (Emeneau 1956; Kuiper 1967; Hock 1975,
1982a; Masica 1976; Kachru 1979; Subbarao, Viswanath Rao, Rao and Saxena
1989). The quotative acquired several other functions, such as reason marker
or purpose marker, in the course of time (see chapter 6 for more comprehensive
discussion).
The position of occurrence of the quotative to the right of the embedded
clause is ‘consistent’ and ‘harmonic’ in terms of unmarked sentence structure
of verb-final languages. However, some Indo-Aryan languages, such as Hindi-
Urdu, Punjabi and Kashmiri only have the initial complementizer (IC), and
many other Indo-Aryan languages, such as Konkani, Marathi, Oriya, Bangla,
Assamese, have an IC in addition to having an FC. According to Bayer (2001:
29), “the IC-clause, although theta-marked, falls outside the pattern of argu-
ment licensing in head-final languages.” Such occurrence may be the result
of borrowing from an external source and it does obey the universal FOFC
(Final-over-Final Constraint) proposed in Holmberg (2000). We shall discuss
this in detail in chapter 6, keeping in view a discussion in Biberauer et al.
(2009) concerning the occurrence of the complementizer (IC or FC) and the
polarity question expression (Davison 2007a) in some Indo-Aryan languages.
2.5 Parametric variation and SALs

In this section we shall briefly discuss some of the parameters that SALs
opt for. These include: null pronominals and their relation to the absence
of an overt object with transitive verbs, relatively free word order, marked
absence of superiority effects and the absence of Tense and Person Harmony
in complement clauses.
2.5.1 Null pronominals

Languages that have a rich subject–verb agreement or morphologically uniform
inflectional paradigms (Jaeggli and Safir 1989) permit null pronominals (or pro
drop, as it is generally called) optionally in a sentence. Most of the Indo-Aryan
languages, Munda languages, some Tibeto-Burman languages such as Mizo,
Hmar, Aimol and Paite, and Dravidian languages except Malayalam exhibit a
rich agreement system. A pronoun (pro) functioning as a subject, an object, an
indirect object or an oblique object is freely dropped in these languages.
Tibeto-Burman languages such as Manipuri, Rabha, Tangkhul do not exhibit

subject–verb agreement, and they have morphologically uniform inflectional
paradigms in the sense that agreement markers are not present in all persons
and numbers. Hence, they permit null pronominals of the subject in a simple
as well as in an embedded sentence. In some Tibeto-Burman languages, such
as Tenyidie (Angami), “pro-drop [the occurrence of null pronominals] of the
embedded subject is obligatory. If an overt pronoun is present, it cannot be
coreferential with the matrix subject as in [(20) below]” (Subbarao 2000: 102–
103).
Tenyidie (TB)
(20) dapheŋ- ei [proi ilɔŋ- nu vɔ - tyo- -di] aki pu
Dapheng- nom Shillong- to go- fut- that me told
‘Dapheng told me that he would go to Shillong.’
(Subbarao 2000: 102–103)
The absence of the embedded subject is indicated by pro in (20).

The pro-drop parameter also permits a transitive verb not to have an overt
direct object both in SALs and in (Indian) English. The following dialog from
Hindi-Urdu is illustrative:
Hindi-Urdu (IA)
Speaker A:
(21) kyā āp nei us- koj cit.t.hı̄k likh- dı̄k
pol you erg he.obl-dat letter write- gave
‘Have you written a letter to him?’
Speaker B:
(22) hã˜ (mãı̃ ne)i proj prok likh- dı̄k
yes (I erg) write- gave
Literally: ‘Yes, (I) wrote.’
A sentence corresponding to (22) in Indian English is given in (23). Sentence

(23) is ungrammatical in Standard British or American English.
Indian English
(23) Yes, I wrote.
2.5.2 Relatively free word order

Lexical constituents in a sentence can freely be moved / scrambled in Indo-
Aryan, Dravidian and Munda languages; such movement is less free in Tibeto-
Burman languages. For example, the constituents of (24) can be scrambled as in
(25)–(26). The perfect aspect marker, -ā, and the past tense marker, thā, together
constitute the auxiliary. We label them as Aux1 and Aux2 for convenience.
Hindi-Urdu (IA)
(24) āp ne kis- ko dekh- ā thā
you erg who.obl- dat see- perf pst
subject verb aux1 aux2
Literally: ‘You whom saw?’
‘Whom did you see?’
The main verb along with a u x 1 may be fronted:
(25) ?dekh- ā kis- ko thā āp ne

see- perf who.obl- dat pst you erg
verb aux1 aux2
The direct object kis-ko ‘who.obl-dat’ may be fronted:
(26) kis- ko dekh- ā thā āp ne

who.obl- dat see- perf pst you erg
v e r b a u x 1 a u x 2 su bject
Such movement is due to bringing a specific constituent in a sentence into focus

for the sake of emphasis (see Gambhir 1981 for further details).
Question expressions normally occur in situ and there is no obligatory ques-
tion word movement to the left to the Spec position. Further, there are no supe-
riority effects with regard to question expressions in SALs except in Khasi.
Lasnik and Uriagereka, with Boeckx (2005: 83) define superiority effects as:
“Descriptively, when two Wh-phrases compete for movement to a given site, it
is the ‘superior’ (i.e. higher) Wh-phrase that moves . . .” Interestingly, for exam-
ple, the question expression questioning indirect object can occur in sentence-
initial position, as (27) from Telugu (DR) shows.
Telugu (DR)
(27) evari-ki eppud.u enduku
who.obl.dat when why
indirect object time adverb reason adverb
elāgu ēdi evaru istār-
how what who will give-
manner adverb direct object subject
o nāku teliyadu
comp to me not known
Literally: ‘I do not know to whom, when, why, how, what, who will give not
known.’16
2.5.3 Absence of tense and person harmony in complement clauses

A distinct feature of SALs is that the principle of Tense Harmony between the
tense of the verb of the matrix clause and that of the subordinate clause is not
observed. In other words, the tense of the main clause and the embedded clause
do not have to agree. The matrix verb and the embedded verb may both have
their own independent tenses as a result. Masica (1991: 403) observes: “There
is no sequence of tenses in NIA (New Indo-Aryan languages), the embedded
sentence retains its tense.”
Hindi-Urdu (IA)
(28) saritā- ne kahā thā ki [mãı̃i āp se kal
Sarita.3 pn,s,f- erg said had comp I. 1pn,s,f you with tomorrow
mil.ūngı̄]17
will meet. 1pn,s,f
Literally: ‘Saritai told (me) that Ii will meet you tomorrow.’
‘Sarita had told me that she’d meet me tomorrow.’
Though the matrix verb kahā thā ‘had said’ carries the past tense marker and
the embedded verb mil.ūngı̄ ‘will meet’ carries the future tense marker, the
sentence is grammatical. In English and some other languages, the sequence-
of-tense is observed,18 and extends to time adverbs as well (as in (29) and
(30)). The tense marker and the time adverb are in italics.
English
(29) *? Sarita told me that she’ll meet me tomorrow.
(30) Sarita had told me that she’d meet me the next day.
Unlike English and many other languages, the pronominal forms of the subjects
of the matrix and the embedded clause do not have to be identical in shape in
indirect speech in almost all SALs.19 This is due to the fact that the speaker is
quoting verbatim employing the post-clausal quotative complementizer. How-
ever, significantly in Hindi-Urdu, Kashmiri, Bangla, Oriya and Marathi, the
same phenomenon of tense mismatch and retention of the pronominal forms of
the direct speech are found in spite of the fact that the complementizer occurs
in a pre-clausal position.
2.5.4 Absence of expletive expressions

The English-type languages require a dummy/pleonastic subject such as it with
meteorological predicates or a pleonastic there in sentences with the verb be
denoting location, subject to certain conditions related to definiteness of the
2.6 The effects of language contact 31
entity that is located, etc. In contrast, no SAL has an expletive it/there-type

construction such as it in (31) and there in (32), as in English or French.
English
(31) It is raining in Simla.
(32) There is no one there in the room.
In Hindi-Urdu (IA) bāri ‘rain’ is the subject in (33), and the verb agrees with
bāri ‘rain’ in (33) as is the case in all SALs.20
Hindi-Urdu (IA)
(33) dillı̄ mẽ bāri ho rahı̄ hai
Delhi in rain f,s happen progr f,s pres
‘It is raining in Delhi.’
In the following section we present two instances where intense language con-
tact in the subcontinent results in the addition of a new syntactic phenomenon.
2.6 The effects of language contact

Languages of the four families have been in intense contact with each other for
thousands of years and this has resulted in convergence. We shall present two
cases of convergence between Indo-Aryan, Dravidian and Tibeto-Burman.
2.6.1 Adding a new phenomenon to the grammar

The first case shows how the phenomenon of Backward Control from Telugu
(Dravidian) is added to the grammar of Dakkhini (Indo-Aryan). Telugu (DR)
permits both Forward and Backward Control. An example of Backward Control
in Telugu is given in (34).
However, control structures involving a time expression in the matrix pred-
icate in Telugu do not permit Forward Control, and Backward Control is the
only option, as in (34) (see chapter 7 for more details). The controller mēmu
‘we’ is in the embedded clause and the controllee ∀ occurs in the matrix clause
in (34).
backward control
Telugu (DR)
(34) [S1 [S2 mēmui ikkad.i-ki vacc- iS2 ] ∀ i padi ēl.l.u ayy-in-
we here- to come- cpm ten years happen-pst-
diS1 ]
3s, nm (def agr)
‘It is ten years, since we came here.’
Hindi-Urdu (IA) does not permit Backward Control to occur in sentences

involving time expressions in the matrix clause (35).
backward control
Hindi-Urdu (IA)
(35) [S1 [S2 *hami yahā˜ ā karS2 ] ∀i das sālj huej S1 ]
we here come cpm ten years happened
Intended meaning: The same as in (34)
Hindi-Urdu (IA) permits only a perfect participle in a Forward Control structure

when a time expression such as das sāl ‘ten years’ is the grammatical subject
in the matrix clause as in (36).
forward control
(36) [S1 [S2 PROi yahā˜ ā- ye hueS2 ] ham.ẽi das sālj huej S1 ]
here come- perf pple.obl we.dat ten years happened
Meaning: The same as in (34)
In Dakkhini, when a time expression occurs as the predicate of the matrix

sentence, the conjunctive participle occurs just as in Telugu, when the subjects
of the embedded clause and matrix clause are differently case-marked: “That
is, Dakkhini too permits only Backward Control, which is a non-Hindi like
pattern in such constructions” (Subbarao and Arora 2009: 365).
The following example is illustrative.
backward control
Dakkhini (IA)
(37) [S1 [S2 ham ya-ku ā- keS2 ] ∀ das sāl ho gayeS1 ]
we here-to come- cpm ten years 3p,m happened 3p,m
(Subbarao and Arora 2009: 365)
Thus, we observe that Dakkhini incorporated a new phenomenon of Backward

Control that not only involves having new syntactic structures, but also violates
the rules of the source language, Hindi-Urdu. Backward Control in Hindi-Urdu
yields ungrammatical sentences.21
2.6.2 The non-nominative subject construction

The second case involves non-nominative subjects in Indo-Aryan and Tibeto-
Burman. In Indo-Aryan languages the subject is dative case-marked – in Hindi-
Urdu and Punjabi – or genitive case-marked – in Bangla and Assamese – when
the subject (experiencer) possesses physical ailments or mental states (see
Kokborok is a Tibeto-Burman language spoken in the state of Tripura, where

Bangla is also spoken by a significant majority in the state. Most of the Tibeto-
Burman languages have a nominative subject and not a dative or genitive
case-marked subject. However, Kokborok (TB) has a genitive case-marked
subject. This could plausibly be due to convergence with Bangla, an Indo-
Aryan language.
Bangla (IA)
(38) amar t.hand.a lege- che
I.gen cold feel- pres
‘I have a cold.’
Kokborok (TB)
(39) a- ni kɔngrai tɔng- ɔ
I- gen cold be- pres
(Naphurai Jamatia, p.c.)
The only plausible explanation for the occurrence of a sentence with a genitive
subject could be that the construction is borrowed from Bangla.
There are several other instances of convergence between Dravidian and
Indo-Aryan, between Indo-Aryan and Munda, and between Indo-Aryan and
Tibeto-Burman, some of which will be discussed in this book. Interestingly,
though Khasi (Mon-Khmer) has been in contact with some other non-Mon-
Khmer languages for centuries, our work on Khasi shows that there is no
linguistic convergence. The topic of language contact and convergence needs
an in-depth investigation which has implications for the nature of syntactic
change, which is beyond the scope of the present work.
2.7 Unique features of each language family

In this section we present specific sets of features that uniquely characterize
a language family. We start with Dravidian, then followed by Tibeto-Burman,
which in turn is followed by Munda and then the Indo-Aryan language family.
2.7.1 Dravidian language family

Dravidian languages are the most homogeneous set of languages in the sub-
continent in having syntactic phenomena which exhibit very little syntactic
variation.
Sandhi rules
Dravidian languages have sandhi rules operating between the constituents of
a sentence which results in the hearer’s perception of a sentence as a single
unit without any pause between words. The operation of such sandhi rules is
the result of the rich agglutinating processes that Dravidian languages have
in comparison to all the other families of the subcontinent. In Telugu, “when
the two vowels come together in the case of two separate words, the first
vowel is generally dropped if it is short” (Krishnamurti and Gwynn 1985: 72).
A similar phenomenon of elision is observed in the other literary Dravidian
languages too. For example, in (40) below, from Telugu, the final vowel in
each word except for the last one is deleted/elided, as a result of which the
entire sentence becomes one whole uninterrupted unit. We have indicated the
elided vowels in parentheses. The words perceived as a unit are enclosed in
brackets.
Telugu (DR)
(40) [atan(u)- eppud.(u)- ēm(i)- enduk(u)- an(i)- annād.(u)- ō]
he- when- what- why- quot- said- dub mkr
Literally: ‘I wonder what, when, why he said it.’
‘I wonder what he said when and why?’
It is the clitic -ō in (40) that imparts the interpretation of ‘I wonder’ in Telugu
and the other Dravidian languages.
A similar phenomenon is observed in the other Dravidian languages too. For
example:
Kannada (DR)
(41) [awan(u)- ēn(u)- anta] hēl.ida
he what quot said
‘What did he say?’
(Hemananda Bisembli, p.c.)
In Malayalam (DR) the first vowel of a word is deleted even when it is preceded
by a word ending in a long vowel (Sreekumar, p.c.), a process that Telugu does
not permit.
Malayalam (DR)
(42) [aval.(ə)- ivit.(e)- un.t.- ō- (e)nn(ə) en (i)- kk(ə) ar.(i)- illa]
she- here- be.pres dub mkr quot I- dat know- neg
‘I do not know whether she is here.’
Double Dative Subject construction

Dravidian languages have the Double Dative Subject construction, in which a
sentence has two subjects that are dative case-marked. In Kannada, Malayalam
and Telugu (DR), a dative predicate may assign a single or double dative
case-marking to a subject. Double dative-case marking is permitted in many
Dravidian languages in dative subject constructions expressing inalienable pos-
session and part–whole relationship (Subbarao and Bhaskararao 2004). This
is a feature found only in many Dravidian languages and not elsewhere in the
subcontinent. Thus, the following two types of case marking of the subject
DP are permitted: genitive-dative as in (43) and dative-dative as in (44). The
possessor is genitive case-marked and the possessed – the body part – is dative
case-marked in (43) below.
Kannada (DR)
(43) avan- ai kan.n.j - ige gāyak āgi- de∗i∗j,k
he- gen eye- dat injury happen- agr
‘He got hurt in his eye.’
In Kannada, the possessor as well as the possessed are also both dative case-
marked as in (44).
Kannada (DR)
(44) avani - ige kan.n.j - ige gāyak āgi- de∗i∗j,k
he- dat eye- dat injury happen- agr
(N. Varija, p.c)
The Double Dative Construction is also found in Marathi and Mangalore

Konkani (Indo-Aryan) (Lalita Dhareshwar, p.c.) and this could plausibly be
due to Dravidian influence (see chapter 5 for details).
Expletive light verb with a negative

Another distinct feature of Dravidian is that a finite verb in the affirmative or
a negative may be followed by a light verb ‘to do’ with no semantic content.
No other language family, to the best of our knowledge, has this construction
(Subbarao 2010b).
Kannada (DR)
(45) avanu bara- lilla- a mād.a- lilla
he came- not conj do.pst- not
Literally: ‘He did not come; he did not do.’
‘He did not come at all.’
Malayalam (DR)
(46) avan vannat- um illa (avan) ceytat- um illa
he come.pst- conj neg he do.pst- conj neg
Meaning: The same as in (45)22
(Sreekumar, p.c.)
Such sentences are not permitted in Tamil.

Tamil (DR)
(47) *avan vara- villai avan ceyya- villai
he come.pst- neg he do.pst- neg
Cognitive gerunds
Dravidian languages have a construction (the cognitive gerund)23 in which
the predicate in its gerundival form occurs to the left of the predicate, which
provides a focus interpretation to the gerund. The following examples from
Tamil and Kannada are illustrative.
Tamil (DR)
(48) avan vandadu vantān ān.āl pan.am kon.t.u varaville
he coming (gerund) came but money bring.cpm did not come
‘As for (his) coming, he came, but he did not bring the money.’
(Boologa Rambai Arpudhasamy, p.c.)
Kannada (DR)
(49) awaru baruwud(u) ēnō bandaru, ādare dud.d.u taralilla
he come.gerund as for came but money did not bring
Time expressions and negative and affirmative CPs

When a time expression occurs in the predicate of the matrix clause, the truth-
value of the sentences with an affirmative CP and negative CP is “nearly
identical,” in Telugu (Krishnamurti and Gwynn 1985), Malayalam (Sreekumar
and Sobha Nair, p.c.) and Tamil, as (50) and (51) from Tamil show.
Tamil (DR)
w i t h a n a f fi r m a t i v e c p
(50) nı̄ŋa enka vı̄t.t.u- kku van.tu ren.d.u varusam akiradu
you our home- dat come.cpm two years happened
‘It has been two years since you came to our house.’
w i t h a ne g a t i v e c p
(51) nı̄ŋa enka vı̄t.t.u- kku var- āma pōy- i ren.d.u
you our home- dat come- neg go- cpm two
varusam akiradu
years happened
‘It has been two years since you did not come to our house.’24
(Boologa Rambai Arpudhasamy, p.c.)
To summarize, an extensive use of sandhi rules operating between the con-

stituents of a sentence, the Double Dative Construction, the occurrence of
cognitive gerunds and of an expletive light verb with or without a negative are
some significant features that distinguish Dravidian from the other language
families of the subcontinent.
2.7.2 Tibeto-Burman
Nominalization
The characteristic feature of Tibeto-Burman languages is that in the Gap Rel-
ative, the embedded verb is [-finite] in its nominalized form (Matisoff 1972;
Herring 1991; Bickel 1999; Subbarao and Kevichüsa 1999; Lahaussois 2003).
The nominalizer is also used in the formation of non-finite complement clauses,
as one would expect. Further, it is also used productively in the formation of
compounds.
For example:
Tenyidie (TB)
(52) ke- petha- mie
nozr- teach- person
‘teacher’ (the one who teaches)
(Kevichüsa 1996)
In Manipuri (TB) too the nominalizer is used in the formation of compounds.
Manipuri (TB)
(53) i- mən- na- bə
my- similar- (verb) VREC- nozr
Literally: ‘a person who is similar to me’
‘(my) friend’
(54) lai- yek- pə

image- draw- nozr
Literally: ‘a person who draws images’
‘painter’
(Sarju Devi Thokcham, p.c.)
Internally Headed Relative Clauses (IHRCs)

A distinct and unique feature of the Tibeto-Burman language family is the
occurrence of the IHRC in a subset of languages. The internal head in an IHRC
may or may not carry an overt postposition. In those languages in which the
internal head does not carry an overt postposition, Postposition Incorporation
takes place to make a comitative PP eligible to head a gap relative (see (55)
below, and also see chapter 8 for a detailed discussion).
Incorporation of postpositions in the verb

The incorporation of a postposition in the verb is another unique feature found
in the Tibeto-Burman language family. Such incorporation of a postposition
increases the valence of the [−transitive verb] and makes it [+transitive] (see
In Thadou (TB) for a comitative PP to qualify as head in an EHRC, the
comitative postposition pi and the verbal reciprocal ki are both incorporated in
the verb. The verbal reciprocal and incorporated postposition are in italics in
(55).
Thadou (TB)
(55) thaŋboii - in25 sinema ve øj ai - ki ɔn-
Thangboi- erg cinema (in order to) see 3s- VREC go-
pi- nu- cuj ka kadɔŋmaj - nu aj - hi
with (incorporated postposition) f mkr- def my girlfriend- f mkr 3s- be
‘The girl with whom Thangboi went to see the movie was my girlfriend.’
(Subbarao and Haokip 2009)
(See chapter 8 for a detailed discussion.)

To summarize, the productive use of nominalizations, the occurrence of
the IHRCs and postposition incorporation in the verb are some of the salient
features of the Tibeto-Burman language family.
2.7.3 Austro-Asiatic (Munda)

Absence of the nominal anaphor
Munda languages are unique in indigenously having a verbal anaphor (reflexive
and reciprocal) alone, and no nominal anaphor at all. It is the intransitive marker
that is infixed as a verbal anaphor, which reduces the valence of the transitive
verb. The verbal anaphor, which is monomorphemic, performs a variety of
functions, such as those of a detransitivizer (anti-causative) and a passive.
Kharia (Munda) has an indigenous verbal reflexive d.om (56), and a nominal
anaphor apan ‘self,’ borrowed from Sadani (IA) (Peterson 2006).
verbal anaphor
Kharia (Munda)
(56) proi yo- d.omi − ki- kiyari
they see- VR- middle pst- dual
‘They saw themselves (e.g., in the mirror).’
(Peterson 2006)
Sora (Munda)
(57) al’len ar’- grʔ- tə- biy
we VREC- see- [-pst]- 1p
‘We see ourselves.’
(Anderson and Harrison 2007: 328)
Korku (Munda), however, lost the verbal anaphor and it borrowed the nominal
anaphor from Indo-Aryan.
Incorporation into the predicate

The languages of the Munda group have the phenomenon of incorporation into
the predicate where subject, direct object and indirect object pronominal clitics
are incorporated as agreement markers, just as in Tibeto-Burman languages.
Sora/Savara permits incorporation of nouns and a variety of other elements in
the verb. (See chapter 4 for details.)
In Sora (Munda), in (58), the complex predicate pokunpunŋ ‘stab.belly.knife’
contains the verb stab, the locative PP belly and the instrumental object knife.
Note that all the three parts of the predicate constitute a single lexical item,
a phenomenon unique in Sora amongst the languages of the subcontinent.
The direct object is also incorporated into the verb as the second singular
pronominal clitic am. It is followed by the past tense marker t ēn. All these
segments together constitute the total verbal complex.
do, locative and instrumental pp incorporation

Sora (Munda)
(58) pokunpunŋ- am- tēn?
stab.belly.knife- 2s- 3.pst
‘(Who) stabbed you in the belly?’
(Ramamurti 1931: 25)
Munda languages exhibit possessor agreement. In Santali (Munda) the pos-

sessor agreement marker is incorporated and it “always refers to the object”
(Macphail 1983: 41), and not to the subject, demonstrating subject–object
asymmetry. (See chapter 4 for details.) Kurmali, an Indo-Aryan language
spoken in Bihar, exhibits possessor agreement and it could be due to the influ-
ence of the Munda languages spoken in the area.
The absence of the nominal anaphoric device, rich verb agreement (as dis-
cussed in chapter 4) and the incorporation of various elements in the verb are
some of the distinct features of the Munda languages.
2.7.4 Indo-Aryan
Indo-Aryan languages are heterogeneous in having syntactic phenomena which
exhibit a wide range of syntactic variation. Just to cite an example, the com-
plementizer in one set of languages occurs to the left of the embedded clause,
while it occurs to the right of the embedded complement in another set. While
the right peripheral complementizer in one subset is a form of the verb say
(generally labeled as the quotative), the other subset has a complementizer
that is derived from a pronoun. While the occurrence of the right peripheral
complementizer (the quotative or the one derived from a pronoun) conforms
to a left-branching structure in uniformity with the basic word order of IA
languages, the complementizer that occurs to the left does not conform to a
left-branching structure (see chapter 6 for a detailed discussion with suitable
examples). Indo-Aryan thus is a “less well-behaved language family” in com-
parison to Dravidian and Tibeto-Burman, which are the “most well-behaved
families.”
We shall discuss some distinct features of Indo-Aryan here.26
Word order
It is only in Indo-Aryan that a V2 language such as Kashmiri is found. All the
other languages are verb-final.
Relative clauses
Indo-Aryan languages have three types of relative clauses: relative-correlative
clauses, right-adjoined clauses and VP-adjoined clauses (Dasgupta 1980;
Subbarao 1984; Dayal 1996; Bhatt 2003). Such a wide variety is not found
in either Dravidian or Tibeto-Burman.27
In the gap relative clause too, the positions that can be relativized vary from
language to language. While Bangla, Hindi-Urdu, Punjabi and Kashmiri cannot
have gap relatives in which an Indirect Object (IO) or Oblique Object (OO) is
modified, Gujarati, Konkani and Marathi permit the modification of the IO as
well as of some specific PPs of the OO.
While Bangla, Hindi-Urdu, Punjabi and Kashmiri do not permit the modi-
fication of either the subject or the theme in the non-nominative subject con-
struction, Marathi, Konkani and Gujarati do (see chapter 8 for details).
2.8 Conclusion 41
Non-nominative subject construction

In the non-nominative subject construction, the subject is either dative, genitive
or accusative case-marked in Indo-Aryan languages. Such wide variation is not
found elsewhere in the subcontinent (see chapter 5 for details).
There are some ergative–absolutive languages, such as Hindi-Urdu, Punjabi,
Marathi, Konkani and Kashmiri, and there are some nominative–accusative
languages, such as Bangla, Oriya and Sinhala (see chapter 4 for details).
The embedded subject in the participial (gap) relative may be either genitive
case-marked as in most Indo-Aryan languages, or genitive or nominative case-
marked, as in Oriya (see chapter 8 for details).
Clefts
In a cleft construction, a constituent is focused. No Indo-Aryan language except
Sinhala (see Gair 2003 for details) permits clefts. It is only in Sinhala that they
are found and it could plausibly be due to Dravidian influence. We provide an
example from Sinhala (IA) below in (59).28
The constituent that is clefted, for example ē miniha ‘that man’ in (59),
occurs to the right of the sentence as the final constituent (Gair 2003), as is the
case in Dravidian and Tibeto-Burman languages.
Sinhala (IA)
(59) ı̄ye gunəpālə- t.ə salli dunne ē miniha
yesterday Gunapala- dat money gave-focus that man
‘It was that man who gave Gunapala money yesterday.’
(Gair and Paolillo 1997: 41)
Nominal anaphors
Indo-Aryan languages predominantly exhibit a pattern in which only the nom-
inal anaphor (reflexive and reciprocal) occurs in a sentence, and the verbal
anaphor that occurs in a very few languages has functions that are limited in
comparison to the other language families of the subcontinent (see chapter 3
for details).
2.8 Conclusion
In this chapter, we have provided a brief description of the major typological
features of SALs with a view to enabling the reader to understand some of the
structures available in these languages. We have also discussed some specific
features that are either unique or significant in terms of the structures in a
particular language family. We have provided two instances to demonstrate the
crucial role that syntactic convergence plays in language contact situations.
Though the languages of the subcontinent belong to four genetically different

language families, their syntactic features are almost identical. This is due to the
following: (i) except Khasi (Mon-Khmer), they are all verb-final languages –
except Kashmiri which is V2 – which partially explains their shared features;
and (ii) they have been in close contact with each other for thousands of years.
These common syntactic features provide a strong support for the notion of
“South Asia as a linguistic area” proposed by Emeneau29 (1956), articulated
earlier by Chatterji (1926) and Bloch (1934), and elaborated in Masica (1976)
and others. There are, however, certain features that appear to be language-
family-specific or language-specific. For example, the absence of the nominal
anaphor and the infixation of an intransitive marker to mark the coindexation
of the subject noun phrase with a null element in an argument position – the
sole indigenous device in Munda languages – are not found in any other family
of the subcontinent. Another example is the partial reduplication of the verb
stem, aspect marker or adjective in the formation of polarity (yes/no) questions
in Kokborok (TB) (see the appendix for this chapter on the Website, and also
Subbarao et al. 2010).
In spite of the parametric differences and syntactic features unique to each
language family, the fact that many syntactic processes are shared by the
languages of the subcontinent indicates that there is linguistic unity in diversity.
(See the appendix for this chapter on the Website for a discussion of more
cases of reduplication in SALs.)
3 Lexical anaphors and pronouns in
South Asian languages
3.1 Introduction
The classification of nominal expressions in terms of anaphors (reflexives and
reciprocals), pronouns and referential expressions has been one of the major
areas of attention in the generative tradition, especially in the Government and
Binding framework and the Minimalist Program. Anaphora in simple terms
means ‘backward reference.’ Lust et al. (2000: 885) point out that it is “a
relation between a form and a linguistic antecedent.” The interpretation of the
anaphor is in some way determined by the interpretation of its antecedent (Lust
1986; Wasow 1986). In this chapter, we shall discuss some of the major issues
in the study of anaphors and pronouns with a view to arriving at their typology
in SALs.1 Based on evidence from SALs, we wish to demonstrate that
(i) an anaphor may have a nominative or non-nominative subject as its
c-commanding/subcommanding antecedent;
(ii) the verbal reflexive (VR, hereafter) and verbal reciprocal (VREC, here-
after), which are monomorphemic, perform a variety of functions and are
invariably “subject-oriented”;
(iii) the verbal anaphor2 in SALs provides evidence in support of the implica-
tional universal which states: if the verbal anaphor is coindexed with the
indirect object or oblique object, it should invariably be coindexed with
the direct object;
(iv) long-distance binding is permitted only in non-finite clauses, and is pro-
hibited when either the VR/VREC or a complex form of the anaphor
is present in the embedded clause as this “insulates” the anaphor from
coindexation with a long-distance antecedent;
(v) Blocking Effects are found only in languages without subject–verb
agreement;
(vi) scrambling does not affect anaphor–antecedent relations; and
(vii) all SALs except Marathi (IA) obey Principles A, B and C of the Binding
Theory.
Thanks to Frans Plank and Martin Everaert for their helpful comments on this chapter.
43
44 Lexical anaphors and pronouns in SALs
This chapter is organized as follows: section 3.2 provides a brief description

of the various forms of anaphors in each language family. Section 3.3 deals with
the occurrence of anaphors in different language families. Section 3.4 deals with
the nature of the verbal reflexive/reciprocal. Section 3.5 concerns itself with the
occurrence of lexical anaphors with a dative/genitive subject as antecedent, and
section 3.6 focuses on the absence of the verbal anaphor in the dative subject
construction. Section 3.7 deals with long-distance binding and Principle A.
Section 3.8 shows that what appears to be a case of long-distance binding in
Small Clauses is in fact not long-distance binding. Section 3.9 discusses the
occurrence of an anaphor in a locative PP. Section 3.10 focuses on Principles
A, B and C of the Binding Theory and their validity in SALs. Section 3.11 is
on anaphors and scrambling. Section 3.12 deals with Blocking Effects, while
section 3.13 shows that a subcommanding NP can be an antecedent to an
anaphor. Section 3.14 provides evidence in support of the phenomenon of
Case Copying in Telugu (DR) and Manipuri (TB), and section 3.15 is the
conclusion.
3.2 Form of the anaphor in South Asian languages

In this section we shall discuss the various forms of the nominal and verbal
anaphors in SALs. Binding, in SALs, is expressed either in terms of:
(i) a nominal anaphor (reflexive or reciprocal), or
(ii) a verbal anaphor (a verbal clitic for the reflexive or reciprocal), or
(iii) both nominal and verbal anaphors.
A few Tibeto-Burman languages (Ao and Tenyidie, for example) have a
homophonous form for a pronoun as well as an anaphor. Most of the Indo-Aryan
languages, except Gujarati, Marathi and Sinhala, and some Tibeto-Burman lan-
guages, along with Malayalam, Toda and Korku, have only the nominal form of
the anaphor. All Dravidian languages, except Malayalam and Toda, and many
Tibeto-Burman languages and Khasi (Mon-Khmer) have both the nominal and
verbal form of the anaphor. Most of the Munda languages have only a verbal
device, the nominal device not being present except in a few languages such as
Juang (Patnaik and Subbarao 2000), Ho, Kharia and Korku (Nagaraja 1999).
There are languages in the subcontinent that have both a complex (polymor-
phemic) and simplex (monomorphemic) form of the nominal anaphor. Both
forms are permitted in a “subcategorized position” (Chomsky 1965).
We shall show that reduplication and Case Copying play an important role
in the formation of nominal anaphors in Telugu (DR) and Manipuri (TB).
It is these processes that enable an anaphor to establish coreference with its
antecedent (see 3.14 for more discussion and examples). It is only the simplex
form that permits long-distance binding except in Marathi (IA), in which the
3.2 Form of the anaphor in SALs 45
complex form may also permit such binding. The verbal device is a simplex
(monomorphemic) form in SALs and it blocks long-distance binding.
Before we proceed further, we shall present a brief discussion of some rele-
vant notions related to anaphors and pronouns. Anaphora plays a crucial role in
natural languages. In the generative tradition, the study of anaphors (reflexives
and reciprocals) and pronouns has played a major role. There are three prin-
ciples of the Binding Theory “defined over configurational representations of
sentence structure, because ‘binding’ implies ‘c-command and coindexing’ in
a local domain” (Lust et al. 2000: 3).
Binding principles:
Principle A: an anaphor is bound in a local domain.

Principle B: a pronominal is free in a local domain.
Principle C: an R-expression is free.
(Chomsky 1986: 166)
Some of the notions that are important in the study of anaphora include: local vs.
long-distance binding, c-command, governing category, Functional Complex,
etc. (see Lust et al. 2000: appendix).
Languages that only have a nominal device are the following:
Indo-Aryan: Hindi-Urdu, Kashmiri, Punjabi, Nepali, Bangla,
Assamese and Oriya, just to name a few
Dravidian: Malayalam and Toda
Munda: Korku
Tibeto-Burman: Kokborok and Tenyidie (only the nominal reflexive)
The nominal device consists of either a simplex or a complex anaphor. The
complex form in SALs contains a bipartite structure whose constituents may
or may not be identical.
Languages that have a nominal as well as a verbal anaphor are the following:
Indo-Aryan: Marathi, Gujarati and Sinhala
Dravidian: all except Malayalam and Toda
Tibeto-Burman: Hmar, Mizo, Aimol, Thadou, Zou, Manipuri, Rabha,
Tenyidie (no verbal reflexive, only a verbal reciprocal), Ao and
Bodo
Mon-Khmer: Khasi
Munda: Kharia and Juang (only a verbal reciprocal)
Languages that have an indigenous verbal anaphor alone are the following:
Munda: Ho, Kharia, Mundari, Santali, etc.
A nominal anaphor too is found in Ho, Kharia and Juang. It is borrowed from
Hindi, Sadani and Oriya (IA), respectively.
We shall now provide a description of the form of the anaphor in the four
South Asian language families.
Table 3.1 Forms of the anaphors apne āp / apne + postposition
Complex form Simplex form
Accusative/dative apne-āp-ko ‘self acc’ apne-ko ‘self acc’

Instrumental apne-āp-se ‘self-with’ apne-se ‘self-with’
Locative apne-āp-par ‘self-on’ apne-par ‘self-on’
apne-āp-mẽ ‘self-in’ apne-mẽ ‘self-in’
Possessive ---- apnā ‘self.gen’ = ‘self’s’
3.2.1 Indo-Aryan
In Indo-Aryan, the form of the anaphor is unmarked for person, gender and
number. We provide examples from Hindi-Urdu (IA).
Hindi-Urdu (IA)
The complex form of the anaphor is apne-āp ‘self’s-self + postposition.’
It consists of the oblique form apne followed by the form āp ‘self.’ Further,
the whole complex form of the anaphor apne-āp invariably needs to be followed
by a postposition, and it “has no nominative or ergative -ne form” (Davison
2000: 401). The bare form (devoid of a postposition) of the anaphor apne-āp
‘self’s-self’ stands only for the emphatic form. The emphatic has no simplex
form.
The possessive reflexive form is apnā ‘self’s.’ When followed by a post-
position, apnā has the form apne, and it is referred to as the oblique form in
Hindi-Urdu grammars.
There is a simplex form of the anaphor too, and it is apne + postposition.
Table 3.1 illustrates the complex and simplex forms in different cases.
There are other forms such as khud ‘self’ and swayam ‘self’ followed by
a postposition. The simplex anaphors swayam + PP / khud + PP ‘self + PP’
may also occur in place of apne-āp + PP ‘self + PP,’ where PP stands for a
postposition.
3.2.2 Dravidian
We provide examples from Telugu (DR). The form of the anaphor (reflexive) is
person- and number-dependent. In 1st and 2nd person, the anaphor is formed by
the reduplication of personal pronouns, and in 3rd person by the reduplication
of tanu ‘self’ in singular and tama ‘selves’ in plural. In both the singular and
plural forms, the first part of the anaphor is followed by a postposition, which
in turn is followed by a reduplicated form of the first part (Table 3.2). Thus,
3.2 Form of the anaphor in SALs 47
Table 3.2 Complex forms of the anaphor in DO position in

1st/2nd/3rd persons in Dravidian
Person Singular Plural
1 Acc na-nnu nēnu incl: manala-ni manam

‘I.obl-acc I.nom’ ‘we.obl(incl)-acc we.nom(incl)’
excl: mammali-ni(incl) mēmu(excl)
‘we.obl-acc we.nom’
2 Acc ni-nnu nuvvu mimmali-ni mı̄ru
‘you.obl-acc you.nom’ ‘you.obl-acc you.nom’
3 Acc (masculine/feminine/ tana-ni tanu tama-ni tāmu
[+polite]) ‘self.obl-acc self.nom’ ‘self.obl-acc self.nom’
3 Acc (feminine, intimate/ dāni-ni adi vāt.i (obl)-ni avi
[−animate/−polite]) ‘she/it.obl-acc she/it.nom’ ‘they.obl-acc they.nom’
Table 3.3 Simplex forms of the anaphor in

3rd person in Dravidian
Case Singular Plural
Accusative tana-ni ‘self.obl-acc’ tama-ni ‘self.obl-acc’

Genitive/oblique tana ‘self’s’ tama ‘selves”
(For further details see Krishnamurti and Gwynn 1985.)
the form of the anaphor in the 1st person singular, for example, is na-nnu
nēnu ‘I.obl-acc I.nom,’ ni-nnu nuvvu ‘you.obl-acc you.nom’ in the 2nd person,
whereas it is tana-ni tanu ‘self.obl-acc self.nom’ in the 3rd person. Note that
tana is the oblique form of tanu when followed by a postposition. The second
part of the complex form of the reflexive in Table 3.2 is in the nominative case,
while the first part takes the case of the position in which the reflexive occurs
in a sentence. In Table 3.2, it is taken as the accusative.
The simplex forms of the reflexive in 3rd person are given in Table 3.3. In
Table 3.3 too, it is taken as the accusative.
3.2.3 Tibeto-Burman
We provide examples from Mizo (TB).
In Mizo, the form of the anaphor is person- and number-dependent. For
example, in 1st person singular it is kei-maʔ leʔ kei-maʔ ‘1s-emph and 1s-
emph.’ Table 3.4 gives 3rd person [+human] reflexive pronouns:
Table 3.4 Reflexive pronouns [+human] in the

3rd person in Tibeto-Burman
Case Singular Plural
Accusative a-maʔ leʔ a-maʔ an-maʔ leʔ an-maʔ

3s-emph and 3s-emph 3p-emph and 3p-emph
‘himself/herself’ ‘themselves’
Possessive a-maʔ an-maʔ
3s-emph 3p-emph
‘self’s’ ‘selves”
Abstracted from Lalitha Murthy and Subbarao (2000).
3.2.4 Austro-Asiatic
Munda family
For the Munda family of languages the following statements can be formulated:
(i) Mundari, Ho and Santali have the infix -n- as a verbal reflexive, and -pa-
as a verbal reciprocal. Sora has a verbal reciprocal -əl- (Donegan and
Stampe 2004). It does not have a nominal anaphor.
(ii) In Ho and Santali the VR -n- also functions as an inchoative/anticausative
suffix, as well as a passive marker. Ho also has a nominal anaphor borrowed
from Hindi.
(iii) The verbal reciprocal -pain Ho, Mundari and Santali is infixed in the
verb after the first syllable of the underived form.
(iv) Kharia has a verbal reflexive d.om and a verbal reciprocal kol. It also
has a nominal anaphor borrowed from the Indo-Aryan language Sadani
(Peterson, p.c.).
(v) Juang has no verbal reflexive, only a verbal reciprocal. It has two nominal
reflexives, āpein and āpan.ā; āpan.ā is borrowed from Oriya (IA) (Patnaik
and Subbarao 2000: 842).
Khasi (Mon-Khmer)
Khasi has both simplex and complex forms of the nominal reflexive. The
simplex form for the object position is ya-lade ‘acc/dat-self,’ and the complex
form is da-lade ya-lade ‘instr-self acc-self.’ The reflexive is unmarked for
person, gender and number, just as in Hindi-Urdu. The nominal reciprocal is
iwεi-ya-iwεi ‘one-acc-one’ which means ‘each other.’ Khasi also has a VR -hi
that occurs to the right of the verb stem, and a verbal reciprocal -ya- that occurs
to the left of the verb stem. The occurrence of the verbal reflexive and verbal
reciprocal3 is optional, while the occurrence of the nominal form is obligatory.
As we shall see, this is in contrast to Telugu (DR), Mizo, Thadou, Paite, Zou
3.3 Occurrence of the anaphor 49
and Hmar (TB), etc., where the nominal anaphor is optional and the verbal
anaphor is obligatory (see sentence (11)).
In this section we have presented a brief description of the forms of the
anaphor and in the following section we discuss their occurrence in different
positions in a sentence.
3.3 The occurrence of the anaphor in different language families

We shall now discuss the occurrence of these forms in each language family
taking one or two languages into consideration.
3.3.1 Indo-Aryan
Marathi (Wali 2000: 520) and Gujarati (Mistry 2000: 345–346) have a nominal
anaphor and a verbal anaphor. Sinhala has a verbal form gannəwa ‘take’ as a VR
which “may occur in reflexive contexts to provide or reinforce the reflexivity
and with some verbs, it will enforce a reflexive interpretation on a pronominal
argument of a verb” (Gair and Karunatillake 2000: 725–727). In these three
languages, the functions that the VR performs, the occurrence of the VR in
different syntactic domains and the overt presence of the VR in terms of
its obligatoriness in occurrence are far less in comparison to the other three
language families of the subcontinent. Further, there is no verbal reciprocal in
Gujarati and Marathi. In Sinhala, the verb + gannəwa, the verbal reciprocal,
alone “would suffice to enforce limitations of coindexing to the local domain”
(2000: 748). The authors did not elaborate this statement further.
In Gujarati, the reflexive potā ‘self’ is objective case-marked in (1) and is
coindexed with the subject.
Gujarati (IA)
(1) rāji potāi - ne vagovše
Raj self objective will humiliate
‘Raj will hurt himself.’
(Mistry 2000: 351)
In sentence (2), from Hindi-Urdu (IA), the anaphor may be simplex or complex.
The pronoun use ‘her’ cannot be coindexed with the subject and it has its own
discourse referent.
Hindi-Urdu (IA)
(2) rādhāi ne apne (āp)i,∗ j - ko/ use∗ i,j šı̄še mẽ dekhā
Radha erg self’s selfi,∗ j - dat her∗ i,j mirror in saw
‘Radhai saw / looked at herselfi,∗ j/ /her∗ i,j in the mirror.’
(Davison 2000: 408)
A null form is not possible in coreferential reading.

In Kashmiri (IA) the anaphor may be simplex or complex.
Kashmiri (IA)
(3) mohn chu pān- as/ pannis pān- as parnāvān
Mohan is self dat self- dat teaching
‘Mohan is teaching himself.’
The complex anaphor panun pān ‘self-abs’ in direct object position occurs in
the absolute case in (4), in contrast to Hindi-Urdu and many other SALs.
(4) səlı̄m- an vuch *pān / panun pān ānas manz

Salim erg saw self self-absolutive mirror in
‘Salim saw himself in the mirror.’
(Wali, Koul, Hook and Koul 2000: 483)
3.3.2 Dravidian
Malayalam, Toda and Telugu
To the best of our knowledge, Malayalam and Toda4 are the only Dravidian
languages which have no VR/VREC, and have only the nominal anaphoric
device. Malayalam has a verbal affix -kol..luka. According to Asher and Kumari
(1997: 165): “(V)erb forms carrying kol..luka in Malayalam carry such shades of
meaning as willingness, passive acceptance, permission and the fact that they
often combine with a reflexive pronoun would support the view that reflexivity
is not a dominant feature of the verb form itself.” Hence, -kol..luka cannot be
considered as a verbal anaphor.
Malayalam has two forms of the nominal anaphor – for example, in direct
object position: tann-e tann-e ‘self-acc emph’ or awan-e tann-e ‘he.acc emph’ –
according to Jayaseelan (2000: 122). Jayaseelan (2000: 116) considers the sec-
ond part of the anaphor tann-e as ‘emphatic,’ and not an instance of reduplica-
tion, as it is in Telugu and some other Dravidian languages.
Malayalam (DR)
(5) rāmani tanni/∗ j - e tanne / awan-i/∗ j- e tanne wer.uttu
Raman self- acc emph he- acc emph hated
‘Ramani hated himselfi/∗ j .’
(Jayaseelan 2000: 141–142) (Sentences (112a) and (117a) of
Jayaseelan 2000 are combined as one sentence in (5) above.)
Jayaseelan (2000: 122) observes: “a bare personal pronoun like awan or a bare
reflexive pronoun like tān will not do [in (5) in DO position]. A null form is
also not possible in coreferential reading.”
Telugu (DR)
The nominal anaphor in Telugu is a complex expression formed by a process of
reduplication of the simplex form tan ‘self’ and Case Copying (Subbarao and
Lalitha Murthy 2000). The VR occurs to the right of the main verb. In Telugu
and Kannada (Amritavalli 2000: 53) the occurrence of the VR is obligatory
(barring a few exceptions) when the nominal anaphor occurs in a subcategorized
position.
Telugu (DR)
(6) rādha (tana- ni (tanu)) pogud.u-5 kon- di
Radha self- Acc self (nom) praise- VR- agr
‘Radha praised herself.’
(Subbarao and Lalitha Murthy 2000: 233)
The nominative marker in Telugu is null. The anaphor consists of two parts:
(i) tana-ni, and (ii) tanu. The first part of the anaphor tana-ni carries the
structural accusative case marker -ni. The second part of the anaphor is in the
nominative case, and we have argued elsewhere (Subbarao and Saxena 1987;
Subbarao and Lalitha Murthy 2000) that the nominative case of the subject is
copied onto the second part of the bipartite structure of the anaphor. We have
labeled this as Case Copying. Evidence in support of our claim comes from the
dative subject construction and polymorphemic reciprocals in Telugu (see 3.14).
The following options with regard to the occurrence/non-occurrence of the
nominal anaphor in the nominative subject construction are available in Telugu:
(i) the second part of the nominal anaphor can be dropped leaving the first part
intact, but the first part cannot be dropped leaving the second part alone;
and
(ii) a null form (pro) can occur in place of the nominal anaphor,6 unlike in
Malayalam (Jayaseelan 2000: 122).
Note that the occurrence of the VR is obligatory when the nominal anaphor is
in subcategorized position.
Thus, the following patterns obtain in Telugu:
(i) subject – reduplicated anaphor – Verb – VR – agr
(ii) subject – the first part of the anaphor alone – Verb – VR – agr
(iii) subject – null form – Verb – VR – agr (the reflexive interpretation is intact
without the nominal anaphor).
There are a few predicates which do not permit the VR in Telugu. These
predicates include marcipōvu ‘forget’ (see (7)), gurtu pat..tu ‘recognize,’ cirāku
pad.u ‘be irritated,’ madhana pad.u ‘worry.’
(7) emnı̄šia valla hari tana ni tanu / *tana- ni marci pōyēd.u

amnesia due to Hari self acc self self acc forgot
‘Hari forgot himself due to amnesia.’
Note that the simplex form of the anaphor is not permitted in (7) as there is no
verbal reflexive. This, in our opinion, demonstrates that reduplication plays a
crucial role in anaphoric interpretation in the absence of the VR.
Interestingly, it is the same set of predicates that do not permit a verbal
reciprocal either, and a nominal bipartite reciprocal is required obligatorily, as
there is no verbal reciprocal marker.
Tamil (DR)
In Tamil, just as in Telugu, there is a nominal as well as a verbal anaphor. The
verbal reflexive is kid.u or kol. in written Tamil according to Annamalai (2000).
According to him, the occurrence of the VR is optional. In (8), the verb may
contain kit.u, the VR (in italics), as in pāttukit..tān ‘look at.pst-VR-agr,’ or it may
not, as in pātt-an ‘look at.pst-agr.’ The glossing of kit..t as VR (verbal reflexive)
is ours, not Annamalai’s. He glosses it as aux (auxiliary). Lehmann too (1989:
361) labels it as ‘the auxiliary verb.’
Tamil (DR)
(8) kumāri tannei kan.n.ād.ile pāttu-kit..t-ān/ pātt-ān
Kumar self.acc mirror.in look at.pst-VR-agr look at.pst-agr
‘Kumari looked at himselfi in the mirror.’
(Annamalai 2000: 180)
Just as with the Telugu verb marcipōvu ‘forget,’ the occurrence of the VR is
not permitted with verbs such as ‘remember’ in Tamil.7
(9) kumāri tannei marand-ān

Kumar self- acc forget.pst-agr
‘Kumari forgot himselfi .’
3.3.3 Tibeto-Burman
Mizo
In Mizo, the nominal anaphor is a complex form a-maʔ leʔ a-maʔ 8 ‘himself/
herself and himself/herself.’ It is formed by conjoining two nominal forms of
a-maʔ ‘himself/herself’ with the conjunction leʔ. The VR -in- occurs to the left
of the verb stem. Just as in Telugu, the occurrence of the VR is obligatory when
an anaphor occurs in a subcategorized position (Lalitha Murthy and Subbarao
2000).
When the anaphor occurs in direct object position, the VR detransitivizes the
predicate as a result of which ‘the ergative marker [which occurs obligatorily
with transitive verbs] does not occur with the subject’ (Lalitha Murthy and
Subbarao 2000: 785). That is, the subject is not in ergative case even in a
transitive sentence, when the DO is an anaphor.
a n a p h o r i n direct object p o s i t i o n
Mizo (TB)
(10) zovai (a- maʔ leʔ a- maʔ)i a- in- huā
Zova he- self and he- self 3s- VR- hate
‘Zovai hates himselfi .’
(Lalitha Murthy and Subbarao 2000: 790)
Note that in (10) the nominal reflexive a-maʔ leʔ a-maʔ that occurs in the
direct object position is not morphologically accusative case-marked. However,
it is the occurrence of the reduplicated form that does not necessitate overt
morphological case marking. It is structurally case-marked as will be evident
from (11) in which the monomorphemic form requires overt morphological
accusative case-marking. Such accusative case marking, either overtly as in
(11) or structurally as in (10), creates a paradoxical situation in which, on one
hand, the ergative marker does not occur with the subject as the predicate with
the verbal reflexive is [–transitive], and, on the other hand, the [–transitive]
predicate needs to assign accusative case to the nominal anaphor, a direct
object. Such an anomalous situation poses a problem for which any theory will
find it difficult to provide an explanation.9
The first part of the nominal anaphor a-maʔ ‘himself/herself’ as a simplex
form alone is permitted, and the occurrence of the VR is obligatory in (11) just
as in (10). The occurrence of the accusative marker cu with a-maʔ ‘himself/
herself’ is obligatory in such cases.
(11) lālii (a-maʔi - cu) a- in- that
Lali she-self- acc 3s- VR- kill
‘Lalii killed herselfi .’
(Lalitha Murthy and Subbarao
2000: 790)
The entire nominal anaphor in (10) and (11) may be pro-dropped.

Thus, the following patterns obtain in Mizo (Lalitha Murthy and Subbarao
2000: 793).
(i) subject – reduplicated anaphor – agr – VR – Verb
(ii) subject – the simplex form +acc – agr – VR – Verb
(iii) subject – null form – agr – VR – Verb.
In the indirect object position in Mizo:
(i) the simplex form of the anaphor and the VR occur,
(ii) the anaphor is dative case-marked, and
(iii) the occurrence of the ergative marker with the subject is obligatory.
a n a p h o r i n indirect object p o s i t i o n : e r g a t i v e obligatory

(12) zovai - n a-maʔi - hnēnaʔ loman a- ini - pe
Zova- erg he-self- dat prize 3s- VR- give
‘Zova gave a prize to himself.’
Note that, in spite of the occurrence of the VR with the predicate, the ergative
marker occurs with the subject. This shows that it is the occurrence of the
anaphor in direct object position alone that is instrumental in detransitivizing
the predicate, and not in non-direct object positions.
Manipuri
We shall now discuss the case of lexical anaphors in Manipuri. The occurrence
of the VR clitic je/ce in Manipuri is optional, as indicated by parentheses in
(13). The nominal anaphor in Manipuri is a complex reduplicated form having
a bipartite structure. Just as in Telugu, there is Case Copying of the Case of
the subject onto the anaphor. However, in Manipuri, it is the first part of the
reduplicated structure that carries the Case Copy of the subject. In contrast, it is
the second constituent of the reduplicated structure that carries the Case Copy
of the subject in Telugu.
Manipuri (TB)
(13) aii - na i-sai - na i-sai - bu thagat- (cei )- i
I- nom I-self- nom I-self- acc praise- VR- [−fut]
‘I praised myself.’
(Sarju Devi and Subbarao 2003: 58)
Note that the nominative case marker -na occurs with the subject. The accusative
case marker is -bu. The anaphor is a reduplicated form as in the Dravidian
languages. The first part consists of the anaphor for self in 1st person, namely
i-sa ‘I-self,’ and is followed by the nominative case marker -na. It is the
nominative case marker of the subject that is copied onto the first part of the
reduplicated structure of the anaphor. Note that i-sa in (13) is not an emphatic
form, as the emphatic form of the 1st person is isa-mak. Thus, the phenomenon
of reduplication and Case Copying are both operative in Manipuri too.
The second part of the anaphor carries the accusative case marker -bu in direct
object position. The bipartite structure of the lexical anaphor is schematically
given in (14a) and (14b) as Part I and Part II.
(14a) Part I: Subject pronominal clitic + self + nominative

(14b) Part II: Subject pronominal clitic + self + accusative
In contrast to Telugu and Kannada (DR)

(i) the occurrence of the verbal reflexive in Manipuri is optional, and
(ii) the anaphor with the case copy of the subject occurs as the first part of the
complex structure in Manipuri, while it occurs as the second part in many
Dravidian languages.
The verbal reciprocal marker in Manipuri is -na and its occurrence is obligatory.
It performs the function of a group marker together with the incorporated adverb
min ‘together.’ It imparts the interpretation of doing an act together as a group.
Neither of them can be dropped in (15).
(15) tomba- ga tombi- ga skūl- da cat- min- na- khre

Tomba- conj Tombi- conj school- to go- gp mkr- VREC- pst
‘Tomba and Tombi (as a group) have gone to school.’
(Sarju Devi Thokcham, Alice Konthoujam, Imoba Singh and
H. Surmangol Sharma, p.c.)
Tenyidie, Ao and Rabha

In the TB family, in Tenyidie and Ao of the Naga group and in Rabha, only a
verbal reciprocal is available without there being a verbal reflexive. The VREC
ke occurs to the right of the verb in Tenyidie.
Tenyidie (TB)
(16) uniei - e lešə thu kei - tsə- ya
they (dual)- nom letters write VREC- give- pres
‘They (dual)i write letters to each otheri .’
(Kevichüsa 1996)
The occurrence of the verbal reciprocal and nominal reciprocal are mutually
exclusive in Tenyidie. When the verbal reciprocal occurs, the nominal reciprocal
cannot occur, as the verb is detransitivized in such cases, according to Kevichüsa
(2007).
The verbal reciprocal in Ao is təp and it also indicates collective activity
with the interpretation of together (Coupe 2003: 336).
Khasi
Khasi has a nominal simplex form ya-lade ‘acc/dat-self’ or a complex form
da-lade ya-lade ‘instr-self acc-self.’ It has a monomorphemic VR hi-. It occurs
to the right of the verb and is optional; da-lade ‘instr-self’ too is optional (17)
(Temsen and Subbarao, ms).
Khasi (Mon-Khmer)
(17) u- lemi u- pεit- (hi) (da- lade)i ya- ladei ha- ka- yit
3 m,s- Lem 3 m,s- look- VR instr- self acc- self loc- f- mirror
‘Lemi looked at himselfi in the mirror.’
With psychological predicates too, the VR as well as the nominal anaphor

occurs. The VR hi- occurs to the right of the adjective.
(18) u- tɔm u- šrāy- (hi) ya- lade
3 m,s- Tom 3 m,s- angry- VR acc- self
‘Tom is angry with himself.’
It is interesting to observe how reciprocity is likewise encoded. The verbal

reciprocal is ya-, and it occurs as a prefix on the verb, unlike the VR which
occurs as a suffix. There are three specific patterns that obtain in Khasi.
(i) the nominal and verbal reciprocal can both occur;
(ii) only the nominal anaphor can occur;
(iii) only the verbal anaphor can occur.
Sentence (19) is an example of pattern (i).
(19) ki- kh nnaʔ kiya- doʔ i- wεi- ya- i- wεi
p- child p- VREC- kiss hon- one- acc- hon- one
‘The children are kissing one another.’
The verbal reciprocal performs a variety of functions in Khasi. Some of them

are:
(i) The marker ya- indicates that the activity performed is a group activity.
Hence, we label ya- as a group marker (gp mkr). This is similar to what
Lichtenberk (1985) labels as “collective/group marker.”10
(20) ki- kh nnaʔ kiya- bam ja
p- child p- gp mkr- eat rice
‘The children as a group are eating food.’
It is crucial to mention that the occurrence of the group marker ya- is

obligatory, while the verbal reciprocal may optionally be deleted.11
(ii) The marker -ya- also functions as a hortative marker. We feel that the
hortative too indicates a suggestion for a collective activity, and, hence, is
closely semantically related to the reciprocal. Note that the verbal recip-
rocal also occurs in (21). Thus, these two markers are in close semantic
consonance. Khasi is the only language in the subcontinent where the
verbal reciprocal and the hortative marker are homophonous.
(21) ya- ni- n- ya- lεyt (nɔʔ)
hortative- p- fut- VREC- go descend (vector verb)
‘Let’s go (right away).’
(iii) The verbal reciprocal marker -ya also indicates that the action is performed
with a collaborative effort by the subject. In (22), though the verb exhibits
agreement with the subject in the singular, the interpretation of the sen-
tence is in the plural. Such interpretation, we hypothesize, is contributed
by the collaborative effort marker (cem) ya in (22).
(22) ui - ban ui - ya- knjat ya- ka- bɔrti

m,s- Ban m,s- cem- kick acc- f- bucket
‘Ban and the others are kicking the bucket.’
(collaborative effort to achieve a common goal)
Such markers are found in Kharia (Munda) too.
Munda languages
In most Munda languages, it is only the verbal device that is available as an
anaphor, indigenously. Due to the influence of the superstrate IA languages,
such as Hindi, Bangla and Oriya, however, one does find a nominal anaphor
too. The occurrence of the verbal anaphor is obligatory. It is significant that the
transitive marker that occurs with a transitive verb does not occur when the
verb occurs with a verbal anaphor. It is the intransitive marker that occurs with
the transitive verb that functions as the verbal reflexive.
Thus, in Ho, the intransitive marker, the verbal reflexive and the passive
marker are identical in form, and it is /-n-/. The marker /-d/ that occurs
with transitive verbs cannot occur when the verbal reflexive/reciprocal occurs.
Instead, the intransitive marker, which is homophonous with the verbal reflex-
ive, occurs with the verb. In Mundari (Munda) too, a similar situation obtains.
The non-occurrence of the nominal anaphor is indicated by ø.
Ho (Munda)
(23) proi ar.si- re- m ø nel- ke- ni - a
you mirror- in- 2s see- pst- VR[−tr]- fin
‘You saw yourself in the mirror.’
(Koh and Subbarao, ms)
In Mundari, the VR suffix is -en after consonants, and -n after vowels. It also
functions as an intransitive suffix.
In Ho, Mundari and Santali, the verbal reciprocal -pA- is infixed in the verb,
and it is placed after the first syllable of the underived form. The letter A in
the infix -pA- stands for any vowel, and it harmonizes with the stem vowel that
precedes it. The verb form for ‘hit’ in Santali is dāl. Since the verbal reciprocal
-pA- is infixed in the verb, and the verb is split into two parts, we’ve glossed
da- as ‘hit1’ and -l as ‘hit2.’ The absence of the nominal reciprocal is indicated
by ø.
Santali (Munda)
(24) unkini - kin ø dā- pāi - l- kā- n- a
they dual- dual hit1- VREC- hit2- pst- [−tr]- fin
‘They (two of them) are fighting [with] each other.’
(Minegishi and Murmu 2001: 104)
In Ho (Munda), “[M]any of these reciprocal forms are used as nouns, de-pe-ŋa

‘mutual help,’ e-pe-ra
‘a quarrel’” (Deeney 1979: 57). In Ho, deŋa without
the verbal reciprocal means ‘help (verb)’ and era
means ‘quarrel (verb).’
A verb root together with the same verb in its reciprocal form imparts generic
interpretation in Ho (Munda): “Word pairs are also sometimes formed by using
a (verb) root followed by the same root with a -p-12 (reciprocal) infix, or an -n-
(verbal noun) infix to give a more general or abstract use of the root” (Deeney
1979: 104). For example, deŋa-de-pe-ŋa ‘help-help1-reciprocal-help2’ means
“help in general (including mutual help)” (1979: 104).
Kharia (Munda) has an indigenous verbal reflexive d.om (25), and a nominal
anaphor apan ‘self,’ borrowed from Sadani (IA) (26) (Peterson 2006).
verbal anaphor
Kharia (Munda)
(25) proi yo- d.omi - ki- kiyari
they see- VR- middle pst- dual
‘They saw themselves (e.g., in the mirror).’
(Peterson 2006)
nominal anaphor
(26) proi aina- te apani - te yo- yoʔ- ji
I mirror- in self- acc see- active pst- 1s
‘I saw myself in the mirror.’
(Peterson 2006; the glosses have been
slightly modified)
In Kharia, d.om is also used as passive marker. Peterson 2006 (p.c.) observes:
“with the wholesale importation of the Indo-Aryan reflexive construction with
apan . . . the reflexive use of d.om has receded from this domain and is perhaps in
the process of being lost.” This impending loss is due to superstratum influence
of Hindi and Sadani (IA). Kharia also has a verbal reciprocal kol.
In the following section, we discuss the various functions that the verbal
reflexive/reciprocal performs in SALs.
3.4 Multi-functionality of the verbal reflexive/reciprocal

The verbal reflexive/reciprocal in SALs exhibits many of the properties that the
verbal form does, cross-linguistically. It is often polysemous, and functions as
3.4 Verbal reflexive/reciprocal 59
a detransitivizer (anticausative), as a marker of passive, as a self-benefactive, as

an emphatic, and as an impersonal marker and the like (see Everaert, ms, for a
discussion of data and analysis from other languages). The basic feature that the
VR (in many languages), the passive morpheme and intransitive marker share
is that they all reduce the valence of the predicate by one argument. The VR is
also used in the formation of certain lexical items. All these functions parallel
the functions in languages across the world. Kazenin’s observation that “the
verbal reflexivity always is a morphologically ‘lighter’ device than anaphoric
[nominal] reflexivity” (Kazenin 2001: 918) holds good in SALs as the verbal
anaphor invariably is a simplex form while the nominal anaphor in most cases
is a polymorphemic form.
3.4.1 Direct and indirect reflexive

According to Kazenin (2001: 918), “The type of reflexive in which the Agent is
coreferential with the Patient is called direct. The reflexive which marks coref-
erentiality of the Agent with a participant other than the Patient (Beneficiary,
Recipient) is called indirect. The direct reflexive is a detransitivizing device,
because it conflates the Agent and Patient into a single core NP.”13
Evidence from SALs supports the implicational universal of Kazenin (2001:
918), namely, “if a language allows verbal marking of indirect reflexives, it
allows verbal marking of direct reflexives as well.” Thus, in all SALs, the
verbal anaphor invariably occurs with the subject as its antecedent, if it occurs
with an indirect or oblique object as its coreferential antecedent.
Further, reciprocity is a collective and mutually interactive phenomenon. It
is significant that the marker for the verbal reciprocal also functions as a marker
indicating ‘togetherness’ or collaborative effort, as a group marker, and as a
hortative marker in languages such as Ao, Manipuri (Tibeto-Burman), Khasi
(Mon-Khmer) and Kharia (Munda), all with different genetic affiliations.
We present now the various functions that the verbal reflexive/reciprocal
performs in SALs.
3.4.2 As a detransitivizer/anticausative
(i) Absence of the ergative marker: In Mizo (TB), in the direct reflexive the
ergative marker does not occur when the direct object is coindexed with
the subject, as the VR detransitivizes the verb (see sentences (10) and
(11)).
(ii) Verbal and nominal reciprocals in complementary distribution: In
Tenyidie (TB) the occurrence of the verbal reciprocal detransitivizes the
verb, and, hence, the nominal reciprocal cannot occur when the verbal
reciprocal occurs (Kevichüsa 2007).
(iii) Homophonous form for the verbal reciprocal and anticausative: The VR
and the detransitivizer/anticausative in Dravidian, Mon-Khmer, Munda
and Tibeto-Burman languages are homophonous. In Telugu and Kannada
(DR), and in Mizo and Hmar (TB), the detransitivizer reduces the valence
of the corresponding transitive verb by one argument. This is similar to
the phenomenon of ‘valence-decrease’ in the verb observed in many other
languages, such as Russian and Polish (Haspelmath 1993; Kazenin 2001:
923; Piñón 2001). In Dyirbal (Dixon 1972: 89) too, “[A] reflexive form
functions like an intransitive stem.”
In Telugu (DR), teruc ‘open’ is a transitive verb and, hence, it takes two argu-
ments, whereas terucu-konu ‘open-VR,’ a [−tr] verb, takes only one argument.
Thus, the VR reduces the valence of the verb (see Krishnamurti and Gwynn
1985: 208 for details).
Telugu (DR)
(27) talupu terucu- kon- di
door open- inchoative-14 3s,nm
‘The door opened.’
(Subbarao and Lalitha Murthy
2000: 227)
In Kannada (DR), the intransitive verbs and their corresponding transitive verbs
in Table 3.5 show how the occurrence of the VR (in italics) ‘detransitivizes’
the verb.
Kannada (DR)
Table 3.5 VR as a detransitivizer
[–transitive] [+transitive]
(i) baccit.t.u-kol..l ‘to hide’ baccid. ‘to hide’

(ii) malagi-kol..l ‘to sleep’ malagis ‘to put to sleep’
(iii) mucchi-kol..l ‘to close’ mucch ‘to close’
Amritavalli (2000: 55).
In Mizo and Tangkhul (TB) and Khasi (Mon-Khmer) too, a similar phenomenon
is observed.
Mizo (TB)
(28) konkhāy a- in- hon
door 3s- inchoative- open
‘The door opened (on its own).’
(Lalitha Murthy and Subbarao
2000: 788)
Khasi (Mon-Khmer)
It is the occurrence of the verbal reflexive hi that detransitivizes the transitive

verb plē ‘open [+tr].’
(29) ka- jing- khang ka- plē hi
f- door f- open[+tr] VR
‘The door opened (on its own).’
(Temsen and Subbarao, ms)
Further evidence comes from Mundari, Santali and Ho (Munda) in which

the intransitive marker -n/-en occurs when an anaphor is coindexed with the
subject, and the occurrence of the transitive marker -d is prohibited in such
cases, though the verb may be dyadic (transitive) or tryadic (ditransitive). The
occurrence of the intransitive marker -n/-en with the transitive verb nel ‘see’ in
sentence (23) from Ho is illustrative.
Thus, Dravidian, Tibeto-Burman, Munda and Khasi (Mon-Khmer) languages
have an identical form for the reflexive marker and the anticausative marker, as
with the Russian suffix -sj(a).
Russian
(30) mat’ odela- s’
mother dressed- suffix
‘Mother dressed herself.’ (reflexive)
(31) dver’ otkryla- s’

door opened- suffix
‘The door opened.’ (anticausative)
(Kazenin 2001: 921)
We will argue that all dative predicates are [−tr] (see chapter 5 for full dis-
cussion). The fact that the detransitivizer kon occurs in the dative subject
construction with a transitive verb such as kōs ‘cut[+tr]’ in Telugu (DR) in
(32), turning it to [−tr], lends further support to our claim that kon functions
like a detransitivizer.15
Telugu (DR)
(32) [kapil- ki]i ivāl.a vēluj kōsu- kon- di∗ i,j
Kapil- dat today finger cut[+tr]- inchoative/[−tr]- 3 s,nm
Literally: ‘To Kapil (his) finger got cut today.’
‘Kapil’s finger got cut today (inadvertently).’
Further, the presence of kon in Telugu with a dative subject indicates a non-
volitional act (32).
The verbal anaphor cannot occur in the dative subject construction in
Kannada. Amritavalli (2000: 54) attributes this to the fact that the verbal
anaphor is [−transitive], while Gair and Karunatillake (2000: 726) attribute

such non-occurrence of the verbal anaphor gannəwa in Sinhala (Indo-Aryan)
to “a volitional or intentional implication, so that its occurrence with dative
subject is uncommon, given the usual involitive sense associated with such
subjects.” Since the predicate in the dative subject construction is [−volitional],
the verbal anaphor cannot occur in this construction in Dravidian languages
either.
3.4.3 As a self-benefactive
The marker for the verbal reflexive/reciprocal functions as a self-benefactive in
Telugu and Kannada (DR) and Mizo (Lalitha Murthy and Subbarao 2000: 788),
and Hmar (TB). It is labeled as the ātmanēpada marker in Sanskrit grammars,
and the effect of the action is invariably ‘agent oriented.’ Hence, it (the verbal
anaphor functioning as a self-benefactive) is not found in a passive sentence.
The occurrence of the nominal anaphor is optional.
Telugu (DR)
(33) mālati annam vand.u- kon- di
Malati food cook- self ben- 3 s,nm
‘Malati cooked food for herself.’
Kannada (DR)
ogi- ‘to wash (clothes)’ ogedukol..l- ‘to wash (clothes) for oneself’
(Amritavalli 2000: 55)16
3.4.4 Activities pertaining to body parts

In Telugu, Kannada (DR), Ho (Munda), Mizo and Hmar (TB), possessor argu-
ment suppression takes place when the activity concerns the self’s (agent’s)
body, just as it does in Classical Greek (37).
Kannada (DR)
(34) rāma øi,/∗ j (*tanage) kai tol.edu- kon.d.a
Rama ø (*self-dat) hand wash- inch-agr
‘Ramai washed (hisi,/∗ j ) hands.’
(Amritavalli 2000: 57)
Telugu (DR)
(35) pūjai moham kad.ugu- koni - di
Puja face wash- VR- 3 s,nm
‘Poojai washed (heri ) face.’
In Ho (Munda), the verbal reflexive, the intransitive marker and the self-
benefactive are homophonous (36). -iŋ is the 1st person singular subject agree-
ment marker on the verb.
Ho (Munda)
(36) proi em- en- tan- a- iŋi
I give- [-tr]/VR- progr- fin- 1 s
‘I am giving myself.’
(Burrows [1915] 1980: 74)
The possessor of the body part does not occur in Classical Greek either.
Classical Greek
(37) lou’o- mai t- `
‘ās khe
ir- as
wash- 1s.middle.pres article- acc.p hand- acc.p
‘I wash my hands.’
(Kazenin 2001: 918)
3.4.5 As a marker of passive

Gujarati is the only Indo-Aryan language in which the VR and the passive
marker are identical. The marker is -ā (see Mistry 2000: 345–346 for more
discussion and examples). The marker -n functions as the passive, intransitive
marker and the VR in Ho, Sora and Gorum (Munda).
Ho (Munda)
Direct Object as subject of the passive
In Ho, the suffix -n functions as a [–transitive] marker, a passive marker and a
verbal reflexive. Thus, sentence (38) is ambiguous between a VR reading and
a passive reading.
Ho
(38) proi siŋboŋa- taʔa- re- mi - ema- ke- n- a
you God- to- in- 2s- give- pst- VR/passive- [+fin]
‘You gave yourself to God.’ / ‘You were given to God.’
In Sora, Kharia and Gorum (Munda) too, the VR functions as a passive and
intransitive marker.
Sora (Munda)
(39) ran- n- ēten
crush- [−tr]- 3.pst
‘It was crushed.’
(Ramamurti 1931)
Kharia (Munda)
(40) proi yo- d.omi - ki- kiyari
they see- pass/VR- middle pst- dual
‘They were seen by someone else.’ / ‘They saw themselves (in the mirror).’
(Peterson 2006)
3.4.6 As a verbal emphatic marker

The emphatic marker in Khasi (Mon-Khmer) and Kharia (Munda) is the same
as the VR. Khasi and Kharia are the only languages amongst the SALs that we
know of where the verbal reflexive and emphatic are homophonous, and the
emphatic occurs as a verbal clitic. The emphatic marker hi occurs to the right
of the verb, just as the VR does.
Khasi (Mon-Khmer)
(41) proi ŋai - sōʔ- hi ya- ka-ne ka- sɔpti
1s- sew- emph acc- f.s.-this f.s- dress
‘I sew this dress myself.’
Kharia (Munda)
(42)
oʔ- jom- ta
eat- VR- pres.ind
‘He eats by himself.’
(Malhotra 1982: 178)17
3.4.7 Verbal reflexive with adjectives

The VR-affix occurs with stative adjectives (used as verbs),18 such as sāŋ ‘tall,’
hnūi ‘short,’ and with adjectives such as du ‘proud,’ tak ‘difficult,’ hoi ‘easy’
and their denominals in Hmar (TB). Thus, for example, the adjective for ‘tall’
is in-sāŋ ‘VR-tall’ in Hmar. The occurrence of the VR with adjectives does not
appear to contribute to any nuance of meaning.
Hmar (TB)
(43) hi naupaŋtepa hi kha naupaŋtepa kha nekin a- in- sāŋ
this small boy this that small boy that than 3s- VR- tall
‘This small boy is taller than that small boy.’
Table 3.6 Grammaticalization of the VR in Telugu (DR)
(i) pad.u ‘fall’ pad.u -konu ‘sleep’

(ii) jāru ‘slip’ [−tr] jāru-konu ‘escape’
(iii) anu ‘say’ anu-konu ‘think,’ ‘consider’
Subbarao and Lalitha Murthy (2000: 227).19
Table 3.7 Grammaticalization of the VR in Kannada (DR)
(i) nōd. ‘to see, to look’ nōd.i kol.l. ‘to look after’
(ii) enn ‘to say’ endu kol.l. ‘to think’
(iii) ett ‘to lift’ etti kol.l. ‘to carry’
Amritavalli (2000: 55).
Table 3.8 Grammaticalization of the VR in Mizo (TB)
(i) nei ‘have’ in-nei ‘to marry’

(ii) huwal ‘surround’ in-huwal ‘be engaged’
(iii) hria ‘know’ in-hria ‘believe’ (‘know oneself’)
(iv) ti ‘say’ in-ti ‘to think,’ ‘to consider’ (‘to say to oneself’)
Abstracted from Lalitha Murthy and Subbarao (2000: 789).
The expression for ‘height’ in Hmar is in-sāng-na ‘VR-tall-nozr,’ and for

‘shortness’ is in-hnūi-na ‘VR-tall-nozr,’ where -na is a nominalizer. The VR
does not occur with ɔl, another adjective for ‘easy.’ Whether a VR is permitted
to occur with an adjective or not seems to be lexically determined.
3.4.8 Special lexical items with the verbal reflexive/verbal reciprocal

The VR/VREC is used in Dravidian and Tibeto-Burman languages in the
formation of special lexical items due to a process of grammaticalization of the
VR (Table 3.6).20
In Kannada (DR), there are also pairs of verbs “with a fairly idiosyncratic
(lexicalized) semantic relation between V and V-kol..lu” (Amritavalli 2000: 55);
kol..lu is the VR in Kannada (Table 3.7).
In Mizo (TB), too, a similar phenomenon is found (see Table 3.8).
3.4.9 In taboo expressions

As it is not considered a good omen to say that something (especially, foodstuff)
is finished in the house in Telugu (DR); the intransitive verb nind.u ‘to fill’ is
used with the VR kon (45), although the VR normally does not occur with an
intransitive verb as (44) illustrates.
Telugu (DR)
(44) binde lō nı̄l.l.u nind.u- tunnāyi
pot in water fill[−tr]- pres progr
‘Water is filling in the pot.’
The occurrence of the VR in (45) is obligatory. Such an occurrence/usage of

the VR in (45) appears to be due to the taboo mentioned above.
(45) int.(i)- lō biyyam nind.u konn- āyi

home.obl- in rice fill[−tr] VR- pst
‘The house is filled with rice.’
(Intended meaning: ‘There is NO rice
in the house.’)
3.4.10 Verbal reflexive with verbs

In Hmar (TB), some transitive verbs, such as in-cuk ‘learn,’ in-cɔ̄ k ‘buy,’ in-suo
‘show,’ in-fuʔ ‘encourage,’ and some intransitive verbs, such as in-vāk ‘wander,’
invariably carry the verbal reflexive in-. These verbs do not seem to form any
natural class.
In Telugu (DR), too, verbs such as gnāpakam/gurtu-pet..tu-konu ‘memory-
keep-VR’ – that is, ‘to remember’ – and gnāpakam/gurtu uncu-konu ‘memory
keep-VR’ – that is, ‘to remember’ – obligatorily require the verbal reflexive.
3.4.11 Verbal reflexive as a locality instigator

The presence of the VR blocks long-distance binding in Dravidian, Munda
and Tibeto-Burman (see Table 3.7).21 An anaphoric form which is otherwise
‘antilocal’ becomes ‘local’ when the VR occurs. Dravidian tān/tan ‘self’ and
Marathi (IA) āp-lyā ‘self’ are antilocal and, hence, a local antecedent cannot
bind these. The anaphor tan ‘self’ in (46) in Telugu (DR) and āp-lyā ‘self’
in (47) in Marathi (IA) cannot be coindexed with a subject antecedent in the
absence of the verbal reflexive.
Telugu (DR)
(46) *pragatii tanai - ni tit.t.- indi
Pragati self- acc scold- pst- 3 s,nm
‘Pragatii scolded her*i .’
Marathi (IA)
(47) lili∗ i āplyā- lā∗ i/∗ j haste
Lili self-dat laughs
*‘Lili laughs at self∗ i,∗ j .’
(Wali 2000: 538)
However, sentence (48) shows that the occurrence of the VR makes the nominal
anaphor in (46) local.
Telugu (DR)
(48) pragatii tana- nii,∗ j tit.t.u- koni - di
Pragati self- acc scold- VR- 3 s,nm
‘Pragatii scolded herselfi,∗ j .’
In Marathi (IA), the auxiliary verb ghen.e ‘to take’ functions as a self-benefactive
and, as a VR, it creates “a reflexive context” (Wali 2000: 521): “When the verb
is compounded with ghen.e, however, the use of āpan. [which is otherwise anti-
local] becomes acceptable in the otherwise forbidden local domain” (2000:
521).
Marathi (IA)
(49) lili ni āplyā- lā mārun ghet- la
Lili erg self-dat beat VR- pst
‘Lili beat herself.’
(Wali 2000: 521)
In Sinhala (IA) too, the verb gannəwa ‘take’ occurs with a verb “in reflexive
contexts to provide or reinforce the reflexivity” (Gair and Karunatillake 2000:
725). It makes the anti-local pronoun local. The verb dı̄gatta in (50) is the
compounding of the perfect participial stem with the VR gannəwa ‘take’ (Gair
and Karunatillake 2000: 725).
Sinhala (IA)
(50) gunəpaləi eyāt.əi/∗ j t ǣgi dı̄gatta
Gunapala she/he.dat presents give.pple.+take.pst
‘Gunapalai gave himselfi/∗ j presents.’
(Gair and Karunatillake 2000: 725)
In Hmar and Mizo (TB) too, the bare reflexive amaʔ ‘self’ “cannot take an
antecedent in the local domain – because the VR is not present. It cannot take
a discourse antecedent, too” (Lalitha Murthy and Subbarao 2000: 793). Thus,
(10) and (11) above are ungrammatical if the VR is not present, and only a bare
nominal reflexive occurs.
The data above demonstrate that it is the VR that determines the local domain
for an anaphor, and its presence makes an antilocal anaphor function like a local
anaphor.
Notice that languages belonging to genetically different language families
have a similar lexical item, a verbal anaphor that performs almost identical
functions. Based on this fact, Subbarao (2000) points out that it is not for-
tuitous that such a phenomenon is found in human language. One can hope
that this phenomenon plausibly reflects the mental organization of language
that reflects cognitive capabilities of the human mind. Though functions such
as inchoative, verbal reflexive/reciprocal or self-benefactive or passive appear
outwardly to be unrelated, they all have the feature of transitivity in common –
that is, they are all [–transitive]. It is, therefore, reasonable to hypothesize that
such grammatical facts are grouped together under a single lexical head in
the mental lexicon. It is probably the cognitive capability of the human mind
that enables it mentally to classify phenomena that share cognitive/semantic
commonalities and to assign a single lexical category for such seemingly
unrelated phenomena across languages transcending genetic boundaries and
barriers.22
In this section, we have discussed the various functions that the verbal
anaphor performs in SALs. In the following section we shall discuss the occur-
rence of anaphors in non-nominative subject constructions.
3.5 Non-nominative subjects as antecedents to a lexical anaphor

3.5.1 The dative/genitive subject as antecedent
The dative/genitive subject functions as a c-commanding subject antecedent to
(i) a possessive anaphor, and (ii) a nominal anaphor, in all SALs except Marathi
(Wali 2000: 529).
Antecedent to a possessive anaphor

Hindi-Urdu (IA)
(51) baccõ- koi apnı̄i/∗ j billı̄ dikhāı̄ dı̄j (hai)j
children m,p- dat self’s cat f, s (nom) sight give.perf.f,s be.pres.3s
‘The childreni (have) caught sight/seen self’si/∗ j cat.’
(Davison 2004: 145)
3.5 Non-nominative subjects as antecedents 69
Kashmiri (IA)
(52) kōrii chu panuni bōy pasand
girl.dat is self’s brother like
‘The girli likes self’si brother.’
(Wali et al. 2000: 486)
Maithili (IA)
(53) hari- kẽi apani ghar nahi ch- ainh
Hari dat self’s house not be pres.3hon.non-nominative
‘Hari doesn’t have his own house.’
(Yadava 2004: 257)
Telugu (DR)
(54) rōjā- kii tanai amma ant.ē is.t.am
Roja- dat self’s mother quot is pleasing
‘Rojai likes heri mother.’
Antecedent to a nominal complex/simplex anaphor

The dative subject functions as a c-commanding antecedent to a nominal
anaphor in SALs. The nominal anaphor may be complex or simplex.
Hindi-Urdu (IA)
(55) saritā ko apne (āp) par bharosā thā
Sarita f,s dat self’s self upon confidence m,s was m,s
‘Sarita had confidence in herself.’
According to Wali (2000: 529), in Marathi (IA) “[D]ative subjects cannot

antecede unmarked/bare reflexives with or without the putative VR ghen.e [as
in (56a)]. The reflexive must be a possessive as in (42c) [(56b) below] or a
postpositional NP [Example not provided by the author].”
Marathi (IA)
(56a) lili- lai (āpan.) swatāh∗ i āvad.te/ āvd.un- ghete
Lili- dat self [a bare reflexive] likes/ likes- VR
‘Lilii likes self∗ i .’
(56b) lili- lai (āp-lyā) swatāh-cii bahi āvad.te

Lili- dat self’s sister likes
‘Lilii likes self’si sister.’
(Wali 2000: 529)
However, according to Lalita U. Dhareshwar, Trupti Nissar and many other

Marathi speakers (p.c.) consulted, a nominal anaphor with a structural Case
marker can have a dative subject as its antecedent, contrary to Kashi Wali’s
claim, as in (56c).
(56c) lili- lāi 23 svətāhāi - vər/ āpl-yā- vər ?khatri/višvās nāh

Lili- dat self- on self-obl- on confidence m, s not
‘Lili does not have confidence in herself.’
(Lalita U. Dhareshwar and Trupti Nissar, p.c.)
The occurrence of a reduplicated anaphor is obligatory in Telugu (DR), when

a dative subject is an antecedent, as the VR does not occur in such cases.
Telugu (DR)
(57) kamalai ki [tana- mı̄da tana- ki]i kāpam waccindi
Kamala dat self- on self- dat anger came
‘Kamala got angry [with] herself.’
(Subbarao and Saxena 1987: 126)
The theme position of a dative/genitive subject construction in SALs is nom-

inative case-marked in most of the languages, except in Tamil, Malayalam
(DR), Bangla (IA) and Bodo (TB), where it may be either nominative or
accusative case-marked. We will argue in chapter 5 that the accusative marker
in dative/genitive subject constructions is a specificity marker. The occurrence
of a complex anaphor that is either nominative case-marked or accusative case-
marked is not permitted in Hindi-Urdu, Punjabi and Kashmiri, and none of the
studies carried out so far could provide an explanation.
Hence, in Hindi-Urdu, the predicates yād ānā ‘to remember’ and acchā
lagnā ‘to be pleasing,’ for example, that require a dative subject, do not permit
a reflexive or reciprocal that is either nominative or accusative case-marked.
Hindi-Urdu (IA)
(58) *rām- koi sirf apne āpi acchā lag-tā hai
Ram- dat only self’s self.nom good strike-imperf is
‘Ram likes only himself.’
(Yamabe 1990; Experiencer as subject [as quoted in Davison
2004: 155, sentence (38a)])
(59) rāmi sirf apne (āp)- ko acchā lag-tā hai

Ram.nom only self’s self dat good strike-imperf is
‘Ram likes only himself.’
(Yamabe 1990; Theme as subject [as quoted in Davison
2004: 155, sentence (38b)])
In Kashmiri (IA) a nominative case-marked anaphor in the dative subject con-

struction is not permitted (Wali et al. 2000: 486).
Kashmiri (IA)
(60) kōrii chu ləd.k/ (*pāni / *panun pāni ) pasand
girl.dat is boy.abs self.nom self’s self.nom like
‘The girl likes the boy/*herself.’
(Wali et al. 2000: 486)
However, certain verbs, such as bōzun ‘to find/see’ or khɔš karun ‘to like, make
happy’ permit a nominative case-marked complex anaphor (Wali et al. 2000:
486–487).
(61) mei chu n panun pāni bozn yivān
I.dat be not self’s self.nom find come.pres pple
‘Ii am unable to find myselfi .’
(Wali et al. 2000: 487)
In Telugu (DR), a nominative case-marked simplex form of the anaphor tanu

‘self’ is not permitted, if it is coindexed with a dative case-marked subject. It
can only have a discourse antecedent. A VR is not permitted in (62), as the
subject is dative case-marked.
Telugu (DR)
(62) kāvyai ki tanu ant.e∗ i,j cālā is.t.am
Kavya dat self.nom quot.cond a lot pleasing
‘Kavyai likes *herselfi /herj a lot.’
In contrast to most of the SALs, Bangla (IA), Malayalam, Tamil (DR) and Bodo
(TB) are the only four languages that we know of that permit an accusative
case-marked anaphor in a genitive/dative subject construction.
Bangla (IA)
(63) gopal- er nije ke bhalo lage
Gopal- gen self acc good feels
‘Gopal likes himself.’
The question that arises now is: is there any principled reason for such a
restriction on the occurrence of nominative case-marked or accusative case-
marked anaphors in Hindi-Urdu, Punjabi (IA) and Telugu (DR)? Let us first
discuss the case of the accusative case marking. Dative/genitive predicates are
[–transitive]. Hence, the predicate in such constructions cannot assign/check
the accusative case of the theme.
The fact that Tamil, Malayalam (DR) and Bangla (IA) permit an accusative
marker with the theme in the dative/genitive subject construction is not a
counterargument against the claim that the predicate is [–transitive] in such
constructions. We will argue in chapter 5 that this marker should be treated as
a specificity marker, and, therefore, it is not problematic.
Let us now look at the non-permissibility of nominative case marking of

the anaphor in (58) and (60). There does not seem to be any reason for why a
nominal complex anaphor cannot occur in the theme position. In fact, referential
expressions and pronouns in nominative case do occur in the theme position
of a non-nominative subject (64) (see chapter 5 for details). The finite INFL
checks the nominative case of yah bāt ‘this matter’ in (64) in Hindi-Urdu.
Hindi-Urdu (IA)
(64) rām ko yah bāt acchı̄ lagı̄
Ram dat this thing f,s, nom good f,s appeared f,s
‘Ram liked this matter.’
Davison (2004: 155) points out that there is “a language-specific constraint

that reflexives cannot have nominative case.” That is why sentence (58) is
ungrammatical.24
Several studies have attempted to relate the role of agreement to the syntactic
positions in which an anaphor occurs (Picallo 1985; Anagnostopoulou and
Everaert 1996; Woolford 1999; Everaert 2001). It is argued that anaphors do
not occur in syntactic positions construed with agreement unless agreement is
anaphoric. This is referred to as the Anaphor Agreement Principle.
Thus, nominal subject anaphors as well as nominal object anaphors that are
not specified for phi-features (person, number, gender-features) are not licensed
in languages with object agreement. Following this, languages such as Hindi-
Urdu, Kashmiri, Marathi, qualify for these two conditions, and they do not
permit nominal subject or object anaphors (see sentence (58) from Hindi-Urdu
and sentence (60) from Kashmiri).
Further, there are languages such as Albanian, Basque, Greek and Georgian
that have subject/object agreement and have anaphors in subject/object posi-
tions (Everaert 2001). A similar situation obtains in two Tibeto-Burman lan-
guages, Hmar and Mizo (see sentence (10) from Mizo). The anaphor is specified
for phi-features, and, hence, it is permitted in object position. In Kashmiri, cer-
tain verbs permit a nominal anaphor in object position (see sentence (61)). This
needs to be studied further.
We now consider sentences with passive subjects.
3.5.2 Passive subject as antecedent

We demonstrate below that in Hindi-Urdu, Kashmiri and Marathi (IA) and
Telugu (DR), a passive subject can antecede an anaphor, while in Punjabi (IA)
it is not permitted for one set of speakers, and also in Malayalam (DR) such
sentences are questionable.
Indo-Aryan
The subject in a passive sentence corresponds to the direct object of an active
sentence. In Hindi-Urdu, “the object of an active transitive verb is normally
marked with the dative-accusative postposition if it has human reference,” as in
(65a) (Davison 2000: 413). The object of a transitive verb, which becomes the
derived subject in a passive sentence, may be: (i) nominative case-marked as in
(65b), or (ii) dative-accusative case-marked retaining the original postposition
ko of the active sentence, as in (65c).
(65a) cor- ne ādmı̄- ko is sar.ak- par nahı̃˜ lūt.- ā

thief- erg man- dat this street- on not rob- perf
‘A thief did not rob a man on this street.’
(65b) [yah ādmı̄i ] apnı̄i,∗ j sar.ak- par lūt.- ā nahı̃˜ ga- yā
this man self’s street- on rob- perf not go- perf
‘[This man]i was not robbed [on hisi,∗ j own street].’
(65c) [is ādmı̄i - ko] [apnı̄i,∗ j sar.ak- par] lūt.- ā nahı̃˜ ga- yā
this man- dat self’s street- on rob- perf not go- perf
‘[This man]i was not robbed [on hisi,∗ j own street].’
(Davison 2000: 413)
In Kashmiri (66) and Marathi (67) too, the derived passive subject can antecede
the possessive reflexive.
Kashmiri (IA)
(66) aslam- asi āv pannisi bāg-as- manz lāyn
Aslam- dat pass self.dat garden- in beat.inf
‘Aslami was beaten in self’si garden.’
(Wali et al. 2000: 487)
Marathi (IA)
(67) lili- lāi kāl (āp-lyā) swatāh-cyāi kholit d.āmbla gela
Lili- dat yesterday self’s room.in dumped pass
‘Yesterday Lili was dumped in self’s room.’
(Wali 2000: 529)
In Punjabi the passive subject cannot antecede an anaphor for one set of speakers
(Bhatia 2000: 661), as in sentence (68): “In some dialects of Punjabi all oblique
subjects including instrumentals and locative subjects are permitted to corefer
with a reflexive, and for these dialects [(69) below] is grammatical” (Bhatia
2000: 661).
p a s s i v e s u b j e c t a s a n t e c e d e n t t o a n a n a p h o r : no t p e r m i t t e d
Punjabi (IA)
(68) *mun.d.ei tõ apn.e∗ i,∗ j / o’i de kamre dı̄ ku
jı̄ gavācı̄
boy instr self/ he of room of key f,s lose.pst.f.s
*‘The boyi (inadvertently) lost the key to hisi room.’
(Bhatia 2000: 661)25
Note that (68) is a lexical passive while (69) is a syntactic passive with abilitative
meaning.
(69) mε̃- thõ26 apn.ı̄ kitāb pár.ı̄ ná ˜ı̃ gayı̄
I- by self’s book read.perf.f.s not go.pst.f.s
‘I could not read this book of mine.’
(Om Arora, p.c.)
Dravidian
The use of the passive in Telugu is rather infrequent. It is only the possessive
form of tan ‘self’ that can occur in a passive sentence.
Telugu (DR)
(70) avinı̄tiparulaina udyōgastului tamai il.l.a- ki pamp.a- bad.d.-ārui
corrupt officials selves’ houses- dat send.inf- pass-agr
‘The corrupt officialsi were sent to theiri houses.’
In Kannada, too, a similar pattern obtains.

Kannada (DR)
(71) bras.t.a udyōgigal.ui tammai mane- ge kal.uhis- al- pat.t.-arui
corrupt officials selves’ houses- dat send- inf- pass-3 p
‘The corrupt officialsi were sent to theiri houses.’
In Malayalam (DR), the sentence with a passive subject as antecedent to a

possessive anaphor tan-te ‘self’s’ is questionable. That is, the antecedent must
c-command the anaphor, and there is no reconstruction.
Malayalam (DR)
(72) ??tan-tei bhaks.an.am rāman∗ i - āl pākam ceyya- ppet.-t.u
self-gen food Raman- instr cooking do- pass-pst
??‘Self’s food was cooked by Raman.’
(Jayaseelan 2000: 128)
3.6 The dative subject construction and the verbal clitic

Another significant issue that concerns anaphors is that the occurrence of the
VR/VREC is not permitted in the dative subject construction in SALs if the
predicate is formed out of a [–transitive] verb, as in (73) from Telugu (DR) and
(74) Kannada (DR) (see chapter 5 for details).
Telugu (DR)
(73) karun.ai ki tana mı̄da tanai ki kōpam vacc- (*kon)- indi
Karuna dat self on self dat anger come- VR- pst- 3s,nm
‘Karunai got mad at herselfi .’
Kannada (DR)
(74) rāman-igei tanna mēle nēi,j.∗ k kōpa bantu (anta sı̄tej hēl.idal.u)
Ram-dat self on emph anger came that (COMP) Sita said
‘(Sitaj said that) Rami was angry with selfi,j.∗ k ’
Amritavalli (2000: 54) points out: “there is a distinction between the syntac-
tically generated VR that occurs as an anaphor, and the inchoative VR that
occurs as an intransitive (anticausative) marker.” She further points out: “the
experiencer stative construction is an intransitive construction, and that syn-
tactically generated kol..lu [the VR in Kannada] (as against lexical inchoative
kol..lu) cannot occur in intransitive constructions” (Amritavalli 2000: 54).
Thus, the reason for the non-permissibility of the VR is that the dative
predicate itself is [–transitive], and, hence, a VR which is [–transitive] is not
needed. This could be due to a general constraint in language on duplication of
features with the same value.
However, a set of transitive verbs, such as kōs ‘cut,’ tat..t ‘pat,’ ad.d. ‘stop’
and kot..t ‘hit,’ permit the occurrence of the VR in Telugu (DR) in the dative
subject construction. The reason for such an occurrence is that the verbal
reflexive functions as a detransitivizer and it detransitivizes the transitive verb
(see sentence (32)).
3.7 Long-distance binding and Principle A

In this section we discuss whether instances of long-distance binding go against
Principle A of the Binding Theory, and the role that the verbal reflexive plays
in long-distance binding.
3.7.1 Long-distance binding and the simplex anaphor

Most of the SALs permit long-distance binding of a monomorphemic anaphor.
By long-distance binding, we mean the coindexation of an anaphor in a lower
clause with an antecedent in a higher clause outside its minimal clausal domain.
Such coindexation is an apparent violation of Principle A of the Binding Prin-
ciples (Chomsky 1981, 1986). Japanese monomorphemic form zibun ‘self,’
Chinese ziji and Korean casin permit long-distance binding, while their respec-
tive polymorphemic forms zibun-zisin/kare-zisin, ta ziji and caki-casin do not
(see Cole and Sung 1994 for further details). In SALs too, it is the possessive
anaphor or the monomorphemic form of the nominal anaphor that permits long-
distance binding, while a polymorphemic form does not – except in Marathi
(Wali 2000: 534). Further, the occurrence of the verbal anaphor in the embed-
ded clause blocks long-distance binding.
Some of the major features of long-distance binding in SALs are:
(i) all long-distance anaphors are subject-oriented (Chomsky 1986);
(ii) long-distance binding is not permitted from a finite clause – only non-finite
clauses permit it;27
(iii) it is the monomorphemic form that permits long-distance binding, whereas
polymorphemic anaphors do not; hence, reciprocals, which are in general
polymorphemic in SALs, do not permit long-distance binding.28
Cole and Sung (1994) proposed ‘head raising analysis’ to account for long-
distance binding in Chinese. Under this analysis, the anaphor moves to INFL
of the lower clause, and then to the higher clause, where it is coindexed with a
long-distance antecedent.
We present an example from Hindi-Urdu to exemplify the phenomenon of
long-distance binding.
Hindi-Urdu (IA)
(75) ašoki ne lalitāj se [S2 PROj apne- liyei,j cāy banā-ne
Ashok (m) erg Lalita (f) with self- for tea make- to
ko S2 ] kahā
in order said
‘Ashok asked Lalita to make some tea for himself/herself.’
(Subbarao 1971: 191)
In (75), apne-liye ‘for self’ can be coindexed with the embedded subject PRO
of S2 , which, in turn, is coindexed with the matrix subject lalitā, thus permitting
binding of an anaphor which obeys Principle A. However, apne-liye ‘for self’
can also be coindexed with the subject of the matrix sentence ašok which is
a long-distance antecedent. Such interpretation of the anaphor is an apparent
violation of Principle A of the Binding Theory which states that an anaphor must
be bound in its governing category. One of the ways to account for long-distance
binding of the simplex anaphor apne-liye ‘for self’ without violating Principle A
is to move the anaphor by the head-to-head movement rule. INFL consists of the
interpretable features of TENSE and ASPECT and uninterpretable features of
AGR (Chomsky 2001). According to Davison (2001: 57), the monomorphemic
anaphor first moves to the non-finite TENSE of its own clause, and then, by
successive cyclic movement, it moves to the finite TENSE of the next higher
clause. It is then coindexed with the subject of the higher clause. In Cole and
Sung’s analysis of Chinese, the anaphor cliticizes to AGR, but in Davison’s

analysis it cliticizes to TENSE as “AGR plays no essential role in reflexive
binding” (Davison 2001: 59). Davison’s proposal, in our opinion, would hold
for all SALs that have long-distance binding and exhibit agreement.
Though it is monomorphemic anaphors that permit long-distance binding,
Marathi (IA) permits long-distance binding of a polymorphemic anaphor. The
anaphor (āplyā) swatāh-lā ‘self self-acc’ occurs in the direct object position in
sentence (76), and it can be coindexed with either lili ‘Lili’ or šām ‘Sham,’ the
subjects of the matrix and embedded clauses, respectively.
Marathi (IA)
(76) [S1 lilii - ni šāmj - lā [S2 PROj (āplyā) swatāh- lāi,j
Lili- erg Sham- dat ø self self- acc
bockār-uS2 ] dilaS1 ]
pinch-inf allowed
‘Lilii allowed Shamj to pinch selfi,j .’
(Wali 2000: 534)
No study has thus far provided an explanation to account for the long-distance
binding of the polymorphemic anaphor in sentence (76).
3.7.2 Long-distance binding and the verbal anaphor

We shall discuss the cases in which long-distance binding is prohibited when
a verbal reflexive/reciprocal occurs in the embedded clause. We shall take one
example from each language family except Indo-Aryan.29
In sentence (77), in Telugu (DR), when the verbal reflexive occurs in S2 , the
anaphor is coindexed only locally with PRO, which, in turn, is coindexed with
its object controller lalita in the matrix clause, just as in Tamil. The VR is in
italics.
vr occurring in s2
Telugu (DR)
(77) ašōki lalitaj ni [S2 PRO∗ i,j tana kōsam t.ı̄ cēsu- kō-
Ashok (m) Lalita (f) acc self for tea make- VR-
m- aniS2 ] annād.u
imp- comp said
‘Ashoki told Lalitaj to make some tea for *himself/herself.’
If coreference is intended with the matrix subject, the VR cannot occur.
(78) [S1 ašōki lalitaj ni [S2 PROi,∗ j tana kōsam t.ı̄ ceyya-
Ashok Lalita acc self for tea make-
m- aniS2 ] annād.uS1 ]
imp- comp said
‘Ashok told Lalita to make some tea for himself/*herself.’
In Hmar (TB) too, when the VR occurs in S2 , long-distance binding is blocked,

as in sentence (79). In S2 amaʔ ‘self’ is coindexed with the local antecedent
PRO, which in turn is coindexed with lali in S1 , and it cannot be coindexed with
pāri in S1 . Note that the embedded verb carries no subject agreement marker
as the subject of S2 is PRO.
Hmar (TB)
(79) [S1 pārii - n lalij [S2 PROj ama∗ i/j in- that- lo- diŋS2 ]- in a- tiS1 ]
Pari- erg Lali self VR- kill- not- fut- comp 3s- say
‘Parii asked Lalij not to kill herself∗ i/j .’
(Mukerjee, Subbarao and Walia 2003: 111)
In Ho (Munda) too, the verbal reflexive clitic -n blocks long-distance binding

(80). Recall that the nominal anaphor is a null element. The infinitive carries
the agreement marker of the matrix subject.
Ho (Munda)
(80) [S1 somai rāmj [S2 PROj goe- n∗ i/j - tiy a- eiS2 ] kaji- ad.- aS1 ]
Soma Ram kill- VR- to- 3s tell- pst[+tr]- fin
‘Somai told Ramj to kill himself∗ i/j .’
That is, when the VR is present in S2 , the anaphor cannot cross the VR to go
further up, and thus the VR is instrumental in blocking long-distance binding.
It insulates the anaphor from long-distance coindexation. This phenomenon
has been observed elsewhere too (Everaert 1991; Kazenin 2001).
3.8 Small Clauses and long-distance binding

In this section we discuss how long-distance binding in small clauses works.
In almost all SALs, the subject of a small clause can be coindexed with the
subject of the matrix clause. The small clause consists of a subject DP and
a predicative adjective/DP: “[T]here is no tense or aspect morphology and no
verb agreement” (Davison 2000: 420). Verbs such as samajh-nā ‘to understand,’
lag-nā ‘to appear/strike’ and dikh-nā ‘to appear’ in Hindi-Urdu permit small
clauses, while predicates such as soc-nā ‘to think’ and mālūm ho-nā ‘to come
to know’ do not. If the matrix verb is [+transitive], the embedded subject is
accusative/dative case-marked. If it is [–transitive], the embedded subject is
nominative case-marked.
In Hindi-Urdu, the morphologically complex form apne āp ko ‘self’s self-
acc’ or the simplex form apne ko- ‘self-acc’ can both occur in the subject
position of a small clause, if the anaphor is coindexed with the matrix subject
in the nominative case.
3.8 Small Clauses and binding 79
3.8.1 Indo-Aryan
Hindi-Urdu (IA)
(81) pramodi [apne (āpi/∗ j )-/ khud-i/∗ j / swayam-i/∗ j ko par.hākū] nahı̃˜
Pramod self’s self- self- self- acc nerd not
samajh-tā (hai)
understand-imperf is
‘Pramodi does not consider [himselfi a nerd].’
(Davison 2000: 420)
Let us reconsider sentence (58), an instance of ECM in a dative case-marked

construction. Recall that a dative case-marked subject does not permit a nomina-
tive case-marked anaphor in theme/patient position. An accusative case-marked
anaphor is also not permitted in sentence (58).
The ungrammaticality of sentence (58) is due to the facts that: (i) the dative
predicate lagnā ‘to strike / to appear’ is [–transitive] and, hence, it cannot
case-mark the embedded subject accusative; and (ii) a nominative case-marked
anaphor in Hindi-Urdu is prohibited (see the Anaphor Agreement Principle).
In contrast, in Bangla (IA), a genitive case-marked subject ram-er ‘Ram-
gen’ can be the antecedent of the subject anaphor nije-ke ‘self-acc’ of the small
clause. The [+transitive] verb mone ho ‘consider’ in (82) assigns accusative
case to the anaphor.30
Bangla (IA)
(82) ramer nije- ke bhalo mone hɔy
Ram- gen self- acc good considers
‘Ram considers himself good.’
In Marathi (IA) and Oriya (IA) too, an anaphor occurs in small clauses.
Marathi (IA)
(83) lilii (āplyā) swatāh- lāi libral samajhte
Lili herself- acc liberal considers
‘Lili considers herself liberal.’
(Wali 2000: 534)
3.8.2 Dravidian
In Telugu (DR) too, the matrix predicate that is [+transitive] permits a complex
anaphor when the subject antecedent is in the nominative case (84) and a
[–transitive] dative predicate permits “only the simplex form of the anaphor in
nominative case” as in sentence (85): “the matrix verb which is an intransitive
verb cannot assign accusative case to the embedded subject by exceptional case
marking” (Subbarao and Bhaskararao 2004: 178).
Telugu (DR)
With a [+transitive] predicate bhāvincu ‘consider’
(84) mādhuri tana- ni tanu andagatte- gā bhāvistundi
Madhuri self- acc self pretty- adjr considers.3s.nm
‘Madhuri considers herself pretty.’
(Subbarao and Bhaskararao 2004: 178)
With a [–transitive] predicate anipincu ‘appear’

(85) prativād.i- kiı̄ tanu/ *tana- ni tanu cālā goppavād.u-
every.fellow- dat self.nom self- acc self very great.person-
ani/ gā anipistā.d.u
comp adjr appears.3s.masc
‘Every fellow thinks that he is great.’
(Subbarao and Bhaskararao 2004: 178; sentence (85) is a
consolidated version of sentences (15) and (16) from the original)
In Malayalam, only the complex form awan-e tanne-e ‘he-acc emph’ is per-
mitted as the subject of the embedded clause; a simplex form is not permitted
(86).
Malayalam (DR)
(86) rāmani awani - e tanne-e/ *tanni - e oru mid.ukkan
Raman he- acc emph self- acc one clever person
āyi karuti
being considered
‘Ramani considered himselfi a clever man.’
(Sentence (86) is a consolidated version of sentence (48) on
page 131, and sentence (130) on page 144, of Jayaseelan 2000)
A pronoun in the embedded subject position is “disjoint in reference from the

matrix subject” (Jayaseelan 2000: 138).
3.8.3 Tibeto-Burman
In Mizo (TB), a simplex form of the anaphor occurs in copular complements,
and co-refers with the nonlocal subject. In Mizo, verbs such as ‘consider,’
‘believe,’ etc., inherently carry a verbal reflexive, as they invariably carry the
verbal anaphor. The form a maʔ ‘self’ as the subject of the small clause
co-refers with the matrix verb only when the VR is present. Otherwise, it
does not. The form a maʔ refers to a discourse antecedent in such syntactic
constructions, as in sentence (88) (Lalitha Murthy and Subbarao 2000: 803).
In sentence (87), inti ‘consider’ in Mizo consists of -in ‘VR’ and ti ‘say.’ The
VR -in such constructions in Mizo, and kon- in Telugu in anukonu ‘consider,’
is grammaticalized and, hence, cannot be treated as a verbal reflexive.
3.8 Small Clauses and binding 81
ecm with a vr
Mizo (TB)
(87) zovai - n [a- maʔi,∗ j - cu lāmthiam] a- inti
Zova- erg 3s- self- foc dancer 3s- consider
‘Zova considers self a dancer.’
ecm without a vr
(88) zovai - n [a- maʔ∗ i,j - cu liberal] a- hria
Zova- erg 3s- self- foc liberal 3s- know
‘Zova knows self∗ i/j is a liberal.’
In Manipuri (TB) too, the VR occurs in the matrix clause if coindexation of

the small clause subject with the matrix clause subject takes place. Recall that
Manipuri has a polymorphemic nominal anaphor, and a verbal reflexive -ja that
is optional.
When a nominal anaphor occurs in a small clause, the VR -ja occurs on
the matrix verb, where its occurrence is optional (89). However, when the
embedded subject is not coindexed with the matrix subject, the occurrence of
the verbal anaphor is prohibited (90).
Manipuri (TB)
(89) tombii na masai - bu phaza- hayna lau- (jai )- i
Tombi nom self- acc pretty- comp consider- VR- pres
‘Tombi considers herself pretty.’
(90) tombii na laugij bu phaza hayna lau- (*jai )- i
Tombi nom Laugi acc pretty comp consider VR- pres
‘Tombi considers Laugi pretty.’
3.8.4 Austro-Asiatic (Mon-Khmer)

Khasi (Mon-Khmer) does not have sentences with ECM. The finite comple-
mentizer ba ‘that’ occurs in sentence (91).
Khasi (Mon-Khmer)
(91) u- lεmi u- khεiñ ba pro ui - lɔŋ u- ba- bhabrēw
3 m,s- Lem 3 m,s- count that 3 m,s- be 3 m,s- adjr- beautiful
‘Lemi thinks that hei is handsome.’
We did not find any ECM cases in Ho and Kharia (Munda).

3.8.5 Summary
The question that needs to be addressed is: do the cases involving ECM sub-
jects and the coindexation of the lexical anaphor with the matrix subject
involve long-distance binding? It does not appear to be the case. The fact
that the verbal anaphor occurs in Manipuri (TB) with the matrix verb when
an anaphor occurs as the subject of the small clause shows that the subject
of the small clause is the DO of the matrix clause. That is how it gets case-
marked. Therefore, such cases cannot be treated as instances of long-distance
binding.31
In this section we have shown that, in cases involving ECM:
(i) a dative case-marked subject in Hindi-Urdu (IA) does not permit either
a nominative case-marked or accusative case-marked embedded subject
(sentence (58)), while in Bangla a genitive case-marked matrix subject
permits an accusative case-marked enbedded subject as in (82);32
(ii) in Telugu (DR), a dative case-marked subject can be coindexed with only
the simplex form of the anaphor, and not the complex form, while in
Malayalam the complex form of the anaphor is permitted; and
(iii) the coindexation of a lexical anaphor with the matrix subject in ECM
constructions does not count as long-distance binding.
3.9 The reflexive in a locative PP

3.9.1 Locative PP (non-subcategorized)
Neither a polymorphemic form nor the VR is permitted in any of the SALs
we are familiar with, if the locative PP is non-subcategorized.33 The follow-
ing examples from Hindi-Urdu (IA), Malayalam (DR) and Mizo (TB) are
illustrative.
Hindi-Urdu (IA)
(92) rāmi ne [apnei / us∗ i - ke- nazdı̄k] ek sarp dekh-ā
Ram erg self’s 3s.gen- near one snake see-perf
‘Rami saw a snake [near selfi ].’
(Davison 2000: 409)
Malayalam (DR)
(93) rāmani tani - te munn- il oru āna- ye kan.d.u
Ram self- gen front- in one elephant- acc saw
‘Rami saw an elephant in front of selfi .’
3.10 Binding principles A, B and C 83
Mizo (TB)
(94) zova- ni rūl amaʔi,∗ j- bulaʔ a- hmu
Zova- erg snake self- near 3s- saw
‘Zovai saw a snake near selfi,∗ j .’
3.9.2 Locative PP (subcategorized)

In Kannada, Tamil and Telugu (DR), “where it [the locative PP] is a subcate-
gorized phrase – the VR is required” (Lust et al. 2000: 30).
Kannada (DR)
(95) rāmai tannai - mēle tappu hāki- kon.d.a- nu
Rama self.gen- on blame put- VR- agr
‘Ramai put the blame on himselfi .’
In Telugu (DR), a polymorphemic anaphor and the VR occur in such cases.
Telugu (DR)
(96) saritai tana mı̄da (tanui ) nēram mōpu- kon- di
Sarita self.gen on self blame put- VR- agr
‘Saritai put the blame on herselfi .’
Thus, a polymorphemic anaphor and the VR are permitted only if the locative
PP is a subcategorized argument. If it is not, neither of them is permitted.
3.10 Binding principles A, B and C

As mentioned earlier, there are three principles of the Binding Theory “defined
over configurational representations of sentence structure, because ‘binding’
implies ‘c-command and coindexing’ in a local domain” (Lust et al. 2000: 3).
3.10.1 Principles A, B and C of the Binding Theory

In this section, we shall demonstrate that all SALs except Marathi (IA) obey
Principles A, B and C of the Binding Theory. Principle A of the Binding Theory
states: “An anaphor is bound in a local domain.”
The discussion above showed that Principle A is generally obeyed in SALs,
though there is counterevidence from Marathi (IA) (see sentence (76), where
a complex anaphor may have a long-distance antecedent). Thus, the Marathi
case is an instance of a clear violation of Principle A of Binding Theory.
Principle B of the Binding Theory states: “A pronominal is free in a local

domain.” This implies that a pronoun cannot antecede another pronoun within
the same GC, as sentence (97) illustrates.
(97) *Hei saw himi .
SALs obey Principle B. We provide an example from Khasi (Mon-Khmer).
(98) kai - thok ya- ka∗ i,j

she- lied acc- she
‘Shei lied to her∗ i,j .’
Scrambling effects pronoun–antecedent relationship, and hence (99) is not

permitted.
(99) *ya- ka∗ i,j ka- thok

acc- she she- lied
‘Shei lied to her∗ i,j .’
Principle C states: “An R-expression is free” (Chomsky 1995a: 100).

Malayalam and Telugu provide apparent counterevidence to Principle C. A
proper name or a title in a sentence “may be repeated in apparent violation
of Principle C” (Jayaseelan 2000: 163). Such violations are found in Thai and
Vietnamese too (Jayaseelan 2000).
Malayalam (DR)
(100) rāmani wicāriccu [rāmani jayikkum ennə]
Raman thought Raman will win comp
‘Raman thought that Raman would win.’
Jayaseelan (2000: 163) points out: “the repeated proper name seems to behave
like a pronoun in that it apparently obeys Principle B.” Evidence in support of
this claim comes from the fact that a proper name that is an argument of a verb
cannot be coreferential with another argument of the same verb.
(101) ?*rāmani ?*rāmani - e pukazhti

Raman Raman- acc praised
Jayaseelan (2000) further points out that, just as in the case of pronouns, the
above sentence becomes acceptable by adding tanne, the emphatic.
(102) rāmani rāmani - e tanne pukazhti

Raman Raman- acc emph praised
‘Raman praised Raman himself.’
3.11 Anaphors and scrambling 85
Telugu (DR), too, obeys Principle C. A c-commanding pronoun/anaphor is disallowed

for coreference with a proper name (Subbarao and Lalitha Murthy 2000: 270).
Telugu (DR)
(103) *vād.u/i tanui rāmud.(i)i - ni cūsēd.u
he self Ramudu.obl- acc saw
‘*Hei saw Ramudui .’
(Subbarao and Lalitha Murthy 2000: 270;
the glosses have been slightly modified)
(104) rāmud.ui int.i- ki rā- gānē atanu∗ i,j kūrconnād.u

Ramudu home.obl- dat come- right after he sat down
‘Hej sat down right after Ramudui came home.’
In rhetorical questions, however, two proper names in a sentence may be

coindexed.
(105) rāmud.ui rāmud.i-i int.(i)- lō und.aka- inka ekkad.a unt.ād.u
Ramudu Ramudu’s- house.obl- in if not staying- else where will stay
‘If Ramudu does not stay in Ramudu’s house, where else would he stay?’
Since the repeated occurrence of a proper name is permitted in specifically

restricted contexts in Telugu (DR) in sentence (105), and the addition of tanne
makes sentence (102) grammatical in Malayalam, these data do not count as
counterevidence to Principle C.
3.11 Anaphors and scrambling

SALs freely permit scrambling of constituents in a sentence. The crucial issue
with regard to such movement is: is the scrambled DP moved to an A-position
or to an A-bar position? The implications of this question are the following: if it
moves to an A-position, then binding possibilities change after scrambling, as
the scrambled DP in A-position can be an antecedent of an anaphor. If it moves
to an A-bar position, “a ‘scrambled’ phrase within a single clause behaves
as though it were in its original position for purposes of coindexing with an
antecedent” (Davison 2000: 446). For Hindi (IA), Mahajan (1990) holds the
view that it is A-movement, while Dayal (1994) supports the view that it is
A-bar movement and, hence, subject to reconstruction.
In sentence (106), apne ‘self’s’ corefers with Ram, the subject.
(106) rāmi ne apnei/∗ j baccõ- ko šerj dikhāyā

Rami erg self’si/∗ j children- dat tigerj show.perf
‘Rami showed self’si/∗ j children the tiger.’
(Mahajan 1990: 33)
The NP Ram in sentence (107) is scrambled leftwards and the coindexing

relations remain the same, as it is not in the subject position.
(107) rām ne šerj apnei/∗ j baccõ- ko dikhāyā

Rami erg tigerj self’si/∗ j children- dat show.perf
‘Rami showed self’si/∗ j children the tiger.’
However, according to Mahajan (1990) šer ‘tiger’ may also be the antecedent
of apne ‘self’s’ in sentence (107), though it did not move to subject position, as
his dialect does not have a strict subject condition. Dayal (1994) does not agree
with Mahajan (1990) that scrambled NPs can serve as antecedents to anaphors.
She argues that the facts concerning binding do not support Mahajan’s claim,
and her claim is that, as far as binding goes, “scrambling is an instance of A-bar
movement only, not A-movement” (Dayal 1994: 239).
Based on sentences such as (108), Mahajan (1990) argues that scrambling
moves an NP to an A-position, as the DO binds the possessive anaphor apne
baccõ-ne ‘self’s children-ERG.’
(108) mohan- koi apnei baccõ- ne mārā

Mohan- acc self’s children- erg beat
‘Self’s children beat Mohan.’ (OSV)
Dayal’s contention is that sentences such as (108) are not acceptable to most
of the speakers of Hindi, and that Mahajan’s analysis fails to account for the
ungrammaticality, as the scrambled NP mohan-ko ‘Mohan-acc’ should bind
the possessive reflexive, since it is in an argument position. Dayal argues that
if scrambling is treated as A-bar movement, then the scrambled NP mohan-ko
‘Mohan-acc’ cannot bind an anaphor as “the DO would be reconstructed and
the resulting violation of Principle A would account for its ungrammaticality”
(Dayal 1994: 242).
In Telugu (DR) and Mizo (TB), sentences parallel to (108) are ungrammati-
cal, and if scrambling is treated as A-bar movement, the ungrammaticality can
be accounted for.
Telugu (DR)
(109) *pillal(a)- ni tama amma kot.t.indi
children.obl- acc selves’ mother hit
‘*Selves’ mother beat children.’
Mizo (TB)
(110) *zova- cu amaʔ nupui- in a- hau
Zova- acc self’s wife- erg 3s- scolded
‘*Self’s wife scolded Zova.’
3.12 Blocking Effects 87
The interested reader may refer to Mahajan (1990), Dayal (1994) and Kidwai
(2000) for further details.34
3.12 Blocking Effects

Another significant aspect in the study of anaphors is Blocking Effects. Lan-
guages which do not have subject–verb agreement exhibit Blocking Effects.
We shall now explain what we mean by Blocking Effects.
Blocking Effects are “the impossibility of coreference between a distant
antecedent and a pro-form if there is an interceding (potential) antecedent with
distinct features. Such effects have been found in Chinese with long-distance
anaphors” (Lust et al. 2000: 886). Hence, in long-distance binding, coindexing
of an anaphor with an antecedent is blocked in Chinese, for example, if there
is person disharmony between the antecedent and an intervening subject (see
Cole and Sung 1994). Similarly, long-distance binding is blocked when there is
a 1st or 2nd person subject in a clause intervening between the simplex anaphor
and its long-distance antecedent in Malayalam (Jayaseelan 1998).
Furthermore, languages that lack subject–verb agreement exhibit Blocking
Effects, while languages that have subject–verb agreement do not.
In Malayalam (DR), “when it [tān ‘self’] is long-distance bound, every
intervening subject must be 3rd person; otherwise the coindexing is blocked.
As [(111) below] shows, only the non-identity of the person feature of the
intended long-distance antecedent and the intervening subject results in block-
ing” (Jayaseelan 1998: 12).
(111) Johni wicāriccu [Maryj / kut.t.i kal./ *

ān/ *nı̄/ tani - ne snēhikkunn-
John thought Mary children I you self acc love (pres)-
illa ennə]
neg comp
‘Johni thought that Mary / the children / *I / *you do(es) not love selfi .’
In Malayalam (DR), “tān shows Blocking Effects also if it is conjoined with a

first-person or second-person pronoun, or if there is a first-person or second-
person pronoun in the minimal clause” (Jayaseelan 2000: 133).
Malayalam (DR)
(112) *rāmani wicāriccu kris.n.an nin- akkə tan∗ i - ne paricayapped.utt-um
Raman thought Krishnan you- dat self- acc introduce-will
ennə
comp
*‘Ramani thought that Krishnan would introduce self∗ i to you.’
“In Hindi-Urdu, there is no Blocking Effect caused by different PNG features of

multiple antecedents” (Davison 2001: 55). Kannada (DR) (Amritavalli 2000)
and Hmar (TB) (Mukherjee et al. 2004) also do not show Blocking Effects. We
provide an example each from Hindi-Urdu (IA) and Hmar (TB).
Hindi-Urdu (IA): no Blocking Effects

(113) [harek bāt par ravi kā apne- ko dos. denā] mãı̃ pasand nahı̃˜ kartā
each thing on Ravi gen self- dat to blame I do not like
‘Ii don’t like [Ravij ’s blaming self?i/j for everything].’
(From Subbarao 1984a: 54, as discussed in Davison 2001: 55)
Hmar (TB): no Blocking Effects

(114) kennii - in ama∗ i,j,k a-theiŋhil aniti berrij - in a-hriet cu
Kenny- erg self forgot- comp Berry- erg knows def
cerrik a-ŋaitu
Cherry thinks
‘Cherry thinks that Berry knows that Kenny forgot him.’
(Mukherjee, Subbarao and Walia 2004: 11)
To sum up, SALs that exhibit subject–verb agreement have no Blocking Effects
and conform to the generalization made in Cole and Sung (1994: 364): “the
typological property shared by languages that manifest the blocking effect is
the absence of agreement on verbs and other predicates: Chinese and Korean
lack verb agreement and show the blocking effect, whereas Italian and Icelandic
have agreement and lack the blocking effect.”
3.13 Sub-commanding NP as an antecedent

According to Tang (1989), if an antecedent is a subpart of a phrase that
c-commands a form, it is said to subcommand that form. Thus, “[a] reflex-
ive may have an antecedent which does not strictly c-command it, but is a
specifier of a subject NP (sub-command)” (Davison 2000: 66–67).
In (115) in Hindi-Urdu (IA), rām ‘Ram’ is a possessor occurring in the Spec
position of the head DP man ‘mind’ that c-commands the anaphor apne āp
‘self’s self.’
(115) [rāmi - ke man]j - mẽ apne (āp)i/∗ k - se ghrin.ā thı̄

Ram- gen mind- in self’s self- from hatred was
‘Rami hated himselfi .’
(Davison 2000: 417; sentence (51C) of Davison 2000
is slightly modified)
Similar examples are found in other SALs too.

3.14 Reduplication and Case Copying 89
3.14 Reduplication and Case Copying in complex anaphors

In this section we shall briefly focus on the composition of reduplicated
anaphors in Telugu (DR) and Manipuri (TB).
We have claimed earlier that the complex anaphor in Telugu (DR) and
Manipuri (TB) is the result of reduplication and Case Copying. The nominative
or dative case marker of the subject is copied onto the second part of the anaphor.
The dative subject construction (DSC) provides further support for the proposal
concerning Case Copying on the anaphor in Telugu. The dative case marker
of the subject in italics in sentence (116) is copied onto the second part of the
bipartite structure of the anaphor. Recall that neither a verbal reflexive nor a
verbal reciprocal is permitted in a DSC.
Telugu (DR)
(116) mamatai ki tana mı̄da tanai ki kōpam vaccindi
Mamata dat self on self dat anger came
‘Mamatai got angry with herselfi .’
Reciprocals in Telugu provide further evidence in support of our proposal

concerning Case Copying. The second part of the bipartite structure of the
reciprocal carries the null nominative case marker of the nominative subject in
sentence (117) and the dative case marker of the dative subject in (118).
nominative subject
(117) vāl..lu okal.l.a- tō okal..lu debba lād.u- kon- t.unnā- ru
they (nom) one- with one (nom) fight- VREC- progr- m.p
‘They are fighting with each other.’
dative subject
(118) vāl..la- ki okal.l.a- tō okal..la- ki sariggā pad.adu
they- dat one.gen- with one.gen- dat well get along.neg
‘They do not get along well with each other.’
Reciprocals in Manipuri (TB) too, have a similar phenomenon of Case Copying.

The nominative case marker na occurs on the first part of the reciprocal, while
the accusative marker bu occurs on the second part of the reciprocal in DO
position (119), and the dative case marker occurs in IO position (120). The
occurrence of the VREC as well as of the nominal reciprocal is obligatory.
anaphor in do position
Manipuri (TB)
(119) makhoi- na ama- na ama- bu thagat- na- rammi
they- nom one- nom one- acc praise- VREC- pst
‘They each praised the other.’
anaphor in io position
(120) makhoi- na ama- na ama- da khudopot pi- na- rammi
they- nom one- nom one- dat gift give- VREC- pst
‘They gave gifts to one another.’
(Sarju Devi and Subbarao 2003: 69)
It should also be mentioned that there are two other forms of the reciprocal,
ama ga ama ga ‘each other’ and makhoi masel ‘each other’, and there is no
Case Copying in them (Sarju Devi and Subbarao 2003).
Both in Manipuri (TB) and Telugu (DR), reduplication is productively used
to create a morphological complex that refers to an anaphor, and it is Case
Copying and a lexical case-marker assignment that are used to case-mark the
constituent parts of the bipartite structure of the anaphor. There is, however,
a minor difference: while Case Copying of the subject takes place on the first
part of the reduplicated structure of the anaphor in Manipuri, it is on the second
part of the reduplicated structure that Case Copying takes place in Dravidian.
In Sarju Devi and Subbarao (2003), we pointed out that Case Copying is a
phenomenon that belongs to syntax, and reduplication is a process that belongs
to compound-formation strategies of language. It is, however, the interaction of
these two phenomena that results in the formation of anaphors in two genetically
unrelated languages.
Malayalam (DR) has two forms of the complex anaphor as in (i) and
(ii) below:
(121a) tann- e tanne

self- acc emph
(121b) awan- e tanne
self- acc emph
Jayaseelan (2000) treats the second part of the complex anaphor as an emphatic
in Malayalam (DR), unlike the Telugu and Manipuri cases. Sentence (122) is
illustrative:
Malayalam (DR)
(122) rāmani tanni/∗ j - e tanne / awani/∗ j - e tanne wer.uttu
Raman self- acc emph he- acc emph hated
‘Ramani hated himselfi/∗ j .’
(Sentence (122) is a consolidated version of sentences (112)
and (117) on pages 141–142 of Jayaseelan 2000)
Note that the second part of the anaphor carries the emphatic marker -e.
Jayaseelan treats the second part of the anaphor as an emphatic, and not as
a reduplicated form (Jayaseelan 2000: 142).
3.15 Conclusion 91
3.15 Conclusion
This chapter first provides a brief description of the various forms of the
anaphor in each language family. SALs have a nominal simplex and complex
anaphor and/or a monomorphemic verbal anaphor (reflexive/reciprocal). The
complex form of the anaphor in some SALs is the result of reduplication
and case copying, and in others it is formed by compounding a simplex form
with an emphatic. As for the nature of the verbal reflexive, which is invariably
monomorphemic, we have shown that the VR, the passive and the anticausative
(inchoative) reduce the valence of a transitive predicate by one argument. The
verbal reciprocal, too, performs several other functions, such as group marker,
hortative marker and collaborative effort marker.
SALs present a gradation with respect to the nature of anaphoric binding.
The IA languages are on one end of the scale with their insistence on the
occurrence of the nominal anaphor, and the polysynthetic Munda languages
are on the other end of the scale with their insistence on the occurrence of the
verbal anaphor alone, while the Dravidian languages (except Malayalam and
Toda), most of the Tibeto-Burman languages and Khasi (Mon-Khmer) fall in
between with a nominal as well as a verbal device for anaphoric binding. The
verbal anaphor in Dravidian and Munda appears to be indigenous. However,
as Toda (DR) does not have a VR, and Toda retained most of the features of
Proto-Dravidian (Peri Bhaskararao, p.c.), it could be the case that the verbal
anaphor in the rest of the Dravidian languages is the result of a subsequent
grammaticalization process, and is not indigenous.
With regard to the VR in Tibeto-Burman, one might conjecture that early
Tibeto-Burman had the device, but in course of time it got lost in some lan-
guages. In-depth research is needed before any firm conclusion can be drawn.
Munda languages also use a nominal device borrowed from superstrate lan-
guages, and Korku (Munda) has totally lost the verbal device. It is to be noted
that Marathi (IA) and Sinhala (IA) have the VR, and it could be due to contact
with Dravidian. The occurrence of the VR in Gujarati functioning as a VR
and a passive marker, as in Munda languages, is hard to explain. It could be
an indigenous development. We have also shown that the Anaphor Agreement
Principle rules out the occurrence of nominal subject anaphors in Hindi-Urdu,
Kashmiri and Marathi, which have object agreement. However, the occurrence
of such anaphors in Mizo and Hmar (TB), which have subject as well as object
agreement, cannot be explained by the aforementioned Principle, and further
research needs to be done in this area.
A non-nominative subject can antecede an anaphor. However, a verbal
anaphor is not permitted in a non-nominative subject construction, as the pred-
icates in such constructions are inherently [–transitive], and, hence, the VR –
which is [–transitive] – cannot co-occur. SALs (except Marathi) obey the three
binding principles. Complex (polymorphemic) anaphors do not in general per-

mit long-distance binding (Marathi [IA] being the only exception) but long-
distance binding of a simplex anaphor is permitted. We have shown that the
presence of the VR blocks long-distance binding, as it ‘insulates’ the sen-
tence from coindexing with a long-distance antecedent. Thus, polymorphemic
anaphors and the VR share the common property of blocking long-distance
binding. We have shown that the antecedent–anaphor relationship remains the
same after scrambling in all of the SALs where this has been investigated here.
From the limited study on Hindi anaphors, it appears that scrambling is an
instance of A-bar movement only, not A-movement. An in-depth study of this
phenomenon in other SALs may shed light on this issue. SALs conform to
the generalization that Blocking Effects are found only in languages without
agreement, and a subcommanding NP can be an antecedent to an anaphor.
Finally, keeping in view Martin Everaert’s observation that the term
“anaphor” implies referential status, which implies being an argument / DP-
status, we feel that a better label to characterize the verbal anaphor may be
verbal anaphoric clitic.35
4 Case and agreement
4.1 Introduction
SALs exhibit a variety of agreement patterns in the verb, ranging from no
agreement on one hand to agreement with arguments (subject, direct object,
indirect object, etc.) and non-arguments (oblique objects) on the other. The
study of agreement in these languages becomes more fascinating coupled with
the fact that there is incorporation of arguments, non-arguments, quantifiers,
intensifiers, Negative Polarity Items operating in these languages. We shall
provide two examples to demonstrate the two ends of the spectrum: one from a
language with no agreement, and the other from a language with rich agreement.
In Manipuri (TB), the verb exhibits no agreement at all with any argument
(1), while in Hmar (TB), the verb exhibits rich agreement with the subject and
incorporates the Negative Polarity Item (NPI) la ‘yet’ (2).1
Manipuri (TB)
(1) əyhak cen-ø niŋ- ŋi
I hon run-inf want- [−fut]
‘I want to run.’
Hmar (TB)
(2) pro cainā ram hmun pou pou- ah ka- la- fe- nɔ
I China land place all- loc 1s- yet-(NPI) go- not
‘I have not yet gone to all the places in the land of China.’
The study of agreement has been one of the areas of interest in descriptive
grammars. However, in recent years, agreement in relation to case has received
considerable attention, and has become one of the focal areas of the Government
and Binding framework, and its offshoot, the Minimalist Program of Chomsky
(1995a, 1995b, 2000 and 2001). Agreement can be viewed as a syntagmatic
relationship that is established between a lexical element (a noun phrase or an
adverbial PP) and a functional category, such as tense or aspect. The agreement
markers that are manifested on the functional categories are in terms of Person,
Number and Gender, which are generally termed as the PNG or “phi-features.”
93
As we shall demonstrate, the lexical category that triggers agreement may be

postpositionally case-marked, or it may be in the nominative case and, thus,
it is parametric. As Benmamoun (2006: 141) puts it: “dependencies between
functional categories [such as Tense, Aspect] and the lexical elements [a noun
phrase or an adverbial PP], they enter into licensing or checking relations
[which] require c-command and/or specifier–head configurations. This is, for
example, the standard case of case and agreement.” In the Government and
Binding framework, case on an NP/DP plays a very significant role. Thus, each
DP must be either structurally case-marked or inherently case-marked. If a
DP is not case-marked, the derivation is rejected by the Case Filter (Chomsky
1981: 49). In this framework, structural case is assigned either by a functional
category, such as AGR (agreement), or by a lexical category, such as a transitive
verb or an adposition (preposition or postposition). The case assigner (AGR or
a verb/adposition) must govern the NP. Thus, “NP is nominative if governed
by AGR. NP is objective [accusative] if governed by V . . . transitive. NP is
oblique if governed by P [pre- or postposition]” (Chomsky 1981: 170).
Inherent case can be labeled as lexical case or semantic case. For example,
certain predicates assign a dative/genitive case to the DP in subject position, in
SALs and many other languages (see chapter 5 for details). Such assignment
depends on the nature of the predicate (verb or adjective), and is labeled inherent
case assignment.
In the Minimalist Program of Chomsky (1995b, 2000 and 2001), case is
not assigned. Lexical DPs carrying case features are ‘Merged,’ and case is
checked/valuated by the corresponding functional head. The structural case
features are subsequently deleted. In chapter 5 on non-nominative subjects, we
shall discuss inherent case assignment in the SALs. In this chapter we mainly
focus our attention on structural case and agreement in SALs.
In this chapter we provide a typology of agreement patterns in SALs, and
demonstrate that the occurrence of a lexical (postpositional) case marker blocks
the occurrence of an agreement marker on the predicate (verbs, adjectives, as
well as participles in some languages), and, hence, it is parametric. We shall
discuss adposition incorporation in applicative constructions in Tibeto-Burman
languages, and show that the incorporation of an adposition has a ‘transitivizing
effect’ on the [–transitive] host verb. We shall also discuss long-distance agree-
ment in Indo-Aryan, Austro-Asiatic, Dravidian and Tibeto-Burman language
families. Evidence from relative clauses and clefts from Tibeto-Burman will be
presented to show how the presence of the ergative case marker and agreement
directly correlate with each other.
This chapter is organized as follows. Section 4.2 deals with the typology
of agreement in SALs. Section 4.3 discusses adposition incorporation and its
effect on the transitive nature of the host verb. Section 4.4 discusses the nature
of polysynthesis found in Sora (Munda). Section 4.5 deals with cases involving
long-distance agreement (LDA). Section 4.6 focuses on the role of the principle
of Pronominal Strength Hierarchy in agreement. Section 4.7 deals with ECM
cases and LDA. Section 4.8 shows how a functional category such as agreement
of a language influences the patterns of agreement of a language belonging to
another family in language contact situations. Section 4.9 presents evidence
from Mizo and Hmar (TB) where the occurrence of ergative and agreement
markers is mutually dependent. The final section concludes the chapter.
4.2 Agreement typology of SALs

We shall provide a typology of agreement patterns in SALs to examine how the
presence or absence of a postpositional case marker affects agreement. SALs
exhibit four distinct patterns with regard to agreement:
Type A: Agreement features are ‘tied in,’ or linked to the bare NP in
nominative case. Agreement is blocked by the presence of a post-
positional case marker, as in most of the Indo-Aryan and Dravidian
languages.
Type B: Multiple sets of agreement features are manifested in the verb,
some of which are linked to the nominative, others to the dative,
genitive, locative and so forth. Thus, the presence of a lexical case
marker does not block agreement of that NP with the INFL of the
verb in another set of languages such as Kurmali, Maithili, etc.
Type C: The verb is either polysynthetic or polysynthetic-like in nature
as various types of arguments are represented in the verb, either
as clitics or as noun or postposition incorporation. The presence or
absence of a postposition marker is of no consequence for agreement
in languages such as Mizo, Hmar, Paite, etc.
Type D: There is no subject–verb agreement at all, as in most of the
Tibeto-Burman languages.
Finally, we demonstrate that the presence or absence of a lexical case marker
vis-à-vis agreement is parametric, and, thus, cannot be treated as a universal.
4.2.1 Postpositionally case-marked DP blocking agreement

In all Dravidian languages (except Malayalam) and in some Indo-Aryan lan-
guages (Hindi-Urdu, Punjabi and Kashmiri), the subject in the nominative case
triggers agreement. However, a postpositionally case-marked NP cannot trigger
agreement. When the subject is postpositionally case-marked ergative as in (3),
in Hindi-Urdu, the verb exhibits agreement with the theme (see chapter 5 for
details).
It might show default agreement (no agreement with any constituent in the
sentence) in Hindi-Urdu and Punjabi (IA):
(i) if the theme or patient is postpositionally case-marked with an accusative/

specificity marker as in (4);
(ii) when the passive subject and patient/theme are postpositionally case-
marked as in (5); and
(iii) in lexical passives where there is no overt subject as in (6).
In sentence (3), the theme is in nominative case, and the subject is ergative
case-marked; in (4) the patient is accusative case-marked with a specificity
marker.
theme–verb agreement
Hindi-Urdu (IA)
(3) lər.kõ nei yah kitābj par.h- ı̄∗ i/j
boys, m,p erg this book, f,s read- pst, f,s
‘The boys read this book.’
(Davison 1991: 2)
postpositional case marker blocking agreement
The patient lar.kı̄ ‘girl’ in (4) is accusative case-marked (Porizka 1963;

Pandharipande and Kachru 1976; Mahajan 1990; Davison 1991).
Note that the default agreement marker in Hindi-Urdu is homophonous with
the masculine singular agreement marker.
(4) rām nei lar.kı̄j ko dekh- ā∗ i/∗ j

Ram erg girl acc see- pst.def agr
‘Ram saw a girl.’
default agreement
The agent pulis ‘police’ and patient cor ‘thief’ in (5) are both postpositionally
case-marked in the passive in Hindi-Urdu.
(5) pulisi ke dwārā corj ko pakr.ā gayā∗ i,∗ j

police by thief acc was caught.def agr
‘The thief was caught by the police.’
In lexical/capabilitative passives, the verb is [–transitive], and the subject

is marked by the postposition se ‘by,’ and hence the verb exhibits default
agreement.
default agreement – lexical passives

(6) hami se yahā˜ bait.h-ā nahı̄˜ jā- tā∗ i
we by here sit-perf not go- imperf.def agr
‘We can’t sit here.’
When the subject is case-marked dative/locative, the verb agrees with the theme
as in (7) in Kannada (DR).
Kannada (DR)
(7) nanagei mūvaru hen.n.uj makkal.u iddāre∗ i/j
I.dat three female children be-pst.3 p.human
‘I have three daughters.’
(Sridhar 1990: 133)
In Tamil (DR), the verb exhibits default agreement in the dative subject con-
struction when the subject is dative case-marked, and the theme too is postposi-
tionally case-marked by a specificity marker (accusative). Default agreement is
due to the fact that the experiencer kumār ‘Kumar’ and the patient rājā ‘Raja’
in (8) are both postpositionally case-marked, and hence agreement is blocked.
Thus, the predicate pit.ikk ‘like’ exhibits default agreement; hence, there is no
agreement with any NP in the sentence. The default agreement marker in Tamil
is in the neuter.
theme in the accusative

Tamil (DR)
(8) kumār- kku rājāv- ai.p pit.ikk- um
Kumar- dat Raja- acc like- 3p.s.neuter, def agr
‘Kumar likes Raja.’
(Lehmann 1989: 184)
4.2.2 Postpositionally case-marked DP triggering agreement

In some languages, the presence of a lexical case marker does not block agree-
ment of the case-marked NP with the INFL of the verb. Thus, the presence
of a lexical case marker is of no consequence to verb agreement. There are
two subsets in this type. In subset 1 the verb exhibits agreement only with one
DP; in subset 2 the verb may exhibit agreement with two or three arguments
(DPs). Gujarati (IA) belongs to subset 1 and Maithili and Kurmali (IA) belong
to subset 2.2
Subset 1
In Gujarati, a postpositionally case-marked direct object triggers agreement
(Mistry 1998) as in (9). Ergative case-marked subjects cannot trigger agree-
ment. Thus, in (9), though both the ergative subject and the direct object are
postpositionally case-marked, it is the direct object (patient) that triggers agree-
ment, and not the subject.
Gujarati (IA)
(9) kišori - e rājj ne pajav- y- o∗ i,j
Kishor- erg Raj, m acc harass- perf- m
‘Kishor harassed Raj.’
(Mistry 1998: 426)
Sentence (9) demonstrates that there is subject–object asymmetry with regard

to agreement in Gujarati.
The light verb in Gujarati may also exhibit agreement with the theme madad
‘help,’ which is nominative case-marked as in (10). While a case-marked direct
object can trigger agreement, a case-marked subject cannot.
(10) kišori - e rājj ne madadk kar- y- i∗ i,∗ j,k

Kishor- erg Raj,m acc help,f do- perf- f
‘Kishor helped Raj.’
(P. J. Mistry, p.c.)
According to Mistry (1998), Gujarati (IA) has the marker -ne, which functions
as an object marker as well as a specificity marker. The object marker acts as a
barrier to the verb agreement in (11), just as in Hindi-Urdu and Punjabi (IA),
while the specificity marker -ne does not in (9).
(11) kišor- (n)e3 kāgal ne ad. - v- ũ ha-

Kishor- ergative letter object mkr touch- desiderative- def agr be-
t- ũ
perf- def agr
‘Kishor wished to touch the letter.’
(Mistry 1998: 429)
Subset 2
Maithili (IA) and Kurmali (IA) are spoken in Bihar, northern India, where
the verb exhibits agreement in honorificity with an NP that is postpositionally
case-marked. We shall first discuss the case of Maithili (IA). The verb exhibits
agreement with indirect object ahā˜ ‘you’ dative case-marked by ke, in (12).
postpositionally case-marked io
Maithili (IA)
(12) hami ahā˜-kẽj pāik de- l- ahũj
I you.hon-dat money give- pst- agr.hon
‘I gave you money.’
(Singh 1979)4
Unlike in most of the SALs, the verb exhibits agreement in honorificity with a
dative subject – hun-kā ‘he’ in (13) – or with a passive subject as in (14).
agreement with dative subject

(13) hun- kāi bhukh lag- l- ain(h)i,∗ j
he 3.hon- dat hunger feel- pst- 3.hon [non-nominative]
‘He felt hungry.’
(Yadava 1999: 145; subscripts have been added)
agreement with passive subject

(14) hunkā- sai i cit.hij likh- al ge-
he 3.hon- by this letter write- perf pple neutral passive.AUX-
l- ain(h)i,∗ j
pst- 3.hon
‘This letter was written by him.’
(Yadava 1999: 146; subscripts have been added,
and glosses slightly modified)
In Maithili, the verb exhibits agreement with an adjunct (an ablative PP) which
carries an overt lexical case marker as in (15). The ablative PP in (15) is
hamara sɔ∼ ‘from me,’ and the verb agrees with the ablative object, which is
a subcategorized argument.
agreement with an ablative pp

(15) oi ı̄ vastujatj hamarā sɔ∼k le- l- aka∗ i,∗ j,k
he these things me from take- pst- agr
‘He took these things from me.’
(Singh 1979)
Moreover, in Maithili (IA), there are instances of triple agreement markers on

the verb, as Yadava (1999) reports. In (16), the verb exhibits agreement with
(i) the subject ham ‘we’; (ii) tohar ‘your’; the possessor of the direct object,
and (iii) the direct object bābuji ‘father.’ The subject agreement marker is -i,
the marker for the possessor of the direct object is -au, and the direct object
agreement marker is -nh in (16). Note that agreement with possessor is in terms
of honorificity.
(16) hami toharj bābu-jik - ke dekh- al- ii - auj - nhk

I.nom your 2.midhon father- 3.hon acc see- pst- 1- 2.midhon- 3.hon
‘I saw your father.’
(Yadava 1999: 149)
Yadava (1999: 149) comments: “This triple agreement, is, however, heavily
constrained, in that it permits only the combination of 1.nh [non-honorific] and
2.mh [mid-honorific] and 3.h [honorific] referents.”
To summarize, the following are the characteristic features of Maithili
agreement:
(i) though there is no overt ergative marker, the verb exhibits agreement with
direct object, and an indirect object, and the possessor of direct object too;
(ii) unlike in most other SALs, the verb exhibits agreement with a dative
subject, a passive subject, an object and an ablative postpositional phrase;
(iii) the verb exhibits triple agreement markers, a phenomenon rarely found in
SALs.
The verb exhibits long-distance agreement too, with the possessor of the direct
object of the embedded clause (see section 4.5).
We shall now discuss the case of Kurmali, an Indo-Aryan language spoken
in Bihar in northern India, in which the verb exhibits multiple agreements, with
the possessive pronoun of the subject (possessor) and the subject as well. In
(17) below, the 2nd person possessor agreement marker is -un ‘2 plural,’ and
of the subject is -o ‘3 singular.’
possessor agreement
Kurmali (IA)
(17) toharj kalam-tii gir- l- oi - unj
your.gen.2p pen.def.nom.3 s.f- fall- pst- 3s- 2p
‘Your (2p) pen fell down.’
(Mahto 1989)
The verb in Kurmali also exhibits agreement with the subject and a dative
case-marked indirect object as in (18). The subject agreement marker in (18)
is -o and the indirect object marker is -u. The subject and the indirect object
trigger agreement though thay are lexically case-marked.
indirect object agreement

(18) rāmi - e to-kej ekti chhuā-ø de- l- oi - uj
Ram- nom you-dat.2s one girl-acc.3s.f give- pst- 3s- 2s
‘Ram gave you (singular) a girl.’
(Mahto 1989)
It is worth mentioning here that in Magahi, an eastern Indo-Aryan language like

Kurmali and Maithili, the verb exhibits multiple agreement markers. In (19),
the verb agrees with the subject ham ‘I’ and the DO to(h)ra ‘you’ (honorific).
Note that the subject and DO in (19) are not postpositionally case-marked.
Magahi (IA)
(19) ham to(h)rā dekh- l- i- o
I you (h) see- pst- 1s- 2p
‘I saw you (honorific).’
(Verma 1991: 132)
From a typlogical point of view, Indo-Aryan languages discussed thus far fall
into two major categories with regard to agreement: languages such as Hindi-
Urdu, Punjabi, Marathi, Gujarati come under Category I where the verb agrees
only with one argument, and languages such as Maithili and Kurmali come
under Category II where the verb exhibits agreement with two, or sometimes
even three, arguments.5
In Shina of Skardu, an Indo-Aryan language too, the verb agrees with the
dative case-marked experiencer.
Shina of Skardu (IA)

(20) mo- rei tho mus.j amot.- usi,∗ j
me- dat your word forgot- 1s
‘I forgot your word.’
(Hook 1990: 78)
Unlike in most of the IA languages, in Shina, an ergative subject may alternate

with a dative subject, and the verb exhibits agreement with the subject in the
ergative or dative case.
(21) mu- sui / mo- rej brak pasham- usi,j

I- erg I- dat mountain see- 1s
‘I see the mountain.’
(Hook 1996: 172–174)
A similar pattern is also found in obligative constructions in Hindi-Urdu (IA)

spoken in and around Delhi.
In Kashmiri (IA), a postpositionally case-marked NP can trigger agreement.
The verb exhibits agreement with the ergative case-marked subject tse ‘you
(2s)’ and the absolutive case-marked direct object. The markers respectively
are -th and -u.
Kashmiri (IA)
(22) tsei von- uj - thi timan apuzj
you.erg told- m,s- 2s them.dat lie.abs.m,s
‘You told them a lie.’
(Wali and Koul 1997: 250)
The ergative case-marked subject in 1st and 3rd person does not trigger agree-
ment. However, if it is pro-dropped, the verb must exhibit agreement with the
pro-dropped subject (see Wali and Koul 1997: 250). In Punjabi, “(I)n perfec-
tive tense the 1st and 2nd person subjects are not overtly marked with the
postposition ne” (Bhatia 1993: 170).
4.2.3 Polysynthetic and semi-polysynthetic languages

The third group of languages falls into two major categories: polysynthetic and
semi-polysynthetic languages. Polysynthetic languages are those where “verbs
must include some expression of each of the main participants in the event
described by the verb (subject, object and indirect object)” (Baker 2001: 111).
In this section, we first present evidence from Sora, a Munda language spoken
in Orissa, which is polysynthetic in nature. We then discuss the nature of
incorporation in Sora in terms of the directionality of the incorporated element.
We then present evidence from another set of languages which are not fully
polysynthetic, but they exhibit the features of polysynthesis. The verb may or
may not include an expression of each of the main participants obligatorily in the
events described by the verb. We label them as semi-polysynthetic languages.
These include some Tibeto-Burman and Munda languages, and the Mon-Khmer
Khasi. We shall in this context discuss the nature of adposition incorporation
in Hmar and Paite (TB).
The polysynthetic Sora language

In Sora (Munda), in (23), the complex predicate pokunpunŋ ‘stab.belly.knife’
contains the verb stab, the locative PP belly and the instrumental object knife.
Note that all the three parts of the predicate constitute a single lexical item, a
phenomenon unique in Sora amongst the languages of the subcontinent. The
direct object is also incorporated into the verb as the second singular pronominal
clitic am. It is followed by the past tense marker tēn. All these segments together
constitute the total verbal complex.
do, locative and instrumental pp incorporation

Sora (Munda)
(23) pokunpunŋ- am- tēn?
stab.belly.knife- 2s- 3.pst
‘(Who) stabbed you in the belly?’
This is a unique phenomenon found only in Sora in the subcontinent. There

are very few polysynthetic languages in the world. These include Mohawk and
Wichita (Caddoan); Southern Tiwa and Jemez (Tanoan); Nahuatl (Nahuatlan);
Mayali, Nunggubuyu, etc. (Gunwinjguan); Chukchee and Koryak (Paleosi-
berian); Mapuche (Mapuche); and Ainu (Ainu) (Baker 2001: 115).
We provide another example from Sora in which the direct object bɔŋtεl
‘buffalo’ is incorporated into the verb. Nouns, when incorporated, have con-
tracted forms in Sora. Thus, in (24), bɔŋtεl’s shorter form is bɔŋ, which
occurs to the right of the verb. Recall that in Khasi (Mon-Khmer) too – a
non-verb-final language, unlike Sora – the incorporated noun occurs to the

right of the verb.6 As we shall show later, this is due to the fact that Sora
was at one time verb-medial, and in course of time changed to verb-final. The
3rd person plural pronominal subject is incorporated into the verb as a pronom-
inal clitic -ji, and the DO as the third singular clitic -ε in (24).
Sora (Munda)
(24) proi jom- bɔŋj - t- εj - n- jii pɔ?
(they) eat- buffalo- [-pst]- 3s- [−tr]- 3p pol q mkr
‘Will they eat the buffalo?’ or ‘Do they eat buffalo?’
The semi-polysynthetic languages

In this group of languages, pronominal clitics occur when they are coindexed
with postpositionally case-marked NPs. However, the occurrence of a postposi-
tion with an NP is not the deciding factor for the occurrence of the clitic with the
verb. The languages of the Tibeto-Burman and Munda families belong to this
category. These are the languages in which a noun or verb or an adposition is
‘incorporated’ into the verb, and the incorporated element and the verb together
form a complex predicate. Thus, it is the verb that carries a noun or an adpo-
sition. Indo-Aryan and Dravidian languages permit verb incorporation where
the causative suffix is ‘incorporated’ into the verb, thus forming a complex
morphological category. Baker (1988, 1996) provides a detailed discussion of
noun, verb and preposition incorporation. We shall provide data to show that, in
Tibeto-Burman and Munda languages, direct object, indirect object and oblique
object may not be overtly present, and they are incorporated into the verb as
pronominal clitics.
(i) Pronominal clitics in Tibeto-Burman

Languages such as Mizo, Hmar, Aimol, Thadou, Zou and Paite of the Tibeto-
Burman language family exhibit split ergativity in person. The ergative case-
marked subject in Hmar – lali in (25) – triggers agreement. The object
agreement markers for 1st person and 2nd person singular in Hmar (TB) are
-mi and -ce, respectively. The subject agreement marker is -a. The object is
pro-dropped, and is indicated by pro.
Hmar (TB)
(25) lalii - n proj ai - mij - hmu
Lali- erg me 3s- 1s.DO- met
‘Lali met me.’
The occurrence of the indirect object i-koma ‘for you’ is optional in (26), and
the pronominal clitic -ce ‘2 singular’ is encoded in the verb.7
(26) lalii - n (i- komaj ) lekhabu ai - pek- cej

Lali- erg you- for book 3s- give- 2s.IO
‘Lali gave you a book.’
(ii) Pronominal clitics in Munda

Direct object and indirect object (recipient) are encoded in the verb in Munda
languages.
(iii) Direct object incorporation in Ho (Munda)

In Ho, the direct object is pro, and the clitic -iŋ for 1st person singular direct
object is incorporated to the right of the verb. Note too that the adverb sida
‘first’ (in italics in (27)) is also incorporated into the verb tam ‘hit’ to its right.
Ho (Munda)
(27) birsa proi tam- sida- ke- d.- i ŋi - a
Birsa (me) hit- first- pst- [+tr] 1s.DO- decl
‘Birsa hit me first.’
(Burrows [1915] 1980: 89)
The pronominal clitic -iŋ of the indirect object is incorporated into the verb in
(28) in Gorum (Munda).
(28) mo- taʔy- iŋ

2p- give- 1p
‘You gave me.’
(Aze 1973: 249)
(iv) Possessor incorporation: Incorporation of the possessor

of object (theme) and indirect object (recipient)
In Munda and Tibeto-Burman languages, the verb carries the agreement marker
of the possessor of the object alone, not of the subject: “In Sora, possessor
raising is frequently found in such incorporated constructions” (Anderson 2007:
191).
In (29), the theme (direct object) jēŋ as well as the possessor of the theme
-am is incorporated.
Sora (Munda)
(29) pro lēm- jēŋ- t- am
(I) bow- legs- [-pst]- your
‘I bow to your legs.’
In Ho (Munda), the possessor of the recipient is incorporated. In (30), am- ote

‘your field’ is the recipient, which is [−animate]. The possessor of the recipient
is incorporated as -mi in the verb.
Ho (Munda)
(30) jemai aye- ʔa ote ria daʔa amaj ote- ei - em- ad.-
Jema he- gen field gen water your field- 3s- give- pst perf-
mij - a
2s- fin
‘Jema had given his field’s water to your field.’
In Santali (Munda) too, the possessor agreement marker is incorporated and

it “always refers to the object” (Macphail 1983: 41), and not to the subject.
The direct object kuri ‘girl’ in (31) is a referential expression. The possessor
of the direct object -
appears to the right of the verb dal ‘beat.’ Though the
possessive marker -
may ambiguously refer to the subject kora ‘boy’ too, it
is coindexed only with the direct object, and not with the subject, which shows
subject–object asymmetry.
Santali (Munda)
(31) kora- dɔ kuri- (y)e dal- ke- d- e- ti-
- a
boy- topic girl- 3s beat- pst- [+tr]- 3s- gen- 1s- fin
‘The boy beat my girl.’ (*‘My boy beat the girl.’)
(Macphail 1983: 41)
It is the possessor of the object that is incorporated in Tibeto-Burman languages

too. In Hmar (TB), the possessor of the direct object and not of the subject is
incorporated.
Hmar (TB)
(32) zova- n ka- kut a- mi- sɔ:p- pek
Zova- erg my- hands 3s- 1s- washed- o ben
‘Zova washed my hands.’
(Subbarao 1998b)
Thus, in Munda and Tibeto-Burman, there seems to be a restriction on the

incorporation of the possessor of the subject and it is always the possessor
of the direct object that is incorporated. That is, the possessor of the direct
object takes precedence over possessor of the subject in the incorporation of
pronominal suffixes in munda and Tibeto-Burman. In contrast, in the Indo-
Aryan Kurmali, it is the possessor of the subject as well as of the direct object
that is incorporated (see (17), for example).
(v) Incorporation of quantifiers, intensifiers, Negative Polarity

Items, inclusive particles in Hmar, Mizo and Paite (TB) and
Khasi (Mon-Khmer)
Before we proceed further, it is essential to distinguish between agreement and
incorporation. As we mentioned in the introduction to this chapter, agreement
can be viewed as a syntagmatic relationship that is established between a lexi-
cal element (a noun phrase or an adverbial PP) and a functional category, such
as tense or aspect. Hence, it reflects “dependencies between functional cate-
gories” (Benmamoun 2006: 141) and the DPs they agree with. Thus, the DP
and the agreement suffix are in a “c-command and/or spec–head configuration”
(Benmamoun 2006: 141). Incorporation, in contrast, includes Grammatical
Function (GF) changing processes such as Passive, Antipassive and Possessor
Raising, which are “properly analyzed as subclasses of Noun Incorporation”
(Baker 1988: 22). According to Baker, it involves a “particular type of move-
ment” (1988: 22). A causative affix or an adposition or a passive morpheme
may also be incorporated into the verb, consequently creating a syntactic link
between two positions in the phrase marker, and the moved element leaves a
trace (t).
In TB languages the negative, the Negative Polarity Item (NPI) and a quan-
tifier can occur as part of the verbal auxiliary system. In Khasi, an NPI occurs
as part of the agreement phrase. The occurrence of a quantifier and intensifier
as a part of the verbal system too is unique to TB languages such as Hmar,
Mizo and Paite. All such cases involve incorporation of the specific element
into the verb and the element leaves a trace in the original position from where
it moves.
Quantifiers In Hmar, the quantifier vɔŋ ‘everything’ refers only to

[−animate], while the quantifier sεm ‘all’ refers only to [+animate], entities.
Hmar (TB)
(33a) pro án- fāk- vɔŋ
(they) 3 p- eat- everything
‘They eat everything.’
(33b) pro án- fāk- sεm/seŋ
(they) 3 p- eat- all
‘All of them ate.’
Negative and NPIs In many Tibeto-Burman languages, the negative occurs

to the right of the verb, and the Negative Polarity Item (NPI) may occur either
to the right of the verb as in (34), or to the left of the verb as in (35).
Hmar (TB)
(34) pro tājmahal i- hmu- tā- am
(you) Taj Mahal 2s- see- ever (NPI)- pol q mkr
‘Have you ever seen the Taj Mahal?’
(35) pro cainā ram hmun pou pou- ah ka- la- fe- nɔ
(I) China land place all- loc 1 s- yet (NPI)- go- not
‘I have not yet gone to all the places in the land of China.’
Note that the negative c-commands the NPI hierarchically, as it occurs to the
right of the NPI in (35).
Adverbs such as ve ve ‘together, both’ and inclusive particles such as ve
‘too, also’ occur to the right of the verb. In (36) and (37), huŋ- .tšaŋ ‘come and
participate’ is an instance of complex verb formation in Hmar (TB).
Hmar (TB)
(36) pro an- huŋ- .tšaŋ- ve ve
(they) 3 p- come- participate- together
‘They came and participated together.’
(37) nuŋhāk āpui a- huŋ- .tšaŋ- ve

Ms. Apui 3 s- come- participate- too/also
‘Ms. Apui also came.’
Jowai, Pnar and Rymbai (Mon-Khmer), which are also verb-medial, are con-
sidered as “dialects of Khasi” (Mon-Khmer). An adverb is incorporated into
the verb in these languages too.8 Similar incorporation of nouns, adverbs and
several other elements is found in Savara/So-ra (Munda), spoken in Orissa.
Thus, though Savara/Sora belongs to the Munda family, and Jowai, Pnar and
Rymbai belong to the Mon-Khmer family, incorporation is shared between
the languages of these two subgroups and, hence, we consider it an inherited
feature in these languages.
In Manipuri and Bodo (TB), and Khasi (Mon-Khmer), the adverb ‘together’
is incorporated in the verb (see sentences (93) from Manipuri, and (87) from
Khasi, in chapter 8).
In Mizo (TB), the emphasizer adjunct ‘certainly’ is incorporated.
Mizo (TB)
(38) vɔin.ah zova- cu sikul- (ah) a- kal- ŋei- aŋ
today.adv Zova- foc school- to 3 s- go- certainly- fut
‘Zova certainly will go to school today.’
(Subbarao and Lalitha Murthy, ms)
In Khasi (Mon-Khmer) too, the Negative Polarity Item (NPI) pat ‘yet’ is incor-
porated into the verb to the right of the negative.
Khasi (Mon-Khmer)
(39) ka-kayt [ba- la- bām u-khnna?] ka-m- pat- iʔ
f-banana adjr- pst- eat m-child 3 f,s-neg- yet(NPI)- ripe
‘The banana that the boy ate was not yet ripe.’
Since the negative occurs to the left of the NPI in (39), it c-commands the NPI.
We have demonstrated that Munda and Tibeto-Burman languages permit
incorporation of an object and indirect object and possessor of direct and indi-
rect objects; also of quantifiers, Negative Polarity Items and inclusive particles.
Such incorporation is not permitted in Dravidian and Indo-Aryan languages.
Thus, the incorporation of various types of constituents in the verb is a special
feature of the Mon-Khmer, Munda and Tibeto-Burman language families, a
feature not found in Dravidian and Indo-Aryan languages.
4.2.4 Languages with no agreement

There are some Tibeto-Burman languages such as Bodo, Kokborok, Manipuri,
Rabha and Tangkul, in which the verb does not exhibit agreement with any
constituent in the clause.
Rabha (TB)
(40) aŋ/ ciŋ/ naŋ/ u naŋ- pake reŋ- a sɔyra- jɔ
I we you he/she you- with go- purposive promise- pres perf
‘I/we/you/he/she have/has promised to go with you.’
(Subbarao et al., ms)
In many of these languages there is a nominative marker that occurs with the
subject, which is generally optional.
To summarize the above discussion, it is the lexical case marker (P-case or
postpositional case) that blocks agreement in the first group of languages. It is
assigned either structurally, for example the accusative lexical case marker ko in
Hindi-Urdu or ni in Telugu, or the ergative ne in Hindi-Urdu. It may be a lexical
(inherent) case marker as the dative on indirect objects or the dative subject or
the locative subject. In cases in which both the agent/experiencer/possessor and
the theme/patient are case-marked with a postposition, the predicate exhibits
default agreement.
In the second group of languages, a postpositionally case-marked NP can
trigger an agreement.
In the third group, in the first set of languages, in a polysynthetic language
like Sora, arguments are incorporated into the verb, a phenomenon unique to
Sora among SALs. In the second set of languages, the agreement is manifested
as a clitic, and the occurrence of a postposition is of no consequence for an
agreement, as is the case in Tibeto-Burman languages. Though the subject
4.3 Adposition incorporation 109
is ergative case-marked in Mizo, Hmar and Paite (TB), it can trigger an agree-
ment, and the agreement marker is generally a clitic. In this group of lan-
guages, incorporation of an object, indirect object, possessor of direct and
indirect objects and also of quantifiers, intensifiers, Negative Polarity Items
and inclusive particles takes place.
In the fourth group of languages, there is no agreement at all.
The Tibeto-Burman family of languages exhibits varied agreement patterns –
from very rich agreement to none, a phenomenon found in no other South Asian
language family.
In this section, we have presented the various types of agreement patterns
found in SALs. In the following section, we shall consider a specific case of
incorporation of adpositions in Hmar (TB), and demonstrate that such incor-
poration affects the transitive nature of the host verb.
4.3 Adposition incorporation

In this section, we shall discuss adposition incorporation in Hmar (TB), and
show that when an adposition is incorporated, an intransitive host verb is
transitivized. Adposition incorporation is a unique phenomenon found in select
Tibeto-Burman languages such as Hmar, Mizo and Paite. In Hmar, an ergative
marker -n/an occurs with the subject when the verb is [+transitive]. The verb
.tšɔŋ ‘sit’ is [–transitive], and, hence, the subject in (41) is nominative case-
marked, and the nominative case marker in Hmar is null. The oblique object is
comitative case-marked by -le.
Hmar (TB)
(41) lali kei- le cou a- n-9 .tšɔŋ
Lali (nom) I- with only 3s- VR- sit
‘Lali sat only with me.’
(42) lali naŋ- le cou a- n- .tšɔŋ

Lali (nom) you- with only 2s- VR- sit
‘Lali sat only with you.’
In (43) and (44), the pronouns kei ‘I’ and naŋ ‘you’ do not carry any post-
position, as the postposition le ‘with’ is incorporated into the verb. The verb
instead carries the marker pui ‘with.’ Note that le and pui are allomorphs whose
position of occurrence is with the DP and the verb stem, respectively.
p o s t p o s i t i o n le ‘with’ i n c o r p o r a t e d i n t o t h e v e r b
(43) lali- n kei cou a- mi- n- .tšɔŋ- pui
Lali- erg I only 3s- 1s- VR- sit- with
‘Lali sat only with me.’
(44) lali- n naŋ- cou a- n-10 .tšɔŋ- pui- ce

Lali- erg you- only 3s- VR- sit- with- 2 s
‘Lali sat only with you.’
In sentences (43) and (44), the following are worth noticing:

(i) the subject carries the ergative case marker -n,
(ii) 1st person pronominal clitics occur to the left in (43), and 2nd person to
the right of the verb .tšɔŋ ‘sit’ in (44), and
(iii) the pronouns kei ‘I’ in (43) and naŋ ‘you’ in (44) are accusative case-
marked, and they are not comitative PPs, in contrast to in (41) and (42).
Thus, the [–transitive] verb of (41) and (42) is ‘transitivized’ in (43) and (44)
after the incorporation of the comitative postposition le. Since the pronouns kei
‘I’ and naŋ ‘you’ are in derived direct object position in (43) and (44), by virtue
of not having the postposition le ‘with,’ the pronominal clitics mi ‘1 singular
clitic’ in (43) and ce ‘2 singular clitic’ in (44) occur with the verb.
The data in Hmar (TB) can be neatly explained in Baker’s (1988) proposal
concerning incorporation. An X zero (X0 ) level category moves to another X
zero level category to form a morphologically complex category which again is
an X zero level category. Such movement, according to Baker (1988), creates a
syntactic link between the two positions in the phrase marker. Baker’s argument
is that such a process, which is usually treated as a grammatical function (GF)
changing process, can be reduced to incorporation, and hence, there is no need
to have any grammatical function changing rules such as dative shift, passive,
causative, antipassive, etc. Baker discusses three types of incorporation: noun
incorporation, verb incorporation and preposition incorporation. Preposition
incorporation is a process in which a preposition moves to a verb to form a
complex X zero level category. We cite an example of adposition incorporation
from Kinyarwanda, a Bantu language (Kimenyi 1980), as quoted in Baker
(1988: 238). The verb, the preposition in (45) and the preposition incorporated
in (46) are in italics. In (45), the preposition n’i ‘with’ occurs with the NP
ikaramu ‘pen’ and the verb andik ‘write’ occurs alone. In (46), the NP ikaramu
‘pen’ does not have any preposition preceding it. In contrast, the verb is followed
by the adposition iish ‘with.’ The glosses in (45) and (46) have been slightly
modified.
Kinyarwanda (Bantu)
(45) Umwālimui ai - ra- andik- a ibaruwa n’i- ikaramu
teacher agr- pres- write- a letter with- pen
‘The teacher is writing a letter with a pen.’
(46) Umwālimu a- ra- andik- iish- a ibaruwa ikaramu
teacher agr- pres- write- with- asp letter pen
‘The teacher is writing a letter with a pen.’
(Baker 1988: 238)
Let us now go back to the case of postposition incorporation in Hmar. The

postposition le ‘with’ occurring with the DP kei ‘I’ in (41) in Hmar is an X0
level category. In (43), it moves to the verb .tšɔŋ ‘sit’ which is another X0 level
category, and has the form pui ‘with,’ when it occurs with the verb. Thus, le
and pui are allomorphs of the same morpheme. As mentioned previously, once
incorporation takes place, it ‘transitivizes’ an intransitive verb and, hence,
an ergative marker occurs with the subject lali ‘Lali’ in (43) and the object
agreement pronominal clitic with the verb. There are other Tibeto-Burman
languages, such as Mizo (Subbarao and Lalitha Murthy, ms) and Paite, which
exhibit postposition incorporation.
We provide another example of postposition incorporation from Paite (TB).
The PP .tšut.na-aʔ ‘chair-on’ is locative case-marked in (47). In contrast, .tšut.na
‘chair’ is accusative case-marked in (48), and the subject lali ‘Lali’ is ergative
case-marked. The verb .tšut. ‘sit’ in (47) is [−transitive]. In contrast, the predicate
.tšut.-piʔ ‘sit-on’ in (48) is an incorporated structure, and it is [+transitive]. That
is, postposition incorporation in (48) transitivizes the intransitive verb .tšut. ‘sit.’
Thus, aʔ ‘on’ and piʔ ‘on’ are allomorphs whose position of occurrence is
different: aʔ ‘on’ occurs with the noun phrase; piʔ ‘on’ occurs with the verb
stem.
p o s t p o s i t i o n -aʔ ‘o n’ o c c u r r i n g w i t h t h e np .tšut.na ‘c h a i r ’
Paite (TB)
(47) lali t.šut.na- aʔ a- t.šu
Lali (nom) chair- on 3s- sat
‘Lali sat on a chair.’
p o s t p o s i t i o n -aʔ ‘o n’ i n c o r p o r a t e d i n t o t h e v e r b .tšut. ‘s i t ’ a s piʔ

(48) lali- n t.šut.na a- t.šu- piʔ
Lali- erg chair (acc) 3s- sat- on
‘Lali sat on a chair.’
(Khamkhochinthang Khuptong, p.c.)
Thus, postposition incorporation is a phenomenon unique to select Tibeto-

Burman languages, found in no other language family of the subcontinent.
4.4 Incorporation in polysynthetic languages

We shall now examine some cases of incorporation discussed in Baker (1996) in
relation to the polysynthetic nature of Sora (Munda). Recall that Sora (Munda)
is a polysynthetic language in which incorporation of nouns and postpositions
takes place. Baker (1996: 15) states the polysynthesis parameter as follows:
(49) The polysynthesis parameter

Every argument of a head element must be related to a morpheme in the
word containing the head.
A head element is an X0 category in the X-bar system. The word morpheme
refers to agreement morphemes (or pronominal affixes) and incorporated
roots.
According to Baker (1996) the theta criterion is met morphologically rather

than syntactically, as it is the agreement markers alone that can satisfy the theta
criterion.
(50) The theta criterion

For every theta role there must be one and only one argument; for every
argument there must be one and only one theta role
(Chomsky 1981)
According to Baker (1996: 29), in incorporation “the moved morpheme adjoins

to the left of the target morpheme [italics in the original].” This is due to the
fact that, if an element X is adjoined to an element Y, then X must precede Y.
Following Kayne’s generalization that “an adjoining head (S) will invariably
precede the head that it adjoins to” (Kayne 1994: 38), Baker proposes the
following generalization as a universal.
(51) Adjunction
If X and Y are X0 categories and X is adjoined to Y in the syntax, then X
precedes Y in linear order.
(Baker 1996: 29)
The case of Sora (Munda) incorporation becomes interesting in this case. Baker
(1996: 29) cites data from Sora – which permits noun incorporation, as we have
already seen – and the incorporated noun occurs to the right of the verb, not to
the left as the above generalization claims. Let us now examine how this can be
explained. Sentence (52) is a normal structure in which no incorporation has
taken place.
Sora (Munda)
(52) bɔŋtεl- ən- ədɔŋ jom- t- ε- ji pɔʔ
buffalo- noun suffix- acc eat- [-pst]- 3 s- 3p, sub mkr pol q mkr
(Baker 1996: 32)
In (53), the direct object bɔŋtεl ‘buffalo’ is incorporated into the verb. Nouns
when incorporated have a contracted form in Sora. Thus, bɔŋtεl’s shorter form
is bɔŋ in (53).
(53) jom- bɔŋ- t- ε- n- ji pɔʔ

eat- buffalo- [-pst]- 3 s- [−tr]- 3 p, sub mkr pol q mkr
(Baker 1996: 32)
According to Kayne’s generalization in (51), the incorporated element must

occur to the left of the verb, as it must precede the adjoined category. Note
that the object bɔŋ ‘buffalo,’ an X0 level category, in (53) occurs to the right
of the verb, another X0 level category,after incorporation, contrary to Kayne’s
generalization.
Another interesting fact is that, in verb-final languages, in desiderative con-
structions with an object complement, the matrix verb (higher predicate) follows
the embedded verb, as in Hindi-Urdu (IA) and all other SALs, except in Khasi,
which is verb-medial. The desiderative verb cāh ‘wish’ in Hindi-Urdu follows
the embedded verb in (54).
Hindi-Urdu (IA)
(54) ham lad.d.ū khā- nā cāh- te hãı̃
we sweets eat- inf wish- imperf pres.p
‘We wish to eat sweets.’
Though Sora is a verb-final language, the higher predicate morpheme ‘want’

precedes the verb – contrary to the expected pattern – as in (56), just as is the
case in non-verb-final languages such as English and French.
(55) The children want to eat hamburgers.
Sora (Munda)
(56)
εn əd- məl- jom- jεl- yɔ- aj- t- en- ay
I neg- want- eat- meat- fish- all- [-pst]- [−tr]- 1s, sub mkr
‘I don’t want to eat all the fish.’
(Baker 1996: 32)
To recapitulate, we have two issues that deserve an explanation.

(i) Why does the incorporated element occur to the right of the adjoined
element; and,
(ii) why is the order of verbs in desiderative constructions in Sora similar to
the one found in non-verb-final languages?
The late Stanley Starosta (p.c.) comments: “The real explanation is, I think,
diachronic: Munda languages developed some of their morphology, while they
were still SVO, and then did a syntactic shift to SOV in South Asia; then added
some more morphology.” Anderson (2007: 25) points out that evidence from
the archaic process of noun incorporation found throughout South Munda and
comparative evidence suggests that “SVO or even VSO may have been the
original clausal constituent order in pre-Proto-Munda.”
Starosta’s comment gains support from the fact that in (56) in Sora, the
quantifier aj ‘all’ is incorporated in the verb to its right, just as in verb-final
languages such as Mizo and Hmar (TB), in which quantifiers are incorporated
to the right of the verb (see (33a and b) in Hmar). The agreement morphol-
ogy also follows the verb in Sora in (56). Lehman (1973: 57) concludes that
Proto-Austro-Asiatic was VO and non-agglutinative in morphological struc-
ture. Donnegan and Stampe (2004) provide further evidence to show that Sora
had “VO” (head-first) structure.
Based on such facts, Baker (1996) concludes that the order of morphemes
in Sora is a mirror image of that found in polysynthetic languages. Therefore,
“the direction of adjunction performed by head movement is simply reversed
in Sora” (Baker 1996: 33).
Baker (1996) assumes that most of the polysynthetic languages are head-
initial: SVO-type languages. He points out that polysynthetic languages such
as Ainu and Kiowa which permit incorporation have head-final tendencies.
Baker’s assertion gains support from Starosta’s comment that Sora, a Munda
language, which is polysynthetic in nature, was at one time non-verb-final,
and, hence, incorporation is facilitated. Though it has some head-initial ten-
dencies, it has verb-final tendencies too, which include agreement morphology
and incorporation of quantifiers to the right of the verb as in some verb-final
languages. Thus, a diachronic explanation is helpful in resolving apparent syn-
chronic anomalies.
Another issue that is of significant interest in SALs is that of long-distance
agreement which we shall discuss in the following section.
4.5 SALs and long-distance agreement

There are instances of long-distance agreement in some SALs, such as Hindi-
Urdu (Mahajan 1990; Davison 1991; Butt 1993; Singh 1993; Bhatt 2005;
Chandra 2007; Davison 2007b), Punjabi (Subbarao 2001), Kashmiri (Wali and
Koul 1997; Subbarao and Munshi 2000; Bhatt 2005), Kurmali (Mahto 1989)
and Maithili (Yogendra Yadava, p.c.) of the Indo-Aryan family; Mizo and
Hmar (Subbarao 1998b) of the Tibeto-Burman language family; and Telugu
(Subbarao 1991) of the Dravidian family.
Long-distance agreement has attracted considerable attention in recent theo-
ries of syntax. Long-distance agreement differs from agreement in a simple
clause in the following way: agreement is usually a local phenomenon. In
local agreement, an argument in a simple clause triggers an agreement in
the predicate; while in long-distance agreement, an argument in the embedded
clause triggers an agreement in the predicate (verb, adjective or participle) of the
higher clause. This is similar in nature to long-distance anaphora (see chapter 3),
where the antecedent in a higher clause binds an anaphor in a lower clause.
Both agreement and antecedent–anaphor binding are ‘local’ relationships, and

hence both these processes must be accounted for as local in an explanatory
adequate theory of grammar. We have shown in chapter 3 that long-distance
anaphoric relationships can be explained in terms of local binding obeying
Principle A.
Just as with long-distance anaphora, long-distance agreement too is mostly
found in sentences involving nonfinite clauses. This implies that the subject
of the embedded clause is PRO, which is null. Finite clauses in general block
long-distance agreement.11
We shall now present a descriptive account of long-distance agreement fol-
lowed by a discussion of some recent attempts to explicate it.
4.5.1 Indo-Aryan
Long-distance agreement has been the focus of attention in recent discussions
on Hindi-Urdu agreement (Hook 1979; Mahajan 1989, 1990; Davison 1991;
Butt 1993; Singh 1993; Subbarao 2001; Bhatt 2005; Chandra 2007; Davison
2007b). Mahajan (1990) and Davison (1991), to the best of our knowledge, are
the first two studies of long-distance agreement in Hindi-Urdu (IA).
Hindi-Urdu and Punjabi (IA) permit long-distance agreement only when the
embedded sentence contains a verb in the infinitival form. This is also the case
in Maithili (IA). The following examples from Hindi-Urdu and Punjabi are
illustrative.
l o n g - d i s t a n c e a g r e e m e n t w i t h t h e e m b e d d e d o b j e c t sāikil ‘c y c l e ’
Hindi-Urdu (IA)
(57) rām ko [S2 PRO sāikili calā- n- ı̄i S2 ] ā- tı̄i thı̄i
Ram dat cycle drive- inf- f come- imperf.f pst.f
‘Ram used to know how to ride a bicycle.’
n o l o n g - d i s t a n c e a g r e e m e n t w i t h t h e e m b e d d e d o b j e c t sāikil
‘c y c l e ’
(58) rām ko [S2 PRO sāikil calā- n- āS2 ] ā- tā thā
Ram dat cycle drive- inf- m come- imperf.m pst.m
l o n g - d i s t a n c e a g r e e m e n t w i t h t h e e m b e d d e d o b j e c t sāikil ‘c y c l e ’
Punjabi (IA)
(59) prakāš nũ˜ [S2 PRO sāikili calā- n.- ı̄i S2 ] ā- ndı̄i sı̄
Prakash dat cycle drive- inf- f come- imperf pst
‘Prakash used to know how to ride a bicycle.’
Sentences (57) and (59) are instances of long-distance agreement in object

complement clauses, as the embedded clause occurs in the canonical object
position of the matrix clause. Sentences (57) and (58) demonstrate that long-
distance agreement is optional.
Long-distance agreement in Hindi-Urdu

We shall explicate only the Hindi-Urdu examples here. In (57), the embedded
unmarked infinitive calā-nā ‘to drive’ exhibits agreement with the embedded
object sāikil ‘bicycle,’ which is feminine in Hindi-Urdu. The matrix verb in
turn agrees with the embedded infinitival complement clause. In (58), there is
no long-distance agreement.
Let us first consider the case of sentence (57), in which long-distance agree-
ment takes place. Mahajan (1990) and Chandra (2007) account for this by the
movement of the Agr phrase to the Spec position of the necessary functional
projections. The DP/NP is consequently raised into the matrix clause. The two
clauses are merged into one (Wurmbrandt 2003). Davison (1991) treats infini-
tives as clauses with nominative case and the infinitive gets optional number and
gender features from the embedded object (where PRO has no phi-features).
Butt (1993) treats infinitival clauses as NPs, and thus the infinitival clause func-
tions as a theme in (57). The matrix predicate agrees with the theme, which is an
NP. Hence, in Butt’s analysis, sentence (57) is not an instance of long-distance
agreement. In Bhatt’s (2005) analysis, the embedded clause is actually a VP,
a VP without a subject. The infinitive inflection is ‘defective,’ and does not
block the ‘agree’ between the clausal object and the matrix inflection on finite
tense.
In Bhatt’s (2005) analysis, in sentences without long-distance agreement, as
in (58), the nominative DP is not raised, and hence the two clauses are not
merged. The embedded clause is a full TP/CP with a subject. Bhatt (2005)
labels the complement a “restructuring infinitive.”
Davison (2007) provides arguments to show that in sentences with long-
distance agreement the complement object is not raised, contra Mahajan (1990)
and Chandra (2007). She argues that, in long-distance agreement, the verb
has a full clause complement, contrary to Bhatt’s (2005) claim that it does
not. Davison’s argument gains support from the fact that with the ‘begin’-
class of verbs in Hindi-Urdu, as we shall show below, the matrix verb has a
full clausal complement as its subject. Thus, in such clauses, (i) the embed-
ded complement is a full clause; (ii) infinitive inflection is not ‘defective’;
(iii) the embedded complement has a subject of its own; and (iv) long-distance
agreement takes place obligatorily – not optionally – with the embedded sub-
ject, as in (60) and (61) (Subbarao 2001; Subbarao, Hakacham and Sarju
Devi 2007).
subject complements (non-finite)

Hindi-Urdu (IA)
(60) [S2 d.epo se basẽ nikal- nı̄S2 ] šurū ho gayı̄˜
bus station from buses f,p start- inf f,s start be went f,p
‘The buses began to start from the bus station.’
(61) [S2 šādı̄ kı̄ taiyāriyā˜ ho- nı̄S2 ] šurū ho gayı̄˜

wedding of plans f,p be- inf f,s start be went f,p
‘Plans for the wedding got started.’
(Subbarao et al. 2007: 315)
Another significant fact about long-distance agreement is that it is blocked:

(i) when the matrix subject is nominative case-marked, and not ergative case-
marked, in object complements (62), and
(ii) when the embedded subject is genitive case-marked in subject comple-
ments (64) and (65).
The following examples are illustrative.
(i) Matrix subject nominative case-marked in object complements
(62) *rām [S2 PRO sāikili cālā- n- ı̄S2 ]
Ram (masculine) cycle (feminine) ride- inf- f,s
jān- tı̄ thı̄i
know- imperf f,s pst f,s
Sentence (62) is ungrammatical as the infinitive must be in the unmarked

masculine singular form, and not in the feminine singular form. Further,
the matrix verb should exhibit agreement with the matrix subject Ram.
Thus, the grammatical equivalent of (62) is (63).
(63) rāmi [S2 PROi sāikilj calā- n- ā∗ i,∗ j S2 ] jān-
Ram s cycle f,s ride- inf- f,s know-
tāi thāi
imperf m,s pst m,s
‘Ram used to know how to ride a bicycle.’12
Keeping in view sentences as in (62) and (63), Subbarao (2001) and

Subbarao et al. (2007) argue that PRO, the uncase-marked, ungoverned
empty category, according to standard assumptions, needs to be case-
marked to account for long-distance agreement in Hindi-Urdu, as long-
distance agreement with embedded object in object complement clauses
crucially depends on the non-nominative case-marked (ergative or dative)
nature of the matrix subject – hence the ungrammaticality of (62).
(ii) Embedded subject genitive case-marked in subject complements

The infinitive in (64) and (65) does not exhibit agreement with its subject
as its subject is genitive case-marked. Recall that Hindi-Urdu belongs to
the class of languages where a postpositionally case-marked DP blocks
an agreement.
(iii) No LDA in (64) and (65)
(64) [S2 d.epo se basõ kā nikal- nāiS2 ] šurū
bus station from buses f,p gen m,s start- inf m,s start
ho gayāi
be went m,s
‘The buses began to start from the bus station.’
(65) [S2 bārātiyõ kā ā- nāiS2 ] šurū

members of marriage party gen m,s come- inf m,s start
ho gayāi
be went m,s
‘Members of the marriage party started arriving.’
LDA is permitted when the matrix sentence has a predicate adjective.
(66) [S2 anı̄š se subah bāt kar- nı̄S2 ] muškil

Anish with morning talk f,s (noun) do- inf f,s difficult
hotı̄ thı̄
be.imperf f,s be.pst f,s
‘It used to be difficult to talk to Anish in the morning.’
(S. P. Gupta, p.c.)
To recapitulate, the following facts need to be taken into account for a compre-
hensive analysis of long-distance agreement in Hindi-Urdu (IA).
(i) In subject complements LDA is obligatory, while in object complements
it is optional.
(ii) In subject complements the embedded clause does not contain PRO, while
in object complements the occurrence of PRO is obligatory.
(iii) In object complements the matrix subject needs to be either ergative
or dative case-marked; the embedded subject in subject complements is
nominative case-marked.
(iv) In subject complements long-distance agreement is blocked if the embed-
ded subject is genitive case-marked.
(v) The predicate in subject complements belongs to the ‘begin’-class of
verbs, and the matrix verb may contain a predicate adjective too.
Thus, long-distance agreement in Hindi-Urdu (IA) is quite a complex phe-
nomenon, and we have presented only a brief summary of the issues involved.
We shall now discuss the case of Kashmiri (IA).
Long-distance agreement in Kashmiri (IA)

Kashmiri too exhibits long-distance agreement. Sentence (67) is an instance of
long-distance agreement, where both the matrix predicate and the embedded
infinitive exhibit LDA.
infinitival agreement – long-distance agreement
Kashmiri (IA)
(67) rām- an che hamēš yetšmats [pannis necivis
Ram- erg be.pres f always wanted f,p self.dat son.dat
khətr kōri vuchini]
for girls see.inf f,p
‘Ram always wanted to see girls for his son.’
(Subbarao and Munshi, ms)
However, agreement in the matrix clause has the option of not taking place,
though there is agreement of the infinitive with the embedded object. In that
case, the infinitive vuchini ‘to see’ would be in feminine plural exhibiting
agreement with the embedded object kōri ‘girls,’ while the matrix verb yotšmut
chu ‘wanted m,s, be.pres. m,s’ would exhibit agreement with the matrix subject
rām ‘Ram.’ Hindi-Urdu too exhibits such a phenomenon.
infinitival agreement – no long-distance agreement
(68) rām- an chu hamēš yotšmut [pannis necivis khətr
Ram- erg be.pres m,s always wanted m,s self.dat son.dat for
kōri vuchini]
girls see-inf f,pl
‘Ram always wanted to see girls for his son.’
(Subbarao and Munshi, ms)13
In Kashmiri, infinitival agreement takes place even when the matrix subject is
nominative case-marked. Bhatt (2005: 24) comments: “Infinitival agreement
also takes place when long-distance agreement is not even a possibility.” Recall
that infinitival agreement with the embedded object is not permitted in Hindi-
Urdu when the matrix subject is nominative case-marked (62). The matrix verbs
in (69) and (70) agree with the matrix subjects rām ‘Ram’ and zoon ‘Zoon,’
respectively.
(69) rām chu yatsān [kōri vuchini]
Ram.m be.pres.m.s wants-imperf girls see-inf.f.p
‘Ram wants to see the girls.’
(70) zoon che yatsān [kōri vuchini]
Zoon.f be.pres.f wants.imperf girls see-inf.f.p
‘Zoon wants to see the girls.’
(Bhatt 2005: 24)
According to Bhatt, in (69) and (70) the matrix verb agrees with the matrix
subject, as it is nominative case-marked: “Consequently, there is no room for
the infinitival object to display long-distance agreement with the matrix verb.
However, this does not block the object agreeing with the infinitival verb”
(Bhatt 2005: 24).
Long-distance agreement in Maithili

In Maithili (IA), the matrix verb agrees in honorificity with the possessor tohār
‘your’ occurring in the embedded clause. The pronoun tohār ‘your’ is a 2 mid-
honorific possessive form, and the matrix verb carries the honorific agreement
marker -ah.
Maithili (IA)
(71) rām kẽ tohāri gār.ı̄ calaunāi abait ch- ahi
Ram dat your 2 mid hon car to drive comes- agr mid hon
‘Ram knows how to drive your car.’
(Yogendra Yadava, p.c.)14
4.5.2 Tibeto-Burman
We consider the case of long-distance agreement in Mizo (TB) first. In Mizo
predicate adjectives exhibit agreement with the noun. Adjectives in Mizo
carry person and number agreement features just as verbs do, and they are
homophonous with the agreement markers that occur with the verb. Mizo has
no gender agreement. Sentences (72) and (73) are instances of finite subject
complementation, and hence the embedded sentence occurs in the subject posi-
tion of the matrix clause. The 3rd person singular agreement marker is -a,
and the plural marker is -an. In (72), the predicate adjective in S2 carries -an
‘3rd person plural,’ as it agrees with amerikā mite ‘American people,’ the sub-
ject of S2 . The matrix verb carries -a ‘3rd person singular’ as it is a case of
subject complementation, and the entire S2 is the subject of S1 .
No long-distance agreement in Mizo

Mizo (TB)
(72) [S2 amerikā mite cui ani − thau tak- takS2 ]j a- nij
American people def 3 p- fat very- very 3 s- be
‘It is the case that Americans are very fat.’
(Subbarao 2001: 481; a slightly modified version of the
original – the intensifier very is added in the
sentence translation)
Sentence (72) literally means: ‘(That) Americans are very fat is the case.’
Long-distance agreement in Mizo

In (73), subject to subject raising takes place. Hence, the embedded adjective
carries a null (ø) agreement marker. The matrix verb is in 3rd person plural,
as the embedded subject is in the derived subject position of the matrix clause.
Hence, it triggers agreement on the matrix verb (see Subbarao 2001 for further
discussion).
(73) amirikā mite cui ø- thau tak- tak ani - ni
American people def fat very- very 3p- be
‘It is the case that Americans are very fat.’
4.5.3 Dravidian
Dravidian languages (e.g. Kannada, Tamil and Telugu), too, exhibit long-
distance agreement. We now discuss LDA in Telugu. In Telugu (DR), adjectives
carry pronominal suffixes in predicate position, and thus exhibit agreement
with the subject (Krishnamurti and Gwynn 1987). In (74), the adjective in the
embedded clause, manci ‘good,’ carries the 3rd person pronominal suffix vād.u
‘3rd person singular,’ and it agrees with vād.u ‘he,’ the subject of the embedded
clause. Sentences (74) and (75) are instances of subject complementation. The
verb in Telugu carries phi-features. In (74), the matrix verb is un-di ‘be-3rd
person, non-masculine, singular,’ and in (75), it is unnā-d.u ‘be-3rd person, mas-
culine, singular.’ In (74), the matrix verb undi ‘be + 3 nh’ exhibits agreement
with the entire embedded sentence, and hence -di the ‘3rd person, non-human,
singular’ marker occurs with the verb stem. In (75), the matrix verb exhibits
agreement with the embedded subject vād.u ‘he,’ and hence -d.u ‘3rd person,
masculine, singular’ occurs.
no long-distance agreement
Telugu (DR)
(74) [S2 vād.ui manci- vād.ui - lāS2 ]j un- dij
he good- 3 s,m pron suffix- comp be- s,nh
‘It appears that he is a nice fellow.’
long-distance agreement
(75) vād.ui manci- vād.u lā unnā- d.ui
he good- 3 s,m pron suffix comp be.pres- 3 s,m
‘He appears to be a nice fellow.’
In sentences involving superraising cases too,15 there is long-distance agree-

ment (77). In (76), the matrix verb undi ‘be + 3 s, nh’ exhibits agreement with
the embedded sentences, just as in (74).
no long-distance agreement
(76) [S1[S2[S3 nāku vād.ui manci- vād.u- lāS3] unna-
to me he good- 3 s,m pron suffix- comp be-
t.luS2] j kanipistundiS1] j
comp appears 3 s,nh
‘He seems to appear like a nice fellow.’
long-distance agreement
(77) nāku vād.ui manci- vād.u- lā unna- t.lu
to me he good- 3 s,m pron suffix- comp be- comp
kanipistunnā- d.ui,∗ j
appear- 3 s,m
‘He seems to appear like a nice fellow.’
We do not have data on LDA in Khasi (Mon-Khmer) and the Munda languages.
Another interesting aspect of agreement in SALs is Pronominal Strength
Hierarchy found in subject–verb/adjective agreement, which we shall discuss
in the following section.
4.6 Pronominal Strength Hierarchy in agreement

Many SALs exhibit Pronominal Strength Hierarchy in manifesting agreement.
In sentences with conjoined pronominal subjects, the 1st person subject is
‘stronger than’ 2nd and 3rd person subjects, and the 2nd person subject is
‘stronger than’ 3rd person subject for verb agreement. With conjoined subjects,
the verb agreement is in the plural.
Tibeto-Burman languages such as Mizo, Hmar and Paite exhibit split erga-
tivity in person. The verb carries a subject agreement marker. When the
DO/IO/Possessor is in 1st or 2nd person – in Hmar, for example – the pronom-
inal clitics too are coded in the verb. The subject agreement clitic is a- in
3rd person, and the 1st person direct object clitic is mi in (78).
Hmar (TB)
(78) Lalii - n proj ai - mij - hmu
Lali- erg (me) 3 s- 1 s- saw/met
‘Lali saw/met me.’
In sentences with conjoined subjects in Mizo and Hmar, the 1st person subject
has precedence in occurrence as a pronominal clitic over 2nd and 3rd person
subjects, and 2nd person over 3rd person. Hence, when 3rd person subject and
1st person subject occur as conjuncts as in (79), it is the 1st person subject
clitic kan- in the plural that occurs, and not the 3rd person plural clitic an-. The
4.7 ECM cases (Small Clauses) 123
1st person subject as a conjunct does not occur overtly at all, and there occurs
in its place only a null pro. Thus, the 1st person singular null subject pro in
(79) is recoverable only from the verbal agreement clitic kan- in the plural. The
precise details of such recoverability need to be worked out.
(79) lalii le proj naŋk - cou in- t.šuŋ- pui- kanj+plural - ti- cek
Lali and (I) you- only VR- sit- with- 1 p- fut- 2 s
‘Lali and I will sit only with you.’16
In (80), the conjuncts are in 3rd and 2nd person and hence, it is the 2nd person
agreement marker in the plural that occurs with the verb. The 2nd person
subject as a conjunct does not occur at all, and in its place there is a null pro.
(80) lalii le proj in- in- t.šuŋ

Lali and 2 p- VR- sat
‘Lali and you sat.’
The hierarchies in sentences (79)–(80) can be schematically represented as

follows:
(81) 1p > 2p > 3p where > indicates ‘stronger than’ and p indicates person.
Sentences (79) and (80) also illustrate how recoverability of the null pronoun
(pro) takes place. It is only the agreement marker in plural on the verb that
indicates whether pro in (79) refers to a 1st person pronoun or pro in (80) refers
to a 2nd person pronoun.
In Telugu (DR) and Tamil (DR) too, the principle of Pronominal Strength
Hierarchy operates in verb and adjective agreement. Adjectives in predicate
position in Telugu carry pronominal suffixes except in 3rd person masculine
singular where the entire pronoun is suffixed to the adjective. The pronominal
suffixes too exhibit Pronominal Strength Hierarchy (see Subbarao 2001 for
details). In Oriya (IA) as well, the principle of Pronominal Strength Hierarchy
operates in conjoined structures (Neukom and Patnaik 2003).
4.7 ECM cases (Small Clauses) and long-distance agreement

In all the four language families of the subcontinent there are cases involving
ECM (Exceptional Case Marking), in which an embedded subject gets ‘excep-
tionally case-marked’ accusative – except in Kashmiri (IA), Dumi (TB) and
Telugu (DR) – by the matrix verb. In Kashmiri (Bhatt 1999), Dumi (van Driem
1993) and Telugu (DR), the embedded subject is nominative case-marked (see
section 5.2.4 for details).
Our data show that, in the languages in which the agreement clitic occurs,
the clitic of the embedded subject shows up as a clitic on the matrix verb, thus
demonstrating that it is a case of long-distance agreement.
4.7.1 Indo-Aryan
When the matrix subject is ergative case-marked, the matrix verb agrees with
the embedded subject if it is not followed by a postpositional case marker (82),
and the verb exhibits default agreement if it is followed by a case marker, as in
(83) in Hindi-Urdu.
Hindi-Urdu (IA)
(82) atul ne [[yah cı̄z] burı̄] samajh- ı̄ thı̄
Atul erg this thing. f bad. f consider- perf. f be.pst. f.s.
‘Atul had considered this thing bad.’
(83) atul ne [[is cı̄z] ko burā] samajh- ā thā
Atul erg this thing. f acc bad.def consider- perf. m be.pst. m.s
‘Atul had considered this thing bad.’
(Mahajan 1989)
If the matrix subject is nominative case-marked, the matrix verb agrees with
the matrix subject as (84).
(84) atul [[yah cı̄z] burı̄] samajh- tā hai
Atul this thing. f bad. f consider imperf pres
‘Atul considers this thing bad.’
(Bhatt 2005: 32)
A crucial point to note is that there is no optionality of agreement in these

cases, unlike the cases of long-distance agreement discussed earlier; the verb
obligatorily agrees with the embedded subject, which occurs as a ‘derived
object’ at S-Structure. There is no optionality of agreement in Hindi-Urdu or
Punjabi; Bhatt (2005), for example, attributes such non-optionality to an oblig-
atory restriction. He points out: “ECM/raising infinitives involve obligatory
restructuring and with these infinitives we find that long-distance agreement is
also obligatory.” Thus, in Bhatt’s analysis, restructuring is a precondition for
long-distance agreement.
Long-distance agreement in object complements in Hindi-Urdu and the ECM
cases share a common feature in that both are permitted only when the matrix
subject is dative or ergative case-marked.
4.7.2 Dravidian
For a discussion of ECM cases in Dravidian, see chapter 5.
4.7 ECM cases (Small Clauses) 125
Khasi (Mon-Khmer)
In (85) in Khasi, the derived object tεilin ‘Teilin’ is accusative case-marked.
The embedded subject tεilin ‘Teilin’ moves to the matrix object position from
a [+tensed] COMP-clause (see chapter 6 for a discussion of this phenomenon).
Hence, the 3rd person feminine singular agreement marker ka is retained on
the embedded verb wan ‘come’ in (85) (see appendix 2 of chapter 6 on the
Website for a detailed discussion).
(85) pro ŋa-khmi? lnti ya-ka-tεilini ba ti ka-n-wan

1s-expect acc-3f.s-Teilin c o m p 3f.s-fut-come
‘I expect Teilin to come.’
(Temsen 2006)
Munda
In Ho (Munda), the matrix verb carries the direct object agreement marker,
as (86) illustrates. The subject of the embedded clause, añ ‘I,’ occurs in the
derived object position and hence the direct object agreement marker iñ occurs
as a suffix on the matrix verb manatiŋ ‘consider.’ The fact that the embedded
subject agreement clitic occurs on the matrix verb demonstrates that it is a case
of ECM.
Ho (Munda)
(86) ami [añj concor.-] emi - manatiŋ- t- ad.- iñj - a
you me intelligent- 2s- consider- epen- pst perf- 1s- fin
‘You consider me intelligent.’
4.7.4 Tibeto-Burman
We shall now discuss cases involving ECM subjects in Mizo and Hmar. Recall
that Mizo and Hmar are split ergative in person. Subject as well as object agree-
ment markers occur as pronominal clitics in these languages. The pronominal
agreement marker ce of the subject of the embedded sentence appears as a clitic
on the matrix verb in ECM cases. Hence, ce occurs to the right of the matrix
verb.
Hmar (TB)
(87) lalii - n [øj insual- pui] in a- dit- cej
Lali- erg fight- with comp 3s- want- 2s
‘Lali wants you to fight.’
Note that (87) is also an instance of postposition incorporation where pui ‘with’
occurs with the embedded verb insual ‘fight’ (see the discussion in section 3
of this chapter for details).
In Mizo (88), it is min which is the subject of the embedded sentence showing
up as the object clitic of the matrix S. It occurs to the left of the matrix verb.
Mizo (TB)
(88) zovi- n lekhathɔn ø- ziak turin min- duh
Zovi- erg letter write comp 1s- want
‘Zovi wants me to write a letter.’
Note that the subject of the matrix sentence is in 3rd person. Recall that
1st person DO > (‘is stronger than’) 3rd person DO (according to the Principle
of the Pronominal Strength Hierarchy discussed in 4.6).17
To summarize, in all SALs, it is the subject of the embedded clause that
triggers agreement in the matrix verb when it is in the derived object position.
All the examples discussed in (82), (86), (87) and (88) are instances of LDA.
In the South Asian subcontinent, language contact has played a dominant
role, as a result of which phonological, morphological and syntactic conver-
gence has taken place. In the following section we discuss how features of
agreement are borrowed from one language family to the other.
4.8 Language contact and agreement

In contact situations, a functional category, such as agreement, in one language
may converge with the same category in a language belonging to another family.
We shall discuss the case of agreement in Manda (DR) and Malto (DR).
4.8.1 Agreement in Manda (DR)

The first case in point concerns agreement clitics in Manda, a Dravidian lan-
guage, spoken in areas where Mundari, a Munda language, and the Indo-Aryan
Oriya are spoken. Both Manda (DR) and Mundari (Munda) are nominative–
accusative languages. In Manda (DR), the verb agrees not only with the subject,
but also with the DO, IO, OO (Locative), for- dative, and with the dative subject.
Such agreement is untypical of Dravidian languages and is not found elsewhere
in the language family. In (89), the verb exhibits agreement with the locative
object mā-dengata ‘on us,’ and the subject rı̄ ‘rain.’ In Dravidian languages,
agreement with a locative object – or, for that matter, with any PP or DP – is not
permitted. The question that needs to be answered is: how is agreement with a
locative object permitted in Manda, a Dravidian language? We shall attempt to
explain it.
4.8 Language contact and agreement 127
a g r e e m e n t o f o o (l o c a t i v e )
Manda (DR)
(89) inelin māi - dengata pı̄yu rı̄j - dai - t- ij
yesterday us- on day rain- 1 p- pst- 3 s
‘It rained on us yesterday.’
agreement of the dative subject

(90) nin- kai nūmerj ah tai,∗ j - n- ba
you- to fever catch 2s- [-pst]- progr
‘You got fever.’
(Ramakrishna Reddy 1992a: 575)18
In a Dravidian language, when the subject is dative case-marked, the verb agrees
with the theme, which is nominative case-marked (see section 4.2.1 for details).
Note that in (90) the verb does not agree with the noun phrase nūmer ‘fever’
in the nominative case. Instead, it agrees with a postpositionally case-marked
dative subject. Since none of the other Dravidian languages have agreement
of a non-subject with a verb, while Munda languages do, Ramakrishna Reddy
(1992a, 2004) argues that this phenomenon is due to the Mundari (Munda)
influence on the Dravidian Manda.
4.8.2 Agreement in Malto (DR)

The second case concerns Malto, a transplanted Dravidian language in Bihar,
which exhibits agreement in conjunctive participles (Mahapatra 1979). Con-
junctive participles in Dravidian languages, though finite, do not manifest
agreement. On the other hand, participles in some of the Munda languages
exhibit subject agreement. Thus, agreement in conjunctive participles in Malto
(DR) could be attributed to convergence with the Munda languages with
which it has long been in contact. The following example from Malto (DR) is
illustrative.
Malto (DR)
(91) proi bit- a- ka- ri oy- a- ri mand- a- ri
(they) cook- epen- cpm- 3 p take- cpm- 3 p plant- pst- 3 p
‘Having cooked, having taken them, they planted them.’
(Mahapatra 1979: 223)
In (91) note that:

(1) Mahapatra (1979: 223) glosses the morphemes ka- and a- above as perfect.
They both appear to be the allomorphs of the cpm. Considering the meaning
of the sentence he provides, we interpret his perfect marker in the two
embedded clauses as a conjunctive participial marker (cpm). Note that
there is no conjunction connecting the clauses. The absence of a conjunction

marker supports our contention that ka- and a- denote the cpm.
(2) The glosses and sentence translation have been slightly modified.
To summarize the above discussion, in this section we discussed cases involv-
ing agreement with non-subjects in Manda (DR) and the agreement of the
conjunctive participle with the matrix subject, a phenomenon not found in any
Dravidian language, apart from Malto. Since it is the Munda languages that
exhibit non-subject agreement on the verb, it is reasonable to conclude that a
functional category, such as agreement, in one language may converge with the
same category in a language belonging to another family.
In the following section we discuss the intrinsic relationship between case
and agreement.
4.9 Case and agreement in SALs

As mentioned in the introduction, in recent theories of syntax, case and agree-
ment played a very crucial role. Relative clauses, passives and clefts in Mizo
(Subbarao and Lalitha Murthy 1994), Thadou and Hmar (TB) present signifi-
cant phenomena which show how case and agreement are intrinsically linked.
We discuss below constructions involving the relative clause, the passive and
the cleft in Mizo, Thadou and Hmar (TB).
4.9.1 Relative clauses

We shall first consider the case of relative clauses in Mizo. When direct object,
for example, is modified, the Mizo relative clause exhibits two distinct agree-
ment patterns.
Type I: the embedded subject carries the ergative marker and the
embedded verb carries the subject agreement marker (in italics in
(92)).
Type II: the ergative marker as well as the agreement marker are both
absent, as in (93).
Relative clauses
Type I Embedded subject carrying the ergative marker, and
the verb carrying the subject agreement marker in the
embedded relative clause
Mizo (TB)
(92) [zovi- n lekahbu a- lei- tur19 cu] lali- n a- chiar- aŋ
Zovi- erg book 3 s- buy- fut def Lali- erg 3 s- read- fut
‘Lali will read the book that Zovi will buy.’
Type II Absence of the ergative marker and subject agreement

marker in the embedded relative (indicated by ø)
(93) [zovi ø lekahbu ø lei- tur cu] lali- n a- chiar- aŋ
Zovi ø book ø buy- fut def Lali- erg 3 s- read- fut
‘Lali will read the book that Zovi will buy.’
In (92), the embedded subject Zovi ‘Zovi’ carries the ergative case marker -n,
and the embedded verb carries the agreement marker -a. In contrast, in (93),
the embedded subject Zovi ‘Zovi’ does not carry the ergative case marker -n,
and the embedded verb lei ‘buy’ does not carry the subject agreement marker
-a. In Mizo and Hmar, the ergative marker and the agreement marker are
mutually dependent; either they both occur or neither does. Such dependency
in occurrence is also found in passives and clefts in Mizo (Subbarao and Lalitha
Murthy 1994; Prakash 2006).
4.9.2 Passives
Let us look at the passive in Mizo. The passive too exhibits a similar pattern
to that in the relative clause. The passive in Mizo has a bi-clausal structure
(see Subbarao and Lalitha Murthy 1994: 403). Example (94) is a simple active
sentence.
(94) zovi- ni thei- te ai - ei
Zovi- erg fruit- p 3 s- eat
‘Zovi ate (some) fruits.’
We provide below the two types of passive sentences in Mizo.
Type I Embedded passive carrying the ergative marker with

the subject, and the verb carrying the subject agreement marker
The embedded sentence in the passive in Mizo in (95) contains the ergative
case marker, and the verb contains the 3rd person singular agreement marker.
(95) [S2 thei- te cu zovi- n a- ei- inS2 ] an- ōm

fruit- p def Zovi- erg 3 s- eat [+fin]- linker 3 p- be
‘The fruits were eaten by Zovi.’

marker in the embedded passive
In (96), the embedded sentence in the passive in Mizo does not contain the
ergative case marker, and the embedded verb does not contain the 3rd person
singular subject agreement marker (their absence is indicated by ø in (96)).
(96) [S2 thei- tei cu zovi ø ø ei- inS2 ] ani - ōm

fruit- p def Zovi eat [−fin]- linker 3 p- be
Examples similar to (95) and (96) are found in Hmar (TB) too.

Hmar (TB)
(97) [S2 thei- hai cu zovi- n a- fak- inS2 ] an- um
fruit- p def Zovi- erg 3 s- eat [+fin]- linker 3 p- be
(Vanlal Englien, p.c.)

(98) [S2 thei- haii cu zovi ø ø fak- inS2 ] ani - um
fruit- p def Zovi eat [−fin]- linker 3 p- be
Thus, the data from relative clauses in Mizo (TB) and passives in Mizo and
Hmar (TB) demonstrate that case and agreement are intrinsically linked.
4.9.3 Clefts
In clefts too a similar pattern is found in Mizo. Clefts are bi-clausal in Mizo.
Type I Embedded cleft carrying the ergative marker with the

subject, and the verb carrying the subject agreement marker
Mizo (TB)
(99) boŋ- cu [zova- n a- that] a- ni- lou
cow- foc Zova- erg 3 s- kill [+fin] 3 s- be- not
‘It is not the cow that Zova killed.’

marker in the embedded cleft
The embedded verb is [−finite] in the embedded cleft.
(100) boŋ- cu [zova ø ø thah] a- ni- lou

cow- foc Zova kill [−fin] 3 s- be- not
In Hmar too a similar pattern obtains.
Hmar (TB)
(101) boŋ- cu [zova- n a- that] a- ni- noʔ
cow- foc Zova- erg 3 s- kill [+fin] 3 s- be- not

marker in the embedded cleft
The embedded verb is [−finite] in the embedded cleft.
(102) boŋ- cu [zova ø ø that] a- ni- noʔ

cow- foc Zova kill [−fin] 3 s- be- not
We now consider the case of the passive in Thadou. Thadou (TB) also exhibits
a pattern similar to Mizo with regard to relative clauses, passives and clefts
(Pauthang Haokip, p.c.). We present below two sentences illustrating the two
patterns of the passive in Thadou (TB). The passive in Thadou too has bi-
clausal structure. Example (103) is a simple active sentence. The quantifier
themxat ‘some’ follows the head noun.
Note that the singular and plural markers are homophonous in Thadou.
Thadou (TB)
(103) lalal- ini thei themxat ai - nei
Lalal- erg fruit some 3 s- eat
‘Lalal ate some fruits.’
We provide below the two types of passive sentences in Thadou.

The embedded sentence in the passive in Thadou in (104) contains the ergative
case marker, and the verb contains the 3rd person singular agreement marker.
(104) [S2 thei- ho cu lalal- in a- nei?S2 ] a- hi

fruit- p def Lalal- erg 3 s- eat [−fin] 3 p- be
‘The fruits were eaten by Lalal.’

In (105), the embedded sentence in the passive in Thadou does not con-
tain the ergative case marker, and the embedded verb does not contain the
3rd person singular subject agreement marker (their absence is indicated by ø).
(105) [S2 thei- ho cu [lalal ø ø nei?S2 ] a- hi

fruit- p def Lalal eat [−fin] 3 p- be
‘The fruits were eaten by Lalal.’
Similar data can be adduced from other TB languages, such as Paite and Zou.
The relative clause in Mizo, the passive and the cleft in Mizo and Hmar,
and the passive in Thadou provide support for Chomsky’s (2001) view of
linking together the uninterpretable features of goal and probe. They are either
overtly present or absent altogether. That is, case and agreement are intrinsically
linked.20
4.10 Conclusion
This chapter provides an exhaustive account of the various agreement types
available in SALs, bringing into focus several new phenomena which have not
been discussed in any work thus far. SALs can be classified into four major
agreement types:
(1) languages in which a postposition blocks agreement;
(2) languages in which a postposition does not block agreement;
(3) languages which are polysynthetic and semi-polysynthetic;
(4) languages which exhibit no agreement at all.
In the polysynthetic Sora (Munda) language, direct object, locative PP and
instrumental PP are incorporated, a phenomenon unique in SALs. Pronominal
clitics for subject, direct object, indirect object and possessor are incorporated
in Tibeto-Burman and Munda languages. In Munda and Tibeto-Burman lan-
guages, the verb carries the agreement marker of the possessor of the object
alone, not of the possessor of the subject. Thus, though the possessive marker
can potentially refer to either the subject or the object, it is coindexed only with
the direct object and not with the subject – for example, in Santali. Quantifiers,
Negative Polarity Items, intensifiers, emphasizers, group markers and adverbs,
too, are incorporated in Tibeto-Burman and the Mon-Khmer Khasi.
Adposition incorporation of the comitative postposition in Hmar (TB) and of
the locative in Paite (TB) takes place. When an adposition is incorporated, the
host intransitive verb, as a result, becomes ‘transitivized.’ Such incorporation,
as well as the ‘transitivization’ of the host intransitive verb, is a phenomenon
unique to select languages of the Tibeto-Burman family.
4.10 Conclusion 133
We provided a diachronic explanation of why an incorporated element in

Sora, a polysynthetic language, occurs to the right of the adjoining element,
contrary to Kayne’s generalization. Following Starosta’s suggestion, we pointed
out that Munda languages developed some of their morphology while they were
still SVO, and then did a syntactic shift to SOV in South Asia, then adding
some more morphology.
We discussed instances of long-distance agreement from SALs and provided
an account of some of the analyses of Hindi-Urdu (IA). We provided evidence
to show that: (i) PRO needs to be case-marked for long-distance agreement
to take place in object complements in Hindi-Urdu (IA); and (ii) in subject
complements too, long-distance agreement occurs, though the embedded clause
is intact, and no restructuring takes place.
We demonstrated that the Principle of Pronominal Strength Hierarchy is
operative in Dravidian, Indo-Aryan and Tibeto-Burman languages. According
to this principle, subject is stronger than direct object, and direct object is
stronger than indirect object. The cases of long-distance agreement in ECM
constructions were also presented. We have also shown how agreement patterns
‘converge’ in languages belonging to different language families. Based on data
from relative clauses and passives in Mizo, we have also demonstrated how
case and agreement are intrinsically linked.
5 Non-nominative subjects
5.1 Introduction
Non-nominative subject constructions in the languages of the South Asian
subcontinent include ergative, dative, genitive, locative, instrumental (by-
passive) and accusative subjects. Dative subjects are a feature of the South Asian
linguistic area (Emeneau 1956; Masica 1976). They are the most widespread
in Dravidian and in some Indo-Aryan languages. Some Tibeto-Burman and
Munda languages, too, share this feature. The ergative construction is found
only in some Indo-Aryan and some Tibeto-Burman languages, and is absent in
Dravidian and Austro-Asiatic languages.
In this chapter, we discuss the nature of case marking – lexical/inherent
vs. structural, the choice of case on the subject and object in non-nominative
subject (hereafter, NNS) constructions, general trends in SALs, and the vari-
ation by genetic affiliation or subregion. We provide a brief description of
NNSs in SALs first, keeping in view the notion of subject.1 We shall then
discuss some subject properties of NNSs. We argue that (i) the predicate
in a dative subject construction (DSC) is [−transitive] and unaccusative;
(ii) all NNSs except the ergative are inherently case-marked; (iii) such inher-
ent case marking can be done not by an intransitive verb alone, but by the
whole predicate compositionally consisting of a theme or an adjective along
with the [−transitive] verb; and (iv) information concerning agreement should
be available vP-internally (in the lower thematic S) for proper assignment of
inherent case to the NNS. We shall show that the accusative/dative case mark-
ing of the theme in dative/genitive subject constructions in Bangla, Tamil and
Malayalam does not count as counter-evidence to treating the predicate in NNS
constructions as [−transitive].
This chapter is organized as follows. Section 5.1 is an introduction.
Section 5.2 provides a description of the NNS constructions in SALs.
Section 5.3 deals with the domains of occurrence of the NNS. Section 5.4
discusses some subject and non-subject properties of NNS constructions in
SALs. Section 5.5 deals with the stative and non-stative as well as the intransi-
tive nature of the predicate in dative subject constructions (DSCs). Section 5.6
134
discusses the inherent case marking of the dative subject in SALs. Section
5.7 discusses double/multiple dative subject constructions in Telugu and Kan-
nada, which are unique to SALs. Section 5.8 concludes the chapter. Formal
representation and case checking properties of NNS constructions in SALs are
discussed in the appendix of chapter 5 on the Website.
In all SALs, the non-nominative DP occurs in the initial position of a sen-
tence, [NP of S] (Chomsky 1965) or Spec of TP. SALs have seven types of
NNS constructions. The matrix or embedded subject is (i) ergative case-marked,
(ii) dative case-marked, (iii) genitive case-marked, (iv) locative case-marked,
(v) accusative case-marked or (vi) instrumental case-marked and is in the by-
passive or (vii) the embedded subject is accusative case-marked in non-finite
embedded clauses (the so-called Exceptional Case Marking2 [ECM] cases in
Government and Binding theory).
5.2 NNS constructions in SALs

5.2.1 The matrix or embedded subject ergative case-marked
In the nominative–accusative construction in Hindi-Urdu, Punjabi and Marathi
(IA), the verb agrees with the subject in the nominative case. In the ergative–
absolutive construction, the subject carries an ergative marker depending upon
the transitive nature, aspect, modality and finiteness of the verb (Davison 2004;
Hook and Koul 2004; Kachru 1980; Pandharipande 1997; Wali 2004). The
occurrence of the ergative marker in Marathi and Punjabi, in addition, is
dependent on person too. Hence, these languages are labeled as split erga-
tive languages.
In some of the Tibeto-Burman languages, such as Mizo, Hmar and Thadou,
direct and indirect object agreement in ergative constructions is exhibited only
when the direct object is in 1st and 2nd persons, and not in the 3rd person
(cf. section 4.2.3) (Subbarao and Lalitha Murthy, ms). An ergative subject
exhibits all the properties of the subject, for example as a local antecedent in
(1) in Hindi-Urdu and in (2) in Kashmiri (IA), or a long-distance antecedent in
(3) to an anaphor in Hindi-Urdu, and controller of PRO in Marathi (IA) and
Mizo (TB) in (4) and (5), respectively.
as a local antecedent
Hindi-Urdu (IA)
(1) baccõi ne apnı̄i/∗ j billı̄j dekh- ı̄j
children erg self’s cat f, s nom see- f, s perf
‘The childreni saw / looked at self’si/∗ j cat.’
(Davison 2004: 145)
Kashmiri (IA)
(2) mohani - an vuch panun pāni ə̃˜nas manz
Mohan- erg saw self’s self mirror in
‘Mohan saw himself in the mirror.’
as a long-distance antecedent
Hindi-Urdu (IA)
(3) aoki ne lalitāj se [PROi apne liei/j cāy banā- ne ko] kahā
Ashok erg Lalita with self for tea make- inf in order said
‘Ashok asked Lalita to make tea for self.’
(Subbarao 1971: 191; the glosses have been slightly modified)
as the controller of PRO

Marathi (IA)
(4) minii - ni ravij - lā [PROi pakad.- āycā] prayatna ke- l- ā
Mini- erg Ravi- acc catch- inf try do- pst- 3 s,m
‘Mini tried to catch Ravi.’
(Wali 2004: 226)
Mizo (TB)
(5) lalii - n [PROi min- hmu] ai - duʔ
Lali- erg 1 s- meet 3 s- wants
‘Lali wants to meet me.’
5.2.2 The matrix subject or embedded subject may be dative or genitive

case-marked
There are a number of predicates that are used in NNS constructions which
require a subject with the dative or genitive postposition. All the Dravidian
and IA languages such as Hindi-Urdu, Punjabi, Kashmiri, Marathi, Nepali,
Gujarati and Sinhala have the dative subject construction, while Assamese and
Bangla (IA) have the genitive construction with psychological predicates (see
section 5.3 for examples). Oriya (IA) has both the dative and genitive subject
construction. The predicates, for example, include intransitive verbs such as:
honā ‘to be,’ ānā ‘to come,’ lagnā ‘to appear, to feel,’ sūjhnā ‘to strike,’ honā ‘to
happen’ in Hindi-Urdu; transitive verbs vēyu ‘to drop,’ tōyu ‘to push,’ pat..tu ‘to
catch,’ pōyu ‘to pour,’ pet..tu ‘to put, serve,’ tappu ‘to miss,’ tippu ‘to turn,’ and
intransitive verbs such as vaccu ‘to come,’ pad.u ‘to fall,’ und.u ‘to be,’ agu ‘to
happen’ in Telugu (DR); uth ‘to rise’ in Assamese; and lag ‘to feel’ in Bangla –
just to mention a few. No Indo-Aryan or Dravidian languages have any form
that corresponds to the verb have, and hence, in non-nominative constructions
in these families, it is only the verb be that occurs and not have. Some of the
Tibeto-Burman languages such as Mizo, Hmar and Thadou have two distinct
verbs for be and have, and most of the Tibeto-Burman languages, except Bodo,
Kokborok and Manipuri, do not have any NNS construction (other than the
ergative) at all. Khasi (Mon-Khmer) also has two different verbs for be and
have, and it too does not have any NNS construction at all. We shall demonstrate
later that it is the verb coupled with a noun or an adjective in the verb phrase
that is instrumental in assigning the non-nominative case to the subject.
Dative case-marked subject

The subject of a matrix or embedded sentence may be case-marked dative, and
the verb exhibits agreement with the theme, which is in the nominative case,
in (6) in Hindi-Urdu (IA) and in (7) in Telugu (DR). In (6) kaı̄ acchı̄ kitābẽ
‘several good books’ is the theme, which is in the plural, and hence the verb
agrees with the theme, while manci pustakālu ‘good books’ is the theme in (7),
and hence the verb exhibits agreement with the theme in (7).
alienable possession
Hindi-Urdu (IA)
(6) murlı̄i ko dūkān mẽ kaı̄ acchı̄ kitābẽj milı̄j thı̄˜∗ i,j
Murali m,s dat shop in several good books f,p found f,p be+pst f,p
‘Murali found several good books in the shop.’
Telugu (DR)
(7) mādhurii ki kot.- lō cālā manci pustakāluj dorikēyi∗ i,j
Madhuri dat shop- in very good books p,nh found p,nh
‘Madhuri found very good books in the store.’
In Manipuri (TB), there is no subject–verb agreement.
Manipuri (TB)
(8) manoŋ- də layrik- əmə ləy
he- dat book- one be
‘To him there is a book.’
(Chelliah 1990: 201)
In Marathi (IA), inalienable possession, as well as “existential possession of

relatives (kinship terms), friends, or body parts (including mental faculties such
as intelligence, ignorance and so on) are expressed only through dative posses-
sives and not through postpositional possessives” (Pandharipande 1997: 231).
In Dravidian languages too, a similar situation obtains (see Amritavalli 2004
for Kannada, Jayaseelan 2004 for Malayalam, and Subbarao and Bhaskararao
2004 for Telugu).
inalienable possession
Marathi (IA)
(9) malai don hātj āhet∗ i,j
I-dat two hands p are p
‘I have two hands.’
(10) malai don bhāūj āhet∗ i,j

I-dat two brothers are
‘I have two brothers.’
Genitive case-marked subject

When the subject is dative or genitive case-marked, the theme in most of the
cases carries the nominative marker, and the verb agrees with the theme.
Hindi-Urdu (IA)
(11) raghui kı̄ tı̄n bahnẽj thı̄˜∗ i,j
Raghu m gen three sisters be+pst f,p
‘Raghu had three sisters.’
In Kokborok and Manipuri (TB), there is no agreement manifested, as these

languages lack agreement.
Kokborok (TB)
(12) kɔrmati ni bɔkhɔrɔk sā- ɔ
Kormati gen head pain (verb)- pres
‘Kormati has a headache.’
Manipuri (TB)
(13) ram- gi hui- əmə ləy
Ram- gen dog- one be
‘Ram has a dog.’
‘There exists a dog of Ram.’
(Chelliah 1990: 200)
In Assamese (IA), the subject is genitive case-marked, and the theme is nomina-
tive case-marked, as in (14). The predicate in (14) is a psychological predicate
which depicts the emotional state that the logical subject is in.
Assamese (IA)
(14) tar khɔng uth.i- sil
he/she.gen anger (nom) rise.cpm- pst
‘He/she is angry.’
In Bangla (IA), the subject as an experiencer is generally genitive case-marked.

However, it is dative case-marked in obligative constructions, just as in Hindi-
Urdu and Punjabi.
Bangla (IA)
(15) rina- ke aj pãc t.a dokane jete holo
Rina- dat today five cl shops.loc go.inf was
‘Rina had to go to five shops today.’
(Dasgupta 2004: 130)
In Malayalam and Tamil (DR) too, a similar phenomenon is observed, which

we shall discuss later.
Locative case-marked matrix or embedded subject

The subject may be locative case-marked (16). Such marking indicates tem-
porary possession (Krishnamurti and Gwynn 1985: 85). The logical subject
pratāp ‘Pratap’ in Telugu (DR) is in masculine singular and the theme d.abbulu
‘money’ is in non-masculine plural. Hence, the negative, which is a verb in
Telugu, agrees with the theme d.abbulu ‘money,’ not with the logical subject.
Telugu (DR)
(16) prastutam pratāpi daggara d.abbuluj lē- vu∗ i,j
at.present Pratap near money.3 p,nm be.not- 3 p,nm
‘Pratap does not have any money at present.’
The dative too can occur in sentences where the possessed entity is [+concrete],
and the dative indicates a permanent state of affairs. Thus, (17) is interpreted as
Pratap does not have money at all, and he is poor. Hence, the occurrence of an
adverb such as prastutam ‘at present’ will render the sentence ungrammatical.
(17) pratāp kii (*prastutam) d.abbuj lē- du∗ i,j

Pratap dat at present money.3 s,nm be.not- 3 s,nm
‘Pratap does not have any money (*at present).’
Bodo (TB) has a locative postposition with the experiencer, when a physical
ailment is expressed.
Bodo (TB)
(18) khampha- nao lɯmja- nai dɔŋ- ɔ
Khampha- gen.loc (near) sick (verb)- inf be- pres
‘Khampha has fever.’
The non-nominative marker that occurs with the possessor (logical subject) in
IA languages such as Hindi-Urdu, Punjabi and Marathi depends on the nature
of the object possessed. In Hindi-Urdu, the genitive postposition occurs with
inalienable possession of body parts, kinship and friends, while the locative
ke pās ‘near’ occurs with concrete possession. The locative mẽ ‘in’ occurs
with inherent physical qualities in Hindi-Urdu, while in Bangla the genitive
occurs in such cases (for details, see Verma and Mohanan 1990; Masica 1991;
Bhaskararao and Subbarao 2004). According to Dasgupta (2004: 132), genitive/
locative case marker -r occurs with the experiencer in Bangla (IA).
Bangla (IA)
(19) tomar jɔthet.o aho ache
you.indirect enough courage is
‘Do you have enough courage?’
Bodo (TB) permits a locative with the subject.
Bodo (TB)
(20) bibar- ao mujang mud m- nai dɔng
flower- in good fragrance (verb)- inf be
‘The flower has good fragrance.’
Literally: ‘To fragrance [as verb] well is [there] in the flower.’
5.2.3 Accusative case-marked subject

Sinhala (IA), Manipuri and Rabha (TB)3 are the only SALs we know of where
the subject can be accusative case-marked: “Accusative subjects occur only with
intransitive verbs, and the sense is generally involitive with the implication that
some external force is responsible” (Gair 2003: 791).
Sinhala (IA)
(21) māwə yantaŋ bēruna
I.acc barely escape.pst
‘I barely escaped.’
(Gair 2003: 791)
5.2.4 Subject in by-passive

In SALs, there is a syntactic passive with the passive morphology overtly mani-
fested in the verb, and there is a lexical passive with an intransitive verb carrying
no passive morphology. In both cases, the subject is followed by a postposi-
tion by. While IA languages freely use both of them, in Dravidian languages
the frequency of occurrence of the syntactic passive is low in comparison to
that of the lexical passive. In Tibeto-Burman languages too, syntactic passives
with overt passive morphology are rather low in occurrence, except in Mizo
(Subbarao and Lalitha Murthy, ms) and Hmar. In Khasi and the Munda lan-
guages, there is only the syntactic passive, and the passive marker is the same
as the verbal reflexive and anti-causative (see chapter 3).
In Hindi-Urdu (IA), the patient/theme may be nominative case-marked, and
the verb agrees with it as in (22a), or accusative case-marked as in (22b), and
the verb exhibits default agreement as the agent as well as patient/theme are
both followed by a postposition.
Hindi-Urdu (IA)
(22a) polı̄si ke.dvārā corj pakr.ej nahı̄˜ gaye∗ i,j
police by thieves m,p catch.perf m,p not pass.pst m,p
‘The thievesi werej not caught by the policej .’
(22b) polı̄si ke.dvārā corõj ko pakr.ā∗ i,∗ j nahı̄˜ gayā∗ i,∗ j

police by thieves acc catch.perf not pass.pst.def agr
‘The thievesi were not caught by the policej .’
In Assamese and Bangla (IA), in contrast, the patient/theme is accusative case-

marked and it cannot be nominative case-marked.
Assamese (IA)
(23) ravan.- ak ram- ɔr dwara morua hol
Ravan- acc Ram- by was killed
‘Ravan was killed by Ram.’
(Kakoli Das, p.c.)
In Kashmiri and Sinhala (IA) too, the patient/theme may be dative case-marked.
In (24), in Kashmiri, the patient koryi ‘girl.dat’ “moves to the subject position
in passive” (Bhatt 1999: 215).
Kashmiri (IA)
(24) az av koryi vuch- n (temsin- dasyi)
today came girl.dat see- pass him- by
‘Today the girl was seen (by him).’
(Bhatt 1999: 215)
Colloquial Sinhala (IA) according to Gair (2003: 792) lacks a true passive. The
subject is instrumental case-marked, and the verb carries an involitive verbal
suffix that imparts the passive meaning.
Sinhala (IA)
(25) amma- geŋ sinhəla kǣmə hoňdət.ə hædenəwa
mother- instr Sinhala food well make.involitive.pres
‘Mother (always) makes Sinhala food well.’
(Gair 2003: 792)
In Kharia (Munda) and Khasi (Mon-Khmer) too, the theme/patient may be

accusative case-marked. The passive marker in Kharia (Munda) is d.om as in
(26), and it is homophonous with the verbal reflexive. The verb in Kharia and
Khasi exhibits default agreement, namely a null agreement marker (ø), as the
derived subjects are case-marked, a phenomenon similar to the one found in
Hindi-Urdu (IA).
Kharia (Munda)
(26) pothi- ki- te ter d.om- ki- Ø
book- p- obl give pass- middle pst- default
‘The books were given.’
(Peterson 2006)
Khasi (Mon-Khmer)
(27) ya- ki- sʔiar ø- la- bām da- u- ksɔw
acc- f- fowl default- pst- eat by- m- dog
‘The hen was eaten by the dog.’
(Temsen and Subbarao, ms)
In Dravidian languages, the theme/patient in the passive is invariably nomina-

tive case-marked, and it cannot be accusative case-marked.
In Indo-Aryan languages, the passive has capability interpretation. In
Marathi, for example, according to Pandharipande (1997: 301), “in the spoken
or the informal register of Marathi, the passive construction is used to express
the capability of the agent to carry out the action expressed by the passive verb.
This usage of the passive construction is usually shared by other IA languages
such as Hindi, Kashmiri, Nepali and Punjabi.”
In Marathi, when the postposition kad.ūn ‘by’ is used, the sentence imparts
capabilitative meaning.
Marathi (IA)
(28) mādzyā kad.ūn /*dwārā kām kela gela nāhı̄
I-obl by work do-pst-3s go-pst-3s (pass) neg
‘This work was not done by me.’ (‘I was unable to do the work.’)
In Hindi-Urdu too, the passive expresses capabilitative meaning.
Hindi-Urdu (IA)
(29) rām se per. kāt..e nahı̄˜ ga.ye
Ram by trees m,p cut.perf m,p not pass.perf m,p
Literally: ‘The trees were not cut by Ram.’
‘Ram couldn’t bring himself to cut the trees.’
(Rajesh Bhatt, p.c.)
However, if the agent is not overtly present, the passive sentence does not
impart capabilitative meaning, as in (30).
(30) per. kāt..e nahı̄˜ ga.ye

trees m,p cut.perf m,p not pass.perf m,p
‘The trees weren’t cut.’
There is another passive, labeled as lexical or semantic passive, in which:

(i) the subject carries the by-passive marker, (ii) the verb is [−transitive], and
(iii) the verb does not carry any passive morphology.4
This passive too imparts capabilitative meaning, when the agent is overtly
present as in (31).
(31) is lar.kei se acchı̄ rot.ı̄j bantı̄j nahı̄˜

this boy by good Indian bread made not
‘This boy cannot make good Indian bread.’
When the agent is not overtly present, the capabilitative meaning is not
imparted.
(32) is d.hābe mẽ acchı̄ rot.ı̄j bantı̄j nahı̄˜

this roadside hotel in good Indian bread made not
‘Good Indian bread is not/*cannot be made in this roadside hotel.’
To impart capabilitative interpretation in (32), the overt occurrence of the modal

saknā ‘can’ is required.
Telugu (DR) has a syntactic passive (33), and a lexical passive without passive
morphology (34). In both these constructions, the logical subject is case-marked
by valla ‘by’ and the verb agrees with the argument in the nominative, namely
the theme.
Telugu (DR)
(33) āmei valla ilāt.i panuluj ceyya- bad.- a vuj
she by such things do- pass- neg- nh,p
‘Such things cannot be done by her.’
(34) ilāt.i panuluj āmei valla jarag-a- vuj / kā- vuj

such things she by happen-not-.3 p,nm happen-not.3 p,nh
‘Such things cannot be done by her.’
The lexical passive in Telugu (DR), like in Hindi-Urdu (IA), does not impart
capabilitative meaning when the agent is not overtly present.
Telugu (DR)
(35) ilāt.i panuluj ı̄ ūl.l.- ō jaraga-vuj / kā-vuj
such things this town- in happen-not happen-not
‘Such things are not done in this town.’
The theme cannot be case-marked accusative/dative in Telugu (36), unlike

in Hindi-Urdu, Kashmiri (Bhatt 1999: 215), Bangla (IA) and some other IA
languages, as well as Kharia (Munda) and Khasi (Mon-Khmer). It is always in
the nominative case as in (33). Sentence (22b) from Hindi-Urdu has its theme
case-marked accusative.
Telugu (DR)
(36) *pōlı̄sula- valla dongala- ni/ ki pat.t.u- bad.a- lēdu
police- by thieves- acc dat catch- pass- be.not
‘The thieves were not caught by the police.’ (intended meaning)
In Telugu (DR) too, just as in Malayalam, the passive “is not usually used in
conversation. It has tended to be confined to official language and the language
of journalism” (Asher and Kumari 1997: 268).
Just as in Telugu, in Malayalam too, “passives (including nonfinite forms)
appear without an agent” (Asher and Kumari 1997: 269).
Malayalam (DR)
(37) ivid.e pat.t.usāri-kal. vilkkappet.unnu
here silk saree-p sell-pass-pres
‘Silk sarees are sold here.’
(Asher and Kumari 1997: 269)
5.2.5 Embedded subject is accusative/dative case-marked

The embedded subjects apne āp ko ‘himself’ in (38) in Hindi-Urdu (IA) and
gopal in (39) in Bangla (IA) are accusative case-marked, and tem ‘he’ in
Kashmiri in (40) is dative case-marked, due to Exceptional Case Marking
(ECM). Note that, in ECM, the embedded subject is case-marked by the matrix
verb.
Hindi-Urdu (IA)
(38) hareki insān apne āpi ko hoiyār māntā hai
every person self’s self acc intelligent consider pres
‘Every person considers himself to be intelligent.’
Bangla (IA)
(39) orita gopal ke bhalo mone kɔre
Sarita Gopal acc good thinks
‘Sarita considers Gopal to be good.’
Kashmiri (IA)
(40) me vuch- na tem- is alakh pyāvān
I.erg saw- neg he- dat beat fall-imperf prt
‘I did not see him being beaten.’
(Bhatt 1999: 198)
In Nepali (IA), the embedded subject is dative case-marked (Poudel 2007: 32)
too, and in Mizo (TB), the accusative case marker with the derived object occurs
in the Exceptional Case Marking (ECM) construction.
Mizo (TB)
(41) zovi- n [amah- cu lāmthiam] a- inti
Zovi- erg she.self- acc dancer 3 s- considers
‘Zovi considers self a dancer.’
The derived object in an ECM construction may, however, be nominative case-

marked in Kashmiri (IA), Dumi and Tiwa5 (TB) and Telugu (DR).
(i) In Kashmiri (IA) the embedded subject is nominative case-marked, as in
(42), when the matrix verb is in the past tense (Bhatt 1999: 219). Bhatt
labels it as “Exceptional Nominative Case Marking” where the embedded
infinitival subject “occupies Spec-Agr OP,” and the matrix verb agrees with
the subject of the lower infinitival clause. In (42), ti indicates the trace of
the moved element swa ‘she’ from S2 .
Kashmiri (IA)
(42) tem vuch swai [S2 ti dohay bat khyav- ānS2 ]
he.erg saw.f,s she.nom daily rice.acc eat imperf
‘He saw her eating rice every day.’
(Bhatt 1999: 219)
If the matrix verb is in the present tense, the sentence is ungrammatical

(43a).
(43a) *ba chus swa dohay bat khayv- ān vuchān

I.nom pres.1 m,s she.nom daily rice.acc eat imperf see.imperf
‘I see her eating rice every day.’
(Bhatt 1999: 219)
However, if the object clitic occurs on the matrix verb, the sentence is
grammatical (43b) (Sadaf Munshi, p.c.).
(43b) bai chusi - anj swaj dohay bat khyav- ān

I.nom pres.1 m,s- 3 s she.nom daily rice.acc eat- imperf
vuchān6
see.imperf
‘I see her eating rice every day.’
Thus, the case marking in an ECM construction is sensitive to the tense of

the matrix clause in Kashmiri (IA), if the pronominal clitic is not present.
Kashmiri is the only language we know of where such restriction exists.7
(ii) In Dumi and Tiwa (TB) too, the embedded subject is nominative case-
marked.
Dumi (TB)
(44) aŋ- mō- liŋ- ə- m m- a a- dūkhut- ə ʔe
I- (nom)- sit- walk about- 1 s- nozr he- erg MS- see- 1 s rep
‘He said he saw me hanging around that place.’
(van Driem 1993: 192 – rep: reported speech particle, i.e., hearsay
evidential; MS: marked scenario)8
In Telugu (DR), the embedded subject may be accusative or nominative

case-marked.
Telugu (DR)
(45) mādhuri arun.- (ni) pāripō.tund.agā cūsindi
Madhuri Arun- acc run away.imperf saw
‘Madhuri saw Arun running away.’
Thus, Kashmiri (IA), Dumi and Tiwa (TB) and Telugu (DR) are the four
languages that we know of where, in the ECM construction, the embedded
subject is nominative case-marked. Further, the case marking in an ECM
construction in Kashmiri is sensitive to the tense of the matrix clause and
the agreement clitic of the embedded subject on the matrix verb.
5.3 Domains of occurrence of NNSs in SALs, and their nature

There are several studies that have discussed the domains in which the NNS
occurs (see Kachru 1970, 1990; Masica 1976, 1991; McAlpin 1976; Sridhar
1976; Mohanan and Mohanan 1990; Shibatani and Pardeshi 2001).
Domains of occurrence of the dative subject (adapted from Subbarao and
Bhaskararao 2004):
(a) psychological states and emotions;
(b) physiological and mental ailments;
(c) natural phenomena pertaining to the body;
(d) perceiver of visual and auditory actions;
5.3 Domains of occurrence of NNSs in SALs 147
(e) to express possession and kinship;

(f) subject of predicates expressing obligation and necessity (desideratives);
(g) to denote a recipient;
(h) acquisition of knowledge or a skill;
(i) part–whole relationship (single and double dative case marking).
Psychological states and emotions

The dative/genitive subject occurs in sentences with psychological predicates.
Hindi-Urdu (IA)
(46) pallavii ko bahut khuı̄j huı̄∗ i,j
Pallavi dat very happiness happened
‘Pallavi felt very happy.’
The predicate in Telugu (DR) as in (47), contains verb und. ‘be’ when the
psychological predicate is an adjective, and the verb exhibits default agreement.
Telugu (DR)
(47) ı̄ pāpalai - ki santōs.amj- gā undi∗ i,∗ j
these babies- dat happiness- adjr is (def agr)
‘These babies are happy.’
Bodo (TB) has a genitive subject construction with psychological predicates.
Bodo (TB)
(48) bi- ha jɯbɯr raga jɯng- nai dɔng- ɔ
he- gen very anger set fire- inf be- pres
Literally: ‘to him a lot of anger is to set fire / to ignite.’
‘He is very angry.’
Physiological and mental ailments

The subject is dative or genitive case-marked when it possesses physiological
or mental ailments. The verb agrees with the theme (the closest conjunct) in
Hindi-Urdu (IA), and with the union of both conjuncts in number and gender
in Telugu (DR).
Hindi-Urdu (IA)
(49) pratimāi ko [zukām aur khā˜sı̄]j ho gayi thı̄∗ i,j
Pratima f,s dat [cold and cough]f happen went pst f
‘Pratima caught a cold and cough.’
Telugu (DR)
(50) pratimai ki [daggu- u jalubu- u rend.u- u9 ]j
Pratima f,s dat [cough- conj mkr cold- conj mkr both- emph]
unnāyi p∗ i,j
are
‘Pratima caught both a cough and a cold.’
Kokborok (TB) and Bangla (IA) have a genitive subject in such sentences.
Kokborok (TB)
(51) a- ni kɔngrai tɔng- ɔ
I- gen cold be- pres
Bangla (IA)
(52) amar t.hand.a lege- che
I.gen cold feel- pres
Natural phenomena pertaining to the body

The subject is genitive case-marked in Hindi-Urdu (IA) and dative case-marked
in Telugu (DR), when any event or occurrence that pertains to the body takes
place.
Hindi-Urdu (IA)
(53) us ādmı̄i ke sāre bālj jhar. gayej
that man gen all hair fall went
‘That man lost all his hair.’
Telugu (DR)
(54) ā manis.ii ki anta juttuj - u10 ūd.- i pō- yindij
that man dat so much hair- quantifier mkr fall- cpm go- pst
‘That man lost all his hair.’
Perceiver of Visual and auditory perceptions

The subject of a predicate denoting visual and auditory perceptions is dative
case-marked. In Hindi-Urdu, the predicates sunāı̄ denā ‘to be audible’ and
dikhāı̄ denā ‘to be visible’ require a dative subject, whereas the verbs sun-
nā ‘to hear’ and dekh-nā ‘to see’ require a nominative or ergative subject,
depending on the aspect of the verb.
Hindi-Urdu (IA)
(55) rāt ko mahimā ko t.hı̄k dikhāı̄ nahı̄˜ detā
night at Mahima dat well visible not give
‘Mahima cannot see well at night.’
Verbs such as sun-nā ‘to hear’ and bhūlnā ‘forget’ take a nominative case-
marked subject. In present-day Hindi-Urdu, a few speakers have started using
such verbs with a dative case-marked subject, as in (56).
(56) jor se boliye mere ko11 ū˜cā suntā hai

loud with speak (polite) I (obl) dat high hear pres
Literally: ‘Please talk loudly, I hear high (loud).’
‘Please talk loudly, I can’t hear well.’
(Rama Kant Agnihotri, p.c.)
Peter Hook (p.c.) informs me that in Shina (IA) too, a verb such as to forget,
which takes a nominative subject, permits a dative subject to occur, just as in
Hindi-Urdu (IA).
The verb kanipincu ‘to be visible’ takes a dative subject in Telugu.
Telugu (DR)
(57) pramodi ki rātrul.l.u sariggā kanipinc- a- du∗ i
Pramod dat nights properly visible- not- 3 s,nm (def agr)
‘Pramod cannot see well at night.’
To express possession and kinship

The subject is genitive case-marked when a kinship relationship is expressed
in Hindi-Urdu, Punjabi, Bangla (IA) and Kokborok (TB); and dative case-
marked in Dravidian languages and in Kashmiri (IA). In IA languages, the
case marking of the subject depends on the nature of the object possessed.
In Dravidian languages, except in cases involving temporary possession, the
possessor is always dative case-marked. In sentences expressing possession, all
SALs of the subcontinent have be as the only verb,and do not have the verb to
have, except a few Tibeto-Burman languages such as Mizo, Hmar and Paite, and
the Mon-Khmer language Khasi, which have both be and have. Mahajan (2004)
labels have as an oblique case incorporator, as languages that have the have-
construction do not have a genitive/dative case-marked subject. Jayaseelan
(2007: 37), in a similar vein, argues: “when the dative Case incorporates into
‘be,’ we get ‘have.’”
The verb agrees in phi-features with the theme possessed.
Hindi-Urdu (IA)
(58) rādhāi kı̄ tı̄n bahnẽj thı̄˜∗ i,j
Radha f,s gen three sisters f,p were f,p
‘Radha had three sisters.’
Kashmiri (IA)
(59) mohani - as chuj dukānj / bōyj
Mohan- dat has [is] shop brother
‘Mohan has a shop/brother.’
Kannada (DR)
(60) nana-i ge mūvaru hen.n.u makkad.uj iddāre∗ i,j
I- dat three female children be.nonpst.3 p.hum
‘I have three daughters.’
(Sridhar 1990: 133)
In Kokborok (TB) too, the genitive occurs with the logical subject, just as in
Hindi-Urdu, and the occurrence of the verb to be depends on the presence
of the classifier with the possessed. If the classifier is not present with the
possessed entity, the verb tɔng ‘to be’ must be overtly present as in (61), and if
the [+human] classifier is present, then the occurrence of the verb tɔng ‘to be’
is optional as in (62).
Kokborok (TB)
(61) ni- ni thakhuk b.sk tɔng
you- gen brothers how many be
‘How many brothers do you have?’
(62) ni- ni thakhuk khɔrɔk b.sk (tɔng)

you- gen brothers cl [+human] how many be
‘How many brothers do you have?’
In Hindi-Urdu (IA), the locative ke pās ‘near’ is used to indicate concrete

possession, as in (63), except with time expressions, as in (64).
Hindi-Urdu (IA)
(63) un logō˜i ke pās bahut paisāj thā∗ i,j
those people near a lot of money m,s was m,s
‘Those people had a lot of money.’
(64) un logō˜i ke pās utnā t.āimj bilkul nahı̄˜ thā∗ i,j
those people near that much time m,s at all (NPI) not was m,s
‘Those people did not have that much time at all.’
In Telugu (DR), the locative marker daggara ‘near’ occurs with the subject
to indicate temporary possession. Recall that ki, the dative case marker, also
occurs in sentences denoting possession, as in (17) above. However, there is
a difference: “while the occurrence of the dative [case marker] denotes ‘per-
manent or long-term possession,’ the occurrence of the locative [case marker]
denotes ‘temporary possession’” (Subbarao and Bhaskararao 2004: 172).
Telugu (DR)
(65) mā.kui mugguru panivāl.l.uj unnāruj kānı̄ prastutam māi
we (excl).dat three servants are but right now our (excl)
daggara iddare- e unnāru∗ i,j
near two- emph are
‘We’ve three servants but right now we’ve only two.’
Manipuri (TB) has the locative postposition occurring with the subject in the
non-nominative subject construction. The subject is in the non-nominative
case and is marked by the postposition manak-ta ‘near at.’ The occurrence
of non-nominative subjects in Tibeto-Burman languages is rather infrequent.
Most of the languages do not have the non-nominative subject construction
at all, in contrast to Indo-Aryan and Dravidian languages, where the use of
non-nominative subject construction is a dominant one.
Manipuri (TB) also makes a distinction in terms of which postposition occurs
for temporary and permanent possession, just as in Telugu (DR).
Need or necessity
The dative subject also occurs in sentences denoting need or necessity in Hindi-
Urdu, Punjabi (both IA), Bodo, Garo (TB), Telugu and Kannada (DR).
Hindi-Urdu (IA)
(66) ramyā ko ek kitāb cāhiye
Ramya dat one book needed
‘Ramya wants a book.’
Bodo (TB)
(67) khampa n gəikher naŋɯu
Khampa dat milk want
‘Khampa wants some milk.’
Garo (TB)
(68) aŋ- na i- ko naŋ- no- a
I- dat this- acc need- fut- ?
‘I will need this.’
(Burling 2004: 122)
Kannada (DR)
(69) nana- ge idu is.t.a illa
I- dat this liking neg
‘I don’t like this.’
(Sridhar 1979: 101)
Obligation is expressed with a non-nominative case-marked subject in IA lan-

guages (see (15) from Bangla [IA]). In contrast, in Dravidian languages, the
subject is nominative case-marked in such cases.
Malayalam (DR)
(70) nı̄ pōyē tı̄rū
you go-emph must
‘You really must go.’
(Asher and Kumari
1997: 307)
To denote a recipient
The dative subject also denotes a recipient.
Hindi-Urdu (IA)
(71) mujhei yahã˜ ek cit.t.ı̄j milı̄∗ i,j
I.dat here one letter f,s found f,s
‘I got a letter.’
(Montaut 2004: 193)
Telugu (DR)
(72) ā abbāyilai - ki rōd.d.u mı̄da oka pustakamj dorikin-di∗ i/j
those boys- dat road on a book nm,s got-nm,s
‘Those boys found a book on the road.’
Acquisition of knowledge or a skill

A dative subject also denotes acquisition/knowledge of a skill or talent by the
possessor, and the verb in such cases is ā ‘come’ in Hindi-Urdu (IA) and vac
‘come’ in Telugu (DR).
Hindi-Urdu (IA)
(73) is panjābi lar.kı̄i ko bharat nāt.yamj ātā∗ i,j thā∗ i,j
this Punjabi girl dat classical Indian dance come was
‘This Punjabi girl used to know the classical Indian dance.’
Telugu (DR)
(74) vāl.l.a ammāyi- ki sangı̄tam vaccu
their daughter- dat music comes
‘Their daughter has the knowledge of music.’
Part–whole relationship (single and double dative

case marking)
In sentences with a DP consisting of ‘part–whole’ relationship, either a dative
as in (75), or a locative as in (76), occurs in Hindi-Urdu (IA). The whole –
prasād ‘Prasad’ – and part – man ‘mind’ and sar ‘head’ – are invariably linked
by the genitive postposition.
genitive–dative
Hindi-Urdu (IA)
(75) [prasād ke man]i ko gahrı̄ cot.j pahuncı̄∗ i/j
Prasad, m gen mind dat deep wound, f reached, f
Literally: ‘Prasad’s mind got hurt badly.’
‘Prasad’s feelings were hurt (deeply).’
(Om Arora, p.c.)
genitive–locative
(76) [prasād ke sar]i mẽ gahrı̄ cot.j āı̄∗ i/j
Prasad, m gen head in deep wound, f came f
‘Prasad got hurt badly in the head.’
(Om Arora, p.c.)
In contrast, in Kannada, Malayalam and Telugu (DR), a dative predicate may

assign a single or double dative case marking. Double dative case marking is
permitted in these languages in dative subject constructions expressing inalien-
able possession and part–whole relationship (Subbarao and Bhaskararao 2004).
This is a feature found only in Dravidian languages, and not elsewhere in the
subcontinent. Thus, the following two types of case marking of the subject
DP are permitted: genitive–dative as in (77) and dative–dative as in (78). The
possessor is genitive case-marked and the possessed – the body part – is dative
case-marked.
Kannada (DR)
(77) avani - a kan.n.j - ige gāya(v)k āgi- de∗ i∗ j,k
he- gen eye- dat injury happen- agr
In Kannada and Telugu (DR), the possessor, as well as the possessed, are both
dative case-marked.
Kannada (DR)
(78) avani - ige kan.n.j - ige gāya(v)k āgi- de∗ i∗ j,k
he- dat eye- dat injury happen- be
See section 5.7 for further details. The Double Dative Subject construction in
Dravidian is similar to the double subject construction discussed in Japanese
by Shibatani and Pardeshi (2001).
5.4 Some subject and non-subject properties of the

NNS construction
5.4.1 Some subject properties of NNSs in SALs
We have shown earlier the subject properties of ergative subjects. We shall now
consider the other non-nominative subjects, and demonstrate that, though they
are case-marked with a postposition, they exhibit some properties of subject,
and some of non-subject.
There are several tests that can be used as diagnostics to test the subjecthood
of NNSs (see Sigurdsson 2004). We discuss four tests here:
(1) NNSs as antecedents to anaphors;
(2) NNSs as controllers of PRO;
(3) NNSs and subject-oriented verbs;
(4) phonological evidence from the final complementizer (FC) in Marathi.
NNSs as antecedents to anaphors

Perhaps there is not a single study on NNSs that does not include a discussion
of NNSs as antecedents to anaphors. In Hindi-Urdu, for example, the dative
subject is an antecedent to an anaphor (Davison 2004). In (79), kamzor insān
is a dative case-marked subject, and it c-commands the anaphor apne āp par
‘self’s self on.’
Hindi-Urdu (IA)
(79) kamzor insāni ko apne- āp par gussāj ā- tā∗ i,j hai∗ i,j
weak human being dat self+gen- self on anger come- imperf pres
‘A weak human being gets angry at himself.’
In Telugu (DR), the dative subject is an antecedent to an anaphor. The verbal

reflexive cannot occur in such cases.
5.4 Subject and non-subject properties 155
Telugu (DR)
(80) kāvyai - ki tana mı̄da *(tana-kii ) jālij put.t.indi∗ i,j / *put.t.u-
Kavya- dat self on self-dat pity was born- 3 s,nh be born-
kon- di∗ i,j
VR- 3 s,nh
‘Kavya pitied herself.’
In Hindi-Urdu (IA), Saxena (1985) first pointed out that, in the DSC, the
possessive anaphor alternates with a simple possessive pronoun (see Gurtu
1992, too).
Hindi-Urdu (IA)
(81) mohani ko apnei / uskei/j mā˜- bāp kı̄ yād
Mohan.ms dat self’s his (3s gen) mother- father gen memory f,s
ā- yı̄
come- pst.f,s
‘Mohani remembered self’si/∗ j /hisi/j parents.’
(Saxena 1985)
Thus, dative subjects bind a pronoun and anaphor, or just an anaphor, while
nominative subjects bind an anaphor.
Mohanan and Mohanan (1994: 175) provide an example to show that the
possessive pronoun in a DSC refers to the nominative NP, and not to the dative
subject.12
(82) anui ko nı̄nāj uskı̄∗ i,j bastı̄ mẽ dikhı̄

Anu dat Nina.nom.f,s her neighborhood in appear.perf.f,s
‘Anui saw Ninaj in her∗ i,j neighborhood.’
(Mohanan and Mohanan 1994: 175)
Further, examples similar to (82) from Hindi-Urdu include (83), where the
possessive pronoun refers to a discourse antecedent, and not to the dative
subject.13
Hindi-Urdu (IA)
(83) us lar.kı̄i ko us ke∗ i,j bare mẽ socnā cāhiye thā
that girl dat her about to think needed was
‘That girl should have thought about her/*herself.’
Unlike with possessive pronoun, a nominal anaphor cannot alternate with a

personal pronoun in the DSC.
(84) us lar.kei,∗ j ko apne āp pari,∗ j / us par∗ i,j bahut bharosā hai
that boy dat self-gen self on him on very trust is
‘That boyi,∗ j has confidence in himselfi,∗ j /him∗ i,j .’
Bangla (IA) has the possessive reflexive nije-r ‘self’s’ which requires a
c-commanding antecedent. In Bangla, the possessive pronoun tar ‘his/her’
is coindexed with the genitive subject just like the possessive reflexive nije-r
‘self’s,’ in specific contexts.
Bangla (IA)
(85) bou.t.ii - r nijeri / tari bap- er bar.i- r kɔtha
bride.cl- gen self’s her father- gen house- gen thought
mone por.lo
remembrance fell
‘The bride remembered self’s / her parents’ home.’
Just as in Hindi-Urdu, so in Bangla too, it is not the case that a possessive

pronoun co-refers with the non-nominative subject in all contexts. It has a
discourse antecedent alone as the only choice in some specific contexts. The
subscript j in (86) refers to a discourse antecedent. The possessive anaphor
nijer, in contrast, uniquely refers to the genitive subject (Shukla Basu, p.c.).
(86) oi chelei - t.a- r nijeri opor / or∗ i,j opor khub bhɔra ache
that boy- cl gen self on/ him on very trust is
‘That boy has a lot of confidence in himselfi,∗ j /him∗ i,j .’
In Telugu (DR) too, a possessive pronoun cannot co-refer with a non-nominative

subject for most of the speakers.
Telugu (DR)
(87) kāminii ki tanai,∗ j / āme∗ i,j samasyalu ardham avv-
Kamini dat self’s/ her problems understanding happen-
a- vu
neg- nm,p
‘Kamini does not understand her [own] / her [someone else’s] problems.’
Based on the data presented above, it is evident that a dative/genitive subject

can antecede a possessive anaphor, or a possessive pronoun in Hindi-Urdu and
Bangla (IA), and Telugu (DR), in most cases. In contrast to with a possessive
pronoun, a nominal anaphor cannot alternate with a personal pronoun in the
DSC. The reason for this is that a possessive pronoun is not a subcategorized
argument, while the nominal anaphor in the cases we discussed above is.
Further, when an anaphor occurs in a subcategorized position in Kannada and
Telugu (DR), “the VR is required” (Lust et al. 2000: 30). These facts show that
subcategorization does play an important role in anaphoric binding.
The next question is: how does one account for the occurrence of the pos-
sessive pronoun for the other set of speakers? Let us consider the case of
Hindi-Urdu. We briefly present Davison’s (2004: 155) argument: a DP which
is a subject can bind only a reflexive or reciprocal and not a pronoun. Since the
dative DP “locally binds a pronoun” in (81) above, “it is not a subject.” Thus,
the dative DP in (81) is “either a subject binding a reflexive or not a subject,
binding a pronoun.” In the Minimalist Program, the subject moves to the Spec
position of a Tense Phrase (TP) due to the requirement of the EPP (Extended
Projection Principle).14 The EPP in simple terms means: every sentence must
have a subject. Hence, Davison (2004: 155) argues that a dative DP may or
may not move to Spec/TP depending upon whether it is an antecedent to an
anaphor or a pronoun.
NNSs as controllers of PRO

Just as nominative subjects can be controllers of PRO, NNSs can be too. PRO
is a null element that occurs in subject position, and, according to standard
assumptions, it is uncase-marked and ungoverned. In (88), the infinitival com-
plement, and in (89) the conjunctive participle, have PRO as their embedded
subject. PRO is coindexed with the matrix subject, which is a dative subject in
(88) and a locative subject in (89).
Hindi-Urdu (IA)
(88) harek baccei ko [PROi tair-nā] acchā lagtā hai
every child dat swim-to good feel pres
‘Every child likes to swim.’
(89) harek baccei mẽ [PROi yah bāt sun kar] jān mẽ jān ā gayı̄
every child in this news hear cpm life in life come went
‘On hearing the news every child became cheerful.’
In Bangla (IA) too, the genitive subject is the controller of PRO.
Bangla (IA)
(90) ɔb baccai - r- i [PROi sãtar kat.- te] bhalo lage
all child- gen- emph to swim cut- imperf pple good strikes
‘All children like to swim.’
According to Yoon (2004: 266), NNSs can control PRO in a subject-oriented

adjunct clause in Korean. In SALs too, a similar situation obtains in a nomina-
tive subject or non-nominative subject construction.
Hindi-Urdu (IA)
(91) [PROi is bāt kā patā cal- te hı̄] lar.kiyõi ko
dat this news gen knowledge go- right after girls dat
ghabr.āhat. huı̄
panickiness happened
‘As soon as PROi (the girls) came to know of this, the girlsi got panicky.’
Telugu (DR)
(92) [PROi ı̄ sangati teliya- gānē] ammāyili-i ki kangāru
dat this news know- right after girls- dat panickiness
put.t.indi
was born
A significant feature of Dravidian languages is that an NNS in an adjunct clause

can be the controller of a null subject in the matrix clause. The null element in
the matrix clause is denoted by ∀ (see chapter 7 for more details). The matrix
predicate in (93) takes a nominative subject. Hence, we have glossed the null
element ∀ as nom (see chapter 7 for a discussion of this kind of coindexing of
a null matrix subject with an embedded subject, which we label as ‘backward
control’).
Telugu (DR)
(93) [[ammāyilai - ki ı̄ sangati teliy- a gānē] ∀i kangāru pad.d.āru]
girls- dat this news know- as soon as nom panicky fell
In contrast, in IA languages that we know of, such sentences are not permitted.
Hindi-Urdu (IA)
(94) *[lar.kiyõi ko is bāt kā patā cal- te hı̄] ∀i
girls dat this news gen knowledge go- as soon as nom
ghabr.ā gayı̄˜
panicky went
Bangla (IA)
(95) *[baccai - r jɔr aa matro- i] ∀i u- e por.lo
child- gen fever come as soon as- emph lay- cpm fell
‘As soon as the child had fever, he lay down.’
Sentences (94) and (95) are grammatical only if the matrix subject is present
and the embedded subject is not present. That is, Hindi-Urdu and Bangla
permit only Forward Control, and not Backward Control (see chapter 7 for
more discussion).
In conjunctive participial clauses too, the dative subject can be a controller
of PRO.
In Telugu (DR) and Kharia (Munda), the dative subject controls PRO.
Telugu (DR)
(96) [sı̄ta kui [PROi ā māt.a vin- i] kōpam vaccindi]
Sita dat that matter hear- cpm anger came 3s,nm
‘Having heard that matter, Sita got angry.’
(Lalitha Murthy 1994)
Kharia (Munda)
(97) [etwa- tei [PROi u-ki ya? haleit yo- kon] lebui la?- ki]
Etwa- obl these of condition see- cpm love epm- pst
‘Etwa, seeing their condition, felt compassion.’
(Peterson 2006; PRO in (97) was added for ease of exposition)
Tibeto-Burman languages such as Bodo, Rabha and Kokborok that have been
in intense contact with Indo-Aryan languages have a non-nominative subject
construction. The only language that we know of that has no apparent con-
tact with an Indo-Aryan language but has a non-nominative construction is
Manipuri. Evidence from Khasi (Mon-Khmer) cannot be adduced, as Khasi
does not have constructions involving an NNS.
We shall now consider passive sentences. The subject of a passive sentence
is the controller of PRO in Hindi-Urdu (IA) in (98).
Hindi-Urdu (IA)
(98) baccõi se [PROi yah bāt sun kar] cup rah- ā nahı̄˜ gayā
children by nom this news hear cpm quiet be- perf not go+pst
‘The children could not keep quiet on hearing the news.’
Davison (2004: 146) points out that Hindi-Urdu “does not allow a new gram-
matical subject to be created by a process like passive,” as the ungrammaticality
of (99) indicates.
(99) *[PRO∗ i/∗ j ghar badal kar] usi - ko apnı̄i/∗ j d.āk

home change[−tr] cpm he/she- dat self’s mail
pahũc-ā- yı̄ nahı̄˜ ga-yı̄
arrive-cause- perf.f,s not go-perf.f,s
[PRO∗ i/∗ j having moved], he/shei couldn’t be forwarded self’s∗ i/∗ j mail.’
(Grammatical as ‘Because I moved, I couldn’t forward him/her/my mail.’)
(Davison 2004: 146; emphasis in the original)
Just as a nominative subject can occur in the position of PRO, so can an NNS
in a conjunctive participle. The embedded predicate kōpam vacc ‘anger come’
in Telugu and the other Dravidian languages, cāhi-era ‘needed’ in Nepali and
tsakh khas ‘anger climb’ in Kashmiri require a dative subject. PRO is glossed
here with the case marker that would normally occur with a lexical subject.
Telugu (DR)
(100) mamatai [PROi kōpam vacc- i] vel.l.i pōyindi
Mamata dat anger come- cpm left
‘Having gotten angry, Mamata left.’
Nepali (IA)
(101) [PROi pāni cāhi- era] ui nadi tira ga- yo
dat water need- cpm he river toward go- pst
‘Needing water, he went to the river.’
(Ichihashi-Nakayama 1994 as quoted in Bickel 2004: 81)
Kashmiri (IA)
(102) [PROi tsakh khas- ith] tul lar.kani or
dat anger climb- cpm lifted boy.erg noise
‘After the boyi got angry, hei raised hell.’
(Bhatt 1999: 196)
In Kashmiri (IA) and Telugu (DR), the matrix subject (controller) can be a
dative subject too.
Kashmiri (IA)
(103) [PROi tsakh khas- ith] āv salim- as vadun
dat anger climb- cpm came Salim- dat crying
Literally: ‘Having become angry, crying came to Salim.’
‘Having become angry, Salim began to cry.’
(Aadil Kak, p.c.)
Telugu (DR)
(104) [PROi kōpam vacc- i] ānti ki ēd.upu vaccindi
dat anger come- cpm Shanti dat crying came
Literally: ‘Having become angry, crying came to Shanti.’
‘Having become angry, Shanti began to cry.’
The fact that PRO occurs in a case-marked position in (103) and (104) shows
that PRO is case-marked (see chapter 7 for details). In contrast, in Hindi-Urdu
(Davison 2004), Oriya (Beermann and Hellan 2002: 45) and Punjabi, PRO
cannot occur in the subject position of an embedded conjunctive participial
clause which has a predicate that takes a dative subject.
Hindi-Urdu (IA)
(105) *rādha [S2 PRO gussā ā karS2 ] bāhar cal- ı̄ ga- ı̄
Radha dat anger come cpm outside walk- pst go- pst
Intended meaning: ‘Having felt angry, Radha went outside.’
Punjabi (IA)
(106) *baccā [S2 PRO pu’kkh lag keS2 ] ro rı́ā ai
child dat hunger feel cpm cry prog pres
Intended meaning: ‘Having felt hungry, the child was crying.’
(Sandeep Gupta, p.c.)
Oriya (IA)
(107) *mote jara ho- i ousadha khā- il- i
I.dat fever happen- cpm medicine eat- pst- 1,s
‘Having had fever I took medicine.’
(Beermann and Hellan 2002: 45)
In Bangla (IA), PRO cannot occur in a position where the genitive subject
occurs. Bayer (2004: 56) observes: “In perfective participial clauses in which
a PRO subject is required, the nominative is replaced by PRO [as in (108)], but
the genitive cannot be [as in (109)].”
Bangla (IA)
(108) [PRO hẽe- hẽe] ram amake bolchilo je ---
NOM laughing laughing Ram me told that ---
‘Constantly laughing, Ram told me that . . . ’
(109) *[PRO hãi peye] ram amake bolchilo je ---
GEN laugh having.gotten Ram me told that ---
Intended meaning: ‘Constantly laughing, Ram told me that . . . ’
(Bayer 2004: 56)
However, as Probal Dasgupta (p.c.) points out: “PRO can occur in subject
position of an (embedded) conjunctive participial clause provided that both
that clause and the main clause have a genitive/dative subject.”
Bangla (IA)
(110) [S2 PRO hɔt.at rege giy- eS2 ] amar khub matha dhore gælo
gen suddenly angry become- cpm I.gen very head hold.cpm went
‘Having become suddenly angry, I got a headache.’
A dative subject can be a controllee in an infinitival clause in Telugu (DR)

(Subbarao and Bhaskararao 2004: 176), while it cannot be in Hindi-Urdu,
Kashmiri and Bangla (IA).
In Telugu (DR), the predicate jvaram rāvad.am ‘getting fever’ requires
a dative subject in embedded subject position and PRO occurs in that
position.
Telugu (DR)
(111) mallika [S2 PRO krindat.i nela ı̄ t.aimu lō jvaram rā-
Mallika dat last month this time in fever come-
vad.amS2 ] gurtu cēsu kon- di
inf remember do VR- 3snm
‘Mallika remembered getting fever at this time last month.’
Hindi-Urdu (IA)
(112) *mallikā ne [S2 PRO pichle mahı̄ne is vaqt bukhār ā- nāS2 ]
Mallika erg dat last month this time fever come- Inf
yād kiyā
memory did
Intended meaning: ‘Mallika remembered getting fever at this time last month.’
Jayaseelan (2004: 235) demonstrates that a dative subject in Malayalam too

can control PRO, but points out: “PRO can be controlled also by a non-subject;
therefore, control of PRO is not a good test of ‘subjecthood.’” He prefers to have
a test where “the dative NP can be a PRO” (Jayaseelan 2004: 235). Sentences
(100)–(102) above from Telugu, Nepali and Kashmiri qualify Jayaseelan’s
criterion, and hence we conclude that a dative subject passes this subjecthood
test.
NNSs and subject-oriented verbs

Compound verbs in SALs are a verb + verb combination in which the first verb
is the main verb and the second verb is a vector or an auxiliary verb (Hook
1973). Vector verbs are ‘subject-oriented.’ Thus, vector verbs are permitted with
“both ergative and dative subjects, as well as nominative subjects” (Davison
2004: 147). Davison uses the occurrence of vector verbs in the non-nominative
subject construction as a piece of evidence to demonstrate that non-nominative
subjects behave like nominative subjects for a set of subject-oriented vector
verbs. The thrust of her argument is the following: a vector verb occurs in
sentences with a nominative subject as well as those with a non-nominative
subject. Hence, non-nominative subjects behave like nominative subjects in
sharing this specific property. She draws her evidence (114a) from Hindi-Urdu
(IA) to substantiate her claim. We provide additional evidence from Hindi-Urdu
(114b), Bangla (IA) and Telugu (DR). Such evidences can be adduced from
other SALs too.
t h e v e c t o r v e r b bait.hnā ‘ t o s i t ’ w i t h a nominative s ubject

Hindi-Urdu (IA)
(113a) kuch ritedār subah- subah ghar ā bait.he
some relatives morning- morning house come sat
‘Some relatives came to our place early in the morning.’ (The speaker is
expressing his unhappiness by using the vector verb bait.hnā ‘to sit.’)
t h e v e c t o r v e r b jānā ‘ t o g o ’ w i t h a nominative s ubject

(113b) kuch ritedār subah- subah ghar ā gaye
some relatives morning- morning house come went
‘Some relatives came to our place early in the morning.’ (The speaker is
expressing his unhappiness by using the vector verb jānā ‘to go.’)
t h e v e c t o r v e r b bait.hnā ‘ t o s i t ’ w i t h a non-nominative s ubject

(114a) mujhe us- par krodh ā bait.h- ā
I.dat he- on anger come sit- perf
‘I couldn’t help getting angry at him/her.’
(Davison 2004: 147)
The vector verb bait.hnā ‘to sit’ suggests something was done which had bad
consequences, perhaps unintended, not done on purpose.
t h e v e c t o r v e r b jānā ‘ t o g o ’ w i t h a non-nominative s ubject
(114b) mujhe acānak cot. lag gay- ı̄
I.dat suddenly injury strike go- perf
‘I got hurt suddenly.’
Davison (2004: 147) comments: “For the speakers who accept sentences such
as [(114a)], these auxiliaries cut across subject case possibilities.” A similar
comment holds for (114b) too.
In Telugu (DR) as well, vector verbs such as kurconu ‘to sit,’ cāvu ‘to die,’
pō ‘to go’ are nominative subject-oriented, but they may freely occur with a
non-nominative (dative and locative) subject too.
t h e v e c t o r v e r b caccu ‘ t o d i e ’ w i t h a nominative s ubject
Telugu (DR)
(115) proddunna- proddunna appulavāl.l.u int.i mundu vacc- i
morning- morning lenders (nom) house in front of come- cpm
caccēru/ kūrconnāru
died (vector verb) sat (vector verb)
‘The moneylenders are sitting in front of the house early in the morning.’
(The speaker is expressing his anger by using the vector verbs caccu
‘to die’ / kūrconu ‘to sit.’)
t h e v e c t o r v e r b caccu ‘ t o d i e ’ / kūrconu ‘ t o s i t ’ w i t h a
non-nominative s u b j e c t w i t h a d a t i v e s u b j e c t
Telugu (DR)
(116) panimanis.i ki proddunn(a)- e jvaram vacc- i
servant dat in the morning- emph fever come- cpm
caccindi/ kūrcondi
died (vector verb) sat (vector verb)
‘The servant got fever right in the morning.’ (The speaker is unhappy
about it.)
w i t h a l o c a t i v e su b j e c t
(117) vād.i- daggara inta d.abbu und.- i cacc- i
he.obl- near so much money be- cpm die (vector verb)- cpm
evari- ki ēmi lābham
who- dat what profit
‘Who gets benefited in spite of the fact that he has so much money?’
In Bangla (IA) too, a similar phenomenon is observed. The vector verb in

(118) and (119) is ja ‘to go,’ and it is nominative subject-oriented in (118), and
genitive subject-oriented in (119).
t h e v e r b ja ‘ t o g o ’ w i t h a no m i n a t i v e s u b j e c t
Bangla (IA)
(118) projit bar.i- te col- e gælo
Projit house- to go- cpm went
‘Projit went home.’
t h e v e r b ja ‘ t o g o ’ w i t h a g e n i t i v e s u b j e c t
(119) projit- er ɔukh kor- e gælo
Projit gen ill do- cpm went
‘Projit fell ill.’
(Probal Dasgupta, p.c.)
The evidence presented above clearly demonstrates that there is a set of vector
verbs that are nominative as well as non-nominative subject-oriented.
We discuss below another kind of evidence for the subject property of the
dative subjects from Marathi, in which the complementizer can be reduced.
Phonological evidence from the final complementizer

in Marathi
Another kind of evidence for the subject property of dative subjects comes
from Marathi, in which the complementizer can be reduced due to the process
of the phonological attrition of the first vowel a- of the complementizer asə.
The post-sentential COMP in Marathi (IA) has two alternative forms: asə
and -sə. The latter, a result of the elision of the initial vowel, occurs only when
the subjects of the matrix and embedded sentence are identical. In (120a), the
subject of the matrix clause is a dative subject and the notional subject of the
embedded complement is a nominative subject. Phonological attrition of a- in
the complementizer asə takes place in (120a), which demonstrates that both
the dative subject and nominative subject have identical properties of a subject.
s u b j e c t o f t h e m a t r i x a n d e m b e d d e d c l a u s e i d e n t i c a l : sə o c c u r s
(120a) ma- lāi [CP [S2 PROi mumbai- lā dzā- wa-S2 ] səCP ]
I- dat nom Mumbai- to go- subjunctive- COMP
wāt.ate/wāt.ata
feel.pres s
‘I feel like going to Mumbai.’
Note that such deletion of the vowel is not permitted if the subjects are non-
identical. In (120b) the subject of the matrix clause is ma-lā ‘I-dat’ and of the
embedded clause is tyā-ne ‘he-erg’ and they are not identical.
s u b j e c t o f t h e m a t r i x a n d e m b e d d e d c l a u s e not i d e n t i c a l
(120b) ma- lāi [CP [S2 tyā- nej mumbai- lā dzā- wa-S2 ] ase/
I- dat he- erg Mumbai- to go- subjunctive- COMP
∗
səCP ] wāt.ate/wāt.ata
COMP feel.pres s
‘I think he should go to Mumbai.’
(Kashi Wali, p.c. to Prashant Pardeshi)
To summarize the above discussion, non-nominative subjects behave like nom-

inative subjects as: (i) antecedents to an anaphor, (ii) controller of PRO, and
(iii) subjects for a set of subject-oriented verbs. Finally, phonological evidence
from the FC in Marathi provides further support to our claim.
5.4.2 Some non-subject properties of NNSs in SALs

There are some non-subject properties of the non-nominative subjects that we
shall present now.
Agreement
It is a well-attested fact that in most of the SALs, except in Manda (DR)
(Ramakrishna Reddy 1992b) and Maithili (IA) (Subbarao 2001), the NNS does
not trigger agreement on the verb. However, agreement cannot be considered
as a viable test for subjecthood, as verbs in SALs agree with non-subjects too
(see chapter 4).
The case of modals

We now provide evidence from Telugu (DR) (Subbarao and Bhaskararao 2004)
which shows that non-nominative subjects lack some properties that nominative
subjects possess. The modal gala ‘can, might’ in Telugu permits epistemic and
deontic meaning, when the subject is in the nominative case. However, when the
subject is dative case-marked, the capabilitative meaning is absent, and only the
possibility interpretation is permitted (Subbarao and Bhaskararao 2004: 179).
We reproduce the argument below from Subbarao and Bhaskararao (2004),
omitting some minor details.
The modal gala has a capability/possibility interpretation in nominative–
accusative constructions (121), and there is no corresponding capability inter-
pretation in the dative subject construction (122). Thus, in (122), the dative
subject construction, gala has the interpretation only of possibility, and not of
capability.
m o d a l w i t h a no m i n a t i v e su b j e c t : possibility , a nd capability
Telugu (DR)
(121) vād.ui ı̄ sangatuluj rēpu telusu- kō- galad.ui,∗ j
he this news tomorrow know- VR- can/might
‘He might/can find out this news tomorrow.’
m o d a l w i t h a no n - n o m i n a t i v e s u b j e c t : possibility , a nd
not capability
(122) vād.i- kii ı̄ sangatuluj rēpu teliya- gala- vu∗ i,j
he.obl- dat this news (p) tomorrow known- might/*could- 3 p,nh
‘He will/might get to know this news tomorrow.’
*‘He can get to know this news tomorrow.’
While the modal gala with capability meaning in (121) has a corresponding
negative form (123a), the modal with possibility interpretation does not have a
corresponding negative form (123b).
(123a) vād.u ı̄ sangati rēpu telusu- kō- lē. d.u

he.nom this news tomorrow know- VR- neg.can.3 s,m
‘He cannot find out this news tomorrow.’
(123b) *vād.i- ki ı̄ sangatulu rēpu teliya- lē.vu

he.obl- dat this news (p) tomorrow known- neg.can.3 p,nh
Intended meaning: ‘He will not find out this news tomorrow.’
Further, in the passive voice too, the modal gala ‘can’ behaves differently
from the sentence in active voice. In active voice the modal gala ‘can’ does
not overtly occur in Telugu when the negative occurs as in (124). It is the
agreement on the negative that imparts the capabilitative interpretation. With
such interpretation in (124), there is no corresponding passive sentence with

modal interpretation, as the ungrammaticality of (125) shows.
(124) vād.u ilāt.i cetta panulu ceyya lēd.u
he such useless deeds do cannot.3 s,m
‘He can’t do such useless things.’
(125) ?*vād.i- cēta ilāt.i cetta panulu ceyya- pad.a- lēvu
him- by such useless things do- passive- neg.can.3 p,nh
‘Such useless things cannot be done by him.’
Telugu uses a lexical passive in such cases (126). A lexical passive does not carry
the overt morphology that a passive predicate carries; it imparts the meaning
of a passive, as the verb in such cases is [−transitive]. Thus, in syntactic and
lexical passives, the predicate is invariably [−transitive].
(126) vād.i- valla ilāt.i cetta panulu kā- vu/
he.obl- by such useless deeds cannot.happen- 3 p,nh
jaraga- vu
will.not.happen- 3 p,nh
‘Such useless things cannot be / will not be done by him.’
Evidence similar to Telugu (DR) can be adduced from Hindi-Urdu (IA) too,
from the difference exhibited by the modal sak-nā ‘can.’ Just like the modal
gala ‘can, might’ in Telugu, saknā ‘can, might’ in Hindi-Urdu (IA) has a capa-
bility and a possibility interpretation in nominative–accusative constructions
(127); there is no corresponding capability interpretation in the dative subject
construction (128). It has only the possibility meaning.
Hindi-Urdu (IA)
(127) ren.u yah bāt kal malūm kar saktı̄ hai
Renu this thing tomorrow learn do can pres
‘Renu can [capability] /might [possibility] find out this tomorrow.’
(128) ren.u ko yah bāt kal malūm ho saktı̄ hai
Renu dat this thing tomorrow learn be can pres
‘Renu might get to know this tomorrow.’ (possibility)
*‘Renu can find out this tomorrow.’ (capability)
From a semantic point of view too, the dative subject construction differs from
the nominative subject construction with regard to the feature of volitionality. It
is generally agreed that the predicate in the dative/genitive subject constructions
is [−volitional].
Krishnamurti (1975), McAlpin (1976), Klaiman (1979) and Pandharipande
(1990) point out that the predicate in the dative/genitive subject construction
is [−volitional]. Hence, adverbs such as kāvāli ani ‘deliberately,’ kōru-koni
‘desirously,’ i.tam gā ‘willingly,’ kutūhalam gā ‘anxiously,’ ātruta gā ‘eagerly’
and anu kōkund.ā ‘unintentionally’ in Telugu (DR), jān būjh kar ‘deliberately,’
cāh kar ‘desirously,’ utsuktā se ‘eagerly’ and samhal kar ‘carefully’ in Hindi-
Urdu (IA) and icche kore ‘deliberately’ in Bangla (IA), and jān. ke ‘intention-
ally’ in Punjabi (Bhatia 1993: 87) cannot occur in a dative, genitive or locative
subject construction (see papers in Verma and Mohanan 1990 for a discussion
on the non-volitional nature of DSCs).15
Evidence from Telugu (DR) and Hindi-Urdu (IA) clearly shows that, while
nominative subjects and dative subjects do share many structural properties,
they differ with regard to the semantic feature of volitionality.
Coordinate reduction
The issue with regard to NNS constructions and coordinate reduction is whether
it is the morphological identity in terms of case marking on the conjuncts that
counts or the grammatical function that is crucial. In Icelandic, the subject of
the second conjunct can be elided, though it might be differently case-marked
from the subject of the first conjunct. We shall demonstrate that three out of four
SALs we checked permit differently case-marked subjects in the two conjuncts.
Thus, our data support the view that “it is not the morphological identity that
counts, it is the grammatical function” (Zaenen, Maling and Thráinsson 1985).
Bayer (2004: 57–58) shows that, in German and Bangla (IA), conjunction
reduction is not permitted when the subjects of the conjuncts are not identically
case-marked, as in (129). Sentence (129) is ungrammatical because “nominative
and genitive subjects cannot mix in deletion contexts” (Bayer 2004: 57–58).
Bangla (IA)
(129) *ami bar.i theke gelam ar *(amar) kanna peyeche
I home from went and I.gen crying came
‘I left the house and I felt like crying.’
(Bayer 2004: 57)16
Probal Dasgupta (p.c.) provides an example to show conjunction reduction is

permitted in Bangla (IA), though the subjects are not identical. The genitive
subject of the second conjunct is elided.
(130) ami chobi- t.a dekhlam kintu, koi, kanna pelo

I picture- cl watched but where (what) weeping light verb
na to?
not particle
‘I did watch the movie but, well, didn’t burst into tears.’
Sentence (130) shows that the nominative subject of the first conjunct and the
genitive subject of the second conjunct can mix in deletion contexts in Bangla
(IA) and, hence, the genitive subject can be elided.
Mohanan (1994: 131) demonstrates that “in coordination constructions in

Hindi, the gapped element in a coordinate constituent must be identical to the
gapper in case.” In (131), the gapper and gapped elements are both ergative
case-marked, and hence it is grammatical, while in (132) the gapper is nom-
inative case-marked and the gapped element is ergative case-marked – hence
the ungrammaticality.
Hindi-Urdu (IA)
(131) ravı̄ ne khānā khāyā aur ø pikcar dekhı̄
Ravi erg meal ate and erg movie watched
‘Ravi ate his meal and watched a movie.’
(132) ravı̄ ghar gayā aur *ø khānā khāyā
Ravi home went and erg meal ate
‘Ravi went home and he ate his meal.’
(Mohanan 1994: 131)
Thus, Hindi-Urdu requires strict case identity of the subjects in both conjuncts.
In contrast, in Nepali (A), sentences of the type in (133), in which the
first conjunct has a dative subject, and the second a nominative subject, are
permitted.
Nepali (IA)
(133) us.lāı̄i jyādai tirkhā lāg-yo ra ɵi gāū-
he.dat much:focus thirsty perceptible-pst.3s and nom village-
eu-t.ā pas-yo
one-cl enter-pst.3sm
‘He was thirsty and went into a village.’
(Ichihashi-Nakayama 1994, as quoted in Bickel 2004: 81)
Telugu (DR), too, is similar to Nepali (IA) and permits sentences of the type in
(134), as does Nepali.
Telugu (DR)
(134) āvid.a ki ēd.upu rānu- u vaccindi proi kas.t.am mı̄da
she dat crying come- conjn came she.nom difficulty on
āpukonu- u āpukondi
stop- conjn stopped
Literally: ‘Crying came to her, but she even controlled it with difficulty.’
‘She was about to cry (but) she controlled herself with difficulty.’
Based on the fact that Bangla, Nepali (IA) and Telugu (DR) permit non-
identical case-marked subjects in two conjuncts, we can tentatively conclude
that coordination can be used as a piece of evidence to demonstrate that NNSs
behave like nominative subjects in this context.
5.5 Nature of the predicate in DSCs

We shall discuss the stative–non-stative nature of the NNS in section 5.5.1 and
the [−transitive] nature of the DSC in section 5.5.2.
5.5.1 [+/−stative] nature of the verb in NNS constructions

In NNS constructions, the verb may be either stative ((135) and (136)), or non-
stative ((79) and (137)). Krishnamurti (1975) proposes the term ‘en-stative’
(‘entering into a state,’ as in ‘I came to know that’) for one set of predicates
(Sridhar 1979).
[+s t a t i v e ]
Hindi-Urdu (IA)
(135) un baccõ ko yah bāt mālūm thı̄
those children dat this matter known was
‘Those children knew of this.’
Telugu (DR)
(136) ā pillala- ki ı̄ sangati telusu- at.a
those children- dat this matter known- ev mkr
‘It seems that those children knew this thing.’
[ −s t a t i v e ]
Sentence (137) is an example with a [−stative] predicate.17
Telugu (DR)
(137) ānti ki eppud.ū kōpam vastūn.ē unt.undi
Shanti dat always anger coming.emph keeps
‘Shanti keeps getting angry all the time.’
Kachru (1990: 67), discussing experiencer and other oblique subjects in Hindi
(IA), points out that the dative subject occurs in transient psychological states.
She further points out that transient psychological states, beliefs, knowledge,
want, need, etc., can be expressed with active (nominative subject) constructions
too. To summarize, the predicates that take non-nominative subjects may be
either stative or non-stative.
5.5.2 [−transitive] nature of the verb in NNS constructions

Jayaseelan (1990), Pandharipande (1990), Shibatani and Pardeshi (2001),
Subbarao (2001) and Subbarao and Bhaskararao (2004) claim that the
predicate in NNS constructions is [−transitive]. Sigurdsson (2000) and Amri-

tavalli (2004) claim that the predicate in NNS constructions is unaccusative.
In this section, we provide evidence from Hindi-Urdu (IA) and Telugu
(DR), first in support of the claim that the predicate in NNS constructions
is [−transitive]. We also show that the predicate in a DSC or passive is non-
subject-oriented. We shall then examine the putative evidence from Bangla
(IA), Tamil and Malayalam (DR), where the theme is accusative/dative case-
marked, and demonstrate that the predicate in the DSC in those three lan-
guages too is [−transitive], and the accusative marker that occurs with the
theme/patient is a specificity marker.
In support of our hypothesis that the predicate in NNS constructions is
[−transitive], we provide three pieces of evidence: (i) no accusative case mark-
ing of the theme in Exceptional Case Marking instances in DSCs; (ii) the
non-occurrence of complex anaphors in DSCs; and (iii) the non-availability of
the passive in a DSC.
Case marking
In sentences involving Exceptional Case Marking in SALs, the matrix verb
assigns accusative case to the embedded subject and, hence, it carries the
accusative case marker.
(i) The case of Hindi-Urdu (IA)

In Hindi-Urdu, the verb mān-nā ‘to consider’ is [+transitive], and hence šarmilā
‘Sharmila,’ the derived direct object in (138), carries the accusative case marker
ko, while, in (139), the dative predicate lagnā ‘to appear’ is [−transitive], and
hence accusative case marker ko is not permitted.
Hindi-Urdu (IA)
(138) ham sab armilā ko ı̄māndār mānte rahe
we.nom all Sharmila acc trustworthy consider kept
‘We all had been considering Sharmila trustworthy.’
(139) ham sab ko armilā (*ko) ı̄māndār lagı̄
we all dat Sharmila acc trustworthy appeared
‘Sharmila appeared to be trustworthy to all of us.’
Sentence (138) is a case of Exceptional Case Marking and the verb mān-nā ‘to
consider’ is [+transitive]. Hence, the predicate assigns accusative case to the
patient armilā ‘Sharmila.’ In contrast, the dative predicate lagnā ‘to appear’ in
(139) is [−transitive], so it cannot assign accusative case to the theme/patient
armilā ‘Sharmila.’ The derived subject armilā does not permit the accusative
case marker ko to occur in (139).
Further, evidence comes from the non-occurrence of complex/simplex

anaphors in Hindi-Urdu (see Davison 2000 for a detailed discussion). An
anaphor in an argument position subcategorized by the predicate needs to be
case-marked accusative or dative. Hence, if our hypothesis that dative predi-
cates are [−transitive] is correct, a lexical anaphor (either complex or simplex)
cannot be permitted when the dative predicate is the case assigner. Sentence
(140) proves the point.
(140) *ham sab ko apne āp (ko) imāndār lage

we all dat self+gen self acc trustworthy appeared
Intended meaning: ‘We all appeared to be trustworthy to ourselves.’
Imperfect and perfect participial constructions in Hindi-Urdu also support our

contention that the dative predicate is [−transitive] (Subbarao and Bhaskararao
2004). Sentence (141) with an imperfect participle is the result of Exceptional
Case Marking (accusative) of the embedded subject by the matrix verb dekhnā
‘to see,’ a transitive verb.
Hindi-Urdu (IA)
(141) surabhi ne kriti ko nācte.hue dekhā
Surabhi erg Kriti acc dancing saw
‘Surabhi saw Kriti dancing.’
If the matrix sentence has a dative predicate, the embedded subject cannot be
case-marked accusative (142) as the dative predicate is [−transitive].
(142) *surabhı̄ ko kriti ko nācte.hue dikhāyı̄.par..ı̄

Surabhi dat Kriti acc dancing came visible
Intended meaning: ‘Kriti appeared to Surabhi to be dancing.’
In (142), kriti ‘Kriti’ has to be in the nominative case because the dative
predicate dikhāyı̄ par.nā ‘to appear,’ which is unaccusative, cannot accusative
case-mark the embedded subject. Hence, it is nominative case-marked in (143).
(143) surabhı̄ ko kriti nācte.hue dikhāyı̄.par..ı̄

Surabhi dat Kriti. nom dancing came visible
‘Surabhi saw Kriti dancing.’
Similar evidence can be adduced with perfect participles too.
(ii) The case of Telugu (DR)

In the case of Telugu (DR), a similar phenomenon is seen (Subbarao and
Bhaskararao 2004). The verb bhāvincu ‘to consider’ is [+transitive], and can
exceptionally case-mark the embedded subject prasād ‘Prasad’ accusative in
(144).
w i t h a no m i n a t i v e p r e d i c a t e
Telugu (DR)
(144) nēnu prasād ni nammakastud.u- gā bhāvistunnā-nu
I.nom Prasad acc trustworthy- adjr consider.1 s
‘I consider Prasad trustworthy.’
In contrast, the verb anipincu ‘to feel’ is a dative predicate, and it too per-
mits exceptional case marking. If our hypothesis that dative predicates are
[−transitive] is correct, then the embedded subject cannot be accusative case-
marked by a dative predicate. If the embedded subject is nominative case-
marked, the sentence is grammatical. Our prediction turns out to be correct in
(145).
exceptional case marking with a dative predicate
(145) nā-ku prasād (*ni) nammakastud.u- gā anipistunnād.u
I-dat Prasad acc trustworthy- adjr appears.3 s,m
‘Prasad appears trustworthy to me.’
Thus, the evidence from case marking in Telugu (DR) and Hindi-Urdu (IA)
shows that the predicate in NNS constructions is [−transitive]. However,
there appears to be some putative counterevidence to our claim regarding the
[−transitive] nature of dative predicates. In Tamil, Malayalam (DR), a dative
predicate, and in Bangla (IA) and Bodo (TB), a genitive predicate, appear to
assign a non-nominative case to the theme.
(iii) The case of Bangla (IA)

A theme which is [+animate] in Bangla (IA) carries the accusative case marker
as in (146).
(146) tomar18 kake cai
you.indirect who.acc want
‘Whom do you want?
Creative errors made by Bangla learners/speakers of Hindi-Urdu show that the

phenomenon of assigning accusative case marker to the theme is transferred to
Hindi-Urdu, as in (147).
hindi-urdu spoken by a bangla learner/speaker
(147) *āp ko kis ko cāhiye
you dat who acc needed
Intended meaning: ‘Who do you want?’
In standard Hindi-Urdu (IA), in such cases the theme kaun ‘who’ is in the
nominative case, as cāhiye ‘needed’ is [−transitive].
Hindi-Urdu (IA)
(148) āp ko kaun cāhiye
you dat who.nom needed
‘Who do you want?’
Let us first examine the Bangla (IA) data further. In Bangla, the features animacy
and specificity play an important role in the assignment of the accusative case
marker ke to the theme, just as in many Indo-Aryan languages – for example,
in Hindi-Urdu (Mahajan 1990) and in Marwari (IA) (Magier 1987, 1990). The
accusative case marker ke does not occur when the theme is [−definite] and
[−animate].
Bangla (IA)
(149) rina- r kichu bhalo lage na
Rina- gen anything good appear neg
‘Rina does not like anything.’
(150) rina- r kono jini bhalo lage na
Rina- gen any thing good appear neg
Note that the marker ke is not present with the theme kichu ‘anything’ or kono
jini ‘anything,’ which clearly shows that ke is a marker that does not occur
when the theme is [−definite] and [−animate]. If the theme is under focus or
contrastive stress, the marker ke occurs, as in (151), as Probal Dasgupta (p.c.)
points out.
theme under contrastive focus

(151) rina- r kono jini ke- i otti-otti19 bhalo lage na
Rina- gen any thing acc- emph really good appear neg
‘Rina does not really like any thing at all.’
Dasgupta further points out that the correlation between the behavior of the
patient in the experiencer subject sentences (see (149) and (150)) and in the
agent (nominative subject) sentences (see (152) and (153)) is exact. There is
an interaction with animacy and specificity, but that interaction is identical in
the two clause types. Note that the marker ke is not present in (152) and (153),
while it is present in (154), when the theme is under contrastive focus.
t h e m e (i n i t a l i c s ) u n d e r n e u t r a l f o c u s
(152) rina kichu pɔchondo kɔre na
Rina anything liking does not
(153) rina kono jini pɔchondo kɔre na

Rina any thing liking does not
t h e m e (i n i t a l i c s ) u n d e r c o n t r a s t i v e f o c u s
(154) rina kono jini ke- i otti-otti pɔchondo kɔre na
Rina any thing acc- emph really liking does not
‘Rina does not like really anything at all.’
Hence, we can conclude that the marker ke in Bangla in the nominative–

accusative construction and the genitive–accusative construction is a specificity
marker, and not an accusative marker, just like the marker ko in Hindi-Urdu
(IA), which is treated as a specificity marker (Mahajan 1990; Magier 1987,
1990).
In the case of Malayalam and Tamil (DR) too, a similar fact is observed.
(iv) The case of Malayalam (DR)

We now provide evidence to show that the accusative marker -ye in Malayalam
functions as a specificity marker. Sentence (155) is a DSC, and the accusative
marker ye occurs with the theme āna ‘elephant.’
Malayalam (DR)
(155) kut..ti- k’k’ə āna- ye it.am āyi
child- dat elephant- acc liking became
‘The child liked the elephant.’
Interestingly, this construction alternates with a nominative subject construction

(156a). The accusative marker -ye occurs with the theme.
Malayalam (DR)
(156a) kut..ti āna- ye is.t.a- ppet.t.- u
child elephant- acc liking- ?- pst
‘The child liked the elephant.’
When the theme is [−animate] and [−definite], the accusative marker -ye does
not occur (156b).
(156b) en- ik’k’ə oru māŋŋa vēn.am

I- dat one mango.nom want
‘I want a mango.’
Thus, the features animacy and definiteness explicate the occurrence of the
accusative marker -ye, and it is not the transitive nature of the predicate that is
instrumental in its presence.
(v) The case of Tamil

Just like Malayalam, Tamil (DR) too permits an accusative case-marked theme
in a DSC (Paramasivam 1979: 65–66, Lehmann 1989: 184, Schiffman 2000: 37
for Tamil). Lehmann (1989: 184) labels such DSCs as the DAT–ACC pattern.
According to him, the predicates that require this pattern are:
(a) verbs of mental experience, such as teri ‘to know,’ puri ‘to understand’;
(b) verbs of emotional experience, such as pit.i ‘to like’; and
(c) verbs of physical and biological experience, such as paci ‘to be hungry,’
vali ‘to feel pain,’ ari ‘to itch,’ kūcu ‘to feel ticklish.’
Lehmann (1989: 184) treats these predicates as morphologically defective, as
they exhibit agreement in the neuter. This, of course, is expected as there is no
nominative case-marked subject to agree with. Hence, it should be treated as
default case like in many other SALs, such as Hindi-Urdu, Punjabi (IA) and
Telugu (DR).
Tamil (DR)
(157) kumār- ukku inta ūr- ai.t teri.y- um
Kumar- dat this place- acc know.fut- 3s,neuter
‘Kumar knows this place.’
(Lehmann 1989: 184)
In Malayalam too, the accusative marker -e occurs with the theme, just as in
Tamil.
Malayalam (DR)
(158) awal.a- kkə awan- e is.t.appet.um
she- dat he- acc like.fut
‘She will like him.’
(Shobha Nair and P. Sreekumar, p.c.)
Note that in Tamil (DR) too, the features [+animacy] and [+specificity] play
a crucial role in the occurrence of the specificity marker, as (159) and (160)
indicate. In (159) and (160), the theme inta id.am ‘this place’ is [+specific] and
[−animate] and no accusative marker occurs with the theme.
(159) kumār- ukku inta id.am teriyu- m

Kumar- dat this place know- 3 s
‘Kumar knows this place.’
(160) kumār- ukku inta id.am teri- yā- du

Kumar- dat this place know- neg- 3 s
‘Kumar does not know this place.’
According to our language consultants, P. Ananda Mohan, Vasanta Mohan and

R. Nikhil (p.c.), even the [+human] patient rājā ‘Raja’ in (161), or yār-um
‘who.nom-NPI’ which is [−specific] and [+human] in (162), need not be case-
marked by the accusative with the predicates pit.ik ‘to like’ in (161) and teri ‘to
know’ in (162).
(161) kumār- ukku rājā pit.i.kk- um

Kumar- dat Raja. nom like.fut- 3 s, neuter
‘Kumar likes Raja.’
(162) kumār- ukku yār- um teri- yā- du20

Kumar- dat who.nom- NPI know- neg- 3 s
‘Kumar does not know anybody.’
In Bodo (TB), the adjective mɯjaŋ ‘good,’ together with a tense marker,
imparts the meaning of ‘like,’ and this predicate assigns genitive case ha ‘of’
to its subject. Recall that adjectives behave like verbs in many Tibeto-Burman
languages (see chapter 2 for details). The patient in such cases is accusative
case-marked by khɯu.
Bodo (TB)
(163) khampha ha laogi- khɯu mɯjaŋ- mɯn
Khampha gen Laogi- acc good- pst
‘Khampha liked Laogi.’
We do not have further data to show that the accusative marker khɯu is a
specificity marker in Bodo.
In conclusion, though the phenomenon of accusative/dative case marking
of the theme in Bangla (IA), Malayalam and Tamil (DR) seems to suggest that
the predicate in DSCs is [+transitive], we have demonstrated that the marker
that occurs with the theme in such constructions is a marker of specificity and
animacy as Magier (1987,21 1990) and Mahajan (1990) have shown for Marwari
and Hindi-Urdu (IA), respectively. Hence, predicates in NNS constructions are
syntactically [−transitive]. We agree with Alice Davison (p.c.) who points out:
“The issue of whether dative subject verbs are [+transitive] is complicated . . .
the dative/ergative near minimal pairs in Hindi like dekh-nā ‘to see’ and dikh-
nā/dikhāı̄ de-nā ‘to be visible’ are both bivalent, i.e., transitive in argument
structure, though the dative subject version does take intransitive vectors.” She
points out that the fact that dative subject predicates take small clauses and
participial complement clauses shows their transitive nature. Hence, we feel
that one may, with reasonable certainty, conclude that dative predicates are
semantically transitive but syntactically intransitive.
Since the object marker is not an accusative case marker, its occurrence does
not count as counterevidence to the claim that the predicate in non-nominative
subject constructions is [−transitive].
Aissen (2003) discusses DOM, differential object marking, in which some
objects are case-marked, and some others are not, depending upon the semantic
and pragmatic features of the object. Aissen points out that DOM depends on
two features – animacy and definiteness – and they compete with each other
for dominance. While Persian prefers specificity, Hindi chooses animacy. The
case marking of theme in the DSC (dative/genitive subject construction, to be
very specific) in Bangla, Malayalam and Tamil, for example by the accusative,
we have observed, depends on animacy/specificity, independent of transitivity
in the [−NNS construction].
Recall that, in nominative–accusative constructions too, the accusative
marker is associated with transitivity and animacy/specificity, and hence should
be treated as a specificity marker, as it denotes specificity (see Magier 1987,
1990; Mahajan 1990).
Thus, with regard to differential object marking (DOM) in SALs, when
the accusative case marker denoting specificity occurs, the predicate is
[−transitive] in the DSC (dative/genitive subject construction), and the pred-
icate is [+transitive] in the nominative subject construction. Based on these
facts, we propose the following parameter to account for this variation.
The Differential Object Marking (DOM) parameter: When the noun
phrase is accusative case-marked, the object marker is associated with
either transitivity and animacy/specificity in the [+NNS construction],22
or purely animacy/specificity, independent of transitivity, in the [−NNS
construction].
Discussing accusative case in Kannada, Lidz (2006) notes that there are two
kinds of specificity: positional and inherent. Inherent specificity means case-
marked NPs independent of their position in a sentence get a specific interpreta-
tion. Non-case-marked NPs too can get specific interpretation depending on the
position of occurrence, which he labels as positional specificity. The accusative
marking in DSCs in Tamil, Malayalam (DR) and Bangla (IA) comes under
inherent specificity.
Anaphors
The second piece of evidence comes from anaphors in Telugu (DR) (Subbarao
and Bhaskararao 2004).
When the matrix sentence has a nominative predicate and the embedded sub-
ject is coindexed with the matrix subject and ECM takes place, a reduplicated
anaphor can occur as in (164).
Telugu (DR)
(164) mādhuri tana- ni tanu andagatte- gā bhāvistundi
Madhuri self- acc self pretty- adjr considers-3 s.nm
‘Madhuri considers herself pretty.’
However, when the matrix sentence has a dative predicate, only the simplex
form of the anaphor in nominative case can occur, and a reduplicated form in
accusative case is not permitted, as in (165).
(165) pratı̄.vād.i- kı̄ tanu/ *tana- ni tanu goppavād.u ani/

every.fellow- dat self. nom self- acc self. nom great.person COMP
gā anipistād.u
COMP appears
‘Every fellow thinks that he is great.’
A complex anaphor is not permitted due to the fact that the matrix verb which
is [−transitive] in (165) cannot assign accusative case to the embedded subject
by Exceptional Case Marking (ECM), unlike the [+transitive] verb in (164).
That is, NNS predicates cannot exceptionally case-mark and, hence, they are
[−transitive].
Passivization
Bhatt (1999) demonstrates that sentences with a dative subject do not passivize
in Kashmiri. This phenomenon is found in other SALs too.
Kashmiri (IA)
(166) *rām- an vuch [humis lar.k- as yi kūr khar- ni
Ram- erg saw that boy- dat this girl (nom) hate- pass
yiv- ān]
come- imperf
‘Ram saw the boy being hated by the girl.’
(Bhatt 1999: 201)
Dative/genitive predicates are similar to anticausatives, as both are [−transitive]

in nature. Alexiadou, Anagnostopolou and Schafer (2006) point out that anti-
causatives cannot be modified by “by- phrases, agent-oriented adverbs [or]
allow control into purpose clauses.” The passive sentence (166) from Kashmiri
(IA) supports this hypothesis.
We present below further evidence which shows that the predicate in the
DSC behaves like an anticausative. These include: (i) the non-occurrence of
agent-oriented adverbs in DSCs; and (ii) the dative subject as a controllee in
purpose clauses.
5.5.3 Agent-oriented adverbs

Agent-oriented adverbs (in italics in (167)) are not permitted in DSCs in SALs.
This is due to the fact that the predicate in a DSC is non-volitional.
Hindi-Urdu (IA)
(167) *mādhurı̄ ko is bāt par jān būjh kar gussā ā gayā
Madhuri dat this news on intentionally anger came
Intended meaning: ‘Madhuri’s anger went up intentionally at this news.’
In Telugu and the other Dravidian languages as well, such sentences are not
permitted.
5.5.4 The dative subject as a controllee

The dative subject cannot be a controllee in purpose clauses in Hindi-Urdu and
Kashmiri (IA).
Hindi-Urdu (IA)
(168) *urmilāi [PROi dat bhūkh lagne ke liye] rasoı̄ mẽ gayı̄
Urmila (controllee) hunger feel in order to kitchen in went
Literally: ‘*Urmila went into the kitchen in order to feel hungry.’
Kashmiri (IA) also does not permit such sentences.

In contrast, the dative subject can be a controllee in purpose clauses in Telugu
(DR).
Telugu (DR)
(169) d.ākt.aru gārui [PROi dat jvaram tagga- d.āni ki]
doctor polite mkr (controllee) fever become less- in order to
mandu tı̄sukont.unnāru
medicine is taking
‘The doctori is taking some medicine in order to bring hisi fever down.’
Further research into this phenomenon needs to be done.23

To summarize, we have provided several pieces of evidence to show that the
predicate in NNS constructions is [−transitive]. These include: (i) accusative
case marking of the theme/patient; (ii) the non-occurrence of complex anaphors
in NNS constructions; and (iii) the inability of NNS predicates to passivize.
We have also discussed the non-occurrence of agent-oriented adverbs in DSCs
and the dative subject as a controllee in purpose clauses.
5.6 Inherent case assignment in DSCs

It is generally accepted that dative subjects are universally inherently case-
marked, and it is the transitive verb that assigns inherent case. In this section
we wish to argue that an intransitive verb together with theme or an adjective
compositionally assigns non-nominative (dative or locative) case to the subject
vP-internally in the thematic (lower) S.
Jayaseelan (1990) argues that inherent case is assigned compositionally in
the DSC. Davison (2004), too, treats dative case assignment in DSCs as inherent
case. According to her, “dative lexical case does not require checking outside
of VP [as inherent case need not be checked]” (Davison 2004: 153). We further
argue that, for such case assignment to take place, information concerning
agreement too must be available vP-internally (in the lower thematic S).
We now provide evidence in support of our contention concerning inherent
case assignment compositionally by the predicate.
5.6.1 Default agreement and inherent case marking

Consider the following sentences from Telugu (DR) in which the (a) sentences
have a nominative subject and the (b) sentences have a non-nominative subject.
The predicate is identical in all sentences except for agreement. The verb is
un∼unn ‘to be’ and the predicate adjective is kōpam gā ‘angry’ in all the
sentences (Subbarao and Bhaskararao 2004).
(170a) vād.ui kōpam- gā unnā- d. ui
he anger- adjr be- 3 s,m
‘He is angry.’
(170b) vād.ii - ki kōpam- gā undi∗ i
he.oblique dat anger- adjr be- 3 s [def agr]
‘He is angry.’
(171a) āmei kōpam- gā un- dii
she anger- adjr be- 3 s,nm [+animate]
‘She is angry.’
(171b) āmei - ki kōpam- gā undi∗ i
she- dat anger- adjr be- 3 s,nm [def agr]
‘She is angry.’
In (170a), the verb agrees with the subject vād.u ‘he’ in 3rd person singular
masculine. In (170b), the verb does not agree with the subject, as the latter is
dative case-marked, and there is no other DP in the nominative case; kōpam-gā
‘angry’ is an adjective, and there is no agreement of the verb with an adjective.
The verb, therefore, exhibits default agreement, which is 3rd person singular
non-masculine in Telugu.
In (171a), the verb agrees with the subject āme ‘she’ in 3rd person singu-
lar non-masculine [+animate]. In (171b), the verb does not agree with the
subject – as the latter is dative case-marked – and, hence, exhibits default
agreement.
Our claim is: when there is a predicate adjective and the verb exhibits default
agreement, the subject is inherently assigned non-nominative case (dative or
locative) by the predicate compositionally. When the verb exhibits person
agreement in masculine or non-masculine [+animate], the subject is in the
nominative case. That is, the feature [+/−animate] plays a role in case assign-
ment. Alice Davison (p.c.) suggests that default agreement in T should be the
consequence of non-nominative case, rather than the way it is suggested here.24
We leave the issue open for further research.
We observe that the verb phrases in (a) and (b) sentences are identical.
What really distinguishes (a) sentences from (b) sentences is agreement. While
(a) sentences exhibit subject–verb agreement, (b) sentences exhibit default
agreement. The verb un alone cannot assign inherent case to its subject. The
adjective + verb to be + default agreement together play a crucial role in
assigning inherent case to the subject in the (b) sentences. This implies that
information concerning agreement should be available in the lower thematic S
for the proper assignment of inherent case to the subject. Since the occurrence
of the dative case on the subject in our approach is the result of the feature
agreement found in the lower thematic S, it appears that it is not what bottom
to top MERGE/checking would derive.
5.6.2 Complex predicates and verb agreement

In Telugu (DR), there are several complex predicates that can take either a
nominative subject or a dative subject. Such case assignment depends on the
nature of the verbal agreement suffix. We shall demonstrate that the feature
[+/−animate] plays a crucial role. These predicates include:
Telugu (DR)
Noun Light verb Meaning
1. val.l.u ‘body’ cēyu ‘to do’ ‘to become fat’

2. picci ‘craziness’ ekku ‘to climb’ ‘to go crazy’
3. picci ‘craziness’ lēcu ‘to wake up’ ‘to go crazy’
4. badili ‘transfer’ avvu ‘to become’ ‘to be transfered’ or ‘to be shifted’
5. kovvu ‘fat’ ekku ‘to climb’ ‘to become arrogant’
6. kovvu ‘fat’ baliyu ‘to increase’ ‘to become arrogant’
7. t.hār ‘panickiness’ etti pōvu ‘to raise’ ‘to become panicky’
While the verb in the nominative subject construction agrees with the subject,
the verb in the non-nominative (dative subject) construction agrees with the
theme.
nominative subject construction
(172) āvid.ai piccij ekkipōyin- dii,∗ j
she craziness climb- 3 s,nm [+animate]
‘She went crazy.’
dative subject construction
(173) āvid.a kii piccij ekkipōyin- di∗ i,j
she dat craziness climb- 3 s,nm [−animate]
Literally: ‘Craziness climbed on to her.’
‘She went crazy.’
Sentences (172) and (173) differ in the feature animacy alone, with regard to
agreement.
If the agreement marker is coindexed with the subject as in (172), the case that
is assigned to the subject is nominative. If the agreement marker is coindexed
with the theme as in (173), the case that is assigned to the subject is non-
nominative. Hence, for the proper assignment of case to the subject, the complex
predicate plus information concerning animacy in the agr phrase are required.25
We have shown earlier that the predicate in a DSC in Telugu is [−transitive].
There are verbs in Telugu which are transitive in the nominative–accusative
construction, and are also used in the DSC. The set includes: pōyu ‘to pour,’ ‘to
serve,’ ‘to pass something, e.g. urine’; oppu ‘to agree’; vēyu ‘to put,’ ‘to keep,’
‘to serve,’ ‘to wear,’ ‘to take something, e.g. medicine’; pet..tu ‘to put,’ ‘to keep,’
‘to insert’; tappu ‘to miss’; tippu ‘to turn,’ ‘to take someone around’; tōyu ‘to
push’; pat..tu ‘to catch’; ād.u ‘to play’; tagulu ‘to touch’; kot..tu ‘to hit’; mottu ‘to
hit’; kat..tu ‘to tie’; vācu ‘to have a swelling’; kalugu ‘to have’; ekku ‘to climb.’
The verb is used transitively in (174) in a nominative–accusative sentence and
intransitively in (175).
(174) āmei pillala- ki cokkāj vēs- in- dii,∗ j
she.nom children- dat shirt put.on- pst- 3 s,nm [+animate]
‘She put the shirt on the children.’
In (175) the same verb is used in the DSC.

(175) āmei ki calij vēs- indi∗ i,j
she dat cold put.on- pst- 3 s,nm [−animate]
‘She felt cold.’
The theme cokkā ‘shirt’ in sentence (174) is structurally case-marked

accusative, and the accusative marker is null. In contrast, the theme cali ‘cold’
in (175) is nominative case-marked and accusative case-marking on the theme
is not permitted (176).
(176) *āmei ki calij ni vēsin- di∗ i,j

she dat cold acc put.on- 3 s,nm [−animate]
Intended meaning: ‘She felt cold.’
Further, the verb in (174) does not agree with the theme, whereas it invariably
agrees with the theme in the nominative in (175). We have already shown that
the theme and verb together assign theta role to the subject in the DSC. We now
propose that the theme and the verb together assign inherent case to the subject
vP-internally in the lower (thematic) clause in view of the following points.
If it is only the verb that assigns inherent case to the subject in (174),
the question that needs to be answered is: why doesn’t the same verb assign
inherent case to the subject in (175)? It is because of the transitive nature of
the verb in (174) that the theme is accusative case-marked, while it needs to
be explained why the theme in (175) cannot be accusative case-marked, as the
ungrammaticality of (176) shows.
One might say that there are two different sets of transitive verbs which are
homophonous, and a specific set (call it Set A) assigns inherent case while
the other set (call it Set B) assigns nominative case. Such an assertion is
counterintuitive, and it is not clear how such a stipulation can be formulated
and implemented in the grammar of a language. Hence, we conclude that it is
the theme/adjective, together compositionally with the predicate, that assigns
dative case to the subject.
5.6.3 The role of tense

The third piece of evidence concerning inherent case assignment comes from
the role of tense. Jayaseelan (1999: 105) demonstrates that it is not just the
complex predicate alone that is sufficient to assign an inherent non-nominative
case, and that a functional head such as Tense plays a crucial role. A predicate
such as is..tappet. ‘to like’ takes a nominative subject (177a) when the tense
marked is the present, and it takes a dative subject in the past tense (177b).
nominative subject in present tense – grammatical

Malayalam (DR)
(177a) awal. awan- e is.t.appet.unnu
she-nom he- acc like-pres ind
‘She likes him.’
dative subject in past tense – grammatical

(177b) awal.a- kkə awan- e is.t.appet.t.u
she- dat he- acc like-pst
‘She liked him.’
Sentence (178a) with the dative subject in the present tense is ungrammatical,
according to Jayaseelan.
dative subject i n present t e n s e – u n g r a m m a t i c a l
(178a) *awal.a- kkə awan- e is.t.appet.unnu
she- dat he- acc like-pres ind
When the predicate is..tappet. ‘to like’ occurs in past tense, nominative subject
is questionable.
nominative subject i n past t e n s e – q u e s t i o n a b l e
(178b) ?awal. awan- e is.t.appet.t.u
she he- acc like-pst
According to Jayaseelan (1999: 105), the dative case marker is assigned to the
subject “at the point where is..tappet. and the past tense marker are put together
[i.e. MERGE].”
5.6.4 The role of aspect

In Malayalam (DR), the functional category aspect also plays a role in case
assignment. In (179) and (180) the verb is the same, namely pō ‘to go.’ The
aspect marker -ām signifies permission, and it requires a dative case marker
with the subject, as in (179), while the subject in (180) is nominative case-
marked, as the predicate is in present tense, and the aspect marker -ām is not
present (Jayaseelan 1999: 103).
(179) niŋal.- kkə pōk- ām
you- dat go- permission
‘You may go.’
(180) awal. pōk- unnu

she (nom) go- present
‘She goes.’
In Hindi-Urdu (IA) as well, modals play a role in the assignment of dative case.
The modal sak-nā ‘to be able to’ takes a nominative subject, as in (181), while
the modal cāhiye ‘must’ requires a dative subject, as in (182).
Hindi-Urdu (IA)
(181) bacce ghar jā sakte hãı̃
children (nom) home go can pres
‘The children can go home.’
(182) baccõ ko ghar jā- nā cāhiye

children dat home go- inf must
‘The children must go home.’
To summarize, different modals have lexical selection restrictions which are

language-specific.
5.6.5 The role of the verb ‘to come’

In Telugu (DR) and many SALs, the verb ‘to come’ is a verb of motion with
the nominative case. It has an extended sense with a dative subject as a kind
of reanalyzed end-point/goal. It also denotes possession of knowledge or skill,
and the subject who is the possessor of knowledge or skill is case-marked
dative/genitive.
vac ‘ t o c o m e ’ a s a v e r b o f m o t i o n
Telugu (DR)
(183) pel.l.ivāl.l.u vid.idi ki vaccē- r(u)- ā
bridegroom’s.family.nom guesthouse to come- 3p,hon- pol q mkr
‘Has the bridegroom’s family arrived at the guesthouse?’
vac ‘ t o c o m e ’ a s a v e r b i n d i c a t i n g k n o w l e d g e
(184) āme ki hindı̄ bāgā vacc(u)
she dat Hindi well comes
‘She knows Hindi well.’
vac ‘ t o c o m e ’ a s a v e r b i n d i c a t i n g s k i l l ( ( 7 4 ) r e p e a t e d h e r e )
(74) vāl.l.a ammāyi- ki sangı̄tam vaccu
their daughter- dat music comes
‘Their daughter has the knowledge of music.’
The verb ‘to come’ is used in Hindi-Urdu, Nepali (IA) and Newari (TB)
(T. K. Kansakar, p.c.), too, as a verb of motion, and to denote possession of
knowledge and skill.
Hindi-Urdu (IA)
ā ‘ t o c o m e ’ a s a v e r b o f m o t i o n
(185) pratimā daftar se āyı̄
Pratima office from came
‘Pratima came back home from the office.’
ā ‘ t o c o m e ’ a s a v e r b i n d i c a t i n g k n o w l e d g e ((73) repeated here)
(73) is panjābı̄ lar.kı̄i ko bharat nāt.yamj ātā∗ i,j thā∗ i,j
this Punjabi girl dat classical Indian dance come was
‘This Punjabi girl used to know the classical Indian dance.’
Note that the verb ā ‘to come’ in Hindi-Urdu (IA), by itself, cannot assign
dative case inherently to its subject as the verb does not have any information
about the nature of activity that it represents. Combined with the theme which
involves knowledge or a skill, the verb ā ‘to come’ forms a complex predicate
and only then is it able to assign dative case to its subject.
Thus, a predicate with subject and other constituents pro-dropped in
Telugu (DR) or Hindi-Urdu (IA) is ambiguous between the nominative sub-
ject and dative subject readings. The following dialog from Telugu (DR) is
illustrative.
Telugu (DR)
(186a) mādhuri rātri ki vastund(i)- ā
Madhuri night dat comes.3 s,nm pol q mkr
‘Will Madhuri come tonight?’
(186b) pro pro vastundi
comes.3 s,nm
‘She’ll come.’
(187a) mādhuri ki telugu vastund(i)- ā
Madhuri dat Telugu comes.3 s,nm- pol q mkr
‘Does Madhuri know Telugu?’
(187b) pro pro vastundi
comes.3 s,nm
‘She knows it.’
A similar ambiguous reading obtains in Hindi-Urdu (IA) and many other

SALs.
Thus, empirical facts from Telugu, Malayalam (DR) and Hindi-Urdu (IA)
strongly support the view that it is not the verb alone that assigns inherent
case to its subject, but the verb together with the theme/adjective, nature of
the modal, or tense, and default agreement that play a crucial role in inherent
case assignment to the subject. And for such case assignment to take place, it
is crucial that information concerning agreement be available vP-internally in
the thematic S in some cases.
At this point, we wish to summarize the discussion so far and present
the salient features of the non-nominative subject constructions in SALs.
As the ergative–absolutive construction stands apart, we exclude it in this
summary.
(i) It is the dative/genitive subject construction that is quite predominant in
occurrence, with a variety of semantic predicates.
(ii) While IA languages have either the dative or the genitive with the subject,
depending on the nature of the predicate, in Dravidian languages it is
mostly the dative that occurs. In possession, the locative occurs in IA

as well as in Dravidian. Bangla and Assamese are the IA languages that
have the genitive in most of their non-nominative constructions. Some
Tibeto-Burman and Munda languages, too, have the non-nominative
construction.
(iii) The occurrence of the case marker is language-specific, and it depends
on the semantic nature of the predicate (Masica 1976, 1991; Sridhar
1979; Mohanan and Mohanan 1990). As Mahajan (2004: 290) puts it:
“the shape of the non-nominative morpheme is lexically stipulated using
lexical linking rules.” He further adds: “notions such as GOAL, POS-
SESSION, CONSCIOUS CHOICE and INTERNAL ABILITY play a
crucial role.” The use of capital letters indicates that the notions are
semantic predicates. To this we may add notions such as OBLIGATION,
NECESSITY, EMOTIONS, DESIRE, etc. It should be underscored that
this is not an exhaustive list.
(iv) The non-nominative subject lacks agentive theta role, and hence the
predicates are [−volitional].
(v) The NNS construction is predominantly found in verb-final languages,
though it is also found in some non-verb-final languages such as Ice-
landic, Russian and Finnish.
(vi) The predicate in a non-nominative construction is [−transitive], and,
hence, it does not have the capacity to mark the theme as accusative. The
apparent accusative marker that occurs in the dative and genitive subject
constructions in some languages is a specificity marker, not an accusative
case marker.
(vii) The theme in such cases gets nominative case from the Tense (INFL) of
the clause.
(viii) It is the semantic nature of the predicate that determines what type of
case marking the logical subject gets.
(ix) Agreement and features such as [+animate] play an important role in
inherent case assignment.
(x) As Jayaseelan (1990) first argued, inherent case is compositionally
assigned. It is assigned vP-internally in the thematic S, and it is the
verb together with the theme/adjective, nature of the modal, or tense, or
COMP and default agreement that play a crucial role in inherent case
assignment to the subject.
(xi) It is the [+/−finite] nature of the COMP together with the matrix verb
to be that is instrumental in assigning nominative or dative case to the
matrix subject, and such assignment has to be done compositionally.
In the following section we discuss the nature of the Double Dative Subject
construction in Telugu and Kannada (DR).
5.7 Double/multiple Dative Subject 189
5.7 Double/multiple Dative Subject constructions

Malayalam (Sobha Nair, p.c.), Kannada (N. Varija, p.c.) and Telugu (DR)26
have a Double Dative Subject construction that occurs in sentences express-
ing inalienable possession, locational possession and part–whole relationship
(Subbarao and Bhaskararao 2004: 191–194).
A noun phrase case-marked by the dative or locative can be modified by a
genitive DP. In (188) in Telugu, (189) in Kannada and (190) in Malayalam,
the genitive modifies the DP/NP. The part, ‘hand’ in English, is dative case-
marked (in italics) in (188)–(190). Let us label it as the Single Dative Subject
construction.
single dative subject construction
Telugu (DR)
(188) vād.i cēti ki debba tagil- in- di
his (he.gen) hand dat injury hit- pst- 3 s,nm
‘He got hurt on his hand.’
Kannada (DR)
(189) avana kan.n.- ige gāya(v) āgi- de
his (he.gen) eye- dat injury happen- be
‘He got hurt on his eye.’
Malayalam (DR)
(190) avanre kai- kku katam parri
his (he.gen) hand- dat injury happened
There are corresponding sentences in which the whole – ‘he’ in English –
and the part – ‘hand’ in English – are both dative case-marked (in italics) in
(191)–(193).
Let us label it as the Double Dative Subject construction.
double dative subject construction
Telugu (DR)
(191) vād.i- ki cēti ki debba tagil- in- di
he.obl- dat hand dat injury hit- pst 3 s,nm
Kannada (DR)
(192) avan- ige kan.n.- ige gāya(v) āgide
he- dat eye- dat injury happened
‘He got hurt on his eye.’
(Swathi P. G. Bhat and Shivarama Bhat Padikkal, p.c.)
Malayalam (DR)
(193) avan- ə kai- kku katam parri
he- dat hand- dat injury happened
The sentences in (191)–(193) reflect a ‘part–whole’ relationship. Thus, in sen-

tences in (188)–(190), the DP in the genitive/possessive case stands in ‘whole’
relationship to the DP that is dative/locative case-marked, which is in ‘part’
relationship. In sentences (191)–(193) the ‘whole’ as well as the ‘part’ are
dative case-marked.
In the Single Dative Subject construction in (188)–(190), an adverb cannot
occur between the genitive case-marked DP which represents ‘whole’ and the
dative case-marked DP which represents the part, as shown in (194) in Telugu
and in (195) in Kannada (DR).
Telugu (DR)
(194) *vād.i- ivāl.a cēti ki debba tagil- in- di
his- today hand dat injury hit- pst- 3 s,nm
Kannada (DR)
(195) *avana indu kan.n.- ige gāyav āgide
his today eye- dat injury happened
In contrast, in the Double Dative Subject construction in (191)–(193), an adverb

can occur to the right of the possessor (‘whole’), which is dative case-marked,
as illustrated in (196)–(197).
Telugu (DR)
(196) vād.i- ki ivāl.a cēti ki debba tagil- in- di
he.obl- dat today hand dat injury hit- pst- 3 s,nm
‘He got hurt on his hand today.’
Kannada (DR)
(197) avan- ige indu kan.n.a- ige gāya(v)- āgide
his- dat today eye- dat injury- happened
‘He got hurt on his eye today.’
Further, a possessive reflexive can modify the possessed NP in the Double

Dative Subject construction, but not in the Single Dative Subject construction.
The second piece of evidence, to show that the dative case-marked possessor
and the body part constitute two independent constituents, comes from the fact
that both the constituents of the double subject can be independently questioned.
5.7 Double/multiple Dative Subject 191
dative case-marked possessor questioned

Telugu (DR)
(198) evari- ki cēti ki debba tagil- in- di
who- dat hand dat injury hit- pst- 3 s,nm
Literally: ‘To whom did it get it hurt to his hand?’
‘Who got hurt on his hand?’
dative case-marked possessed questioned

(199) vād.i- ki ekkad.a debba tagil- in- di
he- dat where injury hit- pst- 3 s,nm
Literally: ‘To him where (in which part) did it get hurt?’
‘At which place (where) did he get hurt?’
The third piece of evidence for this construction comes from particles in
Telugu. The focus particle aitē ‘as for,’ the inclusive particle kūd.ā ‘also’
and -ē ‘the emphatic particle’ can be inserted after the possessor. We provide
only one example with the particle kūd.ā ‘also.’
(200) vād.i- ki kud.ā cēti- ki debba tagil- in- di

he- dat also hand- dat injury hit- pst- 3 s,nm
‘He too got hurt on his hand.’
The fourth piece of evidence comes from the contrastive focus clitic (CFC) -ā27
in Telugu.
t h e ‘w h o l e ’ q u e s t i o n e d
(201) vād.i- k(i)- ā cēti- ki debba tagil- in- di
he- dat CFC hand- dat injury hit- pst- 3 s,nm
‘Is it he who got hurt on his hand?’
t h e ‘p a r t ’ q u e s t i o n e d
(202) vād.i- ki cēti- k(i)- ā debba28 tagil- in- di
he- dat hand- dat- CFC injury hit- pst- 3 s,nm
‘Is it on his hand that he got hurt?’
Thus, these arguments indicate that, in the Single Dative Subject construction,
the genitive case-marked DP (the ‘whole’) and the dative/locative case-marked
DP (the ‘part’) together form a single constituent, while in the Double Dative
Subject construction, the two inherently case-marked DPs are separate con-
stituents. Subbarao and Bhaskararao (2004) show how inherent case is assigned
in sentences (188)–(190), and (191)–(193). Since the genitive case-marked DP
and the dative case-marked DP together function as a single constituent in
sentences in (188)–(190), the predicate compositionally assigns dative case to
the whole DP vP-internally in the thematic S.
In (191)–(193), where the DPs behave as independent DPs, inherent dative

case is assigned to the initial DP that represents ‘the whole relationship,’ and
to the DP that represents the ‘part relationship,’ vP-internally. That is, the pred-
icate assigns inherent case simultaneously to two DPs. The theme is attracted
by the probe to have its structural nominative case erased and to delete the
uninterpretable features of the probe. Subbarao and Bhaskararao (2004) point
out that the mechanics of inherent case assignment to two DPs simultaneously
needs to be worked out, as these too might require multiple spec positions in
Spec TP.29
5.8 Conclusion
This chapter provides a description of the NNS constructions in SALs, and
presents a discussion of the domains of occurrence of the NNS. It also focuses
on some subject and non-subject properties of these constructions. The stative
and non-stative, as well as the intransitive, nature of the predicate in NNS
constructions are also discussed.
We have argued that: (i) the verb in the DSC is intransitive; (ii) all DSCs
are inherently case-marked; (iii) such inherent case marking cannot be done by
an intransitive verb alone, but by the whole predicate compositionally, consist-
ing of a theme or adjective along with the [−transitive] verb; (iv) information
concerning agreement should be available vP-internally (in the lower thematic
S) for proper assignment of inherent case to the non-nominative subject; and
(v) the accusative/dative marker that occurs with the patient/theme in NNS con-
structions is a specificity marker and, hence, does not count as counter-evidence
to the claim that dative predicates are [−transitive]. A brief summary of the
four major analyses that account for the NNS constructions is also presented.
Finally, the Double Dative Subject construction that occurs in sentences that
express inalienable possession was discussed.
Appendix – formal representation of NNSs

There have been four major analyses to account for the non-nominative con-
structions in general, and the dative subject construction in particular, in SALs.
These include:
(i) Jayaseelan’s (1990, 2004) analysis;
(ii) Bhatt’s (1999) analysis;
(iii) Davison’s (2004) analysis; and
(iv) Subbarao and Bhaskararao’s (2004) analysis.
(See the appendix of chapter 5 on the Website for details.)
6 Complementation
6.1 Introduction
Complementation1 has been a topic of intense study, in SALs as well as many
other languages, starting with the work of Rosenbaum (1967). In this chapter,
we discuss the complement constructions in SALs, which are of great empirical
and theoretical significance in view of the fact that the complementizer may
occur:
(i) exclusively in clause-initial position (labeled as left peripheral or pre-
sentential or initial complementizer [IC]), as in many Indo-Aryan (Kash-
miri, Hindi-Urdu, Kurmali and Punjabi) and Munda languages (Ho,
Santali, Kharia and Korku), and the non-verb-final Mon-Khmer Khasi;
or
(ii) exclusively in clause-final position (labeled as right peripheral or post-
sentential or final complementizer [FC]), as in most of the Dravidian
languages and all Tibeto-Burman languages, the Munda language Mundari
and the Indo-Aryan Sinhala; or
(iii) in both clause-initial and clause-final positions, as in Eastern Indo-Aryan
Nepali, Assamese, Bangla and Oriya, or Western Indo-Aryan Gujarati
and Marathi, or in the transplanted language Dakkhini Hindi-Urdu in the
South of India.
In this chapter, we shall show how the heterogeneous distribution of comple-
mentizers in SALs correlates with question markers, clefts and subordinate
relative clause markers. Several issues with regard to the movement of clauses
and operator scope, too, will be discussed.
An embedded clause may occur as a complement either in the subject or
object position of a predicate (verb/adjective) of the matrix clause, or the
next higher clause. The former is a case of subject complementation and the
latter, a case of object complementation. It is the complementizer that links
the embedded clause to the next higher clause. It may occur to the left of the
embedded clause as in (1) in Hindi-Urdu (IA), or to the right of the embedded
clause as in (2) in Kokborok (TB). Thus, (1) instantiates the case of the IC
ki ‘that’ in Hindi-Urdu, and (2) the FC hn-i ‘that’ in Kokborok (TB); hn-i
193
194 Complementation
‘say + cpm’ is called a quotative, as the complementizer is a grammaticalized

form of the verb ‘say.’
Hindi-Urdu (IA)
(1) ham ne sunā thā [CP ki [S2 ramyā dillı̄ pahũnc gayı̄S2 ]CP ]
we erg hear. perf pst that (IC) Ramya Delhi reach went
‘We had heard that Ramya reached Delhi.’
Kokborok (TB)
(2) akuŋ [CP [S2 khumti naithɔkS2 ] hn-iCP ] sa- kha
Akung Khumti beautiful say-cpm that (FC) tell- pst
‘Akung said that Khumti is beautiful.’
Marathi and Gujarati (IA) have: (i) two FCs derived from a pronoun: in Marathi
(IA), for example, the FC is derived from asa/asə ‘like this,’ as in (3), and in
Gujarati it is evũ/em ‘like this, this way, thus’; and (ii) another FC, namely
te ‘that’ in Marathi (Wali 2006: 210) as well as Gujarati (IA) (Cardona 1965;
Masica 1999; P. J. Mistry and Trupti Nissar, p.c.).
Marathi (IA)
(3) [CP [S2 tū tithe kā dzātes?S2 ] asaCP ] ti- ne malā witšārle
you there why go.2 s,f quot she- erg I.dat ask.pst.3 s, neuter
She asked me, “Why do you go there?”
The complementizer is a functional category like Tense and Aspect, and is

normally a ‘grammaticalized form’ that is devoid of any semantic content.
According to Noonan (1985: 47), “complementizers typically derive from pro-
nouns, conjunctions, adpositions or case markers, and, rarely, verbs.”
Interestingly, most of the SALs and many other languages beyond the South
Asian subcontinent (Lord 1976) have a COMP that is normally a form derived
from the verb ‘to say,’ and it is labeled as the ‘quotative’ (Emeneau 1956;
Kuiper 1967; Hock 1975, 1982a; Masica 1976; Kachru 1979; Subbarao et al.
1989).
The initial or final complementizer may acquire several other functions in
the course of time, in addition to performing as a subordination linker, which
we shall discuss in this chapter.
Verb-final languages in general have a complementizer that occurs in the
final position to the right of the embedded clause, which conforms to the Head
Direction parameter (see chapter 2).
The position of occurrence of the complementizer in many Indo-Aryan and
Munda languages is in the initial position to the left of the embedded clause.
This demonstrates that such occurrence is marked, and it does not conform to
the Head Direction parameter. Some SALs (Indo-Aryan, for example) have the
IC as well as the FC. Some central and north Dravidian languages do not have
the quotative (FC) (Masica 1976: 189).
The generalization that non-verb-final languages have an initial complemen-
tizer (IC), and verb-final languages have a final complementizer (FC) holds
good only for the non-verb-final language Khasi, which has an IC, and the
literary verb-final Dravidian languages and all Tibeto-Burman languages in the
subcontinent, which have only an FC. Some Munda and Indo-Aryan languages
have both an IC and FC. Hence, they do not conform to the cross-linguistic
generalization that verb-final languages have an FC. It is quite possible that
some Munda languages lost the quotative FC as a complementizer due to the
influence of the major superstrate languages such as Hindi and Sadani. The
absence of an FC in Hindi-Urdu, Kashmiri and Punjabi, too, could be due to
the influence of Persian, which has only an IC ke.
It is worth noticing that some Munda languages, for example Ho, Santali
and Juang, have a post-sentential quotative (complementizer): some languages,
like Ho and Santali, have the quotative functioning as a purposive marker,
and some languages, like Kharia and Ho, have the quotative functioning as a
reason marker as well. These functions in different languages of the Munda
family indicate that the quotative must have been in use as subordinate linker
(complementizer, to be specific) in all Munda languages at some stage, but
subsequently must have been lost in some languages.
In this chapter, we discuss several issues that concern complement clauses
in SALs. First, we discuss the nature of the quotative, reported speech vs.
indirect speech with regard to the phenomena of person agreement and tense
harmony. We focus our attention on the position of occurrence of an FC-clause,
the nature of the left and right peripheral COMP, the position of occurrence of
an IC-clause and the Case Resistance Principle of Stowell (1981). We present a
detailed analysis of the parametric variation among SALs in the expression of
wide and narrow scope devices. Several issues that concern the problems with
the Rightward Extraposition analysis, the direction of c-command in sentences
with Negative Polarity Items, COMP agreement (Juang, Mundari) and, finally,
the reanalysis of the initial complementizer resulting in syntactic changes due
to convergence are also discussed.
We shall also focus our attention on one of the crucial issues that concern
complementation: that is, is there any correlation between the semantic nature
of the predicates – such as factive or non-factive – the position of occurrence
and the choice of the complementizer? We demonstrate that there is.
This chapter is organized as follows. Section 6.2 discusses the position of the
COMP and some crucial functions that it performs in SALs. Sections 6.3 and 6.4
deal with structural differences between an IC and an FC. Some notable issues
include: how the subject of a finite CP clause gets case-marked by the matrix
verb and whether such marking violates the Phase Impenetrability Condition
196 Complementation
of Chomsky (2001) (see the Website); and COMP agreement with the matrix
subject (see 6.3.4). Section 6.5 focuses on the position of occurrence of an
IC- and an FC-clause and the Case Resistance Principle (CRP). Section 6.6
deals with wh scope and the factors which permit and limit wide wh scope.
Section 6.7 provides arguments against Rightward Extraposition in finite CP
clauses and discusses some issues related to it. Section 6.8 concludes the
chapter. The Appendix of this chapter presents a universal constraint labelled
as the Final-over-Final Constraint (FOFC) and discusses how the diachronic
changes that took place in the position of occurrence of the complementizer in
SA language contact situations strictly obey this universal constraint.
Some other crucial issues related to complementation in SALs are also
discussed separately on the Website. These include: (i) formal syntax as a tool
for explicating a typological distinction between the English-type languages
and the Korean- and Kashmiri-type languages; (ii) the direction of c-command
and the occurrence of the Negative Polarity Items; and (iii) syntactic reanalysis
in language contact situations, to demonstrate how the IC of Hindi-Urdu (IA)
is reanalyzed in Dakkhini Hindi-Urdu (IA) due to contact with Telugu (DR).
6.2 The position of occurrence of the COMP and its functions

6.2.1 The position of occurrence of the COMP
First we shall present a brief discussion of the position of occurrence of the
COMP in SALs. We have mentioned above that SALs have an IC as well as
an FC.
Category 1a: the quotative FC derived from the verb ‘to say’
All the Dravidian and the Tibeto-Burman languages, IA languages – Marathi,
Nepali, Sinhala – and the Munda languages – Mundari, Ho and Juang – have
the quotative FC derived from the verb ‘to say.’
Category 1b: the FC derived from a pronoun

Gujarati and Marathi (IA) have an FC derived from a pronoun.
Category 2: only an IC
Indo-Aryan languages such as Hindi-Urdu, Kashmiri and Punjabi, Khasi (Mon-
Khmer) and some Munda languages have only an IC.
Category 3: an IC as well as an FC
Assamese, Bangla, Gujarati, Konkani, Marathi, Nepali, Oriya (IA) and Juang
(Munda) have an IC as well as an FC.
6.2 Position of occurrence of COMP 197
A brief comment on the COMP in Sanskrit (Old Indo-Aryan) and Dakkhini

(IA) is in order.
(i) Sanskrit initially had an IC iti, which was later reanalyzed as an FC
quotative. The FC iti, just like the quotative in Dravidian, is used in
CAUSE/PURPOSE to name or label persons or things, with question
words, with onomatopoeia, etc. (see Hock 1982a for a detailed discussion).
(ii) Colloquial Dakkhini Hindi-Urdu (Dakkhini, hereafter) lost the IC ki ‘that’
as an IC, inherited from the source language Hindi-Urdu, due to intense
contact with Telugu (DR). It has acquired the FC quotative bol ke ‘having
said,’ which is a calque from Telugu; bol ke ‘having said’ is used in the
same way as the FC quotative ani in Telugu, due to ‘extreme convergence’
(Subbarao and Arora 1989). The IC ki of Hindi-Urdu (IA) has been
syntactically reanalyzed in Dakkhini to perform several other functions,
which we discuss in section 6.8. Dakkhini also has an FC, kate, literally
‘(they) say,’ which is derived from the verb kah ‘say’ (see sentence
(32)).
We shall now provide some examples illustrating the occurrence of the IC and
FC from the different language families of the SALs.
In Khasi (Mon-Khmer) (4), a verb-medial language, the IC ba occurs to the
left of the embedded clause (S2 ) and the embedded clause occurs to the right
of the VP of the matrix clause.
Khasi (Mon-Khmer)
(4) u- kmεn u- m- ŋεyt [CP ba [S2 ki- brēw ki- hyar patεŋ
3m- Kmen 3m- neg- believe IC p- humans p- descended
na- ki- šrē?S2 ]CP ]
abl- p- monkey
‘Kmen does not believe that human beings descended from monkeys.’
An example of an IC clause in Ho (Munda) is in (5).
Ho (Munda)
(5) somai ur.ū- ke- d.- a [CP ci ramj ø∗ i/j suku- n- tan- aCP ]
Soma think- pst- tr- fin IC Ram self like- VR- pres- fin
‘Somai thought that Ramj likes himself∗ i/j .’
In some languages (e.g., Oriya and Marathi), the IC and FC may occur in
the same sentence, where the IC precedes and the FC follows the embedded
complement.
198 Complementation
Oriya (IA)
(6) se kɔt.ak jibɔ boli bhabu thila je kintu gad.i nɔ thi.ba-
he Cuttack will go quot was thinking COMP but train not be.inf-
ru jai parila nai
due to could go not
‘He was thinking of going to Cuttack but he could not go as there was
no train.’
(P. P. Mahapatra and Prakash Patnaik, p.c.)
Marathi (IA)
(7) anū mhant.e [kı̄ [[tı̄/mı̄ hindı̄ šikel] (asa/ mhan.ūn)]]
Anu say.pres.3s IC she/I Hindi learn.fut.3s FC/ FC
‘Anu is saying [that she will learn Hindi].’
Malayalam (Hany Babu and Sobha Nair, p.c.), Tamil (Arulmozhi, p.c.) and
Telugu (DR) have two FCs occurring in a row, as in (8) and (9). For example,
the FC in Tamil enru ‘having said,’ is a grammaticalized form of the verb en ‘to
say.’ The other FC that can follow it is solli ‘having told,’ a grammaticalized
form of the verb soll ‘to tell.’ We label them as FC1 and FC2.2
Tamil (DR)
(8) nı̄ vara.villai enru (solli) nān vara.villai
you came.not FC1 (REASON) FC2 I came.not
‘I did not come because you did not come.’
(Arulmozhi, p.c.)
FC1 enru has reason interpretation in (8). FC2 can occur only if FC1 has
reason interpretation; and FC2 can be optionally dropped, while FC1 cannot
be in Tamil or Telugu.
A similar situation obtains in Malayalam too.
Malayalam (DR)
(9) nı̄ varinn.illa ennu (para
u) nān vann.illa
you came.not FC1 (REASON) FC2 I came.not
‘I did not come because you did not come.’
(Sobha Nair, p.c.)
Thus, the phrase structure of CP in SALs is either:

(a) [CP IC [S2 - - -S2 ]CP ], as instantiated in (1)
(b) [CP [S2 - - -S2 ] FCCP ], as instantiated in (2)
(c) [CP [CP IC [S2 - - -S2 ]] FCCP ] CP ], as instantiated in (6) and (7)
(d) [CP [CP [S2 - - -S2 ] FC1CP ] FC2CP ], as instantiated in (8) and (9)
We have given examples of complement clauses with their position of occur-
rence from the four language families.
The complement clauses in Kashmiri (IA) deserve special mention, as

Kashmiri is the only language in the subcontinent to have – like Dutch and
German – its head in the final position despite being a V2 language.
In Kashmiri, the COMP ki/zi is left-peripheral, just as in Hindi-Urdu and
Punjabi (IA). Kashmiri is a V2 language in which the verb occurs in the second
(V2 ) position in the matrix as well as the complement clause – unlike in German,
a V2 language, where the verb occurs in the V2 position in the matrix clause,
and in the final position in a complement clause (Bhatt 1999: 83). The vector
verb dyut ‘gave’ occurs in the second (V2 ) position in (10a), as well as in (10b)
in the embedded clause.
v2 (in italics) in a simplex clause

Kashmiri (IA)
(10a) rāth dyut lar.k- an tswat.h dār- yith
V2
yesterday gave boy- erg trash (noun) throw- cpm
‘Yesterday, the boy threw (away) the trash.’
v 2 (i n i t a l i c s) i n s 2 i n a c o m p l e x s t r u c t u r e
(10b) tem dop ki [S2 rāth dyut lar.k- an
V2
he.erg said comp yesterday gave boy- erg
tswat.h dār- yithS2 ]
trash throw- cpm
‘He said that the boy threw (away) the trash yesterday.’
(Bhatt 1999: 83)
Thus, there is no asymmetry between the root and embedded clause in Kashmiri
(IA) complements,3 unlike in German, which exhibits asymmetry.
Bhatt and Yoon (1991) and Bhatt (1999) show that the FC or IC functions like
a subordinator linking the main clause and the embedded clause. In this context,
the idea that CP resolves into mood and subordination, proposed by Bhatt
and Yoon (1991) and Bhatt (1999), is worth considering, and is presented in
appendix 6.1 on the Website.
In this section, we have provided data that demonstrate: (i) the positions of
occurrence of the IC and FC; and (ii) the occurrence of double complement
clauses in SALs. We have shown that, unlike German, Kashmiri retains the
V2 order in the main as well as subordinate clauses. We show in appendix 6.1
how the separation of subordinator and mood as separate categories helps to
explicate the absence of asymmetry in root and embedded complements in
Kashmiri.
The IC and FC perform several functions other than just as complementizers,
which we present now.
200 Complementation
6.2.2 Functions that the IC and FC perform in SALs

We shall briefly discuss some functions that the IC performs in SALs, followed
by the functions that the FC performs.
The initial COMP (IC)

The IC ki ‘that’ in Hindi-Urdu and Marathi (Pandharipande 1997: 11), ke in
Punjabi (Bhatia 1993: 7) and Gujarati (IA), and ki in Kashmiri co-occurring
with a negative function as disjunctive markers in alternative questions.4
as a clausal disjunctive marker
Hindi-Urdu (IA)
(11) vah billı̄ hai ki billā?
that female cat is or male cat
‘Is it a female cat or a male cat?’
Marathi (IA)
(12) to ã˜mbā khāto ki sāntra (khāto)
he mango eats or orange eats
‘Does he eat a mango or (does he eat) an orange?’
(Pandharipande 1997)
In Kashmiri (IA), there are two forms of the IC – the COMP ki and zi ‘that.’
Only ki occurs, and not zi, as a disjunctive marker, provided it is accompanied
by the negative n ‘not’ (Kak ms). Together, ki ‘that’ and n impart the meaning
of ‘or.’ Note further the yes/no question marker -ā occurs with V2 .
as a phrasal disjunctive marker in alternative questions
in kashmiri
Kashmiri (IA)
(13) su čhā pādšah ki- n/ *zi- n bēcvun
he is.yes/no qm king IC- neg IC- neg beggar
‘Is he a king or a beggar?’
(Kak, ms)
In Kharia (Munda), the IC no functions as a yes/no question marker (Peterson

2006).
The IC functions as a subordinating linker linking the main clause and the
embedded clause with the interpretation of English then in Hindi-Urdu and
Kashmiri (IA).
Hindi-Urdu (IA)
(14) ham št.ešan pahunc hı̄ rahe the ki [S2 gār.ı̄ chūt. gayı̄S2 ]
we station reach emph progr were IC train leave went
‘We were about to reach the station, then the train left.’
Kashmiri (IA)
(15) b drās ki/zi su av
I left IC he came
‘As soon as I left, he came.’
(Kak, ms)
Hindi-Urdu, Punjabi and Kashmiri employ an expletive IC with a relative

pronoun in non-restrictive clauses.
Kashmiri (IA)
(16) bad.šah [yus ki akh jān pādšah o:s] o:s akh ke:šur
Badshah who IC one good king was was one Kashmiri
‘Badshah, who was a good king, was a Kashmiri.’
(Kak, ms)
Hindi-Urdu (IA)
(17) ašokā jo ki ek mahān rājā the bar.e dayālu the
Ashoka who IC a great king was very kind was
‘Ashoka, who was a great king, was very kind.’
In Gujarati (IA) too, the IC ke occurs in relative clauses, but it precedes the
relative pronoun (P. J. Mistry, p.c.), whereas in Hindi-Urdu, Kashmiri and
Punjabi (IA), the IC follows the relative pronoun.
Gujarati (IA)
(18) mān.as ke je letter lakhe che te ā story jān.e che
man IC who letter is writing he that story knows pres
‘The man who is writing a letter knows this story.’
Thus, the IC functions as a disjunctive marker, as a subordinating linker with the

interpretation of then, and as an expletive IC in non-restrictive clauses in some
Indo-Aryan languages, and as a yes/no question marker in Kharia (Munda).
The final COMP (FC)

We have discussed above some of the functions that the IC performs. We shall
now discuss some of the functions that the right peripheral performs in SALs.
Almost all the Dravidian languages, all the Tibeto-Burman languages, some
IA languages, and Munda languages have an FC which is a grammaticalized
form of the verb to say. It is labeled as the quotative. Vedic Sanskrit had an
initial complementizer iti ‘thus,’ which was later reanalyzed as an FC (Hock
1975, 1982a). Middle Indic languages too have an FC. The quotative FC in
Sanskrit acquired several other functions in the course of time (see Emeneau
1956; Kuiper 1974; Hock 1975, 1982a).5
202 Complementation
We mention below a few of the functions that the quotative (FC) performs in
SALs (see Kachru 1979, Subbarao et al. 1989 and Patnaik 1998 for a detailed
description).
(i) As a complementizer
In Gujarati (IA), Malayalam (DR) and Mizo (TB), the embedded clause with
an FC may occur to the left of the matrix subject.
Gujarati (IA)
(19) tεno bhāı̄ pās thayo εm te bolyo
he younger brother pass be.pst FC he said
‘He said that his younger brother passed.’
(Lambert 1971: 205, as quoted in Masica 1991: 403)
Malayalam (DR)
(20) [CP [S2 kut.t.i uraŋŋiS2 ] ennəCP ] amma paraŋŋu
child slept say.cpm (FC) mother said
‘The mother said that the child slept.’
(Hany Babu 1997)
Mizo (TB)
(21) [CP naktuk cu colhni a- ni- aŋ tiʔCP ] zova- n
tomorrow def holiday 3 s- be- fut comp Zova- erg
a- hria
3 s- know
‘Zova knows that tomorrow will be a holiday.’
The FC, a quotative, in Gujarati is -εm (19), in Malayalam ennə (20) and in
Mizo tiʔ (21). The FC, derived from the verb to say, is: (a) a participial form
in Dravidian, Indo-Aryan languages, Mundari, Santali (Munda) and Kokborok
(TB); (b) an infinitival form in Manipuri (TB); and (c) a stem form of the verb
in Mizo and Hmar (TB).
(ii) As a reason marker

In all Dravidian languages, in Mundari (Munda) (Tikkanen 1987: 312, 2001:
1120), Ho (Munda) (see Deeney 1979: 81) and Sinhala (Gair 2003: 87), the
quotative is used with the interpretation of ‘because.’
Mundari (Munda)
(22) senog- jan mente rag- tan- a- e
go- pst- say- abl/instr (=converb) weep- def.pres- fin- 3s, subj
‘. . . because [literally: ‘saying’] he went, he is weeping.’
(Tikkanen 2001: 1120)
Sinhala (IA)
(23) tāttage upandinə kiyəla api kēk hæduwa
father’s birthday quot we cake made
‘We made cake on father’s birthday’ (i.e., ‘since it was his birthday,’ or
‘realizing it was his birthday’).
(Gair 2003: 807)
In Sanskrit too, the FC iti has the interpretation of because.
Classical Sanskrit (IA)

(24) pro vaidēšikah. asmi iti pr.cchāmi
stranger am FC am asking
‘Since I am a stranger, I ask (you) . . . ’
(Hock 1982a: 42)
(iii) As a purpose marker

The quotative FC is a conjunctive participial form of the verb to say in
Kokborok, just as in the Dravidian languages.
Kokborok (TB)
(25) aŋ khumti nɔ nai- na hn-i phai- kha
I Khumti acc see- inf say-cpm (quot) come- pst
‘I came in order to see Khumti.’
In Santali (Munda), “The quotative marker mεnte and the postposition

ləgit/lagat ‘for’ can signal purpose as well” (Neukom 2001 [ms]).
Santali (Munda)
(26) i
- dɔ nɔtε- ge maraŋ hɔr- thεn guti- k’-a
I- topic here- focus big person- dat serve- middle indicative
mεnte orak’- khɔn- i
odok- hij- uk’- kan-
purpose mkr house- abl- 1s come out- come- middle- imperf-
tahεkan- a
copula.pst- ind
‘I was coming from home in this direction, to serve under a big person.’
(Neukom, ms)
(iv) As an identifier
Just as in Sanskrit (Hock 1982a), the quotative is used in naming and labeling in
Marathi, Assamese (IA), the Dravidian languages and in some Tibeto-Burman
languages (Subbarao et al. 1989).
204 Complementation
Marathi (IA)
(27) tyā gāwāt widzay mhan.ūn/ asa ek mulgā malā bhet.la
that town. loc Vijay quot (identifier) that one boy I.dat meet.pst.3sm
‘In that town I met a boy named Vijay.’
Telugu (DR)
(28) indira anē maniši ikkad.i-ki vaccindi
Indira quot (identifier) person here-to came
‘A person named Indira came here.’
Out of more than thirty functions that the quotative performs in SALs
(Subbarao et al. 1989), we have discussed only a restricted set due to limi-
tations of space. Though both IC and FC function as complementizers, there
are differences in their syntactic behaviour, which we discuss in the following
section.
6.3 Structural differences between an IC and an FC

6.3.1 An FC cannot be dropped
We shall now discuss the differences in the occurrence of the COMP in terms
of its presence or absence in an elliptical answer to a question, to distinguish
between an IC and an FC. We shall show that the FC is more closely tied in or
attached to the embedded clause than the IC is.
For a question in (29) in Telugu (DR), for example, the answer must contain
the FC as in (30). Sentence (30) is ungrammatical in Telugu if the FC ani is not
present.
Telugu (DR)
(29) mādhuri ēmi (ani) andi
Madhuri what comp said
‘What did Madhuri say?’
(30) tanu pei.i.i cēsu konnānu *(ani)
self marriage do VR.pst.agr comp
‘that she got married’
In Sema (TB) and Dakkhini (IA) too, the FC, the quotative, cannot be dropped
in an elliptical answer.
6.3 Structural differences – IC and FC 205
Sema (TB)
(31) pa- ye kla- va *(pi)
he- nom wed- pst FC
‘that he got married’
(Achumi 2000; Achumi and Subbarao 2005)
Dakkhini (IA)
(32) unõ bajār ku jāte *(kate)
they market to are going FC
‘(that) they are going to the market’
(Harbir Arora, p.c.)
The data demonstrate that the FC is closely tied in or attached to the embedded
clause.
We now demonstrate that, in contrast, the IC cannot occur in sentence-initial
position in an elliptical answer in Hindi-Urdu (IA).
Hindi-Urdu (IA)
(33) mādhurı̄ ne kyā kahā
Madhuri erg what said
‘What did Madhuri say?’
(34) (*ki) mãı̃ ne šādı̄ kar lı̄
IC I erg marriage do took
‘that I got married’
In Kashmiri (IA), the elliptical answer with ki/zi is grammatical, although not
very commonly used. Further, the COMP in an elliptical answer is also used to
provide emphasis to the answer (Kak, ms).
elliptical answer
Kashmiri (IA)
(35) (ki/zi) burhān mūd
IC Burhan died
‘that Burhan died’
(Kak, ms)
In Ho (Munda), in answer to a question in (36), the IC ci may or may not occur

(37).
Ho (Munda)
(36) soma cinā kaji- tad.- a
Soma what say- pres- [+fin]
‘What does Soma say?’
206 Complementation
(37) (ci) ay e? dai akachaka tan- a

comp he win surprise pres- [+fin]
‘That he won is surprising.’
In Bangla (IA), the IC may be retained, or may be omitted.
Bangla (IA)
(38) (je) o bie koreche
COMP she marriage has done
‘(that) she got married’
Thus, to recapitulate, in an elliptical answer, while an FC cannot be dropped,

the presence of an IC is not obligatory. The data above show that the IC is
not as closely tied into the embedded clause as the FC is. We provide further
evidence in support of our claim.
6.3.2 No pause permitted with FCs

According to Dayal (1996: 37), “in a finite complementation structure a pause
can intervene quite naturally after ki [the IC ‘that’ in Hindi-Urdu] and before
the subordinate clause.” Such a pause is not permitted between the quotative FC
ani and the preceding subordinate clause in Telugu (except to impart narrow
scope interpretation as in (104)) and the other Dravidian languages, or in
Bangla, Dakkhini (IA) and Tibeto-Burman languages. This fact provides further
evidence to demonstrate that the FC is more closely attached to the embedded
clause than the IC. The next piece of evidence comes from the fact that the IC
and the FC are not mutually interchangeable, which we discuss now.
6.3.3 Complementary distribution of ICs and FCs

Bayer (2001: 15) points out that the IC je and FC bole in Bangla (IA) are in
complementary distribution, and they cannot be ‘swapped.’ Thus, their posi-
tions are fixed. Our data indicate that this generalization holds good for all other
SALs that have an IC and an FC.6
6.3.4 COMP agreement with the matrix subject

In a few Munda languages, the complementizer shows agreement with the
matrix subject. Such agreement only takes place in languages with an FC, and
we do not know of any South Asian language with an IC where such agreement
takes place. Hence, we consider this a distinguishing factor between an IC and
an FC.
The pronominal clitic of the matrix clause subject cliticizes onto the right of
the COMP (39) in Mundari (Munda). The post-sentential COMP in Mundari is
a quotative form of the verb to say. The first person subject agreement marker
-
occurs to the right of the FC mente.
Mundari (Munda)
(39) proi [CP somaj - hiju?- aka- n- a mente-CP ]
i,∗ j ayum-
I Soma- come- asp- [−tr]- [+fin] comp- 1 s hear-
ke - d- a
asp- pst- [+fin]
‘I heard that Soma came.’
(Osada 1992: 67; the glosses have been slightly modified)
In (39) the subject agreement marker cliticizes onto the right of the preverbal
constituent. The embedded S plus the COMP is the preverbal constituent in
(39), and hence the agreement marker occurs to the right of the COMP mente.
In Juang (a Munda language spoken in Orissa) too, the final complementizer
gamɔjɔ agrees with the subject of the matrix clause. The complementizer
gamɔjɔ is the quotative form of the verb to say.
Juang (Munda)
(40) nii amte [rabij ɔpɔmanian nei,∗ j - gamɔjɔ] nei - gamɔ
we you-acc Ravi was insulted 1 pn- comp 1 pn- told
‘We told you that Ravi was insulted.’
(Patnaik 1999)7
6.3.5 Nature of ICs and FCs

There is general agreement that the FC, derived from the verb ‘to say’ in
SALs, is more ‘verb-like’ than the IC. Bayer (2001) has articulated this posi-
tion strongly. Bayer (2001) and Bal (1990) consider the IC an operator. Bhatt
(1994) and Davison (1995: 2) view it as a conjunction, and the IC ki is not
treated as a complementizer. Mohanty (2002) argues that it is an “emphasizer.”
According to Dwivedi (1994), ki- clauses are licensed like coordinate con-
stituents. Both Bal (1990) and Mohanty (2002) claim that the FC boli is the
only overt complementizer in Oriya. We shall not go into further details.
Bayer (2001: 14) presents evidence that there could be a lexical trigger to
distinguish FC-clauses from IC-clauses.
6.3.6 Syntactic and semantic selection

The IC je ‘that’ and the FC bole in Bangla, for example, “differ with respect
to their property of being semantically selected” (Bayer 2001: 15). ICs “can be
208 Complementation
selected by factive verbs, attitude verbs, verbs of perception (in their epistemic
reading) as well as by nominals.” FCs on the other hand are restricted in various
ways. They do overlap in the semantic selection of verbs, but, while certain
predicates, such as to know, to think, to hear, etc., permit both an IC and an
FC, predicates such as to see, to be possible, to be likely, etc., do not (Bayer
2001: 15).
Bangla (IA)
(41) [ram kolkatta- y jacche bole] {janlam/ bhablam/ šunlam/
Ram Kolkata- loc goes quot knew-1 thought-1 heard-1
*dekhlam/ *ɔšombhob}
saw-1 impossible
‘That Ram is going to Kolkata [I knew / I thought / I heard / *I saw/*is
unlikely].’
(42) {janlam/ bhablam/ šunlam/ dekhlam/ ɔšombhob} [je [ram kolkatta-

knew-1 thought-1 heard-1 saw-1 impossible IC Ram Kolkata-
y jacche]]
loc goes
‘[I knew / I thought / I heard / I saw / It is unlikely] that Ram is going to
Kolkata.’
(Bayer 2001: 15)8
6.3.7 ICs and FCs and question expressions

Both the IC and FC usually permit a question word to occur in the embedded
complement, though the IC je in Bangla (IA) does not. Bayer (2001: 17) points
out: “another important difference between bole [the FC in Bangla] and je [the
IC in Bangla] is that bole must be underspecified for the feature [+wh], whereas
je is always negatively specified as [−wh].” Thus, “the indirect question is
compatible” with an FC, while it is not permitted with an IC in Bangla as
(43) and (44) show. Sentence (43) “shows that the indirect question ke ašbe is
compatible with [the FC] bole, while it is incompatible with [the IC] je,” as in
(44) (Bayer 2001: 17). Thus, (44) is grammatical, if je is not present.9
Note that the FC in (43) is optional.
Bangla (IA)
(43) ram ke aš- be (bole) jiggeš koreche
Ram who come- will quot question made
‘Ram asked who will come.’
(44) ram jiggeš koreche (*je) ke aš- be
Ram question made IC who come- will
‘Ram asked who will come.’
However, in Hindi-Urdu (IA) an IC is permitted in such cases, as in (45), and

in Marathi (IA) an FC mhan.ūn/asa is permitted as in (46) (Bayer 2001: 17). In
Marathi, the IC as well as the FC may occur in the same sentence, as in (46).
Hindi-Urdu (IA)
(45) tum jānte ho [ki [us ne kyā kiyā]]
you know pres comp he/she erg what did
‘You know what he/she did.’
Marathi (IA)
(46) mohan ne witšārle [kı̄ [sudhā kut.he gelı̄] {asa/ mhan.ūn}]
Mohan agentive asked IC Sudha where went FC quot
‘Mohan asked where Sudha went.’
Thus, whether an IC-clause permits a question word is language-specific, and

hence parametric. Bayer’s contention is that an FC and an IC serve “distinct
functions.” An FC has a verbal feature, and an IC has the feature of an operator,
in view of their distinct sources of origin. While the FC, a quotative, is derived
from the verb to say, the IC ki comes from a question word. Masica (1999:
149) points out that the Bangla initial complementizer je has its origin in the
Sanskrit complementizer yat ‘that.’10
6.3.8 Linear order of the COMP (IC and FC) and yes/no (polarity)
question words: a mirror image relationship
Rizzi (1997) proposes that CP has a much finer “articulated structure,” and is
really a series of projections consisting of four possible functional heads such
as Force, Topic, Focus, etc. Davison (2007a) following Rizzi’s idea suggests
that FCs are in a different projection from ICs. She discusses the case of ICs
in languages such as Hindi, Punjabi and Kashmiri, and argues that the IC ki
‘that’ in Hindi-Urdu is a marker of subordination in the FORCE projection.
Davison (2007a) presents a correlation between the positions of occurrence of
the COMP and the yes/no (polarity) question marker in Indo-Aryan languages.
She demonstrates how the complementizer precedes the yes/no question marker
in sentences with IC, while it follows it in sentences with FC (for details, see
Rizzi 1997; Davison 2007a; and the appendix 1 at the end of this chapter).
Languages that have an FC have a sentence-final polarity question marker
and languages that have an IC have a sentence-initial question marker. Further,
while the yes/no question marker precedes the FC, it follows the IC. Thus, there
is a mirror image relationship in the order of occurrence of these, which is as
follows:
210 Complementation
A. Languages with FC
pol q mkr – FC
B. Languages with IC
IC – pol q mkr
Our analysis of the SALs shows that this correlation holds good in all the SALs
that we worked on.11
Biberauer, Newton and Sheehan (2009) argue that the changes that occurred
in contact situations in South Asian languages systematically obey a univer-
sal labeled the Final-over-Final Constraint (FOFC) proposed in Biberauer,
Holmberg and Roberts (2007, 2008). We shall present a brief discussion of this
in the appendix at the end of this chapter.
6.3.9 The semantic nature of the matrix predicate and the form of the FC
In Manipuri (TB), Gujarati and Marathi (IA), it is the semantic nature of the
matrix predicate that determines the form of the FC that occurs.
The case of Manipuri (TB)

There are two FCs derived from the verb to say in Manipuri: hay-bə ‘say-
inf’ and hay-nə ‘say-adv mkr.’ Bhat and Ningomba (1995: 436) point out:
“Complements ending in hay-nə refer to a statement whereas the ones ending
in hay-bə refer to a fact (a state, process or action); because of this difference,
the use of these complementizers involves certain pre-suppositional difference.”
For example:
Manipuri (TB)
(47) tombə- nə lak- kəni hay-nə ta- y
Tomba- nom come- nonfut comp hear- nonfut
‘I heard that Tomba would come.’ (unplanned coming)
(48) tombə- nə lak- kəni hay-bə ta- y
Tomba- nom come- nonfut comp hear- nonfut
‘I heard that Tomba would come.’ (planned coming)
(Bhat and Ningomba 1995: 436)
Bhat and Ningomba (1995: 437) further point out: “according to [47] Tomba’s
coming might not actually have been programmed, whereas according to [48],
it is considered a settled fact.”
It appears that the complementizer hay-bə ‘say-inf’ occurs with factive
predicates such as khaŋ ‘to know,’ niŋamakta ‘to regret,’ pharakta ‘to become
odd,’ and the complementizer hay-nə ‘say-adv mkr’ occurs with non-factive
predicates such as thajə ‘to believe,’ khan ‘to think,’ khaŋ ‘to guess,’ oythok
‘to be possible’ (used with a negative suffix) (Bhat and Ningomba 1995: 438).
The following examples are illustrative.
w i t h t h e f a c t i v e p r e d i c a t e khaŋ ‘k n o w ’
(49) [cp məhak- nə cak ca-y hay-bəcp ] əy khəŋ- ŋi
he- nom rice eat-nonfut FC I know- nonfut
‘I know that he ate rice.’
w i t h a v e r b o f c o m m u n i c a t i o n hay ‘t o s a y ’
(50) əy [cp pro cat-kani hay-nəcp ] ha-y
I go-fut say-adv say-nonfut
‘I said that I will go.’
It is interesting to note that the predicate niŋsəŋ has the interpretation of

‘remember’ while occurring with the COMP haibə, and it has the interpretation
of ‘remind’ while occurring with the COMP hainə. In (52), niŋsəŋ ‘remember’
is a factive predicate which presupposes the truth of the complement, while the
predicate niŋsəŋ ‘remind’ is not.
(51) [CP məhak- ca- gədə- bə- ni hay-bəCP ] əy niŋsəŋ- ŋi

he- eat- fut- inf- copula say-inf (FC) I remember- nonfut
‘I remembered that he should eat.’
(52) [CP məhak- ca- gədə- bə- ni hay-nəCP ] əy- nə niŋsəŋ- ŋi

he- eat- fut- inf- copula say-adv (FC) I- adv remind- nonfut
‘I reminded (him) that he should eat.’
(Bhat and Ningomba 1995: 439–440)
Sarju Devi (2006) also discusses the differences in the occurrence of the two
complementizers in Manipuri in some detail. However, from the data and
analysis that are provided in Bhat and Ningomba (1995) and Sarju Devi (2006),
it is not clear whether the generalizations that they made would clearly hold.
Further research needs to be done on this issue.
The case of Gujarati (IA)

In Gujarati (IA), an FC -em occurs with lakh ‘write’-type verbs and the FC
-te occurs with jān. ‘know’-type predicates. P. J. Mistry (p.c.) points out that
Dayashankar Joshi attempted to correlate this variation with factive/non-factive
distinctions in the matrix predicate. S2 stands for an embedded clause in (53)
and (54).
212 Complementation
Gujarati (IA)
the f c -em w i t h a no n - f a c t i v e p r e d i c a t e
(53) rāj [[S2]- em] lakhe che
Raj FC writes
‘Raj writes that . . . ’
the f c -te w i t h a f a c t i v e p r e d i c a t e
(54) rāj [[S2]- te] jān.e che
Raj that knows
‘Raj knows that . . . ’
The case of Marathi (IA)

In Marathi, factive predicates with the FC -te do not permit Wide Scope reading,
whereas non-factive predicates with the FC -te permit Narrow Scope reading
(Wali 2006: 210, and also see p. 227 for a discussion). Thus, Gujarati and
Marathi share the feature of having the homophonous FC -te with factive
predicates.
To recapitulate, in this section we have shown that there are structural dif-
ferences between an IC and an FC, which include:
(i) while an FC cannot be dropped in an elliptical answer, the presence of the
IC is not obligatory;
(ii) while a pause is permitted between the IC and the embedded clause, a
pause is not permitted between the embedded clause and the FC;
(iii) ICs and FCs are in complementary distribution, and they are not mutually
interchangeable;
(iv) the matrix subject agreement marker on the complementizer is found only
in languages with an FC; and
(v) the IC is like an operator and, hence, is incompatible with question words
in Bangla, while the FC, a participial/infinitival form of the verb in most
of the SALs, is verb-like and, hence, there is no such restriction.
We have shown that there is a neat correlation between (I) the occurrence of
an IC or an FC and the position of the yes/no question marker in SALs, and
(II) the semantic nature of the matrix predicate and the form of the FC in
Manipuri (TB) and Gujarati and Marathi (IA).
6.4 Some crucial issues concerning FCs and ICs

6.4.1 Finiteness and case marking
In Khasi (Mon-Khmer) and Assamese (IA), the subject of a [+finite] CP
clause gets case-marked by the matrix verb when the complementizer is overtly
6.4 Some crucial issues – FCs and ICs 213
present. In many SALs and languages across the world the subject of an embed-
ded clause cannot be case-marked by the matrix verb as the complementizer
(IC/FC) blocks such case marking. However, in a few SALs, either with an
IC as in Khasi (Mon-Khmer) or an FC as in Assamese and Bangla (IA), the
embedded subject does get case-marked by the matrix verb.
The case of Khasi, an SVO language

Recall that Khasi has only the initial complementizer ba. The subject of the
embedded clause at a deeper level is lin ‘Lin,’ and the verb of the embedded
clause j ɔp ‘win’ is [+finite] in (55). The embedded subject lin ‘Lin’ gets
accusative case-marked by the matrix verb kwa? ‘want’ in (55).
Khasi (Mon-Khmer)
(55) u- lam u- la- kwa? [DP ya- ka- linDP ]i [CP ba ti ka-
3m,s- Lam 3m,s- pst- want acc- 3 f,s- Lin comp 3 f,s-
n- jɔpCP ]
fut- win
‘Lam wanted Lin to win.’
(Temsen and Subbarao 2004; Temsen 2006)
The question that arises is: how does the embedded subject get case-marked,
when the initial complementizer ba ‘that’ is overtly present? Since in (55) it
does, it is a clear violation of the ‘Tensed-S Condition’ (proposed in Chomsky
1973: 238) and the PIC (the Phase Impenetrability Constraint – Chomsky 2001:
12). Temsen and Subbarao (2004) and Temsen (2006) provide three pieces of
evidence to demonstrate that the embedded subject moves out of S2 and is in
the object position of the matrix clause (see appendix 6.2 on the Website for
further details).
The case of Assamese and Bangla (IA)

In Assamese (IA), in the IC je- ‘that’ clause in (56):
(i) only a pronoun occurs as the subject of the embedded clause, and
(ii) an anaphor cannot occur as the matrix subject and the embedded subject
are not clause-mates.
w i t h a n I C je ‘t h a t ’-c l a u s e
Assamese (IA)
(56) mala- yei xopun- ot dekhile [CP je [S2 taii / *nijo- kei ud.i-
Mala- nom dream- in saw that she self- acc fly-
asilS2 ]CP ]
be.pst progr
‘Malai saw in her dream that shei was flying.’
214 Complementation
In contrast, when the FC quotative buli occurs, the subject of the embedded
clause surfaces as nijo- ke ‘self-acc,’ as the matrix subject and embedded subject
become clause-mates. The anaphor gets accusative case-marked by the matrix
verb.
w i t h a n f c buli ‘t h a t ’-c l a u s e
(57) malai - ye nijo- kei xopun- ot [CP ti plεn- ɔr ud.i-thoka buliCP ]
Mala- nom self- acc dream- in plane- in was flying quot
bhabhile
thought
‘Mala thought that she was flying in a plane in her dreams.’
(Subbarao 1984b)
In Bangla (IA) too, a similar pattern obtains.

Bangla (IA)
(58) kɔmolai nije- kei plen-e kor-e ur.che bole šɔpno dekhechilo
Kamala self- acc by plane was flying FC dream saw
‘Kamala dreamt that she was flying by plane.’
(Subbarao 1984b)12
In Subbarao (1984b), the occurrence of the accusative case-marked anaphor

nijo-ke ‘self-acc’ in Assamese in (57) and nije-ke ‘self-acc’ in Bangla in (58)
was attributed to a violation of the Tensed-S Condition of Chomsky (1973).
In more recent theoretical terms, this can be viewed as a violation of the PIC
(the Phase Impenetrability Constraint – Chomsky 2001), as the anaphor needs
to be valued for its accusative case by the matrix verb bhab ‘to think’ in
Assamese and dekh ‘to see’ in Bangla in a mutual c-command relationship
with it.
This means that the DP has to move out of the CP, and not only does such
movement violate the PIC, but also there is no apparent motivation for such
movement to a higher clause. In conclusion, case marking of the embedded
subject by the matrix verb takes place in select languages with an IC (as in
Khasi) and with an FC (as in Assamese and Bangla).
6.4.2 The IC and Principle A violation

Marathi (IA) permits a long-distance antecedent for a complex anaphor, which
is a violation of Principle A of the Binding Theory (Wali 2000: 534, and see
also chapter 3). In Marathi (IA), in sentences with an IC, the anaphor in the
embedded clause may have an antecedent in the matrix clause, which is a clear
violation of Principle A. In (59), the antecedent subject anū ‘Anu’ in the matrix
clause is coindexed with the complex anaphor āplyā-la ‘self-acc’ and āplyā
swatāhā-lā ‘self-acc.’
Marathi (IA)
(59) anū-i lā wāt.ta [CP kı̄ [S2 madhūj ne āplyā- lai/∗ j *swatāhā-
Anu (f)- dat feels IC Madhu (m) erg self- acc self-
lā/ āplyā swatāhā- lai/∗ j phasawlaS2 ]CP ]
acc self- acc deceived
‘Anui feels that Madhuj deceived heri /*himselfj .’
(Pandharipande 1997: 210; the glosses have been slightly modified)13
In this case the PIC seems not to have been violated, as there is no movement of
a DP from the embedded clause, but the question remains whether the elements
in the CP phase are ‘visible’ to a higher antecedent in the matrix clause for
establishing antecedent–anaphor relationship. They do not seem to be.
6.4.3 Subject to Subject Raising

The case of Subject to Subject Raising in Maithili (IA) deserves attention.
Yadava (2007) shows that the subject of a tensed clause with the IC je in
Maithili (IA) raises to the empty subject position of the matrix clause, when the
seem-type of non-factive predicates occur in the matrix clause. The predicates
include lagnāi, bujhenāi and pratit honāi with the interpretation of ‘seems,
appears, looks like,’ etc. (Yadava 2007: 144). Yadava demonstrates that, after
such movement, the DP-trace t, the empty category, obeys Principle A of the
Binding Theory. The following examples are illustrative.
Maithili (IA)
(60) [IP [DP e] lagait aich [CP je [IP ahã˜ ahā ghar nahi jā sakab]]]
(it) seems IC you today home not go can
‘It seems that you cannot go home today.’
(61) [IP [DP ahã˜i ] lagait aich [CP je [IP ti ahā ghar nahi jā sakab]]]
you seems IC today home not go can
‘You seem not to be able to go home today.’
(Yadava 2007: 144)
In support of his claim, Yadava (2007: 154) provides evidence which shows that
the matrix verb must agree with the raised/derived subject. Recall that Maithili
has a rich system of verb agreement (see chapter 4 for details).
To summarize the above discussion, the PIC is violated in verb-final lan-
guages with an FC and non-verb-final languages with an IC.
6.4.4 Direct vs. indirect speech

In SALs, whether the COMP is an IC or an FC, a quote is repeated verbatim
along with the IC or FC, and no person harmony or tense harmony between
216 Complementation
the matrix clause and the embedded clause is required (Masica 1991: 403;
Subbarao 2001).
Languages with an FC
In Sinhala, just as in other SALs, the sentence can be interpreted “as either
a direct or indirect quotation with different meanings” (Gair 2003: 806–807).
The ambiguity arises due to the fact that the quotation is repeated verbatim. The
subscripts ‘-i’ and ‘-j’ refer to the subject of the matrix clause and the speaker,
respectively.
Sinhala (IA)
(62) [(mat.əi,j wad.uwædə kərannə bǣ) kiyəla] nimali kiwwa
I.dat carpentry do.inf can’t quot Nimal say.pst
‘Nimal said “I can’t do carpentry”’ or ‘Nimal said that
I [speaker] can’t do carpentry.’
(Gair 2003: 806–807)
In Kannada (DR) and Sema (TB) too, a similar ambiguity arises.
Kannada (DR)
(63) [(nānui,j bombāige hōde) anta] ammai hēl.idaru
I Bombay.dat go.pst.1s quot mother say.pst.3hon
“‘I [= mother] went to Bombay,” mother said.’
‘Mother said that I [= speaker] went to Bombay.’
(Sridhar 1990: 45)
Sema (TB)
(64) atolii nɔ he [(nii,j ye dilli lo wu- ni-) pi]
Atoli nom wrote I nom Delhi to go- fut- quot
‘Atoli wrote that she’d go to Delhi.’/‘Atoli wrote, I [speaker]
would go to Delhi.’
(Achumi and Subbarao 2005)
In Manipuri (TB) (Bhat and Ningomba 1995: 441–442) and Angami (TB)
(Kevichüsa 2007), there is no person harmony in such cases.
Languages with an IC
Though Hindi-Urdu has an IC that is not a quotative, the direct quote is repeated
verbatim and, hence, the sentence in (65) contains the quotation with an IC
preceding it.
Hindi-Urdu (IA)
(65) rešmāi ne likhā thā [ki (mãı̃i,j dillı̄ jā rahı̄ hũ˜ i,j )]
Reshma erg wrote had comp I Delhi go progr am
‘Reshma wrote that I [= Reshma] would be going to Delhi’ or ‘Reshma
wrote that I [= speaker] would be going to Delhi.’
However, Khasi (Mon-Khmer), a verb-medial language, differs from the other

SALs that we know of. Person harmony is obligatory, and hence a direct quote
with no change in person in the embedded subject position co-refers with only
the matrix subject, and not the speaker. Tense harmony is not needed.
Khasi (Mon-Khmer)
(66) ka- barii ka- la- thɔʔ [ba (proi kai - n- lεit ša delhi)]
3s, f- Bari 3 s, f- pst- write IC 3 s, f- fut- go to Delhi
Literally: ‘Bari wrote that she (Bari) will go to Delhi.’
A direct quote refers only to the speaker of the utterance.
(67) ka- barii ka- la- thɔ? ba ŋa∗ i,j - n- lεit sha- delhi
3s, f- Bari 3 s, f- pst- write IC 1 s, f- fut- go to- Delhi
‘Bari wrote that I [the speaker, not Bari] will go to Delhi.’
(Temsen 2006)
In the case of direct quotes, the SVO Khasi (Mon-Khmer) differs from the other
SALs, and it is similar to English to the extent that it observes person harmony
but not tense harmony.
Creative errors in Indian English

In Indian English, ‘creative errors’ occur when a speaker uses the direct quote
with the that complementizer while reporting someone else’s statement, as in
(68). The verbatim quote “I am not well” is embedded in the matrix clause with
the IC that, and this leads to misinterpretation for the hearer. The subscript ‘k’
is coindexed with the speaker.
(68) Ramani told Rashmij that Ii/∗ j/k [Raman or the speaker] am not well.
The embedded subject I may ambiguously refer either to the speaker of the
utterance or to Raman.
To summarize the above discussion, both IC- and FC-clauses share the
following features:
(i) case marking of the embedded subject by the matrix verb takes place in
Assamese and Bangla with an FC and Khasi with an IC;
(ii) the occurrence of the polymorphemic anaphor in a finite complement
clause is permitted with an antecedent in a higher clause, violating Prin-
ciple A of the Binding Theory;
218 Complementation
(iii) a verbatim quotation is repeated, and no person harmony or tense harmony

with the matrix clause is required in the embedded clause, except in
Manipuri and Angami (TB) with an FC, and Khasi (Mon-Khmer) with an
IC; and
(iv) the lack of person and tense harmony in complement clauses is reflected
in Indian English too.
In the following section we discuss issues related to the canonical position of
occurrence of an IC- and an FC-clause.
6.5 Position of occurrence of an IC- and an FC-clause and the Case

Resistance Principle (CRP)
One of the crucial issues that concern ICs and FCs is: what is the canonical
position in which these clauses occur? Are there any restrictions on the specific
position of their occurrence? Do ICs and FCs behave the same way with regard
to their mobility? Are there any universal principles that govern or prohibit
their occurrence in a specific position?
6.5.1 FC-clauses
First we consider FC-clauses. Recall that all FC-clauses have the quotative, a
grammaticalized form of the verb to say, as complementizer. In Marathi (IA)
and Gujarati (IA), there is also a different set of FCs derived from a pronoun.
In Marathi it is asə/asa ‘like this’ (literally), and in Gujarati it is em/evũ ‘this
way’ (literally) (P. J. Mistry and Trupti Nisar, p.c.).
The canonical position of the FC quotative clause in SALs is the same as the
position in which the direct object occurs. It may stay in situ in its canonical
(clause-medial) position14 in the matrix clause as in (69), or may occur to the
right of the verb of the matrix clause as in (71), or to the left of the verb of
the matrix clause as in (73), illustrated in (70), (72) and (74), respectively,
from Mizo (TB). Such occurrence in the three positions is found in Dravidian,
Indo-Aryan and Tibeto-Burman – except in Rabha – languages that we have
studied.
fc in situ
(69) [S1 - - - [CP S2 COMP] VP1 S1 ]
Mizo (TB)
(70) zova- n [CP naktuk cu colhni a- ni- aŋ tiʔCP ] a- hria- aŋ
Zova- erg tomorrow def holiday 3 s- be- fut FC 3 s- know- fut
6.5 Position and Case Resistance Principle 219
To the right of the verb of the matrix clause

(71) [S1 - - - VP1 [CP S2 COMP]S1 ]
(72) zova- n a- hria [CP naktuk cu colhni a- ni- aŋ tiʔCP ]
Zova- erg 3 s- know tomorrow def holiday 3 s- be- fut FC
To the left of the verb of the matrix clause

(73) [S1 [CP S2 COMP] - - - VP1 S1 ]
(74) [CP naktuk cu colhni a- ni- aŋ tiʔ CP ] zova- n a- hria
tomorrow def holiday 3 s- be- fut comp Zova- erg 3 s- know
Rabha (TB) is the only language that we know of where an FC-clause with a
quotative as the complementizer can either remain in situ, or occur to the left
of the matrix verb, but not to the right, as in (75).
Rabha (TB)
(75) *bharoti (be) kani- bamɯn [CP te u- (be) riba- nɔ n-eCP ]
Bharati nom tell- pst today she- nom come- fut FC
‘Bharati told that she would come today.’
(Subbarao et al., ms)
6.5.2 The position of occurrence of the IC-clause

Let us now look at IC-clauses. The positions of occurrence of the IC-clause are
restricted in comparison to the positions of occurrence of the FC-clause.
(i) Clause-final position

An IC-clause can occur in SALs to the right of the matrix verb in clause-final
position, as in Khasi (Mon-Khmer), a non-verb-final language. In (76), the
embedded complement occurs to the right of the matrix verb ŋεyt ‘to believe.’
Khasi (Mon-Khmer)
(76) u- kmεn u- m- ŋεyt [CP ba [S2 ki- brēw ki- hyar patεŋ
3m- Kmen 3m- neg- believe IC p- humans p- descended
na- ki- šrēʔS2 ]CP ]
abl- p- monkey
(Temsen 2006)
220 Complementation
(ii) Clause-medial (canonical)15 position

It was first noted in Subbarao (1974, 1984a) that in Hindi-Urdu an IC-clause
“cannot occur [in situ in its canonical object position] unless it has the head
noun to the left of its structure. Such a complement [without the head] must
be extraposed to the right of the VP. We shall call this the ‘Complement Push
Constraint’” (Subbarao 1974, 1984a: 103).
Hindi-Urdu (Davison 1994; Dayal 1996) and Bangla (Bayer 2001) also show
that an IC-clause cannot stay in situ in its canonical position of occurrence, and
it has to occur/move to the right of the VP of the matrix clause.
Bayer (2001: 19) mentions that Turkish ki-clauses and Persian ke-clauses
also obey the same restriction. Thus, it appears that there is a general restriction
on the occurrence of an IC-clause in its canonical position.
One might explain this phenomenon in terms of Stowell’s Case Resistance
Principle (CRP), which states that a tensed clause cannot occur in a case-marked
position and, hence, is moved from that position. The constraint proposed in
Subbarao (1984a) is similar in spirit to Stowell’s CRP.
We shall provide some examples from Hindi-Urdu (IA), Bangla (IA) and
Khasi (Mon-Khmer), in which an initial complement clause cannot occur in its
canonical position.
a complement clause occurring in clause-internal position is

not permitted
Hindi-Urdu (IA)
(77) *ham ne [CP ki [S2 ve ā rahe hãı̃S2 ]CP ] abhı̄ abhı̄ sunā
we erg IC they come progr pres just now heard
‘We just now heard that they are coming.’
(Subbarao 1974, 1984a: 103)
Bangla (IA)
(78) *chele t.a [CP je [S2 or baba aš- beS2 ]CP ] šuneche
boy cl IC his father come- fut heard
Intended meaning: ‘The boy heard that his father will come.’
(Sentence extrapolated from Bayer 2001: 14)
Khasi (Mon-Khmer)
(79) *u- kmεn [CP ba [S2 ki- brēw ki- hyar patεŋ na- ki- šrēʔS2 ]CP ]
3 m- Kmen IC p- humans p- descended abl- p- monkey
u- m- ŋεyt
3 s,m- neg- believe
6.5 Position and Case Resistance Principle 221
complement clause occurring in clause-medial (canonical)

p o s i t i o n w i t h a lexical head ( n p c o r r e l a t e )
When headed by lexical head nouns such as yah baat / yah khabar / yah
samācār ‘this news’ in Hindi-Urdu (IA), an IC-clause can occur in clause-
internal position, as the head noun can be assigned structural case by the matrix
predicate and, hence, there is no violation of the Case Resistance Principle.
cp with a lexical head np in canonical object position
clause-internally
Hindi-Urdu (IA)
(80) ham ne [NP [NP yah samācārNP ] [CP [S2 ki ve ā rahe
we erg this news IC they come progr
hãı̃S2 ] CP ] NP ] abhı̄ abhı̄ sunā
pres just now heard
‘We just now heard (the news) that they are coming.’
(Subbarao 1974, 1984: 103)
Such occurrence with a lexical head noun in clause-internal position neatly
correlates with the occurrence of the ki ‘that’ complement clause with a head
NP in elliptical answers. In elliptical answers too, the lexical head noun must
be overtly present as in (82), or else the sentence is ungrammatical.
speaker a
(81) kōmal ne kaun sı̄ bāt chupāyı̄
Komal erg which matter hid
Literally: ‘Which matter did Komal hide?’
speaker b
(82) yah bāt ki usne šādı̄ kar lı̄
this thing IC she- erg marriage do took
Literally: ‘The thing (news) that she got married’
To conclude, an IC-clause without a lexical head cannot occur clause-internally

in canonical position, while one with a lexical head can.
We shall now look at infinitival complements. They can occur clause-
internally in their canonical object position in SALs. Since they behave like
NPs in Hindi-Urdu (Subbarao 1984a), their occurrence in clause-internal posi-
tion is predicted by the Case Resistance Principle. Dayal (1996: 26–27) treats
non-finite complements as gerunds, and observes that such occurrence does
not violate the Case Resistance Principle, as they lack [+tense] feature. Butt
(1993) also treats them as NPs.
(iii) IC-clause in clause-initial position

We shall now discuss the occurrence of the IC-clause in clause-initial position.
An IC ki-clause in Hindi-Urdu (IA) or an IC je-clause in Bangla (IA) cannot
occur/move to sentence-initial position of the matrix clause.
222 Complementation
Hindi-Urdu (IA)
(83) *[CP ki [S2 ve ā rahe hãı̃S2 ]CP ] ham ne abhı̄ abhı̄ sunā
IC they come progr pres we erg just now heard
‘We just now heard (the news) that they are coming.’
(Subbarao 1974, 1984a: 103)
Bangla (IA)
(84) *[CP je [S2 or baba aš- beS2 ]CP ] chele t.a šuneche
IC his father come fut boy cl heard
Intended meaning: ‘The boy heard that his father will come.’
(Bayer 2001: 14)
We will now consider the case of Khasi (Mon-Khmer) complement clauses.
Khasi, a non-verb-final language, also does not permit a CP clause with the
COMP ba to occur in clause-initial position. Recall that none of the Indo-Aryan
languages which are verb-final with an IC permit this option. In (85) the order
is VSO.
Khasi (Mon-Khmer)
(85) [CP *ba [S2 ka- n- lεyt ka- monS2 ]CP ] u- kmεn
verb subject
IC 3f,s- fut- go 3f, s- Mon 3m,s- Kmen
u- ɔŋ ya- ŋa
3m,s- said acc- I
‘“That Mon will go,” Kmen told me,’ ie., ‘Kmen told me that Mon will go.’
(Temsen 2006)
Change in word order to SV(O) in S2 does not render the sentence grammatical,
as (86) shows.
(86) [CP *ba [S2 ka- mon ka- n- lεytS2 ]CP ] u- kmεn
subject verb
IC 3f,s- Mon 3f,s- fut- go- 3m,s- Kmen
u- ɔŋ ya- ŋa
3m,s- said acc- I
‘“That Mon will go,” Kmen told me,’ i.e., ‘Kmen told me that Mon will go.’
We do not have any explanation as to why Khasi, a non-verb-final language,
does not permit this option of having a complement clause in clause-initial
position.
If the COMP ba is not overtly present, SVO order is permitted, as (87) shows.
(87) [S2 ka- mon ka- n- lεytS2 ] u- kmεn u- ɔŋ ya- ŋa

3f,s- Mon 3f,s- fut- go 3m,s- Kmen 3m,s- said acc- I
subject verb
‘“That Mon will go,” Kmen told me,’ i.e., ‘Kmen told me that Mon will go.’
Such sentences as (87) are permitted in Hindi-Urdu too, if the IC ki is not overtly
present. Subbarao (1984a: 108) treats sentences parallel to (87) in Hindi as cases
of Sentence-Lifting (Ross 1973).16
In this section we have shown that:
(i) an FC-clause can occur either to the left or right of the matrix clause
(except in Rabha [TB]);
(ii) an IC-clause cannot occur clause-internally in any of the SALs, including
the non-verb-final Khasi language, and is obligatorily moved rightwards;
and
(iii) an IC-clause cannot occur to the left of the matrix clause in any SAL,
including Khasi, which is a non-verb-final language.
We have not addressed the following questions due to lack of substantial data
and evidence from SALs: why are FC-clauses more mobile than IC-clauses,
and why are IC-clauses restricted in their mobility? Can such mobility or non-
mobility be linked to the presence or absence of the dummy NP (NP correlate)?
Bayer (2001) attributes the mobility of the FC-clauses to the absence of the
NP-correlate in the matrix clause. We wish to leave this issue open for further
study.
In the following section we discuss issues related to the scope of wh-
expressions in complement clauses in SALs.
6.6 Narrow Scope and Wide Scope

In this section we discuss wh scope in finite and non-finite complement clauses
and the factors which permit and limit wh scope. There is a variety of syn-
tactic and some phonological, as well as morphophonemic, features which are
correlated with scope relations.
Note that, in SALs: (i) the question word occurs in situ; (ii) it can be freely
moved/scrambled clause-internally; (iii) there is no obligatory movement of the
question word to the Spec position; (iv) more than one q-expression can occur
in a row in any position in a sentence; (v) there are no superiority effects; and
(vi) wh-expressions are clause-bound, and they obey locality conditions.
The scope of a question word in a complement clause may be limited to the
embedded clause alone (Narrow Scope), or it may extend to a higher clause
(Wide Scope). In general, finite clauses with an IC are islands for semantic
scope, and those with an FC are not. For example:
narrow scope
Hindi-Urdu (IA)
(88) anu jānnā cāhti hai ki kis- ne kya kharı̄dā
Anu know.inf want pres that who- erg what bought
‘Anu wants to know who bought what?’
‘*What is such that Anu wants to know who bought it?’
(Dayal 1996: 3)
224 Complementation
Note that the scope of the question word is limited to the embedded clause
alone in (88). In contrast, in (89) the scope of the question word extends to the
matrix clause, and it has a Wide Scope reading.
wide scope
(89) kauni tum soc- te ho [ki ti āyegā]
who you think- imperf pres IC will come
‘Who do you think will come?’ (Wide Scope)
(Dayal 1996: 35)
Sentence (88) has an indirect question (Narrow Scope) interpretation, whereas

(89) has a direct question interpretation.
The main difference between English and Hindi-Urdu according to Davison
(1988) is: the questioned NP kise ‘who.dat’ in the Hindi-Urdu is a VP-internal
topic, while in English, in (90), the question word is in pre-sentential COMP
position with a Wide Scope reading.
(90) Whoi did you say that they saw ei ?

(Davison 1988)
There are several factors that affect the scope of a question expression. These
factors include the finite/non-finite nature of the embedded clause, the position
in which the embedded clause occurs (sentence-initial, pre-verbal or post-verbal
in the matrix clause), the nature of the COMP (IC or FC), the use of specific
scope markers, the role of intonation, etc. The ten points represent parametric
variation among languages in the expression of Wide and Narrow Scope. These
are:
(i) the finite and non-finite nature of the initial complement (IC-) clause;
(ii) finiteness of the IC-clause vs. post-verbal position in blocking Wide
Scope;
(iii) nature of the matrix predicate and Comp: whether it permits Wide Scope
or blocks it;
(iv) the role of intonation, pause and morphophonemic criteria;
(v) the position of the FC-clause;
(vi) the position of question expressions and Scope possibilities;
(vii) topicalization and Wide Scope;
(viii) the occurrence of clitics/particles as Scope Inhibitors;
(ix) the use of pleonastic question expressions as Scope markers;
(x) the position of occurrence of the complementizer: initial or final.
We shall discuss these ten parameters, providing evidence from SALs.
6.6.1 The finite and non-finite nature of the COMP

In a finite clause with an Initial Complementizer the question word has scope
only in the embedded clause, and hence it has only an indirect question inter-
pretation (Narrow Scope). In contrast, in non-finite clauses, the scope of the
question word is invariably in the matrix clause. As Dayal (1996: 23) puts
it succinctly, “finite complements [the IC in Hindi] are scope islands for wh
movement. When the complement is non-finite, however, wh movement to
matrix spec is not only possible but necessary.” Such movement takes place at
LF covertly. To the best of our knowledge, this generalization holds good for
all SALs with an IC or an FC.
narrow scope: wh-expression in a finite clause with ic

Hindi-Urdu (IA)
(91) tum jānte ho [ki us ne kyā kiyā]
you know pres that (IC) he erg what do.pst
‘You know what he did.’
w i d e s c o p e : w h - e x p r e s s i o n i n a no n - f i n i t e c l a u s e
(92) [tum [kyā karnā] jānte ho]
you what do.inf know pres
‘What do you know to do?’ NOT
‘You know what to do.’
(Dayal 1996: 23)
Wh-expressions in infinitives in Telugu (DR) too have only the Wide Scope
interpretation as in (93), and for Narrow Scope interpretation the finite com-
plementizer -ō is employed, as in (94).
w i d e s c o p e : w h - e x p r e s s i o n i n a no n - f i n i t e c l a u s e
Telugu (DR)
(93) nı̄ku [ēmi ceyy.ad.am] vaccu
you.dat what do.gerund come
‘What do you know to do?’ NOT
‘You know what to do.’
narrow scope: wh-expression in a finite clause with fc

(94) [vād.u ēmi cēsēd.(u)- ō] nı̄ku telus(u)- ā
he what did- FC you know- pol q mkr
‘Do you know what he did?’
226 Complementation
6.6.2 Finiteness of the IC-clause vs. post-verbal position

in blocking Wide Scope
Based on Hindi-Urdu data, Mahajan (1987) and Davison (1994) claim that
it is the finite nature of the IC-clause that is relevant in determining scope.
In Hindi-Urdu (IA), “non-finite complements precede the [matrix] verb while
finite complements follow it” (Dayal 1996: 23). This holds good for all other
SALs too.
Dayal (1996: 23) argues that “it is not finiteness but the post-verbal position
that is directly responsible for blocking matrix scope of wh in-situ.” Dayal’s
argument is that finite ki ‘that’ clauses in Hindi-Urdu are moved to the right of
the matrix clause due to the Case Resistance Principle of Stowell (1981), and
it is a “position from which wh expressions cannot take matrix scope,” as “the
finite complement [is] a syntactic adjunct” (Dayal 1996: 30).
In Marathi (IA) too, “postposed right branching ki [‘that’] clauses never
allow wide scope” (Wali 2006: 209).
narrow Scope with an ic-clause

Marathi (IA)
(95) mini- la mahit ahe [ki [lili- ni ravi lā kāy dila]]
Mini- dat know pres that Lili- erg Ravi to what gave
‘Mini knows what Lili gave to Ravi.’
‘*What does Mini know that Lili gave to Ravi.’
(Wali 2006: 209)
In Bangla (IA), when the embedded clause occurs to the right of the matrix
verb as in (96), it has only Narrow Scope interpretation (see 6.6.10 for further
discussion).
narrow scope
Bangla (IA)
(96) ora šune.che [ke aš- be]
they heard who come- fut
‘They have heard who will come’ (Narrow Scope of ke)
(Bayer 2001: 24)
To summarize the above discussion, a finite clause with an IC to the right of

the matrix verb cannot have Wide Scope interpretation.
Bhat and Dayal (2007) also reiterate this fact for Hindi (see 6.6.9 too for
further discussion).
6.6.3 Nature of the matrix predicate and COMP

Whether Narrow Scope or Wide Scope is permitted, amongst other things,
depends upon the factive/non-factive nature of the matrix predicate, and the
type of complementizer employed.
An attitude verb think permits only Wide Scope reading with an FC in Bangla
and Marathi (IA) (Bayer 2001: 25). Thus, the non-factive verbs bhab ‘to think’
in Bangla in (97) and vət ‘to believe’ in Marathi in (98) permit Wide Scope
reading.
w i d e s c o p e w i t h t h e n o n - f a c t i v e ‘think’ w i t h a n f c
Bangla (IA)
(97) tumi [[ke bar.i kor- be] (bole)] bhabcho
you who house do- fut FC think
‘Who do you think will build a house?’ (Wide Scope)
(Bayer 2001: 25)
w i d e s c o p e w i t h t h e n o n - f a c t i v e p r e d i c a t e ‘believe’ w i t h a n f c
Marathi (IA)
(98) mini-la [[lili-ni ravi-la kəy dila] asa] və.ta
Mini-dat Lili-erg Ravi-to what gave comp believes
‘What does Mini believe that Lili gave to Ravi?’ (Wide Scope)
In contrast, the factive verb māhit ‘to know’ in Marathi in (99) does not permit
Wide Scope reading when the FC te occurs.
n a r r o w sc o p e w i t h t h e f a c t i v e p r e d i c a t e ‘know’ w i t h t h e f c te
(99) mini-la [[lili-ni ravi-la kəy dila] te] māhit āhe
Mini-dat Lili-erg Ravi-to what gave that knows
‘*What does Mini know that Lili gave to Ravi?’ (Wide Scope not permitted)
‘Mini knows what Lili gave to Ravi.’ (Narrow Scope)
(Wali 2006: 210; meaning added)
The non-factive verb to think does not permit Narrow Scope reading in Bangla.
Hence, (100) is ungrammatical. According to Bayer (2001: 25), “the nar-
row scope interpretation of ke [‘who’] yields a semantic violation” as “the
matrix verb is incompatible with the [+wh] feature of the embedded clause
[in Bangla].” Bayer argues that ke ‘who’ cannot rise outside the scope of the
matrix clause’s attitude verb.
(100) *tumi bhabcho [ke bar.i kor- be]
you think who house build- will
‘*You think who will build the house?’
(Bayer 2001: 25)
228 Complementation
Factive verbs with an IC ki ‘that’ in Kashmiri (IA) also resist Wide Scope as in
(101).
Kashmiri (IA)
(101) mirāi chi khabar [ki [k’eh por mohanan]]
Mira.dat is news IC what read Mohan.erg
‘Mira knows what Mohan read,’ NOT
‘What does Mira know that Mohan read?’
(Wali 2006: 215)
In contrast, the quotative FC mhan.un in Marathi (IA) in (102) (Wali 2006: 212)
and ani in Telugu (DR) in (103) permit Wide Scope interpretation, even when
the predicate is factive.
w i d e s c o p e w i t h a f a c t i v e p r e d i c a t e w i t h mhan.un ‘having said’
(the quotative), an fc
Marathi (IA)
(102) lili ni kon.ā-lā pustak dila mhan.un tulā du’kkha vat.tā
Lili erg who-dat book gave quot (so) you.dat regret
‘Who do you regret that Lili gave a book to?’
(Wali 2006: 212)
w i d e s c o p e w i t h a f a c t i v e p r e d i c a t e w i t h ani ‘that,’ a n f c
Telugu (DR)
(103) lilli evari- ki pustakam iccindi- ani nuvvu bādhā pad.utunnāvu
Lili who- dat book gave- quot you pain fall
‘Who do you regret that Lili gave a book to?’
To summarize: (i) a non-factive verb permits Wide Scope reading with an IC as

in (98); (ii) the FC te does not permit Wide Scope reading with a factive verb
in Marathi as in (99); and (iii) a factive predicate, in contrast, permits Wide
Scope reading when the FC is a quotative in Marathi (IA) and Telugu (DR), as
in (102) and (103), demonstrating that it is the nature of the COMP that takes
precedence over factivity in Wide Scope interpretation.
6.6.4 The role of intonation, pause and morphophonemic criteria

The role of intonation, pause
Rise in intonation and the occurrence of pause may result in Scope changes.
In Telugu (DR), a verb such as to think permits both Wide and Narrow Scope
readings. Sentence (104) is an example of a question word having Wide and
Narrow Scope reading.
Recall that Telugu only has an FC. In (104), the embedded complement
occurs in its canonical position.
w i d e a n d na r r o w sc o p e r e a d i n g
(104) mı̄ru [CP [S2 pārt.ı̄ ki evaru vastāru-S2 ] aniCP ] anukont.unnāru
you party dat who will come- FC are thinking
‘Who do you think will come to the party?’ (Wide Scope)
‘You are thinking who’d be coming to the party’ (Narrow Scope)
In Narrow Scope reading, (i) there is a rise in intonation on the embedded verb
vastāru ‘will come,’ and there is a fall in intonation on the matrix verb; and (ii)
there are two distinct pauses – one after the embedded verb, and the other after
the FC-complement.
In contrast, in Wide Scope interpretation, (i) there is no pause after the embed-
ded verb; and (ii) there is a rise in intonation on the matrix verb. Hence, pitch
considerations also play an important role in interpretation (Peri Bhaskararao,
p.c.).
The role of morphophonemic criteria

The FC ani in Telugu (DR) can be dropped when the stem of the matrix
predicate starts with an-, as in anu- ‘to say,’ anukonu ‘to think,’ anipincu ‘to
feel,’ etc.
When the FC is dropped, the sentence in (104) has only Wide Scope inter-
pretation (105).
wide scope reading

(105) mı̄ru [CP [S2 pārt.ı̄ ki evaru vastār(u)-S2 ] øCP ] anukont.unnāru
‘Who do you think will come to the party?’ (Wide Scope), NOT
‘You are thinking who’d be coming to the party’ (Narrow Scope)
This shows that morphophonemic considerations too play a role in Scope

interpretations.
6.6.5 The position of the FC-clause

According to Bayer (2001: 25), if the quotative FC-clause occurs to the right
of the matrix verb (as in (106) in Telugu [DR]), for example, the Wide Scope
reading is lost (just as in IA languages with an IC), while it is retained if it
occurs to the left of the matrix subject as in (107). A similar observation was
made for Tamil (Savio 1991), and by Mohanan (1984) and Jayaseelan (1998)
for Malayalam, according to Bayer (2001: 34).
230 Complementation
fc-clause to the right of the matrix verb

Telugu (DR)
(106) mı̄ru anukont.unnāru [[pārt.ı̄ ki evaru vastār(u)]- ani]
you are thinking party dat who will come- FC
‘You are thinking who’d come to the party’ (Narrow Scope), NOT
fc-clause to the left of the matrix subject

(107) [[pārt.ı̄ ki evaru vastār(u)]- ani] mı̄ru anukont.unnāru
party dat who will come- FC you are thinking
‘Who do you think will come to the party?’
Thus, we observe that the position of occurrence of the CP-clause affects the
scope of the question word.
6.6.6 The position of question expressions and Scope possibilities

The position of a question expression – in either the matrix or embedded clause –
effects scopal possibilities. A question expression occurring in the embedded
clause has Narrow Scope interpretation, while its occurrence in the matrix
clause has Wide Scope interpretation,
In Hindi-Urdu (IA), for example, when the interrogative question marker
kaun ‘who’ occurs in the embedded clause with the matrix verb sun.nā ‘to
hear,’ kaun ‘who’ cannot have Wide Scope reading – as in (108). It only has a
Narrow Scope interpretation.
n a r r o w sc o p e w i t h sun.nā ‘to hear’ a s m a t r i x v e r b

Hindi-Urdu (IA)
(108) tum ne sunā thā [ki kal kaun ā rahā hai]
you erg heard had IC tomorrow who come progr pres
‘You had heard who’d be coming tomorrow.’ (Narrow Scope)
Wide scope with the interrogative question marker

We shall discuss one case each from Hindi-Urdu and Telugu.
(i) The case of Hindi-Urdu (IA)

We will look at the question word kyā and kaun in Hindi-Urdu as scope markers.
When the question word kaun ‘who’ is scrambled to the matrix clause, the
scope possibilities change from Narrow to Wide Scope (see Dayal 1996 for
arguments in support of scrambling).
(109) kauni tum soc- te ho [ki ti āyegā]

who you think- imperf pres IC will come
‘Who do you think will come?’ (Wide Scope)
(Dayal 1996: 35)
According to Mahajan (1987) and Gurtu (1992), the scrambling of the question
word kyā ‘what’ from the embedded clause to the matrix clause also changes
scope possibilities.
kyā ‘what’ o c c u r r i n g i n t h e e m b e d d e d c l a u s e – n a r r o w s c o p e
(110) tum jānte ho [ki us ne kyā kiyā]
you know pres IC he erg what do.pst
‘You know what he did,’ NOT
‘What do you know he did?’
kyā ‘what’ s c r a m b l e d t o t h e m a t r i x c l a u s e – w i d e s c o p e
(111) kyāi tum jānte ho ki us ne ti kiyā
what you know pres that he erg do.pst
‘What do you know that he did?’
(Dayal 1996: 22)
Note that in (109) and (111) the question word is moved one clause up to the
matrix clause. Alice Davison (p.c.) points out that the scope marker kyā in
Hindi-Urdu must be in the next clause up, but the topic in topicalization (see
6.6.7) can be several clauses away from its coindexed pro.
According to her, “topicalization is not subject to the finiteness constraint,”
while scope marking is. The wh-expression kaun in (112a) is in the scope of
the question marker kyā, which occurs in S2 as well as S3 . If it does not occur
in S2 (indicated by ø in S2 ), the sentence is ungrammatical. This shows that the
scope marker must be in the next clause up.
(112a) zayd ne kyā kahā ki mı̄nā kyā / *ø soctı̄ hai ki kaun tum
Zayd erg what said that Mina what think pres that who you
se nārāz hai
with angry is
‘[S1 Who did Zayd say [S2 that Mina thinks [S3 e is angry with you. S3 ] S2 ] S1 ]’
Here, e is an empty category which is a trace. In (112b), the question expression

KISE is in Focus position.
(112b) zayd ne KISE kahā ki mı̄nā soctı̄ hai ki kaun tum se
Zayd erg whom said that Mina think pres that who you with
nārāz hai
angry is
[S1 Who did Zayd say [S2 that Mina thinks [S3 e is angry with you. S3 ] S2 ] S1 ]’
(Davison 1988: 201)
232 Complementation
(ii) The case of Telugu (DR)

In Telugu (DR), the question word ēmi ‘what’ in the embedded clause S2 gives
Narrow Scope reading with a factive verb with -ō as the COMP (113a), or with
a cleft, as in (113b).
ēmi ‘what’ o c c u r r i n g i n t h e e m b e d d e d c l a u s e – na r r o w s c o p e
Telugu (DR)
(113a) prasād ēmi cēsēd.(u)- ō nı̄ku telusu
Prasad what did- comp you.dat known
‘You know what Prasad did.’ (Narrow Scope reading)
When clefted, the final vowel -i of the embedded verb is elided after it vowel-
harmonizes the complementizer -ō to -ı̄. We have labeled -ı̄ as ‘comp.’17
(113b) prasād ēmi cēsin.d(i)- ı̄ nı̄ku telusu

Prasad what did.3, nm pron suffix- comp you.dat known
‘You know what Prasad did.’ (Narrow Scope reading)
To have Wide Scope reading, there are two options:

Option I:
A question word ēmi ‘what’ / elāgu ‘how’ is employed in the matrix clause, as
in (114).
wide scope
(114) prasād ēmi/elāgu cēsēd.(u)- ō nı̄ku ēmi telusu
Prasad what/how did- comp you.dat what known
‘What/how do you know what Prasad did?’ (Wide Scope reading)
Option II:
The question word ēmi ‘what’ occurs in the matrix clause, and the embedded
clause is clefted with or without the question word, or with the question word to
the right of the cleft, as in (115), or with the question word ēmi ‘what’ followed
by the dubitative marker -ō, as in (116). There is no overt complementizer
in (116).
e m b e d d e d c l a u s e clefted , w i t h t h e q u e s t i o n w o r d i n t h e
matrix clause – wide scope
(115) prasād cēs- in- (a-) di- i nı̄ku ēmi telusu
Prasad do- pst- adjr- 3 nm,pron suffix- comp you.dat what known
‘What do you know what Prasad did?’
a cleft with a question word + dubitative marker − wide scope

(116) prasād cēs- in- (a-) di ēmit.(i).ō nı̄ku ēmi
Prasad do- pst adjr- 3nm,pron suffix what.dub mkr you.dat what
telusu
known
‘What do you know what Prasad did?’
Thus, it is the occurrence of a wh-expression kyā in Hindi-Urdu (IA) and ēmi

‘what’ in Telugu (DR) in the matrix clause that affects scope interpretation.
The next subtopic we consider is Topicalization and Wide Scope.
6.6.7 Topicalization and Wide Scope

According to Davison (1988: 182), in a sentence like (117) the questioned
phrase that occurs in the matrix clause imparts Wide Scope reading and it is
“coindexed with a gap or with a pronoun” (in italics in (117)).
Hindi-Urdu (IA)
(117) [tum- ne kisei kahā [ki ei / wo kal āyegā?]]
you- erg who.dat say.perf that 3p,s tomorrow come.fut
‘Whoi did you say [ei will come tomorrow?]’
Davison (1988: 183)
Thus, her approach treats “the subordinate-clause element as a pronoun, either

null (ei ) or lexical [wo ‘he’ in (117)]. It is coindexed with a matrix-clause
argument of the main verb, a type of object subcategorized as an ‘internal’ topic
by the matrix verb, which assigns it the case of a direct object” (Davison 1988:
183). Hence, kise ‘who’ in (117) in the matrix clause gets the dative/accusative
marker. The questioned NP in (117) is a case of the in situ wh strategy indicating
that it originates there, and it is not a moved wh-element. Davison further points
out: “The wh-element that is left dislocated gets wide scope and it is not subject
to locality conditions, and there is the possibility of a resumptive pronoun,
suggesting that we have an (unbounded) pronominal chain, not a wh-moved
variable” (Alice Davison, p.c.).
Unlike in Hindi-Urdu (IA) in (117), Telugu does not permit any question
expression other than ēmi ‘what’ in the matrix clause as scope marker.
(118) [S1 nı̄ku [S2 nası̄m evar(i)- ni cūsind(i)- S2 ] ō ēmi/

you.dat Nasim who- acc saw- dub mkr what
*evar(i).ni telusuS1 ]
who.acc known
‘What do you know whom Nasim saw?’
234 Complementation
Thus, the direct object evari-ni ‘who-acc’ has matrix scope in (118) due to the
presence of the scope marker ēmi ‘what’ in the preverbal position in the matrix
clause. The occurrence of evari-ni ‘who-acc’ as scope marker in the matrix
clause is not permitted in (118).
To recapitulate, we have shown that there are differences between Scope
Marking and Topicalization: (i) Scope Marking can be done only one sen-
tence higher up, whereas there is no such restriction in Topicalization; and
(ii) in Topicalization, the topicalized NP is an argument subcategorized by the
matrix verb, while the in situ wh-element originates in the matrix clause and
subsequently moves to the COMP position (Davison 1988).
6.6.8 The occurrence of clitics/particles as scope inhibitors

With matrix predicates anukonu ‘to think,’ vinu ‘to hear,’ in Telugu (DR),
the question expression occurring in an FC-clause has Wide Scope reading.
However, with a verb such as think both Wide and Narrow Scope readings are
permitted. Sentence (104) (repeated as (119)) is an example of a question word
imparting a possible Wide Scope reading. Note that in (119) neither the yes/no
question marker nor the clitic -ō is present.
wide scope reading possible

Telugu (DR)
(119) mı̄ru [[pārt.ı̄ ki evaru vastār(u)-] ani] anukont.unnāru
‘You are thinking/wondering who will come to the party.’ (Narrow Scope)
The Wide Scope reading is blocked, and the sentence only has Narrow Scope
reading, when the yes/no question marker clitic -ā or the dubitative marker -ō
occurs to the right of the embedded clause, as in (120).
narrow scope reading

(120) mı̄ru [[pārt.ı̄ ki evaru vastār(u)- ā/ ō] ani] anukont.unnāru
you party dat who will come- qm dub mkr FC are thinking
‘You are thinking/wondering who will come to the party’ (Narrow Scope),
NOT
Our claim is: it is due to the presence of these clitics that Wide Scope reading
is blocked in (120). Hence, we may label them as Scope Inhibitors.
In Sema (TB) too, Narrow Scope results from: (i) the presence in the embed-
ded sentence of the dubitative marker keno as in (121), contrasted with (122);
and (ii) the presence in the embedded sentence of the yes/no question marker
as in (123), contrasted with (124) (Achumi 2000).
n a r r o w sc o p e w i t h t h e d u b i t a t i v e m a r k e r keno
Sema (TB)
(121) no- ye [khū hile iγe- ni keno pi] kms - a
you nom who here come- fut dub mkr FC think- progr
‘You were wondering who would come here.’ (Narrow Scope)
w i d e s c o p e w i t h o u t t h e d u b i t a t i v e m a r k e r keno
(122) no- ye [khū hile iγe- ni pi] kms- a
you- nom who here come- fut FC think- progr
‘Who do you think will come here?’ (Wide Scope)
n a r r o w s c o p e w i t h t h e yes/no q u e s t i o n m a r k e r kya
(123) panoŋu- ye [[khū iγe- ni kya] pi] celu
they- nm who come- fut pol q mkr FC heard
‘They heard who will come.’ (Narrow Scope)
w i d e s c o p e w i t h o u t t h e yes/no q u e s t i o n m a r k e r kyā
(124) panoŋu- ye [[khū iγe- ni] pi] celu
they- nm who come- fut FC heard
‘Who did they hear will come?’ (Wide Scope)
(Achumi 2000)
Sema lends further support to: (i) treating the yes/no question marker and the
dubitative marker (as in Telugu [DR]) as Scope Limiting Particles; and (ii) the
proposal that certain clitics act as Scope Inhibitors.
6.6.9 The use of pleonastic question expressions as scope markers

Mohanan (1984), Gurtu (1985), Mahajan (1987, 1990), Davison (1988), Dayal
(1996) and Wali (2006) discussed Narrow and Wide Scope readings of question
words in SALs.
When a question word is used purely as a scope changing device, and it does
not have any semantic content, it is labeled as a pleonastic expression. In SALs
such usage is observed.
We consider the case of Kashmiri (IA) first. According to Wali (2006: 216),
in Kashmiri “propositional ‘believe’ class verbs are subcategorized for not tak-
ing indirect questions. In these, wide scope is achieved by the employment of
a question particle [k’ah ‘what’] while leaving the original wh phrase [kamis
‘whom’ in (125)] in place.” That is, it is the occurrence of question word k’ah
‘what’ in the matrix clause that gives Wide Scope reading. According to Wali
(2006: 217), “it is this k’ah which carries the feature complex of the embed-
ded Wh. The embedded Wh has no semantic content and no interpretation
at LF.”
236 Complementation
Kashmiri (IA)
(125) mohan- as k’ah chu bāsān [CP (ki) [S2 kamis nish chu mirā
Mohan- dat what aux believes IC whom near pres Mira
bihitS2 ]CP ]
sit
‘Who does Mohan believe Mira is sitting near?’
(Adapted from Wali 2006: 217 in consultation with O. N. Koul, p.c.)
When the question word ēmi ‘what’ in Telugu (DR) or kyā ‘what’ in Hindi-Urdu
(IA) occurs in the embedded clause, the sentence has Narrow Scope reading,
as in (126) in Telugu and (127) in Hindi-Urdu (IA).
narrow scope with a question word

Telugu (DR)
(126) nı̄ku nası̄m [S2 ēmi and(i)- ōS2 ] telusu
you.dat Nasim what said.3s,nm- dub mkr known
‘You know what Naseem said.’
Hindi-Urdu (IA)
(127) tum jānte ho [S2 ki usne kyā kiyāS2 ]
you know pres that he.erg what did
‘You know what he did.’
Srivastav (1989: 2)
The question words ēmi ‘what’ in Telugu (DR) and kyā ‘what’ in Hindi-Urdu
(IA) may also be used as Scope Changing devices. That is, their addition will
lead to a change from Narrow to Wide Scope. Srivastav (1989) labels such
markers as pleonastic operators. The occurrence of ēmi ‘what’ in Telugu (DR)
or kyā ‘what’ in Hindi-Urdu (IA) in the matrix clause, together with the question
expression in the embedded clause, imparts Wide Scope reading.
Telugu has the question word ēmi that is homophonous with the scope marker
ēmi ‘what,’ which is used to indicate Wide Scope. We label the question word
as ēmi1 and the scope marker as ēmi2 for clarity of exposition. Note that, just as
with k’ah in Kashmiri (IA), ēmi2 in (128) in Telugu has no semantic content.
w i d e s c o p e w i t h a n a d d i t i o n a l question word ( w h - e x p r e s s i o n )
a s an expletive
Telugu (DR)
(128) nı̄ku [S2 nası̄m ēmi1 and(i)- ōS2 ] ēmi2 telusu
you.dat Nasim what said- dub mkr FC what (expletive) known
‘What do you know what Naseem said?’ (Wide Scope)
Hindi-Urdu (IA)
(129) tum kyā jānte ho [S2 ki usne kyā kiyāS2 ]
you what (expletive) know pres that he.erg what did
‘What do you know what he did?’ (Wide Scope)
Srivastav (1989: 2)
The scope markers ēmi2 ‘what’ in Telugu and kyā ‘what’ in Hindi-Urdu are
different from the interrogative markers. Evidence for this comes from the fact
that the interrogative word ēmi ‘what’ in (126) can be reduplicated – just as any
question word can be in Telugu and in other SALs – as in (130), but the scope
marker ēmi in (128) cannot be reduplicated as in (131).
question word reduplicated: permitted

(130) nı̄ku nası̄m [S2 ēmi1 ēmi1 and(i)- ōS2 ] telusu
you.dat Nasim what what said-3s,nm- dub mkr known
‘You know what (all) Naseem said.’ (Wide Scope)
s c o p e m a r k e r r e d u p l i c a t e d : no t p e r m i t t e d
(131) *nı̄ku [S2 nası̄m ēmi1 and(i)- ōS2 ] ēmi2 ēmi2 telusu
you.dat Nasim what said- dub mkr what what known
Similar evidence can be adduced from Hindi-Urdu too with the reduplicated
question word kyā ‘what.’
In Sema (TB) too, a question expression functions as a scope marker (see
Achumi 2000, and Achumi and Subbarao 2005).
6.6.10 The position of occurrence of the complementizer: initial or final

Whether the COMP occurs in the initial position or final position plays a major
role in scope possibilities. Let us first take into consideration the data from
Bangla (IA).
Sentence (132) does not have any overt complementizer. It has both WS
(Wide Scope) and NS (Narrow Scope) interpretations.
Bangla (IA)
(132) ora [ke aš- be] šune.che WS,NS
they who come- fut heard
(i) ‘They have heard who will come.’ (Narrow Scope of ke)
(ii) ‘Who have they heard will come?’ (Wide Scope of ke)
The question that needs to be answered is: how do we explain the ambiguity in
(132) which has Narrow Scope as well as Wide Scope interpretations?
Before we proceed further, it would be helpful if the facts in Bangla (IA) can
be summarized as follows.18
238 Complementation
Note that (132) is grammatical though the embedded complement is not

case-marked, and it violates the Case Resistance Principle of Stowell. Since
the FC bole can occur in this position optionally as (133) shows, the Wide
Scope interpretation in (132) can be accounted for without any problem.
(133) ora [ke aš- be] bole šune.che WS

they who come- fut FC heard
‘Who have they heard will come?’ (Wide Scope of ke)19
The question that arises now is: how to account for the Narrow Scope inter-
pretation? Probal Dasgupta (p.c.) suggests that we hypothesize that in (134)
we have the IC as a zero complementizer (ZERO COMP) and a ZERO NP
correlate too (see Dasgupta 2007 for details).20
ICs in many Indo-Aryan languages do permit an NP correlate, such as yah
‘it’ or yah bāt ‘this news’ in Hindi-Urdu (IA) (Subbarao 1984a), or he ‘this’ or
hi gos..ta ‘this story’ with factive predicates in Marathi (Wali 2006: 211).
Taking this as a cue, we can say that the zero form in (134) corresponds to
(135), the non-zero form.
(134) ora [ke- ZERO aš-be] ZERO NP šune-che

they who- IC come-fut heard
(ii) ‘*Who have they heard will come?’ (Wide Scope of ke)
(135) ora [ke- je aš-be] šet.a šune-che

they who- IC come-fut DEM (that) heard
(ii) ‘*Who have they heard will come?’ (Wide Scope of ke)
It is the occurrence of šet.a ‘DEM’ that blocks Wide Scope reading in (135).
Thus, Dasgupta’s suggestion of having zero forms enables us to account for
the ambiguity in (132) in a neat fashion. We observe that the final COMP bole
permits Wide Scope reading in (133), while initial COMP je permits Narrow
Scope reading as in (135).
Bayer (2001: 29) provides another explanation in terms of harmony of direc-
tionality. He provides an explanation as to why an in situ IC-clause in (134)
does not permit Wide Scope reading. For the Wide Scope reading of the ques-
tion word, the question expression must move overtly or covertly to the matrix
clause.
According to Bayer (2001: 29), “covert wh-rising from the IC-clause to the
matrix clause is blocked by the fact that the IC-clause, although theta-marked,
falls outside the pattern of argument licensing in head-final languages.” Note
that verb-final languages have an FC. Hence, the position of occurrence of the IC
je- in Bangla, Assamese or Oriya, or ki- in Hindi-Urdu or Punjabi and Kashmiri,
6.7 Against Rightward Extraposition 239
is not ‘consistent’ from the point of view of word order universals. They are
‘exceptionally’ selected as they are not in “harmony with the directionality
pattern that is dictated by the parameter of the language” (Bayer 2001: 29). In
contrast, FC-clauses permit Wide Scope reading as “the FC-clause is selected
according to the parameter of head-finality” (Bayer 2001: 29). Hence, if there
is no other variable that blocks the movement of the wh-expression from the
FC-clause, an FC-clause does permit Wide Scope reading.21
In this section we discussed the interplay of various factors that affect Narrow
and Wide Scope. The presence or absence of the complementizer, for example,
might lead to Wide or Narrow Scope interpretations, as in Bangla. An IC-clause
that occurs to the right of the matrix clause does not permit Wide Scope reading.
Some other factors include: the nature of the matrix verb; the nature of the
complementizer; the position of occurrence of the scope marker; the presence
of clitics; clitics functioning as scope inhibitors; and question expressions as
scope markers. We distinguished between scope marking and topicalization as
scope marking strategies. Finally, we conclude by saying that it is the interplay
of various strategies that yields scope readings.
Finally, as Alice Davison (p.c.) puts it, a typological generalization that
emerges is: “in all the various scope extending methods there are strict con-
straints on wh scope in all SALs, according to finiteness or position. Languages
find different syntactic, lexical and phonological devices for overcoming these
constraints.”
6.7 Arguments against Rightward Extraposition

Finite clauses in SALs occur to the right of the matrix verb. In this section
the focus of interest from a typological point of view is: how do finite clauses
end up in the right adjunction position in SALs irrespective of the position
of the COMP – that is, whether it is an IC or FC? Is such occurrence due to
movement, or are the clauses base-generated? The issue is: where does the
finite embedded clause end up in relation to the matrix clause position? Do
SALs have movement rules?
In Hindi-Urdu, Punjabi and Kashmiri (IA), the finite complement clause
occurs to the right of the matrix verb. In Subbarao (1974, 1984a), Mahajan
(1990) and Srivastav (1991a), to account for such occurrence in Hindi, it was
proposed that there is a rightward extraposition transformation that moves the
embedded clause from its in situ DO position and adjoins it to the right of
the verb of the matrix clause. Davison (1992a) argues for a base-generated
right adjunction approach for the finite embedded clause. Mahajan (1997a)
following Kayne’s (1994) approach, argues that Hindi is an SVO language,
and rightward movement rules are not permitted on theoretical and empirical
240 Complementation
grounds. According to Mahajan (1997a), rightward extraposition of a com-

plement clause, as suggested in Subbarao (1974, 1984a) and Mahajan (1990),
is not tenable, in view of binding-theoretic arguments. Thus, the extraposed
clause under his assumptions must structurally be in a lower position, in view
of variable binding and Condition C effects tests. We shall show that Mahajan’s
analysis too encounters a problem (see appendix 6.3 on the Website for further
details).
An issue that concerns complement clauses in SALs is the direction of
c-command and the position of occurrence of the Negative Polarity Item (NPI).
See appendix 6.4 on the Website for details.
In appendix 6.5 on the Website, we present a detailed discussion of the
syntactic reanalysis of the complementizer in language contact situations.
6.8 Conclusion
In this chapter we discussed the differences between the ICs and FCs; direct
and indirect speech and tense and person harmony; the Case Resistance Princi-
ple and the non-occurrence of a CP clause intra-sententially; scope of question
expressions in complements; arguments against rightward extraposition; direc-
tion of c-command and the NPI; the Tensed S-Condition and the PIC in relation
to the accusative case marking of the embedded subject; COMP agreement; and
the reanalysis of the IC and FC in language contact situations. We discussed
wh-scope in finite and non-finite COMP-clauses, and the factors that permit
and restrict wh-scope. We have shown that there is a variety of syntactic, mor-
phophonemic and phonological criteria that affect Narrow and Wide Scope. We
demonstrated that it is the interplay of a variety of factors that yields Narrow
and Wide Scope readings.
We have shown that the IC ki, a functional category, of the source language
Hindi-Urdu (IA) is reanalyzed as an FC in Dakkhini. It occurs as an FC in
embedded questions, as a clausal and phrasal disjunctive marker, as a linker in
relative clauses, etc.
Appendix
The Final-over-Final Constraint (FOFC) and diachronic change
In this appendix we present a universal constraint labeled as the Final-over-
Final Constraint (FOFC), discussed recently in Biberauer et al. (2009),22 and
discuss how the diachronic changes that took place in the position of occurrence
of the complementizer in South Asian language contact situations strictly obey
this universal constraint.
Appendix 241
Position of occurrence of the complementizer in SALs: from the discus-

sion in this chapter, it is clear that the complementizer occurs in two posi-
tions in SALs. In all Dravidian and Tibeto-Burman languages and in a single
Indo-Aryan language, Sinhala, it occurs only to the right of the embedded
clause (right-peripheral), and in the other Indo-Aryan languages it may occur
either to the right of the embedded clause (right-peripheral) or to the left
(left-peripheral). Thus, the position of occurrence in Indo-Aryan is language-
specific.
In this section we shall mainly focus our attention on COMP position in
relation to the position of occurrence of the Polarity Question (yes/no question)
marker in Indo-Aryan.
Recall that the complementizer is only left-peripheral in Kashmiri, Punjabi,
Hindi-Urdu and Sindhi (Indo-Aryan) and it is only right-peripheral in Sinhala
(Indo-Aryan). Some Indo-Aryan languages, such as Bangla, Assamese, Oriya,
Marathi, Gujarati and Konkani, have a left- as well as a right-peripheral COMP.
Though the right-peripheral COMP is a quotative derived from the verb to say
in all these languages, there are a few languages, such as Marathi, Gujarati
and Konkani, which in addition have a right-peripheral COMP derived from a
demonstrative pronoun.
Davison (2007a) convincingly demonstrates that there is a neat correlation
between the occurrence of the polarity question marker and the complementizer
in complement clauses in the Indo-Aryan languages.
Set 1: Languages such as Hindi, Punjabi, Kashmiri23 and Sindhi have
the polarity question marker occurring in the initial position of the
complement clause in the unmarked word order. These languages
do not have the Final Complementizer. Recall that the Initial Com-
plementizer is not derived from the verb to say.
Set 2: Languages such as Marathi, Nepali, Dakkhini, Hindi-Urdu,
Assamese, Bangla and Oriya which have FC derived from the verb
to say have a non-initial polarity question marker. It occurs either
in final position as in Marathi (1), or in final or medial position as
in Bangla, as in (3) and (4). The following examples from Davison
(2007a) are illustrative:
polarity question word occurring in sentence-final position –

permitted
Marathi (IA)
(1) to kal parat ālā kā(y)?
he yesterday back come.pst.3m,s pol
‘Did he come back yesterday?’
(Pandharipande 1997: 5, quoted from Davison 2007a: 182)
242 Complementation
Table 6.1
Final Polarity Question Marker – Final Complementizer

Initial Complementizer – Initial Polarity Question Marker
polarity question word occurring in sentence-initial

p o s i t i o n – not p e r m i t t e d
(2) *kā(y) to kal parat ālā
pol he yesterday back come.pst.3m,s
‘Did he come back yesterday?’
polarity question word occurring in sentence-final

position – permitted
Bangla (IA)
(3) modhu aš- be ki (na)
Madhu come- fut pol not
‘Will/Won’t Madhu come?’
polarity question word occurring in sentence-initial

p o s i t i o n - not p e r m i t t e d
(4) *ki modhu aš- be
pol Madhu come- fut
(Davison 2007a: 192, attributed to P. Dasgupta, p.c.)
The next issue that is of interest is the order in which the complementizer and
the polarity question word occur in complement constructions.
Davison (2007a) also makes an interesting observation in this regard.
In languages in which the complementizer occurs in final position, the polar-
ity question marker precedes it, and in languages in which the complementizer
occurs initially, the polarity question marker follows the initial complementizer.
Thus, the order of the polarity question marker and the complementizer is
shown in Table 6.1.
The following examples from Marathi (IA) are illustrative:
Marathi (IA)
(5) [[to kal parat ālā kā(y) mhan.un/ asa] rām malā
he yesterday back come.pst.3m,s pol quot/ such Ram I.dat
witšārat hotā]
ask.progr be.pst.3m,s
‘Ram was asking me whether/if he came back yesterday.’
(Davison 2007a: 184)
Appendix 243
In Marathi, the order in which an Initial Comp is followed by a Final Polar-

ity Question Marker is permitted, according to Pandharipande, as quoted by
Davison (2007a: 184).
(6) rām malā witšārat hotā ki [to kal parat

Ram I.dat ask.progr be.pst.3m,s he yesterday back
ālā kā(y)]24
come.pst.3m,s pol
‘Ram was asking me whether/if he came back yesterday.’
In Kashmiri too, just as in Marathi, an Initial Comp IC followed by a Final

Polarity Question Marker is permitted (7).
(7) rām ōs me prItshān (k’ā) sı̄tā āyā vāpas ra:th

Ram was I-dat ask-progr what Sita came-QM return yesterday
‘Ram was asking me whether/if Sita came back yesterday.’
(O. N. Koul, p.c.)
In Hindi-Urdu (IA), for example, the Polarity Question marker follows the IC
initially in the unmarked word order, as in (8). It may occur in the final position
of the embedded sentence too.
Hindi-Urdu (IA)
(8) usne pūchā [ki [kyā tum ā- ōgē]]
3s- erg ask.perf that pol you come- fut
‘He asked whether you would come.’
If the Polarity Question marker precedes the IC, the sentence is ungrammatical.
(9) *usne pūchā [kyā ki tum ā- ōgē]

3s- erg ask.perf pol that you come- fut
‘He asked whether you would come.’
(Davison 2007a: 183)
Thus, there are four types (Types A–D) that are possible with regard to the
order of occurrence of the complementizer and the Polarity Question marker.
Davison (2007: 183), based on data from Indo-Aryan languages, demonstrates
that there is no language in which the initial Polarity Question marker and
final complementizer co-occur – that is, Type D is not attested. The other three
combinations are permissible, as illustrated in Table 6.2.
We shall now present the universal constraint labeled as the Final-over-
Final Constraint (FOFC) discussed in Biberauer et al. (2009), and discuss how
the diachronic changes that took place in the position of occurrence of the
complementizer in South Asian language contact situations strictly obey this
universal constraint.25
244 Complementation
Table 6.2
Type Position of Pol Position of Comp Languages
A Initial Initial only Hindi-Urdu, Punjabi, Kashmiri, Sindhi,

Maithili, Kurmali
B Final/medial Initial and final Marathi, Gujarati, Assamese, Bangla,
Dakkhini Hindi, Oriya, Nepali (plus some
North Dravidian languages, e.g., Brahui)
C Final/medial Final only Sinhala (plus most Dravidian languages)
D Initial Final Unattested in the area
Source: Adapted from Davison (2007a: appendix B) in Biberauer et al. (2009: 14).
Holmberg (2000: 124) proposes a constraint labeled as FOFC which can

account for the permissible and non-permissible configurations in language. If
there is a phrase α, and α is a head-initial phrase, the phrase β that immediately
dominates α must also be head-initial. If α is a head-final phrase, the phrase ß
that immediately dominates α may be head-initial or head-final.
This constraint is formally stated as in (10).
t h e f i n a l - o v e r - fi n a l c o n s t r a i n t (f o f c )
(10) If α is a head-initial phrase and ß is a categorically non-distinct26 phrase immediately

dominating α, then ß must be head-initial [emphasis provided]. If α is a head-final phrase
and ß is a phrase immediately dominating α, then ß can be head-initial or head-final
[emphasis provided].
Biberauer et al. (2009: 14) demonstrate that the FOFC can explicate diachronic changes
too. They observe: “There is good evidence that FOFC constrains the borrowing of final
complementisers in Indo-Aryan. While Indic languages with final Pol heads [Polarity
heads] were free to borrow/develop final complementisers, the languages with initial
Pol [Polarity] were not as the borrowing of a final C would have led to a FOFC violation.
Further evidence that this is what is at stake in Indic comes from typological trends,
which appear to behave in the same way.”
The type that FOFC predicts to be ungrammatical is in fact the type that is
precisely unattested. Table 6.2, adapted from appendix B of Davison (2007a)
in Biberauer et al. (2009: 14), is illustrative.
Note that types A, B and C are attested in Indo-Aryan while Type D is not,
as predicted by FOFC.
Dravidian and Tibeto-Burman languages too adhere to the FOFC discussed
above. The following examples are illustrative.
In Telugu (DR) and Thadou (TB) the Complementizer, a quotative, follows
the Polarity Question expression.
Appendix 245
Telugu (DR)
(11) pel.l.i- ki lalita vastund(i)- ā ani mālati nannu ad.igindi
marriage- to Lalita will come- pol q m quot Malati I.acc asked
‘Malati asked me whether Lalita would come to the wedding.’
Thadou (TB)
(12) [[kicenna- a na- huŋ- diŋ- ham]- tin] thaŋboi- in lalboi
marriage- loc 2s- come- fut- pol q m- comp Thangboi- erg Lalboi
a- tiem- me
3s- asked/invited- fut
‘Thangboi asked whether Lalboi would come to the wedding.’
(Pauthang Haokip, p.c.)
We do not have data on Munda languages.

Thus, FOFC has the potential capability to explicate why certain configura-
tions are found and why some are not found in language.
See the Website for appendixes 6.1–6.5 for a discussion of several other
issues related to complementation in SALs.
7 Backward Control
7.1 Introduction
This chapter brings to the attention of scholars working in the area of language
typology and universals how widespread the phenomenon of Backward Control
is in the languages of the South Asian subcontinent and the explanatory power
it has to explicate a variety of syntactic phenomena. We shall demonstrate that
Backward Control is a parametric choice that Universal Grammar (UG) permits
as an alternate structure or as the only structure in some specific cases in some
SALs.
We shall demonstrate that the finite or non-finite nature of the embedded
verb with case-marking properties plays a crucial role in permitting Backward
Control. On the basis of data from control structures and the conjunctive par-
ticipial construction, we demonstrate that it is the nature of the embedded verb,
and consequently the assignment/checking of either structural nominative case
or inherent case of the embedded subject that permits Backward Control. We
shall demonstrate that many case alternations found in SALs, and ‘creative
errors’ made by non-Hindi-Urdu speakers while speaking Hindi-Urdu (IA),
can be explained by invoking the phenomenon of Backward Control. We shall
further show the effects of Backward Control in language contact situations
in Dakkhini Hindi-Urdu, Bhalavali Bhasha spoken in Mangalore, and Bangla
(IA) spoken in Assam. We shall provide evidence from Subzapuri, a dialect of
Hindi-Urdu, which exhibits Backward Control and discuss the implications of
this for the language–dialect dichotomy.
This chapter is organized as follows: section 7.2 introduces the reader to
the notion of Backward Control while comparing it with Forward Control
which is generally attested in many languages. Section 7.3 presents evidence
in support of the phenomenon of Backward Control in Mizo (TB) and Telugu
(DR). Section 7.4 briefly discusses Polinsky and Potsdam’s (2003) approach
to explicate Backward Control in terms of the movement theory of control
This chapter is a revised version of the plenary lecture delivered at the South Asian Languages
Analysis (SALA) conference, University of Illinois, 2005.
246
7.2 Forward and Backward Control 247
proposed in Hornstein (1999), and section 7.5 concludes the chapter. In the
appendix at the end of the chapter we discuss the Case-marked nature of PRO.
The reader is further referred to appendixes 7.1–7.5 on the Website. The
issue discussed in each appendix is mentioned here.
Appendix 7.1 discusses case alternations that occur in constructions involv-
ing matrix verbs such as ‘to say,’ ‘to tell,’ ‘to mention,’ ‘to send a message,’
etc., in SALs, in terms of the phenomenon of Backward Control, and attempts
to explain them in terms of control theory invoking the phenomenon of Back-
ward Control. Appendix 7.2 discusses time expressions and Backward Control.
Appendix 7.3 discusses cases involving Backward Control in language contract
situations. Appendix 7.4 demonstrates how a syntactic phenomenon such as
Backward Control can be used as a heuristic tool to show whether a language
under consideration should be treated as a dialect or language, Appendix 7.5
discusses cases involving Backward Control in some SALs.
7.2 Forward and Backward Control

The Control Theory attracted considerable attention as one of the modules
of the Government and Binding theory. Control is a phenomenon in which
two syntactically linked clauses share a common subject, which is elided in
either the embedded clause (Forward Control) or the matrix clause (Back-
ward Control). Thus, in control structures: (i) the controller (lexical NP)
and controllee (the null element labeled as PRO in Government and Bind-
ing theory) are coindexed; and (ii) there exists an asymmetric relationship
between the controller and the controllee in which the controller c-commands
the controllee in Forward Control and, in contrast, it is the controllee that
c-commands the controller in Backward Control. The controllee (PRO) in
Forward Control is non-overt and is generally assumed to be uncase-marked and
ungoverned in Government and Binding theory according to standard assump-
tions, though there is evidence to the contrary from languages such as Icelandic
(Sigurdsson 1991), Telugu (Lalitha Murthy 1994) and many other SALs
(Subbarao et al. 2007) to show that PRO is case-marked. In contrast, Back-
ward Control, in which the controllee occurs in the matrix clause and the
controller occurs in the embedded clause, is labeled as an instance of Reverse
Equi (Kuroda 1965; Subbarao and Lalitha Murthy, ms), or Backward Control
(Polinsky and Potsdam 2002, 2003; Subbarao 2004). A general example of
Forward Control is schematically given in (1a ), and one for Backward Control
in (1b ). We use English sentences (1a) and (1b) for illustration.
forward control
(1a) [S1 Johni wanted [S2 PROi to kill a mosquitoS2 ]S1 ]
[−finite]
248 Backward Control
Schematically:
(1a ) [S1 NPi Verb + aux [S2 øi INFINITIVE NPS2 ]S1 ]

[−finite]
backward control
(1b) [S1 øi want [S2 Johni kill a mosquitoS2 ]S1 ]
[+/−finite] [+/−finite]
Schematically:
(1b ) [S1 øi Verb [S2 NPi VERB NPS2 ]S1 ]

[+/−finite] [+/−finite]
In these hypothetical examples, we have marked the empty element as ø, and

the matrix verb in (1b) and (1b ) as [+/−finite], as finiteness of the matrix verb
in Backward Control constructions is language-specific. The crucial point to
be noted is that the embedded verb in (1b) in Backward Control either: (i) is
[+finite] as in (2b); or (ii) may be a predicate that assigns inherent case to its
subject as in (11) below, from Telugu.
The crucial difference between Forward Control and Backward Control is
that in Forward Control the non-finite verb does not assign overt case and,
hence, the subject is overtly null or PRO according to standard assumptions, as
in (1a), while in Backward Control, the embedded subject can be case-marked,
as one would expect if the embedded clause is morphologically finite. It may
also be a non-finite participle but can assign inherent case, as in (11) below.
In recent years, an innovative idea concerning control has been proposed
which views control as raising (Hornstein 1999; and Polinsky and Potsdam
2003).
In Subbarao (2004), we have shown in detail that Mizo (TB), Telugu (DR)
and Assamese (IA) have not only the phenomenon of Forward Control but also
the phenomenon of Backward Control, in which the controller occurs not in
the matrix clause but in the embedded clause. We shall provide a couple of
arguments later to substantiate our claim.
Languages in the subcontinent that exhibit Backward Control are:
Dravidian: Kannada, Malayalam, Tamil and Telugu;
Tibeto-Burman: Mizo, Tibetan, Ao (Mongsen), Kokborok, Bodo, Mao
Naga (Giridhar 1994), Ladakhi (Koshal 1979), Garo (Burling 2004);
and
Indo-Aryan: Assamese, Gujarati (Mistry, 2004:18, p.c.), Nepali,
Bhalavali Bhasha, a.k.a. Bhalavali Marathi (spoken in Mangalore),
Eastern Bangla (spoken in Assam), Dakkhini Hindi-Urdu (spoken
in Andhra Pradesh), Swat-Dir Kohistani (IA), Oriya and Torwali
(IA) (Bashir 2003).
Table 7.1 Finite and

non-finite forms of the verb
‘sit’ in Mizo
t.šu t.šut
sit sit
[+fin] [−fin]
Interestingly, Khasi (Mon-Khmer) and the Munda languages do not exhibit

the phenomenon of Backward Control.
7.3 Evidence for Backward Control

To demonstrate what Backward Control is, we shall present an example each
from Mizo (TB) and Telugu (DR) and provide two pieces of evidence in each
case in support of our claim (for further evidence refer to Subbarao 2004).
7.3.1 Evidence from Mizo (TB)

In Mizo, some verbs exhibit two different forms – one for the finite and the other
for the non-finite – while some verbs are homophonous in finite and non-finite
forms. For example, the embedded verb in (2a) is ‘sit’ and the verb for to sit
in Mizo has two forms – one for the finite and the other for the non-finite (see
Table 7.1).
An example of Forward Control is given in (2a). In (2a), the matrix verb
carries the 3rd person singular agreement marker a- and the subject is ergative
case-marked by -n as the matrix verb duh ‘want’ is [+transitive]. The embedded
verb in its infinitival form does not carry any agreement marker (indicated by ø
in (2a) and subsequently) as infinitives do not exhibit any agreement in Mizo.
PRO indicates the position in which the embedded subject occurs.
Mizo (TB)
(2a) zova- n [PRO t.šutleŋ- ah ø- .tšut] a- duh
Zova- erg bench- on sit [−fin] 3s- want
‘Zova wants to sit on the bench.’
An example of Backward Control is given in (2b). In (2b), the matrix subject

is not overtly present and the DP zova ‘Zova’ occurs as the embedded subject.
∀ indicates the position in which the null element occurs: the matrix subject
position. The nominative marker in Mizo is null and is indicated by ø to contrast
(2a) and (2b) with regard to the presence of the ergative marker in (2a) and the
absence of the ergative marker in (2b).
Table 7.2 Finite and

non-finite forms of the
verb ‘to write’ in Mizo
ziak ziah
write write
[+fin] [−fin]
Mizo (TB)
(2b) ∀ [zova- ø t. šutleŋ- ah a- .tšu] ø- duh
Zova- nom bench- on 3s- sit [+fin] 3s- want
‘Zova wants to sit on the bench.’
Note that in (2b) the matrix verb duh ‘want’ does not carry any subject agree-
ment marker (indicated by ø), as the matrix subject is not overtly present. In
contrast, it is the embedded verb that carries the 3rd person singular agreement
marker a- (in italics) in (2b). The crucial point to which we wish to draw the
attention of the reader is that it is the embedded verb in Backward Control in
Mizo that is [+finite] carrying the agreement marker, and the matrix verb does
not carry agreement features.
With the matrix verb tum ‘to try,’ Forward Control is permitted (3a) for all
speakers, and Backward Control is permitted for most of the speakers (3b) –
though, for a few, it is not. The verb for to write in English has two forms in
Mizo – one for the finite (ziak ‘write’) and the other for the non-finite (ziah ‘to
write’) – just as the verb sit has.
(3a) lali- n [PRO lekhabu ø- ziah] a- tum

Lali- erg book 3s- write[−fin] 3s- try
‘Lali tried to write a book.’
(3b) ?/*[∀ lali- n lekhabu a- ziak] ø- tum

Lali- erg book 3s- write[+fin] 3s- try
‘Lali tried to write a book.’
Let us now go back to (2a) and (2b). Evidence for our claim that the DP zova
‘Zova’ in (2b) occurs as the embedded subject comes from the fact that the
embedded subject does not carry the ergative case marker in such sentences,
as the embedded verb .tšut ‘to sit’ is [−transitive]. Note that the matrix verb is
duh ‘want’ and it is [+transitive], and hence requires the subject of the matrix
verb to carry the ergative case marker -n. The occurrence of the ergative case
marker with zova ‘Zova’ in (2a) illustrates the point. In (2b), the DP zova
‘Zova’ does not carry any ergative marker, as it is not the matrix subject and it
is the embedded subject that overtly occurs. Thus, (2b) is a case of Backward
Control in which only the embedded subject is overtly present and the matrix
subject is absent. We now provide evidence to demonstrate that zova ‘Zova’ in
(2b) is the subject of the embedded clause, and not that of the matrix clause.
Evidence from the transitive nature of the verb

The first piece of evidence comes from sentence (4), in which the embedded
verb is [+transitive]. Recall that in Mizo, a transitive verb requires that the
subject be ergative case-marked. Since the matrix verb in (4) is [+transitive],
the subject of the matrix clause in (4) is ergative case-marked. Sentence (4) is
an example of Forward Control.
forward control
(4) [S1 zovii - n [S2 PRO chiarS2 ] ai - duhS1 ]
Zovi- erg study [−fin] 3s- want
‘Zovi wants to study/read [something].’
In (5a), a case of Backward Control, the embedded verb is [+transitive] and

[+finite], and hence, the DP zovi ‘Zovi’ carries an ergative case marker. The
matrix verb does not carry the 3rd person singular agreement marker a- as
the subject of the matrix clause is not present (indicated by ∀ ). If it does, the
sentence is ungrammatical as in (5b).
b a c k w a r d c o n t r o l : e m b e d d e d v e r b i s [ +t r a n s i t i v e ] a n d [+fin ],
a n d he n c e e m b e d d e d s u b j e c t i s e r g a t i v e c a s e - m a r k e d
(5a) [S1 ∀ i [S2 zovii - n ai - chiarS2 ] ø- duhS1 ]
Zovi- erg 3s- study [+fin] 3s- want [−fin]
If there is no overt matrix subject, no agreement marker is permitted on the

matrix verb (in italics) in (5b).
(5b) [S2 *zovi- n a- chiarS2 ] a- duh

Zovi- erg 3s- study [+fin] 3s- want [−fin]
The ungrammaticality of (5b) is due to the fact that the subject agreement
marker in Mizo can occur on the verb if, and only if, the subject is overtly
present.
Since the DP zovi ‘Zovi’ is the subject of the embedded verb, which is
[+finite], it has to be obligatorily case-marked ergative, or else the sentence is
ungrammatical as in (6).
Table 7.3 Finite and non-finite forms of

some verbs in Mizo
[+finite form] [−finite form] gloss
t.šu t.šut ‘sit’

ziak ziah ‘write’
that thah ‘kill’
thɔk thɔh ‘work’
dɔŋ dɔn ‘receive’
sā sāk ‘sing’
(6) [S2 *zovi a- chiarS2 ] ø- duh

Zovi 3s- study [+fin] 3s- want [−fin]
In (7a), the embedded verb is [−transitive]. Hence, in Backward Control, there

is no ergative case-marking on the embedded subject.
(7a) [S2 zovi sikul- ah a- kalS2 ] ø- duh

Zovi school- to 3s- go [+fin] 3s- want [−fin]
‘Zovi wants to go to school.’
If the embedded subject is ergative case-marked because of the transitive nature

of the matrix verb, the sentence is ungrammatical.
(7b) [S2 *zovi-n sikul- ah a- kalS2 ] ø- duh

Zovi-erg school- to 3s- go [+fin] 3s- want [−fin]
Finite and non-finite forms of the verb

The second piece of evidence comes from the finite and non-finite forms of the
verb in Mizo.1 As mentioned earlier, certain verbs in Mizo have two distinct
forms: one for the finite and the other for the non-finite form (see Table 7.3).
For example, the English verb to kill has the finite form that and a non-finite
form thah in Mizo.
There is another set of verbs where finite and non-finite forms of the verb
are homophonous.
A sentence such as Lali wants to kill Lala can be expressed either with a
finite embedded verb – a case of Backward Control – or a non-finite verb – a
case of Forward Control in Mizo. Sentence (8) is a case of Forward Control
with a non-finite embedded verb, while (9) is a case of Backward Control with
a finite embedded verb.
forward control
(8) lali- n [S2 PRO lala ø- thahS2 ] a- duh
Lali- erg Lala 3s- kill [−fin] 3s- want
‘Lali wants to kill Lala.’
The embedded subject PRO in (8) is not overtly present, and hence the embed-
ded verb does not carry 3rd person agreement marker -a. Its absence is denoted
by ø in (8). The form of the verb is thah, which is [−finite], and the matrix
verb carries the 3rd person agreement marker a- as the matrix subject is overtly
present.
In contrast to (8), in (9) the embedded subject is overtly present and the
matrix subject is not, and hence:
(i) the embedded subject carries the ergative case marker -n as the embedded
verb is [+transitive];
(ii) the embedded verb carries the 3rd person agreement marker a- and the
form of the verb is that which is [+finite]; and
(iii) the matrix verb does not carry the 3rd person agreement marker (denoted
by ø in (9)) as the matrix subject is not overtly present.
backward control
(9) [S1 ∀ i [S2 lali- n lala a- thatS2 ] ø- duhS1 ]
Lali- erg Lala 3s- kill [+fin] 3s- want
‘Lali wants to kill Lala.’
7.3.2 Evidence from Telugu

We shall now present evidence from Telugu (DR). An example of Forward
Control in Telugu is:
Telugu (DR)
(10) [S1 raman.ai [S2 PROi kōpam vacc- iS2 ] int.i- ki vel.l.i pōyēd.uS1 ]
Ramana (nom) (dat) anger come- cpm home- to left
‘Having become angry, Ramana left for home.’
Note that, in (10), the embedded predicate kōpam vacc ‘anger come’ takes a
dative subject, and PRO occurs in the embedded subject position. An example
of Backward Control in Telugu is:
(11) [S1 [S2 raman.ai ki kōpam vacc- iS2 ] ∀ i int.i- ki vel.l.i pōyēd.uS1 ]
Ramana (nom) dat anger come- cpm home- to left
‘Having become angry, Ramana left for home.’
We have used the symbol ∀ to indicate the absence of the matrix subject
coindexed with the embedded dative subject.
We shall now provide evidence in support of the presence of the dative

case-marked subject as embedded subject.
Reduplicated form of the anaphor

The first piece of evidence is based on the occurrence of the reduplicated form
of the anaphor in the dative subject construction. An anaphor in Telugu requires
a c-commanding antecedent and its occurrence strictly obeys Principle A of the
Binding theory (Subbarao and Lalitha Murthy 2001, and see chapter 3). Telugu
has a nominal anaphor as well as a verbal one (kon, which literally means ‘buy’),
and the nominal and verbal anaphor occur together when the anaphor occurs
in a subcategorized position. The nominal anaphor is a reduplicated form of
tanu ‘self’ and the first part of the reduplicated form is morphologically case-
marked for the position in which the anaphor occurs, while the second part of
the anaphor has a copy of the morphological case of the subject. Thus, in (12),
since the subject is case-marked nominative and the nominative case marker
is ø, the second part of the anaphor too is case-marked ø. The first part of the
anaphor is accusative case-marked by -ni, since the anaphor occurs in direct
object position. The occurrence of the verbal reflexive is obligatory, while that
of the nominal reflexive is optional. We’ve indicated this by parentheses in (12).
(12) kamalai (tana-ni-tanui ) eppud.ū pogud.u- koni - t.undi

Kamala self-acc-self always praise- VR- progr- 3 s,f
‘Kamala always praises herself.’
Let us now look at the dative subject construction. The patient/theme in the
dative subject construction in (13) carries the locative marker mı̄da ‘on’ (see
(13) kamalai -ki sı̄ta mı̄da kōpam vaccindi

Kamala-dat Sita on anger came
‘Kamala got angry at Sita.’
When the subject is dative case-marked with ki, it is the dative case marker ki
that is copied onto the second part of the anaphor, and the first part carries the
locative marker mı̄da ‘on’ (Subbarao and Lalitha Murthy 2001). The occurrence
of the verbal reflexive kon in NNS constructions is prohibited (15), as the
predicate is [−transitive] and the occurrence of the reduplicated form of the
nominal anaphor is obligatory in the dative subject construction (14) (Subbarao
and Bhaskararao 2004).
(14) kamala kii tana-mı̄da-tana kii eppud.ū kōpam vastū- unt.undi

Kamala dat self-on-self dat always anger coming- keeps
‘Kamala keeps getting angry at herself all the time.’
(15) *kamala kii tana-mı̄da-tana kii eppud.ū kōpam vacc- kon- t.undi
Kamala dat self-on-self dat always anger come- VR- keeps
Sentence (16) is an example of Backward Control with the dative subject as

the embedded subject.
(16) [S1 [S2 mādhuri ki talanoppi- gā und.iS2 ] ∀ ippud.ē pad.ukondiS1 ]

Madhuri dat headache- adjr having had just now slept
‘Having had a headache, Madhuri slept just now.’
An anaphor that occurs in the embedded clause with the dative subject mādhuri
‘Madhuri’ as its antecedent must carry the dative case marker ki if our claim is
correct. Sentence (17) shows that it does. We’ve coindexed the two occurrences
of the dative case marker ki to show that the second occurrence is a copy of the
subject dative case marker.
(17) [S1 [S2 mādhuri kii tana-mı̄da-tana kii kōpam vacc- iS2 ] ∀ ippud.ē
Madhuri dat self-on-self dat anger come cpm just now
pad.ukondi S1 ]
slept
‘Having gotten angry at herself, Madhuri slept just now.’
Recall that a dative predicate does not permit a verbal reflexive to occur with the
verb as ‘dative predicates’ are [−transitive] (see (15) above, and also chapter 3).
If the dative subject were the subject of the matrix clause, the verbal anaphor
should not occur with the matrix verb. The fact that verbal anaphor does occur
in (18) and (19) proves that the dative subject is the subject of the embedded
clause, and not that of the matrix clause. An empty element (∀ ) is the subject
of the matrix clause.
(18) [S1 [S2 raman.a ki picci ekk- iS2 ] ∀ tana- ni- tanu tit.t.u kon-
Ramana dat mad go- cpm self- acc- self abuse VR-
t.unnād.u S1 ]
be.progr.3 m
‘Having gone crazy, Ramana is abusing himself.’
(19) [S1 [S2 salmā ki santōs.am vacc- iS2 ] ∀ tana- ni- tanu pogud.u-
Salma dat happiness come- cpm self- acc- self praise-
kon- t.un-diS1 ]
VR- progr.3 s,f
‘Having felt happy, Salma is praising herself.’
If the dative subject were the subject of the matrix clause, a verbal anaphor
would not be permitted.
Sentences (20a) and (20b) are ungrammatical as the verbal reflexive is not
present. Recall that the occurrence of the verbal reflexive is obligatory, when a
nominative subject is an antecedent to an anaphor. That is, a dative subject is

not the subject of the matrix sentence.
(20a) [S1 [S2 *raman.a ki picci ekk- iS2 ] ∀ tana-ni-tanu tid.u- tunnā-d.uS1 ]
Ramana dat mad go- cpm self-acc-self abuse- be.progr.3 s,m
(20b) [S1 [S2 *salmā ki santōs.am vacc- iS2 ] ∀ tana-ni-tanu pogud.u-

Salma dat happiness come- cpm self-acc-self praise-
tun-diS1 ]
be.progr-3s,nm
Thus, the ungrammaticality of (20a) and (20b) is a conclusive proof that the
dative case-marked DPs raman.a ki ‘Ramana to’ in (18) and salmā ki ‘Salma
to’ in (19) belong to the lower sentence and the subject position of the matrix
clause is null.
The scope of an adverb

The second piece of evidence comes from the occurrence of adverbs and their
scope. The embedded verbs tala noppi gā und. ‘to have a headache’ in (21) and
jabbu gā und. ‘to be sick’ in (22) take a dative subject and permit an adverb of
duration while the matrix verb in the following examples is not in the scope of
the duration adverb and takes only a nominative subject.
(21) [S1 [S2 mādhuri ki gant.a nunci talanoppi- gā und.iS2 ] ippud.ē
Madhuri dat hour from headache- adjr having had just now
d.ākt.aru daggiri- ki vel.l.indiS1 ]
doctor near- to went
‘Having had a headache for one hour, Madhuri went to the doctor just now.’
(22) [S1 [S2 madhu ki padi ēl.l.a nunci jabbu- gā und.- iS2 ] ∀ ippud.ē
Madhu dat ten years from sick- adjr be- cpm just now
pōyēd.uS1 ]
died
‘Having been ill for ten years, Madhu died just now.’
Neither of the matrix verbs vel..l ‘to go’ in (21) nor pō ‘to die’ in (22) permits
an adverb of duration, and hence it is the embedded verb that is in the scope
of the duration adverb, and the dative case-marked subject occurs to its left in
the embedded clause as its subject. This shows that the dative subject is in its
in situ position as the subject of the embedded clause.
7.4 Backward Control in Malagasy

Discussing the cases involving Backward Control in Malagasy (an Austrone-
sian language spoken in the island of Madagascar), Polinsky and Potsdam
7.5 Conclusion 257
(2003) observe that the Principles and Parameters approach (P and P approach:
Chomsky and Lasnik 1995) is not adequate to handle the cases of Backward
Control as the empty element in the matrix clause in such cases “is not bound,”
as in (23) (Polinsky and Potsdam 2003: 10). They mark the empty element as
PRO and we marked it as ∀ in our discussion so far.
Malagasy
(23) manomboka [mitondra ny fiara Rabei ] PROi
begin drive the car Rabe
‘Rabe is beginning to drive the car.’
The authors make two observations: (i) if PRO is not bound, it gets the arbitrary
PRO interpretation, which is unavailable in Malagasy; it is also the same case in
SALs; (ii) the configuration in example (23) from Malagasy “violates Condition
C of the Binding Theory, which requires that R-expressions like Rabe be
free” (Polinsky and Potsdam 2003: 10). Though the DP in (23) is not free, the
sentence in Malagasy is grammatical, they point out. They conclude: “Since the
arguments against the PRO analysis are essentially independent of the details of
Malagasy syntax, we conclude that P and P quite generally rules out Backward
Control” (Polinsky and Potsdam 2003: 10). They propose an analysis in terms
of Hornstein’s (1999) Movement analysis. We agree with their conclusion
concerning the P and P approach’s failure to account for Backward Control.
However, as we have shown elsewhere (Subbarao 2003), Hornstein’s (1999)
Movement analysis – as well as Manzini and Roussou’s (2000) analysis –
encounters problems in handling the data from SALs. Hence, we leave the
issue open for further investigation.
A significant point made by Frans Plank (p.c.) is that the available evidence
shows that Backward Control is possible only in verb-peripheral languages
(SOV and VSO) and not in verb-medial languages. It is not clear what the
implications of this are for Universal Grammar (UG).
7.5 Conclusion
To conclude, Dravidian languages such as Telugu, Kannada and Malayalam
and Indo-Aryan languages such as Assamese, Nepali, Eastern Bangla and
Subzapuri that have a tensed conjunctive participle permit Backward Control,
while standard Bangla and Kashmiri do not, though they do have a tensed
conjunctive participle. Dakkhini Hindi-Urdu, which has been in contact with
Dravidian languages, permits Backward Control, though it does not have a
tensed conjunctive participle. We have shown that allowing Backward Control is
a parametric choice that Universal Grammar permits, and while some languages
may exploit it, some may not. We have also shown that, in some cases, Backward
Table 7.4 A partial list of languages that permit

Backward Control
Language Language family
Nepali Indo-Aryan
Assamese Indo-Aryan
Eastern Bangla Indo-Aryan
Oriya Indo-Aryan
Gujarati Indo-Aryan
Bhalavali Bhasha Indo-Aryan
Dakkhini (Hindi-Urdu) Indo-Aryan
Swat-Dir Kohistani Indo-Aryan
Torwali Indo-Aryan
Kannada Dravidian
Malayalam Dravidian
Telugu Dravidian
Ladakhi Tibeto-Burman
Mao Naga Tibeto-Burman
Mizo Tibeto-Burman
Control is the only option permitted in a language in some structures, and thus
it may not necessarily be a marked construction, as Table 7.4 illustrates.
7.6 Postscript2
Backward Control is permitted in Mangalore Konkani (IA) while it is not in
Standard Konkani (IA). Note that, in (1) below, the subject of the embedded
clause ammi ‘we’ is nominative case-marked as the embedded predicate yev
‘to come’ does not take a non-nominative (dative) subject.
backward control
Mangalore Konkani (IA)
(1) [S1 [S2 ammii haŋga yev- nuS2 ] ∀i dha vərsəj dzallı̃jS1 ]
we here come- cpm ten years happened
Intended meaning: ‘It is ten years since we came here.’
(Lalita Dhareshwar, p.c.)
A Hindi-like Forward Control structure is also possible in Mangalore Konkani.

Both the CPM and the perfective participle are possible. The subject of the
matrix clause am-kã ‘we-dat’ is dative case-marked as the matrix predicate
dzallı̃ ‘happened’ requires a non-nominative subject in (2) and (3).
Appendix – case-marked PRO 259
forward control structure with a conjunctive participial

marker
(2) [S1 [S2 PROi haŋga yev- nuS2 ] am-kãi dha vərsəj dzallı̃j S1 ]
here come- cpm we.dat ten years happened
‘It is ten years since we came here.’
f o r w a r d c o n t r o l st r u c t u r e w i t h a p e r f e c t p a r t i c i p l e
(3) [S1 [S2 PROi haŋga ai- l εS2 ] am-kãi dha vərsəj dzallı̃j S1 ]
here come- perf pple we.dat ten years happened
‘It is ten years since we came here.’
Mangalore Konkani shares a variety of syntactic features with the Dravidian

Kannada language.
In the appendix below, we discuss a crucial issue regarding the nature of
case-marked PRO which concerns Backward Control as well as conjunctive
participial (CP) clauses in SALs. (For a detailed discussion of conjunctive
participle clauses in SALS, see chapter 9 on the Website.)
Appendix: the Case-marked nature of PRO

Since, in many South Asian languages, it is the conjunctive participial construc-
tion that permits Backward Control and the subject of the conjuctive participial
clause is case-marked, in this section we wish to provide evidence to demon-
strate that PRO is case-marked. Chomsky and Lasnik (1995) argue that PRO
is null case-marked, and such Case is checked by a non-finite T. Based on
evidence from Icelandic, Sigurdsson (1991) argues that PRO is case-marked.
Martin (2001) argues that theories proposed in terms of Binding and Govern-
ment to account for the distribution of PRO are inadequate, and he attempts
to account for it in terms of minimalist assumptions keeping in view the tem-
poral properties of non-finite T. Lalitha Murthy (1994) argues that PRO is
case-marked, and provides evidence from select Indo-Aryan and Dravidian
languages. Based on evidence from non-nominative subjects in Telugu (DR)
and Rabha (TB), dative and genitive subjects in Manipuri, and long-distance
agreement in Hindi-Urdu (IA), Subbarao et al. (2007) argue that PRO is case-
marked. Our argument relies on the fact that non-nominative subjects can occur
as subjects of a CP clause. Since non-nominative case is inherently assigned,
and the non-nominative subject occurs in the same position as PRO normally
occurs, we argue that the position earmarked for PRO is a case-marked position.
Our claim is further strengthened by long-distance agreement in Hindi-Urdu
and Punjabi. We shall argue that, unless PRO is case-marked, facts of long-
distance agreement in Hindi-Urdu cannot be explained.
We present three arguments to show that PRO is case-marked.
(i) PRO in dative case-marked position

In chapter 5, we have shown that PRO can occur in the position of subject of a
CP clause that permits a dative subject. The embedded predicates kōpam vacc
‘anger come’ in Telugu and in other Dravidian languages (in (1)), and cāhi-era
‘needed’ in Nepali (IA) (in (2)) require a dative subject. PRO is glossed here
with the case marker that would normally occur with a lexical subject.
Telugu (DR)
(1) rādha [PRO kōpam vacc- i] vel.l.- i pōyindi
Radha dat anger come- cpm go- cpm went
‘Having felt angry, Radha left.’
Nepali (IA)
(2) [PRO pāni cāhi- era] u nadi tira ga- yo
dat water need- cpm he river toward go- pst
‘Needing water, he went to the river.’
(Ichihashi-Nakayama 1994, as quoted in
Bickel 2004: 81)
In contrast, in Hindi-Urdu (Davison 2004a), Punjabi and Bangla (IA), PRO

cannot occur in the subject position of an embedded conjunctive participial
clause which has a predicate that takes a dative or genitive subject. We provide
an example from Hindi-Urdu. The predicate gussā ānā ‘anger come’ requires
a dative subject.
Hindi-Urdu (IA)
(3) *rādhā [PRO gussā ā kar] bāhar cal- ı̄ ga- ı̄
Radha dat anger come cpm outside walk- pst go- pst
Intended meaning: ‘Having got angry, Radha went outside.’
In such cases, only a predicate such as nārāz honā ‘to be/become angry’ that
takes a nominative subject is permitted, as in (4).
(4) rādhā [PRO nārāz ho kar] bāhar cal- ı̄ ga- ı̄

Radha nom anger come cpm outside walk- pst go- pst
‘Having felt angry, Radha went outside.’
We have shown in chapter 5 that dative case in non-nominative subject con-

structions is a lexical case, and it is inherently assigned.3 Since PRO occurs as
the subject of a CP clause containing a predicate that takes a non-nominative
(dative) subject, PRO must be dative case-marked in (3) and (4). This, in our
opinion, provides conclusive evidence in support of the case-marked nature
of PRO.
Appendix – case-marked PRO 261
(ii) Evidence from Manipuri (TB)

Manipuri, too, strongly supports our claim. In Manipuri, the subject as a pos-
sessor is genitive case-marked (5).
subject is genitive case-marked

Manipuri (TB)
(5) mahak- ki minai yamna lay
he- gen servant many has
‘He has got many servants.’
PRO occurs in a CP construction in a position in which a genitive case-marked

NP occurs.
(6) [PROi minai yamna lay- raga- su] tombai (yum-gi) thabak
gen servant many has- cpm- also Tomba house-gen work
tou-i
do-nonfut
‘In spite of having many servants, Tomba does the work (household chores).’
Note that genitive case is inherently assigned by the predicate to the subject
in Manipuri, and hence the occurrence of PRO in (6) cannot escape being
case-marked genitive.
An experiencer subject is dative/locative case-marked (7).
(7) mangon- da dosa ta-re

he- dat/loc ill-fate fall-nonfut
Literally: ‘Ill-fate fell to/on him.’
‘He had bad luck.’
PRO occurs in a position in which a dative/locative case-marked NP occurs.
(8) [PROi dosa ta- raga- su] mangoni - da

dat/loc ill-fate fall- cpm- also he- dat/loc
nungai- ta- ba lay- te
happy- neg- inf be- neg
‘In spite of having ill fate fallen [falling] on him, he does not regret.’
Dative/locative case too is inherently assigned to the subject by the predicate

in Manipuri, and hence PRO in (8) too is inherently case-marked, just as the
genitive case-marked PRO is in (6).
Thus, it is the nature of the predicate that assigns inherent genitive/

dative/locative Case to the subject in Manipuri.
(iii) PRO triggering agreement: evidence from Ho (Munda), Malto

and Kurukh (Dravidian)
(i) In Ho (Munda), the fact that the conjunctive participle carries the subject
agreement marker and the subject in such cases is PRO demonstrates that
PRO is case-marked (sentence (53) in chapter 9 on the Website).
(ii) In Kurukh (DR) and Malto (DR) too, the CP carries the number agreement
marker (see (54) in Kurukh and (55) in Malto in chapter 9 on the Website).
Further support for our claim concerning case-marked PRO comes from long-
distance agreement in Hindi-Urdu and Punjabi. We demonstrate elsewhere that,
unless PRO is case-marked, facts concerning long-distance agreement in object
complements in Hindi-Urdu and Punjabi cannot be adequately explained (see
8 Noun modification: relative clauses
8.1 Introduction
‘A relative clause is a clause that modifies a phrasal constituent, generally a
noun phrase. We call the noun phrase that is so modified the head of the relative
clause’ (Riemsdijk 2006: 338).1 A relative clause is a subordinate clause in
which the embedded predicate may be [+finite] or [−finite]. Such a non-finite
predicate may be participial or infinitival depending upon the language family
that the specific language under consideration belongs to.
In this chapter, we discuss three types of relative clauses2 that are found in
the South Asian subcontinent. They include:
(i) externally headed relative clauses (EHRCs);
(ii) Relative-Correlative clauses; and
(iii) internally headed relative clauses (IHRCs)
We shall briefly elaborate the above three types:
(i) Externally headed relative clauses (EHRCs)
EHRCs are of two subtypes:
(a) the sentential-type; and
(b) the non-finite type.
(a) The sentential-type are post-nominal relative clauses with a finite
embedded verb, as in Indo-Aryan. They do not have a null
operator.
(b) The non-finite type may either be pre-nominal or post-nominal.
They both have a ‘null (covert) operator’ in the embedded rela-
tive clause. They may either be participial as in Dravidian, Indo-
Aryan, Mon-Khmer and Munda, or infinitival (nominalized) as in
Tibeto-Burman languages.
In fact it is the absence or presence of the covert operator that distinguishes
the sentential-type from the non-finite type, in addition to other distinctive
features such as:
• the [+finite] nature of the embedded verb in the sentential-type and the
[−finite] nature of the embedded verb in the non-finite type; and
263
• the presence of an adposition (a postposition or a preposition) that indi-

cates the grammatical relation of the embedded head with the embedded
verb in the sentential-type and the absence of such adposition in the
non-finite type.
(ii) Relative-correlative clauses are sentential, as in Indo-Aryan and less com-
monly in Dravidian, and they have a finite embedded clause.
(iii) Internally headed relative clauses are nominalized embedded clauses with
a null element in the matrix clause coindexed with an overt head in the
embedded clause. The internal head may be lexically case-marked or
abstractly case-marked in the embedded clause.
One of the significant characteristic features of SALs is the occurrence of
the relative-correlative clauses, which are common only in SOV languages,
according to Downing (1978) and Keenan (1985).
There are several issues concerning relative clauses in general, and those
is SALs in particular. In this chapter, we shall restrict our attention only to a
specific set of issues that includes:
(i) types of relative clauses in terms of their position of occurrence and the
difference, if any, between the pre-nominal correlative structures and the
post-nominal relative clauses;
(ii) the positions in the embedded clause that are relativizable,3 keeping in
view Keenan and Comrie’s (1977) Noun Phrase Accessibility Hierarchy
(NPAH);
(iii) the ‘Strict OV Constraint’ and its adherence in relative-correlative clauses
in Dravidian and Tibeto-Burman;
(iv) the nature of internally headed relative clauses (IHRCs);
(v) potential ambiguity in relative clauses and factors responsible for ambi-
guity resolution;
(vi) evidence in support of the canonical position of the internal head in
IHRCs;
(vii) the role of movement and reduplication of the head in IHRCs;
(viii) implications of adposition incorporation in IHRCs in Hmar (TB); and
(ix) syntactic changes that occur in the formation of relative clauses in lan-
guage contact situations.
This chapter has illustrations of the basic patterns and variation of sentential
relative clauses and EHRCs from Indo-Aryan and Dravidian; IHRCs from
Tibeto-Burman. Further information about less well-known and under-analyzed
languages is provided in the appendixes on the Website.
This chapter is organized as follows: section 8.2.1 deals with externally
headed relative clauses (EHRCs) which are either sentential relative clauses
or participial/infinitival relatives in which a relative clause is formed by
having a covert operator in a specific position in the embedded relative
clause; section 8.2.2 deals with the relative-correlative construction, while

section 8.2.3 deals with internally headed relative clauses (IHRCs) in which a
(null) pro4 occurs in the matrix clause coindexed with an internal head. This
section provides illustrations from SALs. Section 8.3 presents a brief summary
of the relativizable positions in a relative clause, and a brief introduction to
the asymmetries between the pre-nominal and post-nominal relative clauses.
Section 8.4 presents a brief discussion of EHRCs in the four language fami-
lies of South Asia, and the asymmetries found. Section 8.5 deals with IHRCs
in Tibeto-Burman. Section 8.6 demonstrates that for a comitative PP to head
an EHRC or an IHRC, a thematic relation needs to be established between
the predicate of the embedded clause and the head of the EHRC/IHRC, in
terms either of an overt case marker or some marker in the embedded verb to
indicate accompaniment. Section 8.7 deals with the implications of adposition
incorporation in IHRCs in Hmar. Section 8.8 concludes the discussion. On the
Website, appendix 8.1 deals with the relativizable positions in a relative clause
in less well-known languages of the Munda, Khasi and Tibeto-Burman fam-
ilies, while appendix 8.2 discusses the asymmetries between the pre-nominal
and post-nominal relative clauses. Appendix 8.3 presents a brief description of
the relativizable positions in EHRCs (participal relatives) in the less well-known
languages of the Munda, Khasi and Tibeto-Burman language families, while
appendix 8.4 discusses the asymmetries found in Tenyidie (TB) in EHRCs.
SALs exhibit four noun modification strategies that include:
(i) relative clauses (the wh-relative clause-type which are ‘fully sentential’);
(ii) the participial- or nominalization-type with an infinitival form of the
embedded verb, depending on the language family;
(iii) the agentive-nominal-type construction (which corresponds to the -er type
nominals of English);5 and
(iv) the modification of a noun by an adjective.
In this chapter, we shall restrict our attention only to (i) and (ii).
8.2 Types of relative clauses

In terms of the position of occurrence of the head in a relative clause, the
relative clause in SALs exhibits three distinct patterns.
8.2.1 Pattern 1: externally headed relative clauses (EHRCs)

In EHRCs the head is overtly present in the matrix clause. In this category, there
are two subtypes. The relative clause may be finite or non-finite (infinitival or
participial) and the latter may be pre-nominal or post-nominal which is language

specific.
Subtype 1: in subtype 1, the relative clause is a sentential relative clause
in which the embedded relative clause occurs immediately to the right of the
head noun phrase or of the VP of the matrix clause. The embedded verb is
[+finite].
Subtype 2: in subtype 2, the embedded verb is [−finite] (infinitival or par-
ticipial).
8.2.2 Pattern 2: the relative-correlative clause

In this pattern a pronoun or a Demonstrative Phrase in the matrix clause is
coindexed with the relativized noun phrase in the embedded relative clause.
According to Bhatt (2003: 486), in Hindi-Urdu (IA) for example, “there is a
Correlative clause that contains a Relative Phrase . . . The Correlative Clause
is associated with a matrix clause that contains a Demonstrative Phrase.” In
Dravidian languages, it is always a pronoun that occurs in the matrix clause,
and not a Demonstrative Phrase.
8.2.3 Pattern 3: internally headed relative clauses (IHRCs)

This pattern consists of IHRCs in which the embedded verb is infinitival (nom-
inalized) and the head occurs inside the embedded relative clause.
We now provide examples to illustrate each category mentioned above.
Pattern 1: EHRC
These consist of either the post-nominal sentential relative clause or the non-
finite (infinitival or participial) relative clause.
Recall that, according to our classification, Pattern 1 has the post-
nominal relative clause and the EHRC. Let us label them as Subpattern 1
and Subpattern 2, respectively.
Subpattern 1: the post-nominal sentential relative clause

Post-nominal relative clauses may occur either: (a) immediately to the right of
the head noun, which we shall label as NP-adjoined relative clauses; or (b) to
the right of the VP of the matrix clause, which we shall label as the extraposed
(CP-adjoined) relative clause.
Sentence (1) is an example of a post-nominal relative clause where the
embedded relative clause occurs immediately to the right of the head noun.
post-nominal relative clause

Hindi-Urdu (IA)
(1) [[ve log]i [jo zyādā cāy pı̄-te hãı̃]i ] kam so
which people who a lot of tea drink-imperf pres less sleep
pā-te hãı̃
can-imperf pres
‘People who drink a lot of tea sleep less.’
In Subpattern 1, the embedded relative clause may also occur to the right of the
matrix VP, in which case the relative clause is away from the head and is not
adjacent, as in (2). The embedded relative clause is thus considered extraposed
and is right-adjoined to CP.
the right-adjoined to cp or the extraposed relative clause

(2) [ve log]i kam so pā-te hãı̃ [jo zyādā cāy
which people less sleep can-imperf pres who a lot of tea
pı̄-te hãı̃]i
drink-imperf pres
The right-adjoined relative clause together with the head in (1) forms a DP
constituent, while the CP-adjoined or extraposed relative in (2) does not form
a DP constituent, as it is away from its head.
Subpattern 2: EHRC: embedded verb is infinitival or participial

Sentences (3) and (4) are examples of a non-finite externally headed relative
clause (EHRC) and they belong to Subpattern 2 of our classification.
All the SALs have the EHRC, in which the coreferential NP of the embedded
clause is a covert operator which does not occur overtly. We have indicated
it by ø in the examples given later. Thus, if the head noun of the embedded
clause is a covert operator occurring in a specific position, such as Subject (S),
Direct Object (DO), Indirect Object (IO), Oblique Object (OO) or Object of
the Genitive (OGEN) of the embedded clause, and the head occurs only in
the matrix clause, it is called an externally headed relative clause (EHRC),
which is non-finite. It is that specific position in the embedded clause that is
targeted in an EHRC and it is that position that is relativized. Thus, in sen-
tence (3) in Hindi-Urdu (IA), for example, the direct object of the embedded
clause is kapr.e ‘clothes,’ and in (4) in Tenyidie (TB) the subject of the embed-
ded clause is khriesa ‘young man.’ They occur as covert operators and the
positions that are relativized over are the direct object in (3) and subject in
(4). Thus, (3) is an example of Direct Object modification and (4) of Sub-
ject modification. All pre-nominal relatives in Dravidian and Indo-Aryan are
participial, and both pre-nominal and post-nominal relatives in Tibeto-Burman

are infinitival.
e x t e r n a l l y h e a d e d r e l a t i v e c l a u s e (e h r c )
d i r e c t o b j e c t m o d i fi e d
Hindi-Urdu (IA)
(3) [S2 pūjā ke kal ke øi kharı̄de hueS2 ]
Pooja gen (m,p) yesterday gen (m,p) bought (perf pple, m, p)
kapr.ei bahut khūbsūrat the
clothes (m, p) very pretty were (m, p)
‘The clothes that Pooja bought yesterday were very beautiful.’
subject modified
Tenyidie (TB)
e h r c : t h e he a d khriesa ‘y o u n g m a n ’ o c c u r s i n t h e m a t r i x c l a u s e
(4) [S2 øi bulie kemerie se- ke- baS2 ] khriesai - u- e a-zemia
shirt red wear- nozr- progr young man- def- nom my-friend
‘The boy who is wearing a red shirt is my friend.’
(Kevichüsa and Subbarao 1998: 45)
Pattern 2: the relative-correlative clause

An example of the relative-correlative type is provided in (5) below. Such
clauses are predominantly found in SALs and they tend to occur only in
verb-final languages.6 Since the relative-correlative type occurs only as a pre-
nominal relative clause, it is called a pre-nominal relative-correlative clause. In
(5) in Hindi-Urdu (IA), the relative pronoun jo ‘which’ functions as a relative
determiner and ve ‘they’ in the matrix clause is labeled as the correlative pro-
noun or correlate. In Dravidian too the relative pronoun functions as a relative
determiner when followed by a lexical noun such as boy, girl, animal, stone,
etc.
pre-nominal relative-correlative clause

Hindi-Urdu (IA)
(5) [jo log zyādā cāy pı̄-te hãı̃]i [ve]i kam so
which people a lot of tea drink-imperf pres they less sleep
pā-te hãı̃
can-imperf pres
In Indo-Aryan languages there are Multi-Head relative clauses (Bhatt 2003:

491) as in (6), which are invariably the relative-correlative type.
Multi-Head correlative
Multi-Head relative clauses as in (6) are invariably the relative-correlative type
and not extraposed. They generally have the interpretation of a relative clause
where the head is [−definite].
multi-head correlative
(6) [jis- nei joi kar-nā cāhā]i,j [us- nei voj ki-yā]
who- erg what do-gerund/ want.perf he/she- erg that do.perf
For x, y s.t. x wanted to do y, x did y.
Literally: ‘Who wanted to do what, he/she did that.’
(Bhatt 2003: 486)
Multi-Head correlatives are found in all the Indo-Aryan languages (Bhatt 2003:
486) and also in Dravidian Languages.
Pattern 3: internally headed relative clauses (IHRCs)

The third type of clauses are the IHRCs in which the head occurs internally
in the embedded relative clause and it is not overtly present in the matrix
clause. IHRCs are found only in verb-final languages (Keenan 1985).7 All
Tibeto-Burman languages are verb-final and most of them have the IHRC. The
internal head is in italics in examples (7) and (8). The letter ø in (7) and (8a)
indicates the empty category (a null pro) that occurs in the matrix clause, and
is coindexed with the head in the embedded clause. The embedded verb in
an IHRC is an infinitival/nominalized form, as all IHRCs in Tibeto-Burman
languages are the result of nominalization of the embedded clause.
There are two types of IHRCs in Tibeto-Burman, which we shall discuss in
detail in section 8.6. Suffice it to say here that the internal head in an IHRC
may or may not carry a postposition.
Type A: internal Head of an IHRC with an overt case marker (postposition)
Sentence (7) is an example of an IHRC in which the head is an indirect
object and the dative case marker occurs overtly with the head. Note that such
occurrence is prohibited in EHRCs (which have a gap or a relative pronoun).
We might label the languages in which the postposition overtly occurs with the
head as Type A languages, and languages in which the postposition does not
overtly occur with the head as Type B languages.
Type A: Internally Headed Relative Clause (IHRC) with an

overt postposition
When the IO is the head of an IHRC, the dative postposition khaŋ ‘to’ is overtly
present in Sangtam (TB). The head nistar ‘person’ occurs in the embedded
clause.
indirect object modified – dative case-marked io as head of

an ihrc
Sangtam (TB)
(7) [S2 n- n nistar khaŋi šeti t.hraʔ- ba- tsəS2 ] øi (pro)8
you- nom person to letter write- nozr- def
khataŋ tšŋle
very tall
‘The person you wrote a letter to is very tall.’
Type B: Internally Headed Relative Clause (IHRC) with

no overt postposition
indirect object modified – dative case-marker not present

with io as head of an ihrc
Type B consists of an IHRC in which the postposition indicating the grammat-

ical relation of the head noun phrase with the embedded verb is not overtly
present.
IHRC TYPE B
i n d i r e c t o b j e c t m o d i f i e d – i o a s he a d o f a n i h r c w i t h no o v e r t
postposition
Hmar (TB)
(8a) [lala- n bɔŋi (IO) bu a- pek] kha øi zanikhan a- thi
Lala- erg cow food 3s- give det pro yesterday 3s- died
‘The cow which Lala gave food to died yesterday.’
(Kumar 1998)
(8b) lala- n bɔŋ (IO) hnenaʔ bu a- pek- tɔ

Lala- erg cow dat food 3s- give- perf
‘Lala gave food to the cow.’
Note that the dative postposition hnenaʔ ‘to, for’ with the internal head bɔŋ
‘cow’ is not present in (8a), while in a simple sentence (8b) its presence is
obligatory.
The examples provided above show the various types of relative clauses
found in SALs. We shall now discuss the syntax of relative clauses in SALs.
It is in Indo-Aryan languages that relative-correlative clauses are used fre-
quently and freely, “except for Sinhala, which, like neighboring Tamil, has
only the latter [the EHRC]” (Masica 1991: 408).9 In Dravidian languages,
8.3 Grammatical functions accessible 271
relative-correlative clauses are used less frequently. Korku (Munda) and most
of the TB languages do not have sentential relative clauses at all. They only
have a participial or infinitival gap relative.
One of the interesting issues that arises is: what are the grammatical functions
that are accessible to relativization in SALs? That is, what are the positions in
which a covert operator can occur?
In EHRCs (sentential relative clauses) of Pattern 1 and in the relative-
correlative construction of Pattern 2, all positions / grammatical functions
(subject, direct object, indirect object, oblique object, possessor modification
and object of comparison) of the Noun Phrase Accessibility Hierarchy (NPAH)
(Keenan and Comrie 1977) are relativizable, and thus there are no restric-
tions on these SALs. The relative-correlative construction that is found in
SALs is:
the characteristic Indo-Aryan (Old as well as New) . . . where the modifying clause,
marked by a member of the “J”-set of relative pronouns, adverbs, and other words, is
“represented” by a correlative in its role-slot (i.e. basically that of Modifier in the NP)
in the main clause. The correlative is usually identical with the remote demonstrative
in the language (except in Sindhi and Dakkhini, where the correlative so preserves the
form of an earlier IA demonstrative). (Masica 1991: 410)10
To give an idea of the nature of relativization in SALs, we shall present here

examples illustrating the relativizable positions in Indo-Aryan and Dravidian,
and present the rest of the examples pertaining to Mon-Khmer, Munda and
Tibeto-Burman in appendix 1 on the Website.
8.3 Grammatical functions accessible to relativization in sentential

relative clauses
We shall now discuss “the grammatical role [emphasis provided] of the noun
phrase in the relative clause” (Fox 1987: 856) that is accessible for relativization
in accordance with Keenan and Comrie’s Noun Phrase Accessibility Hierarchy
(NPAH).
In relative clauses (Pattern 1 (Subtype 1) and Pattern 2), there are no restric-
tions on the grammatical role accessible for relativization or on the case-marked
nature of an NP (nominative or non-nominative subject) in the embedded clause
in SALs. For example, a dative, genitive or locative phrase can be the head of
the relative clause.
Another feature that deserves mention is that time, place, manner and quantity
adverbs can head a relative clause.
The following discussion applies to both sentential EHRCs and relative-
correlative clauses.
8.3.1 Indo-Aryan
Nominative subject as head
The NP mulgi ‘girl’ is not present in (9) in the right-adjoined relative clause.
Marathi (IA)
(9) ti mulgi [ji ghari geli] ti ithe rāhte
that girl which home went that here lives
‘The girl who went home lives here.’
(Wali 2006: 289)
Sentence (9) demonstrates that in an NP-adjoined relative clause which is

externally headed, an internal head may also be realized in some languages.11
Dative subject as head

The relativized NP jyā-lā ‘who-dat’ is a dative subject, and it is coindexed with
the matrix subject to ‘he’ in nominative case in (10) in Marathi (IA).
Marathi (IA)
(10) jyā- lāi mulaj nastāt toi tyānnāj premāne vāgavto
who- to children not that (he) them-to love.with treat
‘One who does not have children treats them with love.’
(Wali 2006: 291)
An important fact about the relative clause in SALs is that the relativized NP
is generally fronted, if it is a non-subject (DO, IO, OO, etc.).
A comment about the V2 order in Kashmiri (IA) is in order. Recall that
Kashmiri is a V2 language (chapter 2). Bhatt (1999: 83) observes: “In relative
clauses and subordinate clauses V2 is blocked.” Hence, in (11) and (12) the
word order in the embedded clause is SOV and not SVO, while in the matrix
clause it is SVO. The position from which the NP is fronted is indicated by the
trace t, which is coindexed with the relativized NP in (11).
direct object
Kashmiri (IA)
(11) su d.ab [S2 yusi tse ti lobuthS2 ]
subject direct object verb
corr box rel (which) you found
chu k¯mtı̄
is precious
‘The box which you found is precious.’
indirect object
(12) [su naphar] [S2 yas me gər ditsS2 ]
subject subject direct object verb
that person rel (whom) I.erg watch gave
gav gar
v e r b Adv
went home
‘The person (to) who(m) I gave the watch went home.’
Thus, the word order in Kashmiri (IA) in relative clauses exhibits an asymmetry
with that in the main clause.12 In contrast, there is no such asymmetry in
complement clauses (see chapter 6 for a discussion).
Oblique object
All oblique objects can be relativized in IA languages.
locative object
Marathi (IA)
(13) te ghar [dzyā ø- t mı̄nā rāhte]
corr house rel.obl (which) loc Meena live.pres.3s,f
‘The house in which Meena lives . . . ’
(Pandharipande 1997: 88–89)
Comitative
Sentence (14) is an example of the relative-correlative clause in which the
relative clause is left-adjoined to the correlate. The word order is V2 in the
matrix clause and SOV in the embedded clause.
Kashmiri (IA)
(14) [yemis naphras- sı̄¯th’ me sinema vuch]
who.dat (which) man.dat- with I.erg cinema saw
təm’ h’ot makān
he.erg bought house
s u b j e c t v e r b ( v 2) d i r e c t o b j e c t
‘The person with whom I saw the movie bought the house.’
(Sadaf Munshi, p.c.)
Sentence (15) is an example in which the relative clause occurs to the right of
the VP of the matrix clause.
v2 order in the matrix clause and sov in the embedded

cp-clause
(15) tem’ naphr- an h’ot makān [yemis sı̄¯th’ me
that man- erg bought house who.dat with I.erg
subject v e r b ( v 2) o b j e c t s u b j e c t
sinema vuch]
cinema saw
object verb
‘The person with whom I saw the movie bought the house.’
(Aadil Kak, p.c.)
Time adverbs
The relative adverb yelyi ‘when’ is the head of the relative clause in (16) in
Kashmiri and all types of adverbs can head a relative clause in Indo-Aryan,
Dravidian and Munda languages. A few Tibeto-Burman languages also permit
this. The word order is SOV in the embedded relative in (16).
(16) [MP yelyi [[M’ ba [VP par.hāı̄ khatam kar]]]]

when I studies finish do
‘When I finish my studies . . . ’
(Bhatt 1999: 162)
r e l a t i v i z a t i o n o f p o s s e s s o r no u n
Marathi (IA)
(17) tı̄ mulgı̄ [dzi ø- tsa gān.a mı̄ aikla]
corr girl rel poss.3s,neut song.3s,neut I heard.pst.3s,neut
‘The girl whose song I heard . . . ’
8.3.2 Dravidian
Dravidian has only the relative-correlative clauses (Pattern-2-type clauses dis-
cussed in 8.2), which are left-adjoined, and it does not have the other two
varieties, namely NP-adjoined and extraposed relative clauses, as in (2) and
(3) in Hindi-Urdu, and other Indo-Aryan languages. The question word and
the relative pronoun are homophonous, just as in English and French. Any
nominal constituent or an adverbial phrase can be the head of a relative clause
in Dravidian.
In Kannada, Sridhar (1990: 47) observes: “It [the relative-correlative clause]
is more used, however, in writing than in speech, especially to relativize minor
constituents such as location, circumstance, etc.” Asher and Kumari (1997: 58)
point out that in Malayalam too the “little used sentential relativization strategy
is very versatile, in that any constituent of a sentence – including adverbs as

well as NP – can be relativized.” Krishnamurti and Gwynn (1985: 237) point
out: “there are no relative pronouns in Telugu and their part is played by verbal
adjectives” (italics in original). By verbal adjectives, Krishnamurti and Gwynn
(1985: 237) mean participles modifying a noun. Krishnamurti and Gwynn later
observe (1985: 361) that the correlative relative clauses found in Dravidian are
through diffusion from Indo-Aryan and “are imported into Telugu and are used
in formal speech.” Krishnamurti (2003: 448) points out that relative clauses
“are not favoured in spoken varieties of modern standard languages, except for
rhetorical purposes.”
Relative clauses are found in Tamil. Steever provides an example of a relative
clause from Tamil, as in (18).
Tamil (DR)
(18) yār aŋkē mutali vantu cēru- v- ār- ō avar
who there first come.cpm reach- fut- 3s,hon- dub mkr he.nom
t.iket.t.u vāŋkalam
ticket buy.permissive
‘Let whoever reaches there first buy the tickets.’
(Steever 1988: 30)
Steever (1988: 30) glosses -ō as ‘or,’ which is one of the functions that the
dubitative marker performs in several Dravidian languages.
Telugu (DR)
(19) [ēdi kāwāl(i)- ō] adi pat.t.u- ku- pō
what be-wanted- comp that take- VR go-imp-s
‘Take away what you want.’
Krishnamurti (2003: 448)
Post-nominal relative clauses (NP-adjoined or CP-adjoined) are not permitted

in Dravidian, in contrast to in Indo-Aryan. That is, the relative clause in a
relative-correlative construction cannot be ‘extraposed.’
cp-adjoined (‘extraposed’) – not permitted in dravidian

Telugu (DR)
(20) *adi pat.t.u- ku- pō [ēdi kāwāl(i)- ō]
that take- VR go-imp-s what be-wanted- comp
‘Take away what you want.’
All positions of the NPAH are relativizable in Dravidian languages, just as in

Indo-Aryan.
8.3.3 A brief summary of relativizable positions in SALs

We shall discuss in appendix 8.1 on the Website the positions in which the rela-
tive clause occurs in Tibeto-Burman, Munda and Mon-Khmer (Khasi) language
families. We shall now present a brief summary of the relativizable positions
in SALs, abstracted from the data presented above and from appendix 8.1 on
the Website.
(i) As we have shown above, all the Indo-Aryan languages except colloquial
Sinhala permit a finite relative-correlative clause and a right-adjoined and
extraposed relative clause in all the positions of the embedded relative on
the Noun Phrase Accessibility Hierarchy of Keenan and Comrie (1977).
All relative pronouns in Indo-Aryan languages start with j -, and are
different in form from question words, which start with k- in Indo-Aryan
languages.
(ii) There is no distinct form of the relative pronoun as such in Dravidian,
and it is the question word that is used as a relative pronoun, just as
in English and French. Amongst the Tibeto-Burman languages that we
know of, we find that Tenyidie and Sema have indigenous relative pro-
nouns (see sentences (4) and (5) in appendix 8.1 on the Website). Bodo
and Rabha (TB) have the j-type relative pronoun borrowed from Indo-
Aryan (see sentences (1)–(3) in appendix 8.1). In Munda languages,
either a relative pronoun borrowed from Indo-Aryan or the indigenous
question word may be used as a relative pronoun (see sentences (7)–(8) in
appendix 8.1).
(iii) The finite embedded verb in a sentential relative clause carries the ques-
tion/dubitative marker -ō in Dravidian languages, and it can be optionally
dropped in Malayalam (Asher and Kumari 1997: 53) (cf. sentences (2b)
and (2c) in appendix 8.5 on the Website). The marker -ō cannot be deleted
in Tamil, Telugu and Kannada. Tibeto-Burman languages (Tenyidie and
Sema) too use a similar marker, and it cannot be deleted either. No such
marker is found in Indo-Aryan and Munda.
Further, transplanted varieties of Indo-Aryan languages (Dakkhini
(IA), Southern (Mangalore) Konkani (IA) and Bhalavali Bhasha (IA)
in the southern parts of India) also have a marker similar to -ō of
Dravidian.
(iv) The sentential relative clause is used in Tenyidie and Sema (TB) only
when the head is [−definite].
(v) The embedded relative clause in a relative-correlative construction is non-
extraposable in Indo-Aryan, Dravidian, Munda and Tibeto-Burman, in
contrast to the NP [CP-] type.
(vi) The correlative pronoun cannot be dropped in Dravidian and Tibeto-
Burman, while in Ho (Munda) it can be dropped.
8.4 Externally Headed Relative Clause 277
(vii) In Munda languages, the head noun may occur in the embedded as well
as matrix clauses.13
(viii) The relative pronoun in Khasi is ba-, and ba- is used as an adjectivalizer
as well as a complementizer. Following Temsen (2006) we treat ba- as
a relative pronoun. In this role, ba carries the agreement marker of the
head that it modifies (see examples (11)–(18) in appendix 8.1).
8.3.4 Asymmetries in relative clauses in Indo-Aryan

Recall that it is only in Indo-Aryan languages that we have three types of relative
clauses. Some asymmetries are found in the three types of relative clauses in
Hindi-Urdu that were discussed in Subbarao (1974, 1984a), Srivastav (1991b)
and Bhatt (2003). They concern:
(i) the occurrence of the head in the relative clause as well as the main clause;
(ii) a restriction pertaining to the feature of definiteness of the head in the
preposed relative-correlative clauses and post-nominal relative clauses;
(iii) the non-occurrence of bare NPs in relative-correlative clauses; and
(iv) the permissibility of multiple relativization in relative-correlative clauses,
and its non-permissibility in NP-adjoined and postposed relatives (see
appendix 8.2 on the Website for further discussion).
In this section, we have alluded to the asymmetries found in the various types
of relative clauses in Hindi-Urdu. We discuss elsewhere two crucial issues that
are discussed in current literature concerning relative clauses in Dravidian:
(i) the origin and occurrence of relative clauses; and
(ii) the Strict OV hypothesis.
See appendix 8.5 for details.
In section 8.4 below, we shall discuss the nature of the EHRC.
8.4 The Externally Headed Relative Clause (non-finite) (the EHRC)

In this type of an Externally Headed Relative Clause (EHRC), there is a
null (covert) operator that occurs in the embedded clause. Since the rela-
tivized/modified head noun in the embedded clause is not overtly present and
since the head occurs only in the matrix clause, it may be labeled as a non-
finite (participial/infinitival) Externally Headed Relative Clause (EHRC). In
contrast, if the head is not overtly present in the matrix clause, and it occurs
in the embedded clause, this may be labeled as a Non-finite Internally Headed
Relative Clause (IHRC).14
We shall demonstrate that a crucial feature of EHRCs is: the postposition
indicating the grammatical function of the null variable in relation to the
embedded verb is absent universally. Since it is the postposition that denotes
the grammatical function that a noun phrase performs in a sentence, crucial
information with regard to the grammatical function of the noun phrase vis-à-
vis the embedded predicate is not overtly manifested. Hence, it may lead to an
ambiguous interpretation in some EHRCs in SALs. The EHRC is found in all
the four language families of the subcontinent, and it is the preferred strategy
for noun modification in Dravidian, Munda and Tibeto-Burman. In fact, in
almost all TB languages (for example, Hmar, Mizo, Manipuri, Zou, Thadou
and Paite), the EHRC/IHRC is the only strategy that is available. It appears
that the EHRC is the only indigenous strategy in Munda languages, and the
sentential relative is borrowed from Indo-Aryan.15
IHRCs, generally found in verb-final languages, are available exclusively
in some of the languages of the Tibeto-Burman language family. The crucial
issues that involve EHRCs and IHRCs are the following:
(i) grammatical positions accessible for relativization;
(ii) (potential) ambiguous interpretation of EHRCs and IHRCs; and
(iii) factors responsible for ambiguity and disambiguation.
We shall now discuss each of these issues.
8.4.1 Grammatical positions accessible in the EHRC

In general, there are no restrictions on the positions relativizable in the EHRC
on the Noun Phrase Accessibility Hierarchy (NPAH) of Keenan and Com-
rie (1977) in Dravidian, Munda and Tibeto-Burman languages, except in the
comitative case in the EHRC, which we shall discuss later. Indo-Aryan freely
permits an EHRC with imperfect participles. It permits perfect participles of the
intransitive subject (selectively, only with certain predicates) and direct object,
as well as modification of possessor and, in a few languages, the indirect object
and oblique objects (OOs). EHRCs modifying non-nominative dative sub-
jects are permitted in Marathi (IA) marginally (Wali 2004: 245) and Gujarati
(Mistry 2004: 20). Perfect participles in Hindi-Urdu and Punjabi cannot mod-
ify ergative transitive subjects / dative subjects. Dravidian languages freely
permit EHRCs modifying non-nominative subjects. We shall discuss relative
clauses in SALs keeping in view the grammatical functions that are accessible
in EHRCs according to Keenan and Comrie’s (1977) NPAH, which we state
below.
The NPAH constraints are:
(i) a language must be able to relativize subjects;
(ii) the strategy of relativization must apply to a continuous segment of gram-
matical functions / constituents of the NPAH-scale;
(iii) strategies that apply at one point of the NPAH-scale may in principle cease
at any point on the scale.
In order to give the reader a sense of the nature of the EHRC in some SALs,
we present here a discussion of the positions relativizable in the EHRC in
Indo-Aryan and Dravidian, while the examples concerning Munda, Mon-

Khmer Khasi and Tibeto-Burman are presented in appendix 8.3 on the Website.
Indo-Aryan
Subject modification
Subject modification of a [+transitive] verb is permitted in Sinhala, just as in
Dravidian, Munda, Tibeto-Burman and the Mon-Khmer Khasi. The modified
head corresponds to the subject in the participial clause. The gapped subject is
indicated by ø.
Sinhala (IA)
(21) [øi ē wæd.a kərəpu] minissui
that work do.pple men
‘The men who did the work.’
Ergative subject modification: The modified head corresponds to the subject

in the participial clause. Such modification is not permitted in Hindi-Urdu
(except with a limited set of predicates). The verb par.h ‘to study,’ for example,
is ergative case-marked in perfective clauses in Hindi-Urdu (IA) and, hence,
we label the subject of such verbs as an ergative subject. The gapped subject is
indicated by ø.
Hindi-Urdu (IA)
(22) [*øi kitāb par.h- ā huā] lar.kāi
book read- perf pple boy
Literally: ‘The book read boy.’
(See Subbarao 1984a for
a detailed discussion)16
[−transitive] verb modifying subject

The modified head corresponds to the subject of a [–transitive] verb in the
participial clause. The null (covert) operator is indicted by ø.
(23) [øi kone mẽ bait.h- ı̄ huı̄] pyārı̄ baccı̄i

corner in sit- perf pple cute girl
‘The cute girl sitting in the corner.’
There are some [–transitive] verbs kūdnā ‘to jump,’ daur.nā ‘to run,’ tairnā ‘to
swim’ in Hindi-Urdu (IA), which do not permit a perfect participle to modify
their subject “unless the predicate is telic, with an end point represented” (Alice
Davison, p.c.). For details, see Kachru (1980) and Subbarao (1984a).
Dative subject modification

In SALs, many predicates require a dative case-marked subject (see chapter 5).
Such predicates permit the modification of the dative subject in some languages,
and in some they do not.
In Hindi-Urdu (IA), the modification of the experiencer subject is not
permitted.17
Experiencer subject
The modified head corresponds to the experiencer subject in the participial
clause. The gapped subject which is a null (covert) operator is indicated by ø.
Hindi-Urdu (IA)
(24) [*øi bhūkh lag- ā huā] insāni
dat hunger feel- perf pple person
‘The man who felt hungry.’
In Gujarati (IA) and Marathi (IA), the modification of the dative experiencer
subject is permitted.
Gujarati (IA)
(25) [øi patthar vāgelo] vidyārthii
dat rock struck student
‘The student struck by a rock.’
Marathi (IA)
(26) ?[øi bāhuli āvd.- lel- i] mulgii
dat doll like- perf pple- 3s,f girl.3s,f
Literally: ‘The doll liked girl.’
[‘The girl who liked the doll.’]
(Wali 2004: 245)
Thus, Marathi and Gujarati differ from Hindi-Urdu in permitting the modifica-
tion of a dative subject.
Direct object modification

In direct object modification, there is a lot of variation with regard to the case
marking of the embedded subject in IA languages. In IA languages, when the
embedded object is modified, the embedded subject of a transitive verb may
be either (i) ergative (33), or (ii) genitive ((27), (28a) and (29) for example),
or (iii) nominative case-marked (37). Note that the modified head in Bangla
(27), Hindi-Urdu (28a), Kashmiri (29), Oriya (32), Nepali (35), Sinhala (36)
and Dakkhini (37) belonging to Indo-Aryan, and in (38) in Telugu (DR),
corresponds to the direct object in the participial clause. The gapped direct
object which is a null (covert) operator in all these cases is indicated by ø.
The subject of a transitive verb is genitive case-marked in Bangla, Hindi-
Urdu, Punjabi, Kashmiri and Assamese, which indicates that embedded subject
NP acquires the status of a modifier by losing its nominative/ergative case.18
Embedded subject genitive case-marked

Bangla (IA)
The modified head corresponds to direct object in the participial clause in (27).
The null (covert) operator is indicated by ø.
(27) [ramer øi pɔr.- a] boii
Ram- gen read- perf book
‘The book that Ram read. . . ’
(Anupam Das, p.c.)
In (27), the predicate contains a perfect participle, and the embedded subject
is genitive case-marked. Ergative case marking is not permitted in (28a) in
Hindi-Urdu (IA). Though the sentence has the interpretation of a passive in
(28a), the predicate in (28a) is not in the passive.
Hindi-Urdu (IA)
(28a) [salmā- kı̄/ *ne øi likh- ı̄ huı̄] kitābi
Salma- gen erg write- perf pple book
‘The book written by Salma . . .’ (‘The book that Salma wrote . . . ’)
The theme in a passive sentence can be modified in Hindi-Urdu.

(28b) salmā (ke) dwārā/ se likh.ı̄ gayı̄ cit.t.hı̄ yahā˜ par.ı̄ huı̄ hai
Salma by by write.pass letter here lying is
‘The letter written by Salma is lying here.’
In Kashmiri (IA) too, an identical pattern to Hindi-Urdu, as in (28), obtains.

Thus, Kashmiri patterns with many other Indo-Aryan languages.
Kashmiri (IA)
(29) [šı̄la- hnd/*y øi chəl’-mt’] palavi
Sheila- gen/erg wash-perf pple clothes
Literally: ‘Sheila’s washed clothes.’
‘The clothes that Sheila washed.’
(Wali and Koul 1997: 65; Meenakshi Koul
and Aadil Kak, p.c.)
In Hindi-Urdu (IA), in some cases the theme can be modified (30a), if the
embedded predicate is dative case-marked, and in some cases it cannot be (30b),
while in Marathi (31) and Konkani (IA), this is permitted. The predicates cot.
lagnā ‘injury get [literally]’ and bhūkh lagnā ‘appetite feel [literal meaning]’ in
Hindi-Urdu (IA) take a dative case-marked subject, and the theme is cot. ‘injury’
in (30a) and bhūkh ‘appetite’ in (30b). While the former theme in (30a) can be
modified, the latter in (30b) cannot.19 As we shall see later, all the Dravidian
languages freely permit the modification of any non-nominative subject or the
theme, except the passive agent.
The modified head corresponds to the theme in the participial clause in (30a)
and (30b). The null theme is indicated by ø.
t h e m e 20
Hindi-Urdu (IA)
(30a) [rām ko øi lagı̄] cot.i bahut gahrı̄ hai
Ram dat get hurt.perf pple injury very deep is
‘The injury that Ram had is quite deep.’
(30b) *[us bacce ko øi lag- ı̄ ( huı̄)] bhūkhi tez thı̄

that child dat feel- perf pple hunger sharp was
‘The hunger (appetite) which the child felt was sharp.’
In Marathi (IA) in (31), the predicate āvd. ‘like’ takes a dative case-marked
subject and the theme is bāhuli ‘doll’.
Marathi (IA)
(31) [mulgi- lā øi āvd.- lel- i] bāhulii
girl.3s,f- dat like- perf pple- 3s,f doll,3s,f
‘The doll which the girl liked.’
(Wali 2004: 245)
That is, while Marathi (IA) permits the modification of theme in a dative
case-marked sentence, Hindi-Urdu (IA) does it selectively.
Oriya (IA) differs from its east Indo-Aryan neighbor Bangla (IA). The subject
of the embedded sentence is nominative case-marked and it cannot be genitive
case-marked, as in (32). Recall that genitive case marking is permitted in
Bangla, Hindi-Urdu and Punjabi. We shall show that in Dravidian languages, the
embedded subject is nominative case-marked in such cases. Thus, Oriya does
not retain the Indo-Aryan pattern of the genitive case-marked subject, and it
appears to have borrowed the Dravidian pattern of the nominative case-marked
subject, as Dravidian permits such case marking. The nominative marker in
Oriya is null.
Embedded subject nominative/genitive case-marked

The modified head corresponds to the direct object in the participial clause in
(32) in Oriya (IA). The gapped direct object occurs as a covert operator which
is indicated by ø.
Oriya (IA)
(32) [radha øi pɔd.hi thibə] bɔhii
Radha (nom) read.perf pple book
‘The book that Radha read . . . ’
(Prakash Patnaik, Subha Lakshmi Das
and Ahalya Das, p.c.)
Marathi (IA) exhibits two distinct patterns of case marking of the embedded
subject in participial EHRCs: it is either ergative case-marked or genitive case-
marked, with slightly different aspectual interpretations. The theme in (33) is
pustak ‘book.’ The embedded verb is in the perfect participial form.
The modified head corresponds to the direct object in the participial clause in
(33) and (34). The gapped direct object which is a covert operator is indicated
by ø.
Embedded subject ergative case-marked

Marathi (IA)21
(33) [ram- ne øi wāts- lel.a] pustaki
Ram- erg read- perf pple.neut book.neuter
‘The book that Ram read . . . ’
(a simple past reading)
(Prashant Pardeshi, p.c.)
When the embedded subject is genitive case-marked, it is the conjunctive

participial form of the embedded verb that is used. Sentence (34) differs from
(33) in aspectual interpretation.
Embedded subject genitive case-marked

(34) [ram- tsa øi wāts- un dzā- lel.a] pustaki
Ram- gen read- cpm become perf pple.neut book.neut
‘The book that Ram has finished reading . . . ’
(a perfective reading)
(Prashant Pardeshi, p.c.)
Thus, Marathi shares with Bangla, Hindi-Urdu and Punjabi the feature of having
the genitive with the embedded subject (34), but it also may have the ergative
case marking on it, as in (33).
In Nepali (IA), the subject məy ‘I’ is ergative case-marked by le as in Marathi

(IA). The modified head corresponds to the direct object in the participial clause
in (35)–(38) in Nepali, Sinhala, Dakkhini (IA) and Telugu (DR), respectively.
The gapped direct object (a covert operator) is indicated by ø.22
Nepali (IA)
(35) [məy- le hijo øi gare-ko] kāmi
I- erg yesterday do-perf pple work
Literally: ‘the (I did yesterday) work . . . ’
‘The work which I did yesterday . . . ’
(Clark 1977/1989: 185)
In Sinhala (IA), like in Dravidian, the genitive does not occur, just as in Oriya
(IA).
Sinhala (IA)
(36) [siri gunəpālat.ə øi dunnə] potəi
Siri (nom) Gunapala give.pst.adjective book
‘The book that Siri gave Gunapala.’
It is worth noting that in Dravidian languages the embedded subject does not
carry the genitive case marker. Synchronic evidence to show that contact may
induce such change comes from Dakkhini Hindi-Urdu (IA). Dakkhini Hindi-
Urdu, a transplanted variety of Hindi-Urdu (IA) in the southern parts of India,
which has been in intense contact with Telugu (DR) too, does not have the
genitive with the embedded subject, and nor does Telugu have the genitive in
such cases. Data from Dakkhini and Telugu are illustrative.
Dakkhini (IA)
(37) [baccā øi kar- e so] kāmi
boy do- perf pple adjr work
‘The work that the boy did . . . ’
(Arora 2004: 134)
Telugu (DR)
(38) rāmud.u øi cadiv- in- a pustakami
Ram read- pst- adjr book
‘The book that Ram read . . . ’
Thus, the non-occurrence of the genitive with embedded subject in Oriya and
Sinhala (IA) could plausibly be due to Dravidian influence, just as in Dakkhini,
as Oriya and Dakkhini have been in intense contact with Telugu (DR) and
Tamil (DR) with Sinhala.
We have discussed case marking on the subject of the embedded clause in

direct object modification in IA languages. We shall now consider the modifi-
cation of an NP in the other positions of the NPAH.
Indirect object modification

Indirect object modification is not permitted in Hindi-Urdu, Punjabi and Kash-
miri with the embedded verb ‘give.’23
The words kur.ı̄ ‘girl’ in (39) and kūr ‘girl’ in (40) are the indirect objects of
the predicate ‘give’ in Punjabi and Kashmiri.
The modified head corresponds to the indirect object in the participial clause.
The gapped IO (a covert operator) is indicated by ø.
Punjabi (IA)
(39) *øi katāb dittı̄ (hoı̄) kur.ı̄i
book give.perf pple.f.s girl.f.s.
‘The girl to whom the book was given. . . ’
(Bhatia 1993: 61)
Kashmiri (IA)
(40) *rajan øi gər ditsmts kūri
Raj.erg watch give.perf.pple girl
Literally: ‘Raj watch-given girl.’
(‘The girl to whom Raj gave the watch.’)
In contrast, Marathi, Nepali and Sinhala (IA) permit IO modification.

In (41) and (42), mān.sa ‘people’ in Marathi and gunəpālə ‘Gunapala’ in
Sinhala are the indirect objects of the predicates di ‘give’ and du ‘give,’ respec-
tively.
The modified head corresponds to the indirect object in the participial clause
in (41) and (42). The gapped IO (a covert operator) is indicated by ø.
Marathi (IA)
(41) tyā- ne øi āmantran. patrikā dilelı̄ sagl.ı̄ mān.sai
he- erg invitation card give.pst pple all.p.neuter people.p.m
lagnālā ālı̄
wedding.that come.pst.3p.m
‘All the people, whom he had sent (given) invitation cards, had come to
the wedding.’
Sinhala (IA)
(42) [siri øi potə dunnə gunəpāləi ]
Siri book.def give.pst.adjr Gunapala
‘(the) Gunapala to whom Siri gave the book.’
Oblique object modification

Hindi-Urdu, Punjabi, Kashmiri and Bengali (IA) do not permit the modification
of oblique objects.24 In all the examples concerning the modification of the
oblique object, the modified head corresponds to the oblique object in the
participial clause. The gapped PP (a covert operator) is indicated by ø.
In contrast, Marathi (IA), like the Dravidian languages, permits the modi-
fication of the locative phrase in (43), the instrumental phrase in (44) and the
time adverbial in (45) (Pandharipande 1997: 90–91). The modified head corre-
sponds to the locative PP in the participial clause in (43), the instrumental PP
in (44) and the time adverbial in (45). The gapped PP is indicated by ø.
Marathi (IA)
locative
(43) mı̄ øi rāhāt as.lel.a ghari khūp dzuna āhe
I live.pres be.pst.pple.s,n house.s.neut very old is
‘The house in which I am living is very old.’
instrumental
(44) mı̄ øi patra lihilelı̄ pensı̄li mād.ı̄wartšyā kapāt.āt āhe
I letter write pst pple.s.f. pencil.s.f. upstairs of closet.loc is
‘The pencil with which I wrote the letter is in the closet upstairs.’
time adverbial
(45) raghū øi dzanmalelyā wars.ı̄i wad.ilã˜ nnā khūp
Raghu born.pst pple.oblique year.loc father.dat a lot
paisa mil.ālā
money receive.pst.s.m.
‘The father got a lot of money the year Raghu was born.’
The modification of the comitative is not permitted in any Indo-Aryan language.

The modified head corresponds to the comitative PP in the participial clause in
(46), which is ungrammatical. The gapped PP is indicated by ø.
comitative
(46) *[tsor øi tšyā barobar ālelā] polı̄s
thief poss with come.pst pple.3s. m policeman
‘The policeman who the thief came with.’
(Pandharipande 1997: 91)25
Oriya (IA) too permits the modification of the oblique object except the comi-
tative, just like Telugu, in particular, and Dravidian languages in general. The
modified head corresponds to the locative PP in the participial clause in (47)
and the comitative PP in (48). The gapped PP is indicated by ø.
Oriya (IA)
locative
(47) [mũ øi khai thiba] patrai
I eat. perf pple leaf
‘The leaf on which I ate’ (locative PP modification)
‘The leaf which I ate’ (DO modification)
comitative
(48) *[kɔmala øi asi.thiba] lokoi
Kamala come.perf pple person
‘The person with whom Kamala came.’
(Prakash Patnaik, p.c.)
Nepali (IA), too, permits the modification of the oblique object (locative, instru-
mental and ablative) (Pawan Upadhyaya, p.c.).
Nepali (IA), like the Dravidian languages, also permits the modification of
adverbial phrases, destination in (49) and time in (50). The modified head
corresponds to the PP indicating destination in the participial clause in (49),
the time adverbial phrase in (50) in Nepali. The gapped PP is indicated by ø.
Nepali (IA)
destination
(49) [timi hijo øi gae- ko] ghari
you yesterday go- perf pple home
Literally: ‘The (you went yesterday) house . . . ’
‘The house you went to yesterday . . . ’
(Clark 1977/1989: 185)
time
(50) [ma hijo øi ae- ko] belāi
I yesterday come- perf pple time
Literally: ‘The (I came yesterday) time . . . ’
‘The time at which I came yesterday . . . ’
(Clark 1977/1989: 185)
Thus, Marathi, Oriya and Nepali differ from Hindi-Urdu, Punjabi and Kashmiri
in permitting the modification of oblique objects.
The modification of the comitative NP is not permitted in any IA language.
We provide an explanation in 8.6, in terms of the Thematic Eligibility Con-
dition, of why the comitative PP cannot head an EHRC.
Modification of possessor
Possessors can be modified in Hindi-Urdu and Bangla (IA). The embedded
verb kat.nā in Hindi-Urdu and kat. in Bangla (‘to be cut’) is [−transitive]. In
Hindi-Urdu, as Rajesh Bhatt (p.c.) points out, only unaccusative predicates
such as kat.nā ‘to be cut,’ .tūt.nā ‘to be broken,’ can form a participle, and the
possessed NP is an argument of an unaccusative predicate.
In Bangla too a similar restriction holds (Shukla Basu, p.c.). Sentences (51)
and (52) are illustrative.
Hindi-Urdu (IA)
(51) øi dum kat.ı̄ billii
tail cut.perf pple cat
‘The cat whose tail got cut.’
(Subbarao 1984b)
Bangla (IA)
(52) øi læj kat.a ber.ali
tail cut.perf pple cat
‘The cat whose tail got cut.’
Recall that in Hindi-Urdu, while the subject – with a specific set of verbs – and
direct object can be modified, the oblique object cannot be. A similar situation
obtains in some other IA languages too. The fact that the object of the genitive
in Hindi-Urdu can be modified demonstrates that there are breaks in the NPAH,
and it is not a continuum as Keenan and Comrie (1977) claimed.26 We have
shown the positions that are accessible in IA languages and the variation that is
found within the same language family with regard to the positions accessible.
We shall now show the positions accessible for relativization in the Dravidian
language family.
Dravidian
In Dravidian languages, all the positions of the NPAH can be modified. We
provide only a few examples.
In Dravidian languages, the preferred strategy is the participial-type, and the
relative-correlative clause occurs either in formal style or in free relatives. In
most Indo-Aryan languages, both the strategies are available, and it is rather
difficult to pinpoint which strategy is more frequently used in a language, as

there are no studies available in terms of frequency counts.27 The participle in
Dravidian languages carries the finite past tense marker. The marker -a occurs
to the right of the finite tense marker. It is this marker that turns a finite verb to
an adjectival participle. Hence, we label it as adjectivalizer.
Subject modification
In Dravidian languages, there are no restrictions on the transitive nature of the
predicate of the participle in EHRCs. Thus, both [+transitive] and [–transitive]
predicates freely permit the modification of the subject.28 The modified head
corresponds to the subject in the participial clause. The gapped subject is
indicated by ø.
Telugu (DR)
(53) [ivāl.a proddunna øi gōd.a gent- in- a] pillai
today morning wall jump- pst- adjr girl
Literally: ‘Today morning wall-jumped girl . . . ’
‘The girl who jumped the wall this morning . . . ’
In Malayalam (DR), as in other Dravidian languages, “a relative

participle . . . can take the same range of arguments as a finite verb in a sim-
ple sentence. A noun can, in principle, be modified by a plurality of relative
clauses” (Asher and Kumari 1997: 124).
Malayalam (DR)
(54) [niŋŋal.ōt.ə sammεl.anatte kuriccə øi para
a innale ivit.e
you. with conference.acc about tell.pst pple yesterday here
vanna] āl.əi
come.pst pple man
‘The man who told you about the conference and who came here yesterday.’
(Asher and Kumari 1997)
Direct object modification

When the direct object is modified, the subject of the participial clause is
nominative case-marked in all the Dravidian languages, as in (55).29
The modified head corresponds to the object in the participial clause. The
gapped object is indicated by ø.
Telugu (DR)
(55) [mallika øi cadiv- in- a] pustakami bāg- undi
Mallika (nom) read- pst- adjr book good- is
‘The book that Mallika read is good.’
Indirect object modification

Sentence (56) is a simple clause in Tamil, and (57) is an example of the EHRC
in which the indirect object is the head. The modified head corresponds to the
indirect object in the participial clause. The gapped indirect object is indicated
by ø.
Tamil (DR)
simple sentence
(56) taccan van.n.ān- ukku pan.am kut.uttān
carpenter washerman- dat money give.pst.3 s,m
‘The carpenter gave money to the washerman.’
ehrc – io modified
(57) [S2 taccan øi pan.am kut.utt- aS2 ] [van.n.āni ]
carpenter(nom) money give.pst- rp washerman
[iŋe vantān]
here came.pst.3 s,m
‘The washerman to whom the carpenter gave money [came here].’
(The VP iŋe vantān is added to the original sentence from
Asher 1985: 28)
Asher (1985: 28) uses the term ‘rp’ (relative (adjectival) particle), which is
termed an adjectivalizer in this work.
The following points concerning (57) are worth mentioning: (i) the identical
relativized noun phrase in the embedded clause is absent, and it is indicated
by ø, and (ii) it is coindexed with the head noun (Nadkarni 1970; Subbarao
1984b). The suffix -a is the adjectivalizer.
We shall now discuss the modification of oblique objects. The modified head
corresponds to the oblique object (a PP) in the participial clause in sentences
(58)–(61), (63) and (65). The gapped oblique object is indicated by ø.
Oblique object (OO)

Locative The gapped PP, for example, in (58) is ūru ‘town’ + a locative
postposition. Recall that the locative postposition is never overtly realized in an
EHRC, in contrast to an IHRC in some SALs of the Tibeto-Burman language
family, where it is overtly present.
Kannada (DR)
(58) avanu øi hut.t.id- a ūrui idē
he born.pst- adjr town this.emph
‘This is the town where he was born.’
(Sridhar 1990: 57)
Ablative
Telugu (DR)
(59) [pragati øi nı̄l.l.u tecc- in- a] bāvii
Pragati water bring- pst- adjr well
‘The well from which Pragati brought water.’
Instrumental
Malayalam (DR)
(60) amma øi māŋŋa pūl.iya kattii
mother mango slice.pst.rp knife
‘The knife with which mother sliced the mango.’
A time adverb can head an EHRC in Telugu (DR). The head of the EHRC in
(61) is appud.u, which literally means ‘then.’
Telugu (DR)
(61) mı̄ru øi vacc- in- a appud.ui ceppand.i
you come- pst- adjr then (time) tell.hon
‘Tell me when you come.’
Other time expressions such as tarvāta ‘after,’ dāka ‘until,’ lōpala ‘prior to’ and
mundu ‘before’ can also head an EHRC (Krishnamurti and Gwynn 1985:246) in
Telugu – for example, cēs-in-a tarvāta ‘after having done,’ cēsē mundu ‘before
doing’ (1985: 246). A negative participle may also modify a time expression
such as mundu ‘before’ (1985: 246), as in ceyya.ka mundu ‘before not doing,’
that is, ‘before doing.’ The truth value of the sentences with an affirmative CP
and negative CP is identical.30
Comitative (sociative) PP A comitative (sociative) PP cannot be relativized

in a participial relative, when it imparts the meaning of accompaniment, in
Telugu, Kannada or Malayalam. Sentence (62) is a simple clause and in (63)
the comitative PP is the head.
Malayalam (DR)
(62) bābu ez.uttōt.e vannu
Babu letter.with came
‘Babu came with a letter.’
(63) *bābu vanna ez.ut.t.ə

Babu come.adjr letter
‘The letter with which Babu came.’
(Asher and Kumari 1997: 61)31
However, if the comitative PP is a subcategorized argument, an EHRC is
permitted in all Dravidian languages. Sentence (64) is a simple clause, and in
(65) the comitative PP is the head. In (64) and (65), vastādu ‘wrestler’ is a
subcategorized argument of the conjunct verb kustı̄ pat..tu ‘to wrestle.’
Telugu (DR)
(64) mā snēhitud.u oka vastādu tō kustı̄32 pat.t.ēd.u-
our friend one wrestler with (comitative) wrestling (noun) caught
‘Our friend wrestled with a wrestler.’
(65) mā snēhitud.u øi kustı̄ pat.t.- in- a vastādui

our friend wrestling (noun) catch- pst- adjr wrestler
itan(u)- ē
he.prox emph
‘He (proximate) is the wrestler (emphatic) with whom our friend wrestled.’
We discuss the issues related to the comitative PP as head in section 8.6.
Non-nominative subjects as head Dravidian freely permits EHRCs in which

a non-nominative subject can be the gapped NP. This is in contrast to most
of the Indo-Aryan languages in which a non-nominative subject cannot occur
as a gapped noun phrase. The modified gapped head corresponds to the non-
nominative subject in the participial clause. The gapped non-nominative subject
is indicated by ø.
The predicate višakku ‘feel hungry’ takes a dative subject in a simple clause
in Malayalam. Sentence (66) is an example of the EHRC in which a predicate
modifies a dative subject. As expected, the dative case marker does not show
up in (66), as it is an EHRC.
ehrc: dative subject

(66) [øi višakkunna] kut.t.ii
feel hungry.pres.adjr child
‘The child who feels hungry . . . ’
The locative PP in (67) is tand.ri ‘father’ + locative postposition, which is

gapped and is indicated by ø.
ehrc: locative subject

Telugu (DR)
(67) [prastutam øi mūd.u lakšalu āsti unn- a] tand.rii
at present three lakhs property be- adjr father
‘The father who has a property of three lakhs at present . . . ’
A predicate that permits a dative or a locative subject can modify patient/theme

of the clause, as in (68) and (69), respectively. The modified head corresponds
to the patient / theme in the participial clause. The gapped patient is indicated
by ø.
e h r c : p a t i e n t m o d i fi e d i n a d a t i v e s u b j e c t c o n s t r u c t i o n
Malayalam (DR)
(68) [aval.kkə øi karun.a tōnniya] pāvappet.t.avati
she.dat pity feel.pst.rp poor people
‘The poor people for whom she felt pity . . . ’
e h r c : n o m i n a t i v e t h e m e m o d i f i e d w i t h a no m i n a t i v e l o c a t i v e
sub j e c t
Telugu (DR)
(69) [kavita daggira øi unn- a] mugguru panivāl.l.ui
Kavita near be- adjr three workers
‘The three workers that Kavita has . . . ’
It is pertinent to mention here that Hindi-Urdu (IA) does not permit the modi-
fication of the subject (experiencer) in (70), or the theme in the dative subject
construction in (71).33
The modified head corresponds to the experiencer in the participial clause in
(70). The gapped experiencer is indicated by ø.
n o n - n o m i n a t i v e ( d a t i v e ) s u b j e c t m o d i f i e d – no t p e r m i t t e d
experiencer modified
Hindi-Urdu (IA)
(70) [*øi bhūk lag- ı̄ huı̄] baccı̄i
appetite feel- perf pple female child
Literally: ‘The hunger-felt female child.’
‘The female child who felt hungry.’
The modified head corresponds to the theme in the participial clause in (71).
The gapped theme is indicated by ø.
theme modified
(71) [*baccı̄i ko øi lag- ı̄ (huı̄)] bhūk
female child dat feel- perf pple appetite
‘The appetite felt by the female child.’
In this section, we have presented a detailed description of examples of the

EHRC in which the modified head corresponds to different grammatical func-
tions that it performs in the participial clause in Indo-Aryan and Dravidian
languages.
In appendix 8.3, we discuss the case of the EHRCs in Munda, Mon-Khmer

and Tibeto-Burman languages. (See the Website for further details.)
We shall now mention some salient points of EHRCs in SALs with regard
to the accessible positions for relativization in accordance with Keenan and
Comrie’s NPAH.
(i) Subject modification: While subject modification of a transitive verb
is permitted in Sinhala (IA), Hindi-Urdu (IA) does not permit it with
[+transitive] ergative and many intransitive and dative predicates. The
modification of the dative subject is permitted in Marathi, Gujarati (IA)
and in Dravidian. Further, in Dravidian, Tibeto-Burman and Khasi (Mon-
Khmer) and the Munda languages, the modification of [+transitive] as
well as [−transitive] subjects is freely permitted.
(ii) DO modification: When the embedded object is modified, the embed-
ded subject may be nominative/ergative/genitive case-marked, which is
language-specific in Indo-Aryan languages, while it is nominative case-
marked in the other language families. In Sinhala, Oriya and Dakkhini
(IA), the subject is nominative case-marked, just as in Dravidian.
Hmar, Mizo and Thadou (TB) exhibit two distinct patterns, in which
the embedded subject is either ergative case-marked, and consequently
the embedded verb carries a subject agreement marker (as in (21) in
appendix 8.3, in Mizo), or the embedded subject is null case-marked, and
consequently the embedded verb does not carry any agreement marker at
all (as in (22) in appendix 8.3, in Mizo). In Kharia (Munda), the subject
may be nominative or genitive case-marked as in (3) in appendix 8.3. (See
also, section 4.9.)
(iii) Indirect Object: The modification of the indirect object is not permitted in
Hindi-Urdu, Punjabi and Kashmiri (IA), while it is permitted in Marathi
and Sinhala (IA). It is permitted in all other language families.
(iv) Oblique object modification: This is not permitted in Hindi-Urdu,
Punjabi and Kashmiri (IA), while it is permitted in Marathi (IA). The
other language families freely permit such modification. The modification
of the comitative as head is ‘restricted’ in all the language families, but it
is permitted under specific conditions (see section 8.6).
(v) Possessor modification: This is permitted in all languages, although in
a tightly circumscribed set of environments in Indo-Aryan languages.
Such modification in some Indo-Aryan languages shows that the NPAH
proposed in Keenan and Comrie (1977) is violated.
8.4.2 Asymmetries
In appendix 8.2 on the Website we discuss the asymmetries found in various
types of sentential relative-correlative clauses with regard to double or single
8.5 Internally Headed Relative Clause 295
occurrence of the head, the feature of definiteness on heads, and the occurrence
of bare NPs. In appendix 8.4, also on the Website, we discuss the asymme-
tries found in Tenyidie (TB) with regard to the position of occurrence of the
embedded relative when a quantifier or numeral has the head noun in its scope
in EHRCs.
8.5 The Internally Headed Relative Clause (the IHRC)

In SALs, it is only in the Tibeto-Burman language family that IHRCs are found.
According to Cole (1987), IHRCs are present only in those languages whose
basic word order is SOV. Gill (2000) points out that IHRCs may also occur in
languages with SVO order and presents evidence from Riau Indonesian (see
Gill 2000 for further details).
In an IHRC the head NP that is being relativized occurs in the embedded
relative clause, and not in the matrix clause. There is a corresponding null
element in the matrix clause that is coindexed with the internal head.
IHRCs in TB languages fall into two major categories depending upon the
presence or absence of the lexical case marker on the internal head.
Category I (with an overt case marker on the internal head): In TB languages
such as Tenyidie, Sema, Sangtam, Konyak, the case marker overtly occurs with
the internal head. This case marker manifests the case relationship of the internal
head with the predicate.
Category II (with no overt case marker on the internal head): In TB languages
such as Mizo, Hmar,34 Manipuri, the case marker does not occur with the head
in an IHRC, just as in EHRCs universally.
An example of an IHRC of Category I is (72). The head of the IHRC and the
postposition it occurs with are in italics.
Note that the head kutari-pie ‘with the knife’ in Tenyidie (TB) occurs in the
embedded clause, and the matrix sentence carries no head in (72), indicated
by ø.
Tenyidie (TB)
(72) [nɔ kutarii - pie nhasi le- ke-] c- øi u a- se
you knife- instr fruit cut- nozr- dm- def big- very
‘The knife with which you cut the fruit is very big.’
An example of category II is found in Mizo (TB). Note that in (73b) in Mizo

(TB), (i) the head noun lekhabu ‘book’ occurs in the embedded clause to
the left of the embedded verb, and does not occur in the matrix clause, and
(ii) the embedded subject Zovi carries the ergative case marker, as the embedded
verb lei ‘buy’ is [+transitive], and the embedded verb carries the 3rd person
singular subject agreement marker a-. The head NP lekhabu ‘book’ does not
permit the definite marker cu to occur (Lalremzami Chinzah, p.c.) as the internal
head is already [+definite]. In a simple sentence, cu, the definite marker /
accusative case marker, occurs with the direct object, as in (73a).35
Mizo (TB)
(73a) zovi- n lekhabu cu a- lei
Zovi- erg book def 3s- buy
‘Zovi bought a book.’
d i r e c t o b j e c t m o d i fi c a t i o n
(73b) zova- n [S2 zovi- n lekhabu a- leiS2 ] kha a- chiar- aŋ
Zova- erg Zovi- erg book 3s- buy det 3s- read- fut
‘Zova will read the read book which Zovi bought.’
Just as in EHRCs (sentences (21) and (22) in appendix 8.3), there is another
alternative for (73b), as in (74). In (74) the embedded subject does not carry
ergative case, and consequently the embedded verb does not have agreement
markers. We have marked their absence by ø in S2 in (74).
Mizo (TB)
(74) zova- n [S2 zovi-ø lekhabu ø- leiS2 ] kha a- chiar- aŋ
Zova- erg Zovi- ø book ø- buy det 3s- read- fut
‘Zova will read the book which Zovi bought.’
Indirect object
Category I: IHRC with an overt case marker on the IO, the internal head. The
indirect object as head, and the postposition, are in italics.
i o a s h e a d o f a n i h r c (to- d a t i v e )
Sangtam (TB)
(75) [n- n nistar (IO) khaŋ šeti t.hraʔ- ba- tsə] khataŋ tšŋle
you- nom person to letter write- nozr- def very tall
‘The person you wrote a letter to is very tall.’
Category II: IHRC with no overt case marker on the IO, the internal head In
Hmar (TB) the indirect object is case-marked by the dative case marker hnenaʔ
‘to, for’ in a simple clause as in (76).
a simple clause with ditransitive verb-indirect object with

hnenaʔ ‘to, for’
Hmar (TB)
(76) lala- n bɔŋ (IO) hnenaʔ bu a- pek
Lala- erg cow to food 3s- give
‘Lala gave food to the cow.’
8.5 Internally Headed Relative Clause 297
Sentences (77a) and (77b) are examples of a left-adjoined IHRC in which the
internal head is an indirect object. Note that in (77a), (i) the head NP bɔŋ ‘cow’
occurs in the embedded clause as an internal head, (ii) the dative case marker
hnenaʔ ‘to, for’ is not present with the IO, and (iii) the ergative marker -n with
the embedded subject is overtly present, as the embedded verb pek ‘give’ is
[+transitive], and hence, a-, the 3rd person singular agreement marker with the
verb too is overtly present. Note that pro in (77a) and (77b) is obligatorily null
and it is coindexed with bɔŋ ‘cow.’
indirect object modified

IHRC - Pattern I
(77a) [lala- n bɔŋi (IO) bu a- pek] kha proi zanikhan ai - thi
Lala- erg cow food 3s- give det yesterday 3s- died
(Kumar 1998)
In Mizo, Hmar, Thadou and Paite IHRCs, the empty position in the matrix clause
coindexed with the internal head in an IHRC is pro as the matrix verb/adjective
exhibits agreement in number and person with the pro. The 3rd person singular
agreement marker -a on the matrix verb thi ‘died’ shows that the deleted
element is pro which is obligatorily null.
Kumar (1998) observes that, in (77b), the absence of both of them (the
ergative marker with the embedded subject and the agreement marker with the
embedded verb) in the IHRC results in ungrammaticality.
i h r c p a t t e r n i i – no t p e r m i t t e d
(77b) *[lala- ø bɔŋi bu ø- pek] kha proi zanikhan a- thi
Lala- erg cow food 3s give det yesterday 3s- died
Thus, in Hmar (TB), the overt presence of the ergative marker and the
agreement marker in the embedded clause in an IHRC with IO as head is
required, in contrast to in Mizo, where there is an option with regard to their
occurrence.
Oblique object modification

There are two categories of languages with regard to oblique object modifica-
tion: either (i) the internal head must carry the postposition that indicates the
grammatical function of the head, or (ii) the internal head does not carry the
postposition that indicates the grammatical function of the head.
Category I: We shall present examples from those languages in which the
head in the IHRC carries a case marker.
OO as head of an IHRC (instrumental)

The instrumental case marker phasi ‘with’ occurs to the right of the internal
head mit ‘knife’ in Konyak (TB).
Konyak (TB)
(78) naŋ- e mit- phasi peleak watne- con- pu- a yəʔɔŋe lonaŋ ke
you- nom knife- instr fruit cut- use- nozr- def very sharp
‘The knife with which you cut fruit is very sharp.’
OO as head of an IHRC (locative)

The locative case marker -la ‘on’ occurs to the right of the internal head tebəl
‘table’ in Sangtam (TB).
Sangtam (TB)
(79) n- n tebəl la šti keʔp- ba tsə siŋ- nə klpcho
you- nom table on book keep- nozr def wood- with made
‘The table on which you kept the book is made of wood.’
OO as head of an IHRC (ablative)

The case of the ablative in Sema (TB) deserves special mention. In Sema
(Subbarao and Kevichüsa 2005), an IHRC occurs in direct object, indirect
object and oblique object (except the comitative) positions. An ablative PP
can head an IHRC if and only if the internal head is partially repeated as the
external head as in (80). This is a unique phenomenon we found only in Sema,
when the ablative PP heads the IHRC.
It may be noted that the generic possession marker (gpm) is used in Sema
(Subbarao and Kevichüsa 2005), and in many Tibeto-Burman languages, with
kinship terms and possessions (such as ‘well’ or ‘home,’ etc.) which are close
or intimate to the possessor. The generic possession marker in Sema is a-. The
expression for well is zkhikhi. Hence, it carries the marker a- in (80).
In sentence (80), a-zkhikhi ‘well’ occurs only in the embedded clause, and
it has the interpretation of DO as Head of the IHRC. Thus, it imparts the
interpretation that ‘the water is dirty,’ and not ‘the well is dirty.’
do as head of the ihrc
Sema (TB)
(80) nɔ- nɔ a-zkhikhi lɔnɔ az se- keu ti- ye
you- [+tr] gpm-well from water brought- nozr that- [−tr mkr]
mit.he mɔ
clean neg
‘*The well from which you brought the water is dirty.’
‘The water which you brought from the well is dirty.’
(Subbarao and Kevichüsa 2005: 260)
In (80), the NP azkhikhi ‘well’ occurs with an ablative case marker lɔnɔ ‘from.’
Still it cannot head the IHRC, as the PP azkhikhi- lɔnɔ ‘well- from’ occurs only
in the embedded relative clause. Thus, it is the single occurrence of the PP in
the embedded clause alone that rules out the interpretation with the ablative PP
as head, in spite of fulfilling the requirements of both case and word order for it
to be the head. However, the DO az ‘water’ or azkhikhi ‘well’ can potentially
be the heads of the internally headed relative clause; the DO is interpreted as
the head in (80), and not the ablative PP azkhikhi lɔnɔ ‘well from.’
To make an ablative PP the head of an IHRC, there is a specific strategy
that Sema adopts. Under this strategy, the head noun is partially repeated in
the matrix clause. It occurs to the right of the definite marker -u in a position
earmarked for the head noun in an externally headed relative clause. Sentence
(81) is illustrative.
ablative as head of the ihrc

(81) nɔ- nɔ a- zkhikhi lɔnɔ az se- keu zkhikhi
you- [+tr] gpm- well from water brought- nozr well
ye mit.he mɔ
[−tr] mkr clean neg
‘The well from which you brought the water is dirty.’
‘*The water which you brought from the well is dirty.’
The repetition of the noun phrase azkhikhi ‘well’ as zkhikhi is only partial, and
a-, the generic possession marker, is not repeated. Thus, partial reduplication
is a syntactic strategy that Sema adopts to distinguish between IHRCs with DO
and ablative PP as head.
In this section, we have discussed IHRCs with and without an overt case
marker on the internal head. We have noted that an externally headed relative
clause is not permitted with comitative PP as head in Dravidian, Indo-Aryan,
Mon-Khmer and Tibeto-Burman languages. The modification of the comi-
tative PP as head in EHRCs and IHRCs deserves a special discussion and,
hence, it is the topic of discussion of the following section. We shall expli-
cate the reasons for the non-permissibility of sentences with comitative PP
as head in EHRCs in most of the languages, and in IHRCs in a subset of
languages.
8.6 Comitative PP as head

We shall now present a discussion of the modification of the comitative PP.
Though there is considerable variation, as we have noticed, with regard
to the positions that are ‘accessible’ in the Noun Phrase Accessibility
Hierarchy (NPAH) of Keenan and Comrie (1977) amongst South Asian lan-
guages of the four language families (Indo-Aryan, Dravidian, Austro-Asiatic
[Mon-Khmer, Munda] and Tibeto-Burman), the modification of the comitative
PP in EHRCs is not permitted in almost all SALs, except for a few. In the few
languages in which it is permitted, specific syntactic criteria have to be met
to enable the comitative PP to head the EHRC. We shall first elaborate what
these criteria are and then attempt to formulate a generalization that would
neatly account for all these phenomena. We shall also propose a condition
that we shall label as the Thematic Eligibility Condition (TEC) to account for
the facts cross-linguistically, and we wish to demonstrate that if any one of
these syntactic specific criteria is fulfilled, then the requirements of the TEC
are met.
We shall now present a discussion of the modification of the comitative PP.
It is worth recalling that the head of all EHRCs and only a subset of IHRCs
does not carry any adposition that would indicate the thematic relationship of
the head with the embedded predicate, whereas a subset of IHRCs in some
Tibeto-Burman languages does carry an overt case marker. We shall show that
the overt occurrence of the postposition plays a crucial role in the subset of
languages which carries it.
There are restrictions on the occurrence of an EHRC and an IHRC in gap
strategy with a comitative PP as head, when it has the interpretation of accom-
paniment. In those languages in which a PP can head an EHRC, the modification
of the other PPs (locative, ablative, instrumental) is freely permitted, but not of
the comitative PP. The issue we wish to address is: why is there this asymmetry
with regard to the comitative PP alone? Can the reasons, if any, be pinpointed?
What are the syntactic phenomena that are involved in enabling a comitative
PP to head an EHRC?
Unlike most other PPs in the NPAH, for the comitative PP to head an EHRC
or an IHRC, specific syntactic criteria have to be met. We shall examine the
criteria below. We shall demonstrate that once any one of the criteria is fulfilled,
a thematic relationship between the head of the comitative PP and the embedded
predicate is established and that is what enables a comitative PP to head an
EHRC.
First, let us look at a few examples of EHRCs in SALs with the comitative
PP as the head.
Dravidian:
Recall that Dravidian languages do freely permit the modification of oblique
objects (locative, ablative and instrumental), but comitative PP as head is not
permitted in any Dravidian language. For example:
e h r c w i t h a c o m i t a t i v e p p a s he a d
Malayalam (DR)
(82) *bābu vanna ez.ut.t.ə
Babu come.adjr letter
‘The letter with which Babu came’
[That is, Babu came bringing a letter with him.]
Tamil (DR)
(83) *kumār vel.iye pōna umā
Kumar out go.pst.adjr Uma
‘Uma, with whom Kumar went out’
Telugu (DR)
(84) *nēnu vel.l.- in- a saral.a
I go- pst- adjr Sarala
‘Sarala with whom I went’
(Ramarao 2003: 79)36
Indo-Aryan:
Indo-Aryan languages such as Hindi-Urdu, Punjabi, Kashmiri and Bangla do

not permit the modification of an oblique PP as head. Though some of them,
such as Konkani, Marathi, Nepali and Oriya (IA), do permit the modification
of oblique objects, none of them permits the modification of the comitative PP
as head. For example:
Hindi-Urdu (IA)
(85) *karun.a gayı̄ huı̄ lar.kı̄
Karuna go.perf pple.f girl
‘The girl with whom Karuna went’
Marathi (IA)37
(86) *karun.a gel- lel.i mulgi
Karuna go- perf pple.f girl
In Nepali (IA), too, the comitative PP cannot head an EHRC.

Munda and Khasi (Mon-Khmer):
We do not have data from Munda languages on the modification of the comi-
tative PP. Khasi (Mon-Khmer) permits an EHRC with a comitative PP as head
provided the embedded predicate carries an incorporated adverb laŋ ‘together’

(in italics in (87)) and the verbal reciprocal ya (in italics in (87)) which has
the interpretation of a group marker. We shall explicate the reasons for the
permissibility of such sentences later. It is crucial to mention here that, while
the verbal reciprocal in Khasi is optional, the verbal reciprocal functioning as
a group marker is obligatory in (87).
comitative pp with a verbal reciprocal/group marker

Khasi (Mon-Khmer)
(87) ka- khnnaʔi [AP ba u- ban u- ya- wan- laŋ øiAP ]
f,s- child adjr m,s- Ban m,s- gm- come- together
ša- šnɔŋ]
to- village
‘The girl with whom Ban came to the village . . . ’
Tibeto-Burman:
In Tibeto-Burman languages, an EHRC is not permitted with a comitative PP as
head in Manipuri, Tenyidie, Sangtam, Mizo, Hmar and Konyak. For example:
t e n y i d i e (t b ) – e h r c n o t p e r m i t t e d
(88) *nɔ vɔr ke- c- miepuo- u puo cha cha se
you come nozr- dm- person- def 3 s height long very
Intended meaning: ‘The person you came with is very tall.’
Sangtam (TB) – e h r c n o t p e r m i t t e d
(89) *n - n ro- ba- nitsar tsə khataŋ atšŋ e
you- nom come- nozr- person def very tall is
Intended meaning: ‘The person with whom you came is very tall.’
In Tibeto-Burman, there are two sets of languages: let us label them as Set A and
Set B. In Set A the internal head of an IHRC overtly carries the comitative case
marker that establishes a thematic relation with the embedded predicate. In Set
B the internal head of an IHRC does not carry the comitative case marker and,
hence, it is similar to the head in an EHRC in not having an overt postposition
with the head and, consequently, it cannot establish a thematic relation with the
embedded predicate.
Let us consider the case of languages of the Tibeto-Burman family from
Set A in which the internal head carries an overt comitative postposition. These
languages include Ao, Tenyidie, Konyak and Sangtam (TB).
Note that miepuo ‘person’ in (90) in Tenyidie (TB), for example, occurs in
the embedded relative, and is case-marked by ze ‘with.’
t e n y i d i e (t b ) – ihrc w i t h a n o v e r t c o m i t a t i v e – p e r m i t t e d
(90) nɔ miepuo- ze leški- nu vɔ ke- c- u- e mhacha
you person- com school- to go nozr- dm- def- nom cook
si se
well very
‘The person (who) you went to school with cooks very well.’
In Sangtam (TB), too, the case marker te ‘with’ is overtly present with the
internal head nitsar ‘person.’
s a n g t a m (t b ) – o o a s he a d o f a n ihrc w i t h a n o v e r t c o m i t a t i v e –
permitted
(91) n- n nitsar- te ro- ba- tsə khataŋ atšŋ e
you- nom person- com come- nozr- def very tall is
‘The person with whom you came is very tall.’
Let us now consider the languages in Set B where the internal head does
not carry an overt comitative postposition. These languages include Manipuri,
Mizo, Hmar, Thadou, Zou, etc. In this set, a comitative PP cannot head an IHRC,
for example as in Manipuri (TB). The head mi ‘person’ is uncase-marked in
(92) and the sentence is ungrammatical.
m a n i p u r i ( t b ) – o o a s he a d o f a n i h r c (c o m i t a t i v e ) – n o t
permitted
(92) *nahak mi adu lak- pə pha- i
you person det come- inf good- [-fut]
‘The person with whom you came is good.’
However, if the verbal reciprocal marker and an adverb indicating together

occur with the embedded verb, a comitative PP can head an EHRC in languages
such as Manipuri and Dimasa (TB).38 We provide an example from Manipuri.
comitative as head
Manipuri (TB)
(93) tombə- nə lak- min- nə- bə nupi- du pha- i
Tomba- nm come- together- VREC- inf girl- def good- [-fut]
‘The girl with whom Tomba came is good.’
(Surmangol Sharma, N. Pramodini and many other speakers, p.c.)
There are some Tibeto-Burman languages, such as Hmar, Mizo, Paite and
Thadou, in which a postposition is incorporated into the verb (see chapter 4 for
details).
In Thadou, in the embedded verb, the verbal reciprocal ki and the comitative
postposition pi are incorporated in an EHRC, as in (94). The valence of the
[−transitive] verb ɔn ‘go’ increases by one and as a result it acquires the status
of a [+transitive] verb. Evidence for this claim comes from the fact that the
verb ɔn ‘go’ is [–transitive] and, hence, it does not permit an ergative marker
to occur with the subject in Thadou. However, once the incorporation of the
postposition takes place, the embedded subject thaŋboi ‘Thangboi’ is marked
ergative in (94), though the verb is still ɔn ‘go.’ Hence, the reason for the
acquired transitive status of the predicate is due to the incorporation of the
postposition.39 The verbal reciprocal and the incorporated postposition are in
italics.
Thadou (TB)
(94) thaŋboii - in sinema ve øj ai - kii 40 ɔn- pi- nu- cuj
thangboi- erg cinema to see 3s- VREC go- with (ip) f mkr- def
ka kadɔŋmaj - nu aj - hi
my girlfriend- f mkr 3s- be
‘The girl with whom Thangboi went to see the movie was my girlfriend.’
The abbreviation ip in (94) stands for “incorporated postposition.”
(Subbarao and Haokip 2009, 2011)
Example (95) is ungrammatical, as the verbal reciprocal and the incorporated

comitative postposition are not overtly present.
(95) *thaŋboii - in sinema ve a- ɔn- nu- cu ka kadɔŋmaj -

thangboi- erg cinema to see 3s- go- f mkr- def my girlfriend-
nu aj - hi
f mkr 3s- be
Intended meaning: ‘The girl with whom Thangboi went to see the movie
was my girlfriend.’
The issues that need to be resolved are:

(i) Why is it that only an IHRC with a comitative PP as head is permitted in
some Tibeto-Burman languages, such as Tenyidie and Sangtam, when an
overt case marker occurs with the head, while an EHRC is not permitted
in the same set of languages?
(ii) Why is it that an IHRC with comitative PP as head is not permitted in
languages with no overt case marker on the head?
(iii) Why do Khasi (Mon-Khmer), Manipuri and Thadou (Tibeto-Burman)
permit an EHRC with a comitative PP as head, when the verbal reciprocal
and the adverb ‘together’ occur with the embedded verb (see (87) for
Khasi, (93) for Manipuri and (94) for Thadou)?
We now provide an answer to resolve these issues in terms of thematic relations.
Before we proceed, it is pertinent to point out why certain oblique PPs are
permitted to head a non-finite EHRC and why only the comitative PP is not
permitted to head a non-finite EHRC.
Let us now look at some of the Oblique Objects. Locative PP is a subcate-

gorized argument in case of predicates such as keep, place, live, etc. However,
predicates such as sleep, sit, walk, run, etc., are not subcategorized for a loca-
tive object such as bed for sleep, or chair for sit, or road for either walk or
run though it (the locative object) is thematically linked to the predicate of the
embedded clause.
The instrumental PP and the ablative PP are not subcategorized arguments of
the embedded verb, but they too are thematically linked to the (embedded) verb
as the occurrence of the instrumental and ablative PP is crucially dependent on
the nature of the predicate. A predicate such as cut or slice, for example, invari-
ably takes an instrumental PP, and predicates such as get down, alight, bring,
for example, require a PP which is a source or an entity that potentially moves.
In contrast, a thematic relation between the NP of the comitative PP (ez.ut..tə
‘letter’ in (82) in Malayalam [DR], umā in (83) in Tamil [DR], saral.a in (84)
in Telugu [DR], miepuo ‘person’ in (88) in Tenyidie, and nitsar ‘person’ in
(89) in Sangtam [TB]) and the predicates wan ‘come’ in Malayalam (DR), pō
‘go’ in Tamil (DR), vel..l ‘go’ in Telugu (DR), vɔr ‘come’ in Tenyidie (TB),
and ro ‘come’ in Sangtam (TB), respectively, cannot be established, as they
are not essential arguments of the predicate and there is no overt case marker
to establish such a relationship. That is, these arguments are not thematically
related to the embedded predicate.
Recall that there is no comitative case marker overtly present with the head
in the EHRC to indicate the case relationship of the head with the predicate, as
EHRCs universally do not carry overt case markers with their heads. In contrast,
in those languages in which the comitative case marker is overtly present with
the head of the IHRC, there is no problem in establishing a thematic (case)
relationship between the head NP and the predicate. Hence, an IHRC with a
comitative PP as head is permitted only in the set of languages in which an
overt case marker is present as the overt case marker facilitates interpretation.
In those languages in which an IHRC does not carry the overt case marker
with the head (as in Mizo, Hmar, Manipuri and Thadou of Set B), an IHRC with
a comitative PP as head is not permitted, as the thematic (case) relationship
between the comitative PP and the predicate cannot be established.
Let us now discuss the case of the EHRC with a comitative PP as head in
Khasi (Mon-Khmer) in sentence (87), and in Manipuri in sentence (93) and
in Thadou (TB) in sentence (94). We demonstrate that the crucial factors in
permitting the comitative PP as head in these languages are:
(i) the obligatory occurrence of the verbal reciprocal functioning as a group
marker: -ya- in Khasi, -na- in Manipuri and ki in Thadou;
(ii) the verbal suffixes laŋ ‘together’ in Khasi and min ‘together’ in Manipuri;
and the incorporated postposition pi ‘with’ in Thadou with the embedded
verb.
Let us now look at these two factors in detail.

Khasi, Manipuri and Thadou have a verbal reciprocal functioning as a group
marker in the embedded predicate. The occurrence of the verbal reciprocal with
the embedded verb indicates that the subject is in the plural, that is, the head
of the EHRC is not the comitative PP alone, but the comitative PP and the
embedded subject together constitute the head. The verbal reciprocal in these
languages occurs only when there is more than one participant in the subject or
object (antecedent) position of the verbal anaphor. The verbal reciprocal ya and
the adverb laŋ ‘together’ in Khasi and the verbal reciprocal na and the adverb
min ‘together’ in Manipuri, the verbal reciprocal ki in Thadou together with
the incorporated postposition pi impart an interpretation of a group/collective
activity. In addition, the absence of the comitative postposition with the head
and its corresponding incorporation not only enables the comitative PP to head
the EHRC but, as we have seen, it makes the embedded predicate [+transitive]
in Thadou.
Our analysis gains support from the following facts:
(i) nominal or verbal reciprocals require a subject which is not in the singular,
but in the plural (see chapter 3 for a discussion); and
(ii) the occurrence of the adverb laŋ ‘together’ in Khasi, as in (87), and min
‘together’ in Manipuri, as in (93), with the embedded verb indicates that
the subject of the embedded verb is in the plural, as such a suffix cannot
occur with a subject in the singular.41
Thus, though the comitative adposition is not overtly present with the embedded
head NP in Manipuri and Khasi, unlike in an IHRC in some Tibeto-Burman
languages (Tenyidie, Sangtam, Konyak, etc.), the occurrence of the verbal
reciprocal and the adverb ‘together’ with the embedded verb in Manipuri and
Khasi performs the same function as the comitative postposition does in an
IHRC in those Tibeto-Burman languages.
Second, we consider the case of Thadou and the other Tibeto-Burman lan-
guages that permit comitative PP as head in an EHRC. These are the languages
that permit the incorporation of a postposition in the verb. In Thadou in (94), not
only does the verbal reciprocal occur with the embedded predicate but the post-
position pi ‘with’ is also incorporated in the embedded verb. This incorporated
postposition performs the same function as the postposition ‘with’ that occurs
overtly with the internal comitative PP in IHRCs in some Tibeto-Burman lan-
guages such as Tenyidie and Sangtam (TB) (see sentence (90) in Tenyidie and
(91) in Sangtam). Recall that the head of an IHRC in these languages overtly
carries the postposition.
Thus, in Thadou the occurrence of the comitative postposition with the verb
establishes a thematic relationship between the accompanied person and the
embedded verb. A similar phenomenon is found in Hmar and Mizo (TB) too
(Vanlal Englien, p.c.).
Once postposition incorporation takes place, the bare head of the comitative
PP in the EHRC gets syntactically linked to the newly derived verb with the
incorporated postposition.42 In addition, the occurrence of the verbal reciprocal
shows that the agent of the verb performs the action together with someone else,
namely the accompanied person. Thus, the incorporated postposition performs
the same function as the comitative postposition with the head in an IHRC in
some Tibeto-Burman languages such as Tenyidie and Sangtam.
Thus, it is due to the occurrence of these markers – namely,
(i) the comitative postposition with the head in an IHRC, or
(ii) the verbal reciprocal and the adverb ‘together,’ or
(iii) the comitative postposition incorporated in the verb –
that a thematic relationship between the comitative NP as head and the predicate
is established.
Based on the discussion above, the following generalization can be made:
(96) A comitative PP can head an EHRC/IHRC if and only if either:
(i) the head is overtly case-marked comitative, or
(ii) if the head is not overtly case-marked comitative, the embedded verb
carries either the verbal reciprocal, or an incorporated postposition, or
both, to establish a thematic relationship of the bare head NP with the
embedded predicate.
All other comitative heads in EHRCs, and in a subset of IHRCs carrying no

overt case marker, disqualify themselves from being a head by default, as there
is no overt case marker.
Our analysis gains support from evidence with regard to the comitative PP
as head from languages such as Japanese and Korean which are genetically
“unrelated” to SALs. In these languages a comitative PP can head an EHRC
if and only if the adverb with the interpretation of ‘together’ occurs with the
embedded verb (Masayoshi Shibatani, p.c.).
Japanese
(97) [matt- ga *(issyoni) kita] onna- no- ko
Matt- nom together came woman- of- child
‘The girl Matt came with’
(Masayoshi Shibatani, p.c.)
In Korean, too, the occurrence of the adverb hamkkey ‘together’ with the
embedded predicate is obligatory.
Korean
(98) yeongsig- i *(hamkkey)- on yeca
Yeongsig- nom together- come girl
’The girl Yeongsig came with’
(Data provided by Masayoshi Shibatani, p.c.)
Other intransitive verbs such as to swim, to walk, to run, to sit, to crawl, etc.,
also require the presence of the adverb ‘together’ with the embedded predicate
in Japanese and Korean (Matt Shibatani, p.c.), just as in Khasi (Subbarao and
Temsen in press), Manipuri and Thadou, etc. The occurrence of the adverb
‘together’ in Japanese and Korean clearly demonstrates that a comitative PP
can head an EHRC if and only if the predicate can get “thematically related ”
to the embedded predicate.
To summarize the above discussion, we have demonstrated that a thematic
relation has to be established between the predicate of the embedded clause and
the head of the EHRC/IHRC, in terms of either an overt case marker with the
head, or some marker in the embedded verb to indicate accompaniment. That
is, specific syntactic criteria have to be met to make the predicate “thematically
eligible” to accept a comitative PP as head. We propose that each predicate has
to meet the “Thematic Eligibility Condition” (TEC) for it to accept a comi-
tative PP as head. Such eligibility is the result of language-specific syntactic
processes such as Postposition Incorporation, the incorporation of the verbal
reciprocal as a group marker and the adverb ‘together,’ or the incorporation
of the adverb together alone in the embedded predicate, or the overt presence
of the comitative postposition with the head in an internally headed relative
clause.
It is significant that in languages belonging to different language families,
the same kind of verbal marker – namely, a verbal reciprocal – is attached to
the embedded verb to achieve the same result.43
It is also significant that in languages belonging to two different language
families, the same kind of verbal marker – namely, a verbal reciprocal – is
attached to the embedded verb to achieve the same result.
8.7 Implications of Postposition Incorporation

In this section, we shall discuss the implications of postposition incorporation
in IHRCs (for a detailed discussion, see Kumar and Subbarao 2005).
There is an asymmetry with regard to the thematic nature of the embedded
verb in terms of its argument structure and the formation of an IHRC in
Hmar. While EHRCs and IHRCs are both permitted in Hmar, the formation
of the IHRC is permitted if and only if the embedded verb is [–transitive]. In
contrast, there is no such restriction on the formation of the EHRC in Hmar.
This is a feature typical of Hmar and we have not found such a restriction in
the other TB languages, such as Tenyidie, Manipuri, Mizo and Sema, that have
IHRCs.44
Subject modified: An EHRC is permitted with a [+transitive] verb hmu ‘get’
in the embedded clause in Hmar (TB).
8.7 Postposition Incorporation 309
Hmar (TB)
e h r c w i t h a [+t r a n s i t i v e ] v e r b – p e r m i t t e d
(99) [loman hmu] naupaŋpa kha kan in- aʔ a- huŋ
prize get boy DD2 our house- to 3 s- came
‘The boy who got the prize came to our house.’
DD2 : definite determiner.45 A corresponding IHRC is not permitted (100) as

the embedded verb hmu ‘to get’ is [+transitive].
i h r c w i t h a [+t r a n s i t i v e ] v e r b – no t p e r m i t t e d
(100) *[loman naupaŋpa hmu] kha kan in- aʔ a- huŋ
prize boy get DD2 our house- to 3 s- came
However, when the embedded verb is [–transitive], an EHRC and an IHRC

are both permitted. In (101), the NP nuhmeihai ‘girls’ occurs in the main
clause in an EHRC, and in (102) it occurs in the embedded relative clause in
an IHRC.
intransitive verb in the embedded clause – ehrc permitted

(101) [hlo cuŋa in46 t.šuŋ] nuhmeihai cu ka- rol- hai aniʔ
grass loc self ben- sit girl- p DD2 my- friend- p are
‘The girls who are sitting on the ground are my friends.’
intransitive verb in the embedded clause – ihrc permitted

(102) [hlo cuŋa nuhmeihai in- t.šuŋ] hai cu ka- rol- hai aniʔ
grass loc girl- p self ben- sit p DD2 my- friend- p are
‘The girls who are sitting on the ground are my friends.’
Examples (99) and (101) show that EHRCs are permitted independently of the
transitive nature of the embedded verb. However, this is not the case for the
IHRCs. The IHRC in (100) is ungrammatical, while the EHRC in (101), in
contrast, is grammatical as the embedded verb is [–transitive].
Let us now look at Postposition Incorporation in Hmar (see chapter 4 for
discussion of incorporation). We shall now consider applicative constructions
in Hmar and the formation of IHRCs.
Applicative constructions are the ones that permit incorporation of an adpo-
sition (preposition and postposition) of a postpositional phrase (PP) in the verb.
Once the adposition is incorporated, in place of the PP we are left with a bare
NP. That is, the PP is stripped of its adposition due to incorporation that results
in the addition of a suffix to the verb.
In Hmar, we observe that: (i) only intransitive verbs permit Postposition
Incorporation, and incorporation of postposition has a transitivizing effect on
the verb; and (ii) the PP becomes the object of the verb after the adposition is
incorporated, as there is only a bare NP that is left. The bare NP that is left
receives accusative case from the verb.
(103) zova cauki- a a- t.šuŋ
Zova chair- on 3 s- sat
‘Zova sat on the chair.’
After incorporation
(104) zova- n cauki a- t.šuŋ- pui
Zova- erg chair 3 s- sat- loc suffix (on)
‘Zova sat on the chair.’
In (103), (i) the verb .tšuŋ is an intransitive verb and, hence, the subject Zova
does not carry the ergative case marker, and (ii) the PP carries the postposition
a ‘on.’ In contrast, in (104), incorporation of the postposition has taken place,
as a result of which the derived verb acquired the status of a transitive verb. In
(104), due to adposition incorporation: (i) the subject Zova carries an ergative
case marker -n; (ii) the NP cauki ‘chair’ is a bare NP, and is not followed
by the locative postposition; (iii) the verb carries the locative suffix pui to its
right; and (iv) the newly formed transitive verb .tšuŋ -pui assigns structural
accusative case to the object NP cauki ‘chair.’ Incorporation in Hmar, thus,
conforms to Baker’s proposal concerning applicative constructions, according
to which: “a grammatical applicative construction can only occur when the
derived verb assigns accusative case to the NP that is stranded by the movement
of preposition” (Baker 1988: 252).
We shall now demonstrate that this type of Postposition Incorporation in
Hmar has implications for the formation of the IHRCs in that language. We
observe that a [–transitive] verb in the embedded clause permits an IHRC, as
in (105), while a [+transitive] verb does not.
i h r c ( w i t h no i n c o r p o r a t i o n o f t h e p o s t p o s i t i o n i n t h e
embedded clause) – permitted
(105) [zova tabul cu- a kei- le t.šuŋ- na] cu a- lien
Zova table det- on I- with sit- loc adv mkr DD2 3 s- big
‘The table on which Zova sat with me is big.’
However, when incorporation of the postposition le ‘with’ takes place, the

embedded verb .tšuŋ ‘sit’ is transitivized, as a result of which the embedded
subject carries an ergative case marker. It is in such cases that an IHRC is not
permitted due to the newly acquired transitive nature of the verb after Postpo-
sition Incorporation. Sentence (106) is illustrative. In (106), the postposition le
‘with’ is incorporated, and the verb as a result has pui ‘with’ as an incorporating
suffix. The postpositional phrase kei-le ‘with me’ now has the form of a bare
NP kei ‘I,’ which acquires the status of DO in (106).47 The [–transitive] verb
8.8 Conclusion48 311
.tšuŋ is transitivized due to incorporation. Hmar is a split ergative language in

person. First and second person DOs in Hmar thus trigger DO agreement and,
as a result, the verb in (106) carries the object agreement marker min and the
subject, the ergative marker -n.
ihrc (with incorporation of the postposition in the embedded

c l a u s e ) – no t p e r m i t t e d
(106) *[zova- n tabul cu- a min- t.šuŋ- pui- na] a- lien
Zova- erg table det- on 1 s (DO)- sit- with- loc adv mkr 3 s- big
‘The table on which Zova sat with me is big.’
In (105), an IHRC is permitted as the embedded verb is [–transitive]. In contrast,

in (106), the formation of an IHRC is blocked, as the verb is [+transitive] due
to Postposition Incorporation.
To summarize the above discussion, an IHRC while modifying a locative
object is permitted if and only if the verb in the embedded clause is an intran-
sitive verb. Due to adposition incorporation, an intransitive verb becomes
[+transitive], as a result of which an IHRC modifying the locative NP is
not permitted.
One of the crucial issues that concerns EHRCs and IHRCs in SALs is the
potential ambiguous interpretation of these clauses. The potential ambiguity
centers around the DPs as the head in an EHRC/IHRC and the thematic relations
that these DPs bear with the embedded predicate. (See appendix 8.6 on the
Website for more discussion.)
The next issue that concerns IHRCs is the canonical position of the head
in an IHRC. In an IHRC, the head occurs in the embedded clause, whereas,
in an EHRC, it occurs in the matrix clause. In this section, we have presented
three arguments to show that the canonical position of the head in an IHRC is
in the embedded clause, and not in the matrix clause. (See appendix 8.7 on the
Website for a detailed discussion.)
The next topic that we shall discuss is relative clauses and syntactic reanalysis
keeping in view the syntactic changes that took place in relative clauses in
Dakkhini (IA), Mangalore Konkani (IA), Bhalavali Bhasha (IA), Sinhala (IA),
Oriya (IA) and Marathi (IA) due to contact with the Dravidian languages
Kannada, Tamil and Telugu. (See appendix 8.9 on the Website for a detailed
discussion.)
8.8 Conclusion48
In this chapter we provided a detailed analysis of relative clauses, participial
and nominalized EHRCs and IHRCs. With regard to the relative-correlative
clauses, we have shown that: (i) Indo-Aryan languages have relative-correlative,
right-adjoined and extraposed varieties of relative clauses; (ii) Dravidian and a

few Tibeto-Burman languages have only the relative-correlative construction;
(iii) the relative-correlative construction which the Munda languages have is
borrowed from Indo-Aryan. Employing an EHRC of the participial-type or
the nominatized-type is the dominant strategy in SALs. We have discussed
the nature of ambiguity that arises in EHRCs and IHRCs and the factors
responsible for it. Asymmetries observed in the occurrence of the embedded
clause demonstrate that the unmarked word order in relative clauses is the
occurrence of the clause to the left of the head, just as in many verb-final
languages. We have shown that partial reduplication of an NP is used as a
syntactic device to resolve ambiguous interpretation of an IHRC. We provided
an explanation, in terms of subcategorization of arguments and the Thematic
Eligibility Condition (TEC), as to why certain predicates do not permit an
EHRC with a comitative PP as head, while a set of languages that have a
postposition occurring with the comitative PP as head do permit an IHRC.
We have shown that a thematic relation needs to be established between the
comitative head and the predicate by means of a clitic or suffix in the embedded
predicate, if the comitative postposition is not overtly present with the head in
an EHRC or IHRC.
A detailed discussion of IHRCs in Tibeto-Burman languages is also provided,
focusing our attention on Postposition Incorporation. It was demonstrated how
Postposition Incorporation affects the nature of the argument structure of the
predicate and, consequently, how it affects the formation of IHRCs in Hmar
(TB). Arguments in support of the canonical position of the head in IHRCs were
also provided. Effects of language contact in Dakkhini, Konkani and Bhalavali
Bhasha have been briefly discussed, and we have shown how the correlative
marker is reanalyzed as an adjectivalizer in Dakkhini.
Notes
1 I N TRO D U CTION
1. Andamanese and Nicobarese are omitted from this study because of lack of available
syntactic data.
2. Chomsky (1965) used the term subcategorization. Subcategorization, put simply,
means the number of essential arguments that a verb can take in its predicate. As
Lasnik and Uriagereka with Boeckx (2005: 3) put it: “a lexical entry must contain
syntactic information about whether, for instance, a verb is transitive, intransitive,
ditransitive, and so on (so-called subcategorization).” See Chomsky (1965) for fur-
ther details.
3. The feature [−initial complementizer (IC)] in parameter (3) corresponds to [+finite
complementizer].
4. Martin Everaert (personal communication) informs me that in Tanya Reinhart’s Theta
system, it is clearly defined (i) which thematic roles exist, and (ii) how to decide on
what thematic role an argument has (see Reinhart 2002).
See Subbarao and Temsen (in press) for a detailed discussion of Externally
Headed Relative Clauses (EHRCs) with the comitative PP as head in Khasi, and
Subbarao and Haokip (2011) for Thadou EHRCs and Internally Headed Relative
Clauses (IHRCs).
5. Thanks to Alice Davison for her help in the formulation of this parameter.
6. Based on data from noun incorporation in Sora, we demonstrate that there are some
vestiges left of ancestral non-verb-final structures in this language and point out
that only diachronic evidence can resolve the paradox in the order of occurrence of
incorporated elements (see chapter 4 for a detailed discussion).
2 S O U TH A S I A N L ANGUAGE S: A PR E VIEW
1. Due to the limitations of space, we have not focused on Khasi (Austro-Asiatic,
Mon-Khmer). For a detailed study of the features of Khasi, see Temsen (2006) and
Temsen and Subbarao (in preparation).
2. We have also not included Andamanese and Nicobarese in this study due to lack
of sufficient syntactic data and evidence. See Abbi (2006) for a description of the
endangered languages of the Andaman Islands.
3. South Asia, along with the Balkans, is a prime example of the rather rare phenomenon
known as Sprachbund, first mentioned by Trubetzkoy, or “linguistic area” (Emeneau
1956) or “convergence area” (first suggested in Weinreich 1958; see also Hock 1991:
494).
313
314 Notes to pages 18–19
Weinreich (1958) in his paper in Word (14,379) was the first to use the phrase
“convergence area.” The term convergence has been in use ever since to characterize
the phenomenon observed in language contact situations which results in changes at
the level of phonology, morphology, syntax and semantics. It is not clear, however,
who used the specific term convergence first.
4. Pnar, Jowai and Rymbai (Mon-Khmer) are verb-medial too (see Bareh 2007 for
details).
5. Thanks to Rajesh Bhatt for bringing this important paper to our attention.
6. According to Comrie, language typology is concerned with “differences among lan-
guages” and “with the study of variation” (1981: 30–31). The most commonly found
types of languages are verb-final, verb-medial and verb-initial, in which the verb
occurs in the final or medial or initial position of a sentence, respectively. In the stud-
ies on word order typology depending on the position of a verb (head) in a sentence,
certain common structural characteristics amongst languages are abstracted. These
common characteristics are generally called word order universals (see Greenberg
1966 for details). A study of word order universals of SALs demonstrates how sim-
ilar or different the languages are that belong to the four different language families
of the subcontinent.
7. The features discussed are based on the study of the syntax of these languages
and there is no psychological or functional explanation in support of the observed
similarities.
8. There is, however, an exception to this generalization: in Kashmiri, the finite form
of the verb occurs in the second position (labeled as V2 position) in a sentence, just
as in Dutch and German and, thus, it may precede the object as in (i).
Kashmiri (IA)
(i) rāman dits šāmas kitāb
Ram.erg give.pst V2 Sham.dat book
‘Ram gave Sham a book.’
(ii) rām chu šāmas kitāb divān

Ram be.pres V2 Sham.dat book give.progr
‘Ram is giving a book to Sham.’
(Raina 2002: 114)
In (i), the finite form of the verb (V2 ) dits ‘gave’ carries the finite past tense marker,
and in (ii), chu ‘be.pres’ carries the finite present tense marker. Sentences (i) and
(ii) would be ungrammatical if the finite form of the verb occurred in the final position
in the sentence to the right of the object, as (iii) and (iv) illustrate.
(iii) *rāman šāmas kitāb dits

Ram.erg Sham.dat book give.pst
‘Ram gave Sham a book.’
(iv) *rāman šāmas kitāb divan chu

Ram.erg Sham.dat book give.progr be.pres
‘Ram is giving Sham a book.’
(Raina 2002: 114)
Notes to pages 20–26 315
There is also an interesting construction in Assamese (Modhumita Barbora and

Gautam Borah, p.c.), Bangla and Oriya (Probal Dasgupta, p.c.) in which the copula
occurs in a specific set of sentences in the second position, and not in the final
position as all other verbs do. Dasgupta (2003: 384) labels it as “Positive Polarity
Copula” (PPC) construction as it occurs only in affirmative sentences, “avoiding
both interrogation and negation.” Sentence (i) from Bangla is illustrative. It obli-
gatorily places its special-ending-endowed copula in the second position. Such a
sentence is always an “aggressively root” sentence. It is an “edge-of-the-grammar
phenomenon,” as Probal Dasgupta puts it, and, hence, it is excluded in the present
study. Thus, PPC cannot occur in clause-final position.
Bangla (IA)
(i) uni holen birener- i guru
s/he PPC Biren’s emph guru
‘S/he is Biren’s guru (nobody else’s).’
Dasgupta (2003: 38)
It is important to note that not every verb ‘to be’ exhibits this property. The one
used for kinship, for example, normally occurs in sentence-final position.
(ii) rām tomar mastuto bhai hɔy
Ram your maternal cousin brother is
‘Ram is your maternal cousin brother.’
(Dasgupta, p.c.)
Kokborok (TB), which is a verb-final language, exhibits some non-verb-final struc-
tures too, unlike all other Tibeto-Burman languages. See Subbarao, Malhotra and
Barua (2010) for details, and see also sentences (7) and (8) from Kokborok in the
appendix on the Website.
9. Abbi (2001) presents a discussion of several features of verb-final languages.
10. Greenberg’s word order universal 22 is as follows:
Universal 22: If in comparisons of superiority, the only order, or one of the alternative orders,
is standard-marker-adjective, then the language is postpositional. With overwhelmingly more
than chance frequency, if the only order is adjective-marker-standard, the language is prepo-
sitional (Greenberg 1966: 88–89).
11. The verbal reflexive marker -in occurs with a set of adjectives in Hmar (TB) and
such occurrence seems to be lexically determined.
12. Time and place adverbials occur in ascending order in Khasi, a non-verb-final
language.
13. Butt and Geuder (1999) point out that vector verbs, a.k.a. light verbs, are the result
of a type of polysemy, and semantic bleaching is not an issue.
14. See the appendix to this chapter on the Website for a discussion of the reduplication
of some other constituents in SALs.
15. Though echo words are generally considered a part of morphology, they have a
role to play in terms of politeness. While in some contexts the use of an echo
expression might indicate intimacy between the speaker and hearer, in some other
contexts it might signal rudeness that hurts the sentiments of the hearer. Such use
also is an index of the social hierarchy between the speaker and hearer. The use
of echo words in politeness strategies and cultural propriety needs to be worked

out with reference to space, time and interpersonal relationships in each language.
While in Dravidian languages their use might indicate rudeness, indifference and
arrogance on the part of the speaker, in some Indo-Aryan languages (Hindi-Urdu
and Bangla, for example) they can be used between friends and relatives and with
hearers who are younger in age, or lower in the social hierarchy – primarily in
casual conversations (Amit Mitra and Anupam Das, p.c.). No specific statements
can be made at present, as there have been no studies on this issue.
In contexts which require politeness to be adhered to between the speaker and
the hearer, Telugu (Dravidian), for example, uses a [−proximate] non-masculine
pronominal copy adi ‘that’ followed by an inclusive particle -i as in the example
below. The use of an echo word reflects rudeness and arrogance on the part of the
speaker.
Telugu (DR)
mı̄ ammāyi pel.l.i ayyindi mari pārt.ı̄
your [+honorific] daughter wedding occurred so party
adi+i/ gı̄rt.ı̄ ivvar- ā
that (literal)+emph/ echo word give- pol q mkr
and the thing like
‘Your [+honorific] daughter got married. Aren’t you giving a party and
the like?’
Hindi, Punjabi and Bangla (IA) do not permit such usage of a pronominal copy,
and it needs to be worked out whether it is permitted in the other two language
families.
16. Khasi (Mon-Khmer), which is verb-medial, is the only South Asian language that
we know of that strictly observes superiority effects.
17. While Hindi allows for “direct speech” complements as in (28) as reported in the
main text, it also allows for indirect speech complements, too. Thus, one can say:
saritāi ne kahā thā ki [vahi āp se

Sarita. 3 pn,s,f erg said had comp she 3 pn,s,f you with
kal milegı̄ ]
tomorrow will meet 3 pn,s,f
‘Saritai had told me that shei ’d meet you tomorrow.’
18. It is not uncommon to find sentences which do not observe tense harmony strictly
in English (Peter Hook and Colin Masica, p.c.). Peter Hook points out that a Google
search will substantiate this point.
19. There are only two SALs that we know of which are an exception to this gener-
alization. These are Khasi (Mon-Khmer) (Subbarao 2001) and Tenyidie (Tibeto-
Burman) (Kevichüsa 2007).
20. In Khasi, which is a non-verb-final language, too, there are no expletive expressions.
Normally, pro occurs in subject position. However, the NP ka-bnεŋ ‘the heaven’
may also occur in subject position.
Khasi (Mon-Khmer)
proi / kai - bnεŋ kai - khryat
f,s- heaven f,s- cold
Literally: ‘The heaven is cold.’ / ‘It is cold.’
(Temsen 2006)
21. Mangalore Konkani also has structures involving Backward Control as the example
below shows.
[S1 [S2 ammii haŋga yev- nuS2 ] ∀i dha vərsəj dzallı̃jS1 ]
we here come- cpm ten years happened
‘It is ten years, since we came here.’
22. Hany Babu (p.c.) points out that such sentences are not used in his dialect of
Malayalam. This matter of dialect variation in Malayalam is worth investigating.
23. Thanks to Philippe Bourdin (p.c.) for suggesting the label “cognitive gerunds” for
such constructions. According to him, French too has a similar construction.
24. In most of the Indo-Aryan languages we checked, this construction does not exist,
the exceptions being Gujarati (Trupti Nissar and P. J. Mistry, p.c.) and Mangalore
Konkani (Lalita Dhareshwar, p.c.). We do not have any data from the other two
families.
25. Though the embedded predicate ɔn ‘go’ and the matrix predicate hi ‘be’ are
both [−transitive], the embedded subject is ergative case-marked. The reason for
such marking is due to the fact that postposition incorporation in Tibeto-Burman
languages increases the valence of the embedded predicate. Hence, the ergative
marking on the embedded subject. See chapter 4 for a detailed discussion.
26. See Masica (1991) for a brief discussion of the syntax of Indo-Aryan languages.
27. Munda languages have the relative–correlative construction in which a relative
pronoun is used, and a question expression is also used as a relative pronoun in the
formation of relative clauses, as in Dravidian (see Anderson 2008).
28. Clefts are permitted in Dravidian, in some Tibeto-Burman languages and in Khasi
(Mon-Khmer). Amongst the Munda languages, Ho does not permit clefts, and we
do not have information on the other languages.
29. Emeneau (1956) uses the phrase “India as a linguistic area.”
3 L E X I C A L A NA P H O R S A N D P RO N O U N S I N S O U T H A S I A N
LANGUAGES
1. See Lust et al. (2000) for an exhaustive description and discussion of anaphors
and pronouns in fourteen different SALs encompassing four different language
families, a detailed introduction focusing on the issues and a questionnaire. For a
typology of anaphors, see Huang (2000). This chapter greatly benefited from the
papers in Lust et al. (2000).
2. We use the term verbal anaphor as a convenient label for the reflexive or reciprocal
in the absence of a better term. Martin Everaert (p.c.) observes that the term
‘anaphor’ implies referential status, which implies being an argument / DP-status.
The evidence presented below, as Martin Everaert observes, shows “the limitations
of the standard Binding Theory, which is solely focused on arguments, allowing no
room for the concept of a verbal reflexiviser.”
3. The verbal reciprocal -ya- also functions as a group marker. We shall see in chapter 8
that the group marker -ya-, however, is obligatory when it occurs in sentences with
the comitative PP as head. See sentence (87) in chapter 8.
4. We provide an example from Toda (DR):
aθ (ton- ē) tān- n ökši

he self- emph self- acc scratched
‘He scratched himself.’
(Peri Bhaskararao, p.c.)
5. A vector verb such as vēs ‘to drop’ can also occur with the main verb, when a verbal
anaphor occurs.
6. We shall demonstrate later that a null form ( pro) is not permitted in the non-
nominative subject construction, as the verbal anaphor does not occur in such
constructions.
7. To the best of our knowledge, E. Annamalai was the first scholar to have made this
observation regarding such predicates in any Dravidian language.
8. In Lalitha Murthy and Subbarao (2000), we have transcribed the Mizo expression
for ‘self’ as māʔ. Some of the language consultants we interacted with in recent
years feel that ‘self’ should be transcribed as maʔ. Hence, in this work, we have
transcribed it as maʔ.
9. Reinhart and Siloni (2005: 404) argue that “The selection of se [the reflexive clitic in
French] reduces the verb’s ability to check accusative Case [(i) below from French].”
In Mizo, however, the DO needs to have its accusative Case assigned/checked and,
hence, their claim cannot be accepted for Mizo. The letter ti in the example below
stands for trace.
Jeani sei lave ti
John SE washes
‘Jean washes himself.’
10. In the Jowai and Rymbai dialects of Khasi too, the marker ya- functions as a
reciprocal marker as well as a group marker (Bareh 2007).
11. Also see sentence (87) in chapter 8 where ya- the verbal reciprocal, obligatorily
occurs as a group marker in sentences with the comitative PP as head in EHRCs in
Khasi.
12. In our system of analysis, -p- is marked as ‘-pA-’.
13. Evidence for Kazenin’s claim comes from Mizo (TB) in which the ergative marker
does not occur when the anaphor occurs in the direct object position, as the verb
is detransitivized. When the anaphor occurs in a non-direct object position, the
occurrence of the ergative marker is obligatory (cf. sentence (12) from Mizo).
14. An inchoative verb denotes the beginning of an action, state or event. Verbs such
as open, close, thicken, etc., in English are a few examples.
15. Such instances as in (32) are examples of Possessor Ascension as there are corre-
sponding sentences, as below, where the possessive pronoun occurs as a modifier
of the body part:
ivāl.a atan.ii vēluj kōsu- kon- di∗ i,j

today he.gen finger cut[+tr]- inchoative[−tr]- 3 s,nm
‘His finger got cut today (inadvertently).’
16. For more discussion of functions of the VR kol..l and its relation to nominal marking
of coreference, see Lidz (2001a, b and c).
17. As quoted in Anderson (2007: 42).
18. Adjectives behave like verbs in Tibeto-Burman languages. They can take “tense,
aspect and mood suffixes” in Manipuri (TB) (Bhat and Ningomba 1995: 66). They
can take nominalizers and causative prefixes in Tenyidie (TB) (Koulie 2006: 15)
and in Mizo and Hmar (TB). In Kokborok (TB), “the superlative marker, which
normally occurs with adjectives, also occurs with the modal man ‘can’ and is
followed by the nominalizer nai in comparative constructions” (see Subbarao et al.
2010 for details).
19. For a more elaborate list of verbs with konu, see Subbarao and Lalitha Murthy
(2000: 227).
20. These are called intrinsic/inherent reflexive verbs in Germanic/Romance (Everaert,
ms).
21. Lidz (2001c) proposes to account for the distribution of the verbal reflexive in
Kannada in terms of the interaction between lexical aspectual structure and thematic
structure. According to him, there is a mismatch in linking between the verb’s
aspectual decomposition and thematic representation. This, he claims, provides
an explanation for the absence of the VR in NNS constructions, VR blocking
long-distance binding, etc. (2001: 351).
22. Cf. Hopper and Thompson (1980), Kemmer (1993), for similar views.
23. In (56c), in place of the dative subject lili-lā ‘Lili-dat,’ a genitive subject lili-tsyā
‘Lili-gen’ can also occur (Trupti Nissar, p.c.). Lalita Dhareshwar prefers āplya-vərr
‘self-obl-on’ to svətāhā-vər ‘self-obl-on’ with the genitive subject in (56c).
24. Davison (2004: 155) proposes a theory-internal argument, and we quote: “To avoid
this constraint on nominative reflexives, the internal argument or theme [that is,
rām in our (59)] may rise to Spec/TP for the EPP [Extended Projection Principle]
feature, while the dative argument stays within the VP. The reflexive argument is
therefore dative in [(our (59)], bound by the nominative subject.”
There needs to be further investigation into why a nominative case-marked
non-anaphor (R-expression or pronoun) is permitted as theme in the DSC, while a
nominative case-marked anaphor is not.
25. Thanks to Rajesh Bhatt (p.c.) for helpful suggestions.
26. Our language consultants insist on having an aspirated -th in thõ ‘by’ in (69) in
contrast to Bhatia’s (2000: 661) transcription without an aspirated -th in tõ ‘by’ in
(68).
27. A. Usha Rani (p.c.) brought to my notice the sentence below, in which the possessive
anaphor tana ‘self’s’ in a finite complement clause is coindexed with the subject of
the matrix clause. Such coindexing is a violation of Principle A, and is similar to
the violation found in Marathi in (76).
Telugu (DR)
amei [dongaj tanai,∗ j parsu dongilincu- kon i pōyēd.u]
she thieves self’s purse steal- self ben cpm went
ani andi
quot said
‘Shei said that the thievesj stole heri,∗ j purse.’
28. Subbarao (1971) is the first study that discussed long-distance binding in any SAL.
Davison (2001) discusses long-distance binding in Hindi-Urdu in detail. The papers
in Lust et al. (2000) discuss long-distance binding in various SALs.
29. Recall that Indo-Aryan does not have a robust verbal anaphor and, hence, no
evidence can be adduced from Indo-Aryan.
30. The permissibility of occurrence of the accusative case marker in (82) is due to the
fact that in select languages such as Bangla (IA), Bodo (TB), Tamil and Malayalam
(DR), a dative predicate may assign an accusative case marker to the theme in
DSC – cf. (146) from Bangla, (155) from Malayalam, (157) from Tamil and (163)
from Bodo in chapter 5.
31. In the Minimalist Program (Chomsky 1995b) an ECM subject raises to the Spec
Agro for Case checking (Lasnik 1999). Hence, the anaphor occurs in the governing
category of its antecedent.
32. We shall show that a genitive case-marked subject in Bangla permits an accusative
case-marked theme as in (146) in chapter 5, while Hindi-Urdu does not permit an
accusative case-marked theme when the subject is dative case-marked, as in (147)
in chapter 5. The two facts concerning (82) above and (146) in Bangla in chapter 5
might be intrinsically connected. Only further research can shed light on this issue.
33. Rajesh Bhatt (p.c.) speculates that all such cases obligatorily involve a simplex
form.
34. Kidwai (2000) attempts to analyze binding facts of Hindi-Urdu as instances of
XP-adjunction, the discussion of which is beyond the scope of the present work.
35. We did not provide a detailed discussion of reciprocals in SALs. For a detailed
discussion, see Everaert and Subbarao (2011) and Subbarao and Everaert (2011).
4 CA S E A N D AGR E E ME NT
1. An NPI is a syntactic item that strictly requires a negative licensor in the local
domain where the licensing takes place during the course of derivation. See Kumar
(2006) for a detailed discussion of the syntax of the NPI in Hindi.
The NPI is incorporated in Khasi (Mon-Khmer) and in Kokbodok (TB) too,
which we shall discuss later.
2. Thanks to Rajesh Bhatt for suggesting this classification into two subsets.
3. P. J. Mistry (p.c.) labels the marker -(n)e on the subject in (11) as an agentive marker
and points out that there is dialect variation between -ne and -e.
4. See Das (2006) for a discussion of agreement in Hindi-Urdu and its major dialects.
5. Thanks to Rajesh Bhatt for his suggestion in this regard.
6. An example of DO incorporation in Khasi is given below. It may be mentioned
that the function of the clitic -sa in the example is not fully understood. It is
not discussed in any Khasi grammar. We have labeled it as “pm,” where “pm” stands
for “procrastination marker.” Thanks to Juanita War (p.c.) for pointing out that sa
literally means ‘by and by’ or ‘later.’ We may hypothesize that sa, originally an
adverb, is subsequently grammaticalized as a procrastination aspect marker.
Khasi (Mon-Khmer)
ŋi- n- sa- thēd do? ma- ŋi-
1p- fut- pm- buy meat nom- 1p-
aux 1 aux 2 v e r b o b j e c t subject
‘We will buy meat.’
7. For some speakers the occurrence of i-koma ‘for you’ and the pronominal clitic -ce
‘2 singular’ are in complementary distribution (Vanlal Englien, p.c.).
8. Though Rymbai, Jowai and Pnar are generally called dialects of Khasi, the work
of Bareh (2007) demonstrates that there are many syntactic differences between
these “dialects” and Khasi. The adverb yaw ‘market’ in the underived structure is
incorporated into the verb in (i) in Rymbai (Mon-Khmer). Note that it occurs to
the right of the verb stem and is followed by ko, the 3rd person feminine, singular
agreement marker.
Rymbai (Mon-Khmer)
(i) ka- daw lε- yaw- ko
3s,f fut go- market- 3s,f
‘She’ll go to the market.’
(Bareh 2007)
In Pnar (Mon-Khmer) too, incorporation of a DP into the verb takes place. Sentence
(ii) exhibits no incorporation, while sentence (iii) incorporates the DP ja-i
‘rice-
hot.’ Note that the subject agreement marker occurs to the right of the verb in
(ii) while the subject agreement marker -u occurs to the right of V-bar which is
the verb and the first conjunct of the incorporated DP. The first occurrence of the
preposed conjunction wa does not occur in (iii) due to incorporation.
no incorporation with a conjoined dp
Pnar (Mon-Khmer)
(ii) u- jɔni bam- ui [DP w(a)- i ja- i
w(a)- u-
3s- John eat- subj agr and dm rice- hot and- m,s-
sa-pe
DP ]
fruit-mango
‘John ate hot rice and a mango.’
Incorporation of a part of the DP into the verb
(iii) u- jɔni [v̄¯ bam- ja- i
- [DP ui v̄¯ ] w(a)- u- sa-pe
DP ]
3s- John eat- rice- hot- subj agr and- m,s- fruit-mango
‘John ate hot rice and a mango.’
(Curiously Bareh, p.c.)
It is significant that only one conjunct is incorporated into the verb which is a
violation of subjacency. Note that the conjunction marker in Pnar is wa and, when
followed by a vowel, sandhi takes place and the vowel -a gets elided. The particle -u
is a diminutive marker which occurs with items which are very dear to the speaker,
a feature found in Mon-Khmer and some Tibeto-Burman languages.
9. Recall that with some verbs and stative adjectives in Hmar, the occurrence of the
verbal reflexive is obligatory. See chapter 3 for a discussion.
10. Vanlal Englien (p.c.) informs me that, in his dialect, it is not possible to have the
verbal reflexive in (44). Such prohibition could be due to the fact that adposition
incorporation into the verb makes the verb [+transitive] and the verbal reflexive/
reciprocal is [−transitive]. Hence, the verbal reflexive/reciprocal is prohibited as it
is not compatible with the transitive nature of the host verb.
11. The absence of LDA in finite clauses can be explained in recent theoretical terms as
follows: “[T]he embedded clause is a finite CP and given that CPs constitute strong
phases [in the sense of Chomsky 2001] and the Phase Impenetrability Condition
(PIC) of Chomsky (1999), the embedded finite T0 and the other TP-internal material
would be invisible to the matrix finite T0 ” (Bhatt 2005: 17). Since the T0 of S2 is
invisible, the probe–goal relationship cannot be established. A similar argument
can be extended to the antecedent–anaphor relationship too.
12. Sentences (62) and (63) are slightly modified versions of the original. We have used
the past auxiliary as it reflects phi-(PNG) features, unlike the present tense marker.
13. For some speakers in Hindi-Urdu too, agreement on the infinitive might take place
with the embedded object, and the matrix verb might exhibit default agreement. The
default agreement in Hindi-Urdu is the masculine singular form. frij ‘refrigerator’
is in feminine singular and the embedded infinitive cal-nı̄ ‘to start’ is also in feminine
singular. The matrix verb šurū ho jāegā ‘will start’ exhibits default agreement –
that is, masculine singular.
ek ghant.e mẽ frij cal- nı̄ šurū ho

one hour in refrigerator f,s. start inf f,s start become
jāegā
will m,s,def agr
‘The refrigerator will start working in an hour.’
14. Recall that possessor agreement takes place in Hmar (TB) too. In Hmar, the pronom-
inal clitic of the possessor of the DO occurs to the right of the subject agreement
marker (see (32)).
It is worth mentioning here that a possessor that ‘subcommands’ an anaphor in
Hindi-Urdu (IA) can be an antecedent to an anaphor too (see the cases involving
subcommand in chapter 3).
15. Raising is normally to the next higher clause. If raising takes place to a higher
clause, skipping one sentence in-between, it is called Superraising.
16. Sentence (79) is a case of incorporation of the comitative postposition pui ‘with’
in the verb (see Subbarao 2001: 467). (See section 4.3, sentences (47) and (48), for
postposition incorporation in Paite [TB].)
17. Note that the agreement marker on the embedded verb in (87) and (88) is null (ø),
as the subject of the embedded clause is no longer present there; it moves to the
matrix clause, and shows up as object agreement clitic.
18. The glosses have been slightly modified. See Ramakrishna Reddy (1980, 1992a)
for more details.
19. Note that -aŋ and -tur are allomorphs of the future tense marker; -aŋ occurs with
the matrix verb, while -tur occurs with the embedded verb.
20. See sentences (16) and (17) in Mizo (TB) in appendix 3 of chapter 8 on the Website.
In the modification of subject in EHRCs, the embedded subject in (16) and (17)
is not overtly present (indicated by ø). The embedded verb consequently does not
carry subject agreement marker (indicated by ø). This provides further evidence for
the Chomskyan hypothesis that case and agreement are intrinsically linked.
5 NON-NOMINATIVE S UBJECTS
1. This chapter derives its inspiration in content and spirit from the work by the author
in collaboration with Peri Bhaskararao on NNSs in Telugu at ILCAA, Tokyo
University of Foreign Studies, Tokyo, Japan. Crucial analysis and most of the
Telugu examples in this chapter have been taken from Subbarao and Bhaskararao
(2004); hence the source is not always cited. This chapter greatly benefited from
the two volumes on NNSs edited by Peri Bhaskararao and Subbarao (2004) and the
volume edited by M. K. Verma and K. P. Mohanan (1990) on Experiencer Subjects
in South Asian Languages.
2. In sentences involving Exceptional Case Marking, the subject of the embedded
sentence is accusative case-marked by the matrix verb.
3. In Rabha (TB) the accusative marker occurs with the subject to indicate need, or
want.
Rabha (TB)
(i) aŋ- a- be taka- be lagi ca- be
I- acc- emph money- emph want not- emph
‘I really do not want any money at all.’
In Manipuri (TB), an animate subject is accusative case-marked, and the sentence

has a ‘passive-like’ interpretation (Chelliah 1990: 205).
Manipuri (TB)
(ii) əy- bu- nə barton təw- wi
I- acc- foc invitation get- progr
‘I was invited, but not others.’
4. See Mahajan (2000), Kachru (2006), for a detailed discussion of the Hindi
passive.
5. Tiwa is a Tibeto-Burman language spoken in Assam in the northeastern part of
India. The following examples are illustrative.
Embedded subject is accusative case-marked:
(i) āŋ pe- go sinemā ni- wā nugāŋ

I he- acc film see- nozr saw
‘I saw him watching a film.’
Embedded subject is nominative case-marked:

(ii) āŋ pe sinemā ni- wā- ga nugāŋ
I he (nom) film see- nozr cl saw
‘I saw him watching a film.’
(Nath 2009)
6. vuchān ‘see.imperf’ may move leftwards to the right of chus-an ‘pres.1 m,s’ (Sadaf
Munshi, p.c.).
7. Sadaf Munshi (p.c.) provided further data which demonstrate that the occurrence
of the 2nd person agreement clitic of the derived object as a pronominal clitic on
the matrix verb yields grammatical sentences, as (i) and (ii) illustrate.
(i) bai chusi - athj- tsj dohay bat khyav- ān
I.nom pres.1 m,s- 2 pn- you.nom daily rice.acc eat- imperf
vuchān
see.imperf
‘I see you (m,s) eating rice every day.’
(ii) bai chsi - athj - tsj dohay bat khyav- ān

I.nom pres.1 f,s- 2 pn- you.nom daily rice.acc eat imperf
vuchān
see.imperf
‘I see you (f,s) eating rice every day.’
8. The fact that the clitic of the embedded subject occurs on the matrix verb in Kashmiri
(43b) and Dumi (44) shows that this is a case of long-distance agreement.
9. In Telugu (DR) -u is a bound morpheme (a clitic) that performs the function of
a conjunction marker, an emphatic marker, universal quantification marker and a
negative polarity marker. The occurrence of -u is phonologically conditioned, and
its form depends on the preceding vowel of the host (see Dilip, Subbarao and
Umamaheshwar Rao 2011).
10. In Telugu (DR) -u also functions as a marker to impart quantificational interpreta-
tion.
11. In colloquial Hindi the expression mere ko ‘I (obl)-dat’ is often heard, though the
formal expression cited in grammatical descriptions is mujh ko / mujhe ‘I (obl)-dat.’
Since this sentence is a verbatim quote from a native speaker of Hindi, we did not
make any alteration in the sentence. Such usage of mere ko ‘I (obl)-dat’ in place of
the expression mujh ko / mujhe ‘I (obl)-dat’ indicates that there is a syntactic change
in progress. Example (56) was reattested by Rama Kant Agnihotri and several other
native speakers.
12. The verb exhibits agreement in number and gender with the nominative case-marked
noun phrase nı̄nā ‘Nina.’
13. The verb exhibits default agreement as there is no nominative case-marked noun
phrase in (83).
14. EPP in simple terms means that every sentence must have a subject. (See Lasnik
and Uriagereka 2005 for further details.)
15. Masica (1991) prefers the terms neutral (unmarked)/non-volitional (marked) to the
terms volitional and non-volitional that are generally used. Alice Davison (p.c.)
agrees with Masica (1991) and points out that ergative is neutral between the
features [+/−volitional] while dative is invariably [−volitional].
16. Bayer’s sentence has been slightly modified. Our language consultant pronounces
the word for ‘went’ as gelam, and not as gεlam as Bayer transcribes it. Bayer
crosses out the word amar in (129). We have used the more familiar convention of
the parentheses marked by * to show that amar cannot be deleted.
17. Sentence (79) is a similar example from Hindi-Urdu (IA).
18. Dasgupta (2004: 131) points out that toma.r, which is generally glossed as
‘you.gen,’ should be treated as an indirect case form. For a discussion on this,
see Dasgupta, Ford and Singh (2000).
19. Šotti-šotti is a reduplicated form.
20. Even when the patient or theme is in contrastive focus or under emphasis, the
accusative marker is not needed, as below.
Tamil (DR)
kumār- ukku yār- um- e teri- yā- du
Kumar dat who npi- also know- neg- 3 s
‘Kumar does not know anybody at all.’
21. Magier (1987: 192–193) clearly articulated that ko in Hindi does not “convey rela-
tional information” when it occurs with “direct objects,” but “follows a semantic
hierarchy of specificity and animacy that contributes to the overall salience of the
marked object noun.”
22. Thanks to Alice Davison for the formulation of this parameter.
23. See the papers in Bhaskararao and Subbarao (2004) for some discussion on this
issue.
24. Rajesh Bhatt (p.c.) supports Alice Davison’s contention.
25. Though AGR phrase is dispensed with in Chomsky (1995b), our analysis shows
that agreement does play a role in dative case assignment. Further, the fact that
quantifiers, emphatics, negative and negative polarity items are incorporated as
a part of the auxiliary in Tibeto-Burman languages and Khasi (see chapter 4
for an elaborate discussion) shows that the notion of AGR phrase needs to be
incorporated in the theory, and it cannot be dispensed with, as Subbarao (1998)
argued.
26. Tamil (DR), in contrast, has the dative–locative double subject construction
(Parameshwari Krishnamurthy and Rajeshwari, p.c).
27. The marker -ā also functions as a polarity question marker in Dravidian
languages.
28. The marker -a may also occur on all the other constituents in the sentence including
the verb. See Bhaskararao and Subbarao (in preparation).
29. Dakkhini (IA) has borrowed the Double Dative Subject construction (i) from Telugu
(DR). It is not permitted in standard Hindi-Urdu (IA), as in (ii).
Dakkhini (IA)
(i) us koi hāth koi cot.j lag- ı̄∗ i,j
he dat hand dat injury hit- pst f,s
326 Note to page 192
Hindi-Urdu (IA)
(ii) *us koi hāth koi cot.j lag- ı̄∗ i,j
he dat hand dat injury hit- pst f,s
The Dakkhini case shows that the phenomenon of inherent case assignment may
be borrowed from a member of a different language family.
We shall present below data that illustrate the Single and Double Dative Subject
constructions in Mangalore Konkani (IA) (Lalita Dhareshwar, p.c.).
s i n g l e d a t i v e s u b j e c t c o n s t r u c t i o n ( w i t h t h e w h o l e , ta ‘he,’
t h e p o s s e s s o r , i n t h e g e n i t i v e c a s e , a n d t h e p a r t , hātt ‘hand,’
in the dative case)
(iii) tagal- ya hātt- ā- ka lāg- l- ε
he.gen- obl hand- obl- dat hurt- perf- def agr
d o u b l e d a t i v e s u b j e c t c o n s t r u c t i o n ( w i t h t h e w h o l e , ta ‘he,’
i n t h e d a t i v e c a s e , a n d hātt ‘ h a n d ’ i n t h e d a t i v e c a s e )
(iv) tā- -kkā hātt- a- ka lāg- l- ε
he.obl- dat hand- obl- dat hurt- perf- def agr
Just as in Dravidian languages, the Single Dative Subject construction in Mangalore

Konkani also does not permit an adverb to occur between the possessor tagalε ‘his’
in the genitive case and the possessed hātt ‘hand’ in the dative case.
single dative subject construction with an intervening

adverb – not permitted
(v) *tagalε aadzi hāt- aa- ka lāg- l- ε
he.gen today hand- obl- dat hurt- perf- def agr
However, it is possible for an adverb to occur to the right of the possessor that is
dative-case marked as in (iv).
double dative subject construction with an intervening

adverb – permitted
(vi) tā- -kkā ādzi hātt- a- ka lāg- l- ε
he.obl- dat today hand- obl- dat hurt- perf- def agr
Due to limitations of space we have not discussed the nature of non-nominative

subjects in contact languages Dakkhini (IA) and Mangalore Konkani (IA) (see
Arora 2004, and Arora and Subbarao 2004, for a detailed discussion).
Marathi (IA) also has the Double Dative Subject construction (Lalita
Dhareshwar, p.c.). We do not have any information about its occurrence in Goan
Konkani.
6 COMPLEMENTATION
1. Complementation is a process in which a clause is embedded in a matrix (higher
clause) either in its subject or object position. Such embedding depends upon the
argument structure of the matrix predicate – that is, whether the matrix predicate
permits an embedded sentential complement or not. One set of predicates permits
subject complements, while another permits object complements depending on the
nature of the predicate itself.
2. As Alice Davison (p.c.) rightly points out, the FCs in (8) and (9) are “different
from complements” as the first complementizer performs reason function. For a
discussion, see 6.2.2.
3. There is no V2 phenomenon observed in embedded relative clauses and adverbial
clauses in Kashmiri. An explanation in terms of the absence of the mood phrase
in such clauses is provided in Bhatt (1999). See also sentences (5) and (6) in
appendix 6.1 on the Website for an explanation in terms of the etymology of the
complementizer.
4. In Hindi-Urdu (IA), the IC occurs as a disjunctive marker in embedded questions.
Hindi-Urdu (IA)
rekhā ko mālūm hogā ki karan ā rahā hai ki nahı̄˜
Rekha dat known will IC Karan come progr pres or not
‘Rekha would know whether Karan is coming or not.’
5. Recall that Marathi and Gujarati (IA) have another form of the FC. In Marathi it is
asə and in Gujarati it is εm (Masica 1991: 403).
6. As we shall see later, the IC ki ‘that’ of Hindi-Urdu (IA) is reanalyzed as a post-
nominal and post-sentential marker in Dakkhini (IA) and it performs a different set
of functions, such as a disjunctive marker, and as a dubitative marker patterning the
Dravidian complementizer -ō (Subbarao and Arora 1989; Arora 2004). Such a shift
in position, we argue in appendix 6.5 on the Website, is due to syntactic reanalysis.
7. To account for the phi-features of COMP in Juang, Patnaik (1999) proposes that the
quotative construction be treated as a control structure where the complementizer is
treated as the verb of a clause with PRO as its subject. This PRO in turn is coindexed
with the subject of the matrix clause.
8. Although, in Sinhala (IA) and Telugu (DR), the quotative occurs with a verb of
perception such as see, it performs the function of a conditional in Sinhala in (i) and
that of a reason marker in Telugu in (ii). Thus, sentences (i) and (ii) do not go against
Bayer’s claim.
Sinhala (IA)
(i) gihiŋ malli tamange kāmərē innəwa də
go.perf pple brother self.gen room.loc be (animate) q mkr
kiyəla balannə
quot look.inf
‘Go and see if younger brother is in his (own) room.’
(Gair 2003: 807)
Telugu (DR)
(ii) pūlı̄su vād.u vastunnād.u ani cūs- i donga pāripōyed.u
policeman is coming quot see- cpm thief ran away
‘Having seen that the policeman was coming, the thief ran away.’
9. In Bangla (IA), the complementizer je is compatible with a question word inside
the complement clause in specific cases. Probal Dasgupta (p.c.) observes that it
is the explicit and intricate information in the embedded clause that forces an
interpretation of the ‘fact’-type rather than the ‘indirect question’-type and this is
what triggers the je option.
lɔta nije- o janto, amake janiye diyechilo je piknik-e
Lata herself- also knew me- dat informed that picnic-loc
kara ašbe ar kara ašbe na
who will come and who will come not
‘Lata herself also knew, she informed me too as to who will come and
who will not to the picnic.’
10. In Sanskrit (IA), the embedded clause with the IC yat ‘that’ occurs to the right of
the VP of the matrix clause. For example:
vij
ān- āmy’ aham yat [prān.- o
perceive- 1s I.nom that breath- m,s,nom m,s,nom
brahma]
Brahman neuter,nom/acc
‘I understand that breath is Brahman.’
(Chandogya Upanishad 4,10,6, cited by Speijer
1886 [1980]: 359)
11. The following examples are illustrative.
languages with fc
Telugu (DR)
(i) pillalu vaccēr- ā ani amma ad.igindi
children come- pol q FC mother asked
‘The mother asked whether the children had come.’
languages with ic
Hindi-Urdu (IA)
(ii) us ne mujh se pūchā ki kyā tum āoge
he erg me asked IC pol q you come.fut
‘He asked me whether you will come.’
(Davison 2008: 181)
12. Thanks to Suranjana Barua and Shukla Basu for confirming the Assamese and the
Bangla data, respectively.
13. The anaphor swatāhā-lā ‘self–acc’ requires a strict clause-mate condition, and
hence it is permitted if and only if the “antecedent is within the same clause” in
which it occurs (Pandharipande 1997: 210).
14. The canonical position of the complement clause in verb-final languages is the
pre-verbal argument position.
15. In object complement clauses, the complement clause occurs in the ‘object position’
of the matrix clause, and it is called the ‘canonical position’ – that is, the position
in which a complement occurs before it is moved to some other position.
16. By Sentence-Lifting, we mean an embedded complement clause is moved from its
canonical clause-internal position and is placed to the left of the matrix sentence.
The complementizer does not occur with the embedded clause in such cases. It was
John Robert Ross who proposed the term ‘Sentence-Lifting’ (Ross 1973).
17. We labeled -ı̄ as ‘comp’ as it links the main clause with embedded complements.
18. Thanks to Probal Dasgupta (p.c.) for this summary. Dasgupta points out that if the
notion of zero complementizer is invoked, the facts concerning Wide and Narrow
Scope in Bangla can be neatly explained (see Dasgupta 2007).
19. Sentence (133) is not ungrammatical, though the embedded S occurs clause-
internally. The reason for the occurrence of a clause with an FC clause-internally
in languages with an FC is that the FC functions like a case marker shielding its
clause from the CRP.
20. We have slightly modified Dasgupta’s suggestion by replacing his ZERO NP cor-
relate by null DEM. The idea of having a null DEM is plausible as Bangla, like the
other SALs, is a pro-drop language. Under this analysis, there is a zero comple-
mentizer and a null DEM.
21. Taking such facts concerning scope marking in Bangla (IA) into consideration,
Simpson and Bhattacharya (2003: 127) argue that “Bangla actually has obligatory
wh-movement from a basic SVO word order,” and such movement is overt in
Bangla. We do not wish to go into the details as it is beyond the scope of the present
work. For further information, see Simpson and Bhattacharya (2003).
22. Thanks to Rajesh Bhatt for bringing this important paper to our notice.
23. According to Koul and Wali (2006: 144), Kashmiri also has the question marker
-ā which is “added to the finite predicate at the end of all inflections.” It also has a
question marker k’ā which occurs in the initial position, just as in Hindi-Urdu and
Punjabi.
24. Davison (p.c.) rightly observes that: (i) FOFC permits such sentences, as discussed
in Biberauer et al. (2009: 13); and (ii) Rajeshwari Pandharipande (1997) cites such
examples as grammatical. However, Lalita Dhareshwar (p.c.) finds such sentences
ungrammatical. We leave this open for further research.
25. See Biberauer et al. (2009) for a detailed discussion.
26. According to Biberauer et al. (2009: 5), categorial non-distinctness, it appears,
includes categories such as VP, DP, PP, C, inflected auxiliaries and verbs. They
admit that “the notion of ‘categorial distinctness’ is by no means ‘straightforward.’”
7 BACK WA RD C ONT ROL

1. These alternations are labeled as Stem I and Stem II alternations in the descriptions
of the Mizo language.
2. Thanks to Lalita Dhareshwar for providing this information.

3. When case is assigned due to the idiosyncratic, inherent property of a lexical item,
it is labeled as inherent case.
8 NOUN MODIFICATION: RELATIVE CLAUSES

1. Riemsdijk (2006: 338) further adds: “within certain limitations, PPs, CPs, and APs
can also be modified by a relative clause.” As we shall see later, a time or place
expression, or an expression denoting quantity, quality, manner, etc., can be a head of
a relative clause in SALs. In internally headed relative clauses (IHRCs), the internal
head may overtly carry a postposition that indicates the thematic relationship of the
constituent with the predicate (see 8.6 for details).
2. We are thankful to Rajesh Bhatt for his insightful comments and for illuminating
discussions which, in our opinion, have enhanced the quality of this chapter in
terms of new ideas and presentation.
3. For example, as Fox (1987: 856) points out: “I will use the terms ‘subject/object
relative’ to refer to the grammatical role of the NP in the relative clause.” Thus, in
the sentence The book that Sheila bought was expensive, it is the direct object that
is relativized.
4. Evidence in support of having a (null) pro in the subject position of the matrix
clause will be provided in 8.6.
5. See Verma (1971) for a discussion concerning Hindi.
6. South Slavic languages which are non-verb-final have the relative–correlative type
(Bhatt 2003: 491).
7. Khasi, a neighboring language to Tibeto-Burman languages, is verb-medial and it
does not have the IHRC.
8. The empty element in the IHRC in the matrix clause in (7) and (8) is indicated by
ø. The empty element controls agreement of the matrix predicate (verb/adjective).
Hence, we label it as pro though it can never be realized overtly as a pronominal.
We may hence designate it as a null pro.
9. Literary Sinhala, however, has the relative–correlative construction (James W. Gair,
p.c.).
10. There are certain asymmetries in the three types of finite relative clauses, which
we shall subsequently discuss. In Dravidian languages, the correlative is identical
not with the demonstrative pronoun but with the [−proximate] 3rd person personal
pronoun, as in (18) and (19) below.
11. Thanks to Rajesh Bhatt for a helpful discussion.
12. The SOV order is sensitive to a relative head, including adverbial relatives with
‘where,’ ‘when,’ etc., as heads as in (16) below.
13. In Marathi (IA) too, such an occurrence is permitted, as in (9) above, while in
Hindi-Urdu (IA) it is not.
14. It follows that, in a non-finite EHRC or a non-finite IHRC, there cannot be a relative
pronoun available, nor can there be a correlative pronoun.
15. This conjecture needs to be sustantiated with further evidence.
16. A small set of predicates such as par.hā-likhā ‘learned,’ pahnā huā ‘wearing,’ lagāyā
huā ‘wearing,’ sı̄khā huā ‘learned,’ piyā huā ‘drunk,’ in their perfect participial form,
do permit the modification of an ergative subject in Hindi-Urdu (IA). Note that all
such predicates in their perfect participial form manifest a resultant state.
17. The nominative theme in a dative subject construction may be modified in Hindi-
Urdu (Rajesh Bhatt, p.c.) as in (i) below, and Punjabi (Om Prakash Arora,
p.c.).
Hindi-Urdu (IA)
(i) rām ko lagı̄ cot. bahut gahrı̄ hai
Ram dat get hurt.perf pple injury very deep is
‘The injury that Ram had is quite deep.’
Bangla (IA), however, does not permit such modification.
(ii)* rām- er lag.a bætha khub beši chilo

Ram gen get hurt.perf pple injury very more was
‘The injury that Ram had was quite bad.’
(Shukla Basu, p.c.)
It needs to be checked how other Indo-Aryan languages behave in this regard.

18. The occurrence of a nominative case marker (overt or covert) or an ergative case
marker is a typical feature of transitive subjecthood.
19. The semantic/syntactic restrictions, if any, need to be worked out in Hindi-Urdu.
Thanks to Rajesh Bhatt for bringing this fact to our attention and for the data in
(30a) and (30b).
20. Punjabi (IA) permits such modification (Om Prakash Arora, p.c.).
21. Ergative case marking on the embedded subject in an EHRC is also found in Eastern
Shina (IA).
22. In Assamese (IA) too the embedded subject is ergative case-marked.
ram- e por.h- a kitap khon

Ram- erg read- perf pple book cl
‘The book Ram read.’
(Gautam Bora and Ajit Tamoli, p.c.)
23. Some speakers accept sentences in which the IO bacce ‘children’ is modified with
the causative (ditransitive) verb par.hānā ‘to teach’ in Hindi-Urdu (IA).
Hindi-Urdu (IA)
[shaguftā ke urdū par.hā- ye hue] baccei
Shagufta gen Urdu teach perf pple m,p children m,p
‘The children (who were) taught Urdu by Shagufta.’
24. Hindi and Bangla, Assamese and Oriya (IA), however, permit sentences such as
(i) and (iii) in which a locative PP is modified. Such occurrence of locative PP
modification is highly contextualized as the ungrammaticality of (ii) from Hindi-
Urdu and of (iv) from Bangla shows.
Hindi-Urdu (IA)
(i) øi mehandı̄ rac- e hāthi
paste made of a type of leaf put- perf pple hands
‘Hands in which the ‘mehandi’ paste is put.’
(ii) *thailı̄ rakh- ı̄ huı̄ jagah

bag keep perf pple place
‘The place in which the bag is kept . . . ’
(iii) øi alta pɔr-

‘alta’ (a red-colored liquid adorning the feet of girls/women) wear-
a pai
perf pple foot
‘The foot on which ‘alta’ is worn . . . ’
Bangla (IA)
(iv) *thole rakh- a jayga
bag keep- perf pple place
‘The place in which the bag is kept . . . ’
(Shukla Basu, p.c.)
How productive the examples such as (i) and (iii) are, needs to be worked out.
25. The right kind of example for the comitative should be
*[polı̄s øi ā.lel.ā] tsori

policeman come.pst pple.3 s,m thief
‘The thief whom the policeman came with.’
as ‘the thief’ is the head of the comitative PP and not ‘the policeman.’ Cf. 8.6 for
further discussion.
26. Based on evidence from seven IA languages, in Subbarao (1984b), it was shown
that Keenan and Comrie’s NPAH cannot be maintained to be a universal, and there
are exceptions to it. For example, in Hindi-Urdu, Punjabi, Bangla and some other
Indo-Aryan languages (except Marathi, Oriya and Nepali), the Oblique Object (OO)
cannot be relativized while the object of the genitive can be, which is lower than
the OO in the hierarchy.
27. However, in Marathi (IA), it is the participial strategy that is “preferred.” Accord-
ing to Pandharipande (1997: 74), “though Marathi allows the use of relative
participles . . . as well as relative clauses, speakers of Marathi show a clear prefer-
ence for participles over finite relative clauses for communicating the same meaning.
In this respect, Marathi is closer to the Dravidian languages as opposed to the Indo-
Aryan languages, since the speakers of the former show a preference for participles
over relative clauses.” This comment is equally applicable to standard Konkani and
Mangalore Konkani (IA) too (Lalita Dhareshwar, p.c.).
28. In Hindi-Urdu and some other IA languages, most of the intransitive verbs such
as kūdnā ‘jump,’ daur.nā ‘run,’ calnā ‘walk,’ do not permit the modification of the
subject. For example:
Hindi-Urdu (IA)
*āj subah dı̄wār kūd.ı̄ huı̄ lar.kı̄
today morning wall jump.perf pple girl
‘The girl who jumped the wall this morning . . . ’
29. Recall that, in Indo-Aryan Hindi-Urdu, Punjabi, Bangla, etc., the subject is invari-
ably genitive case-marked, while in Sinhala and Dakkhini (IA) it is always nomina-
tive case-marked, and in Oriya (IA) it is either genitive or nominative case-marked.
30. In Telugu and other Dravidian languages, “when a time expression occurs in the
predicate of the matrix clause, the truth-value of the sentences with an affirmative
CP and negative CP is ‘nearly identical’. . . ” (Krishnamurti and Gwynn 1985: 194)
(see section 9.6.1 on the Website for a detailed discussion).
31. See Asher and Kumari (1997: 61–62) and Ramarao (2003) for a detailed discussion
of permissible sentences with other interpretations of the sociative.
32. The construction kustı̄ pat..tu, literally ‘wrestling [noun] catch,’ is a conjunct verb,
and the noun and the verb together impart the meaning of ‘wrestle.’
33. There are some exceptions to this generalization as in sentence (30a) above. The
precise characterization of the predicates permitting such cases in Hindi-Urdu and
other Indo-Aryan languages needs to be worked out.
34. The case marker does not occur when the DO or IO is modified. However, when an
OO is modified, the case marker with the internal head may occur.
35. Thanks to Rajesh Bhatt for a very helpful discussion on this point.
36. Kannada (DR) also does not permit the comitative PP as head in EHRCs.
Kannada (DR)
*hud.uga (ø-jote) banda sušı̄la
boy (ø-with) came.pst.adjr Susheela
‘Susheela, who the boy came with’
(Sridhar 1990: 58)
37. Mangalore Konkani, too, does not permit such sentences.

*karun.a vatsu (gal)-lel- i tsalli
Karuna go- perf pple-f.s girl
Note that gal ‘go’ in the perfect aspectual suppletive form of the verb stem vats is
similar to the suppletive form ga- of the verb stem jā ‘go’ in Hindi-Urdu (Lalita
Dhareshwar, p.c.).
38. An example from Dimasa (TB) is given below.
Dimasa (TB)
aŋ hathai- ha thaŋ- pha- lai- ya- ba masainjik
I market- to go- together- gp mkr / VREC adjr- nozr girl
ani besik
my daughter
‘The girl with whom I went to the market is my daughter.’
(Parichita Langthasa and Ron Kemprai, p.c.)
Note that a ŋ‘I’ in (i) does not carry an ergative marker though the verbal reciprocal /
group marker lai and the adverb pha ‘together’ are incorporated. The reason for
such non-occurrence is due to the fact a verbal anaphor “detransitivizes” a predicate
and, hence, unlike an incorporated postposition, it does not have the capability to
increase the valence of a predicate. We can therefore label a verbal anaphor as a
“valence reducer.” See chapter 3 for further discussion.
39. A similar phenomenon of valence augmenting is observed in Hmar and Paite. See
section 4.3 on Postposition Incorporation in these languages.
40. The verbal reciprocal ki in (94) in Thadou functions as a group activity marker
just as the verbal reciprocal in Khasi and Manipuri. In Thadou, verbs such
as to sit, to walk, to swim, to crawl, to stand and to run require the occur-
rence of the verbal reciprocal / group activity marker ki when the comitative PP
heads an EHRC. Thanks to Pauthong Haokip for a helpful discussion on this
issue.
Note that it is due to incorporation of the postposition pi ‘with’ that the embed-
ded predicate ɔn ‘go’ is transitivized in spite of the presence of the group marker /
verbal reciprocal ki. Recall that the verbal reciprocal detransitivizes a predicate
(cf. chapter 3). In spite of the occurrence of the verbal reciprocal as a group
marker in the embedded predicate, the embedded subject carries the ergative
marker. We hypothesize that the effect of postposition incorporation neutralizes
the “detransitivization effect” of the verbal reciprocal and supersedes its effect.
This hypothesis, however, needs to be empirically tested with more data and
analysis.
41. The following example from Manipuri is illustrative where the subject is in the
plural. Sentence (15) from chapter 3 is repeated below.
Manipuri (TB)
(15) tomba- ga tombi- ga skūl- da cat- min- na- khre
Tomba- and Tombi- and school- to go- together- VREC- pst
‘Tomba and Tombi went to school together.’
42. Recall that postposition incorporation has a “transitivizing effect” on the host verb
that incorporates it. See the case of Hmar and Paite (TB) in chapter 4.
43. In Mising and Bodo (TB), a comitative PP can head an EHRC when the embedded
verb carries the incorporated comitative postposition, which in Mising is zon and
Bodo is phā, as in (i) and (ii), respectively.
Mising (TB)
(i) sita- kə bozar- pɔ g- zon- hom konə- dɔ ŋo- k brmə
Sita- gen bazaar- to go- com- nzr girl- def I- gen friend
‘The girl with whom Sita went to the movie is my friend.’
(Jugen Pegu, Shikha Dele and Darika Pegu, p.c.)
Bodo (TB)
(ii) sita- ya hātha- yau thaŋ- phā hinjāu- sā- yā aŋ-
Sita- gen bazaar- to go- com girl- dm- nom I-
ni binānāu
gen sister
‘The girl with whom Sita went to the movie is my sister.’
(Ganga Brahma, p.c.)
The occurrence of the genitive postposition in (i) with the embedded subject in
Mising is due to superstratum influence of Assamese with which languages such
as Mising, Bodo, Rabha and Dimasa, spoken in Assam, have been in contact for
centuries.
44. This analysis is abstracted from Kumar and Subbarao (2005), and, hence, the source
of the text and each example is not mentioned.
45. There is another determiner DD1 which occures to left of the relative clause and it
is optional. DD1 is not relevant for our discussion here, and hence it is omitted.
46. A set of intransitive verbs and adjectives in Hmar carry the verbal reflexive marker
and such occurrence is lexically determined (see chapter 3 for details).
47. The personal pronoun kei ‘I’ has two agreement markers ka- and -min- both of
which are bound morphemes; ka- is subject agreement marker and -min- is object
agreement marker. For example:
Hmar (TB)
(i) kei- maʔ- cu- an sikul- aʔ lali- cu ka- hmu- aŋ
l- emph- def- erg school- at Lali- def 1 s- see/meet- fut
‘Only I’ll see/meet Lali at school.’
(ii) lali- n sikul- aʔ a- min- hmu- aŋ
Lali- erg school- at 3 s- me (DO) see/meet- fut
‘Lali will see/meet me at school.’
48. This section presents conclusions drawn on the basis of material presented in the
main text of the chapter and the appendices of this chapter in the web material.
Bibliography
Abbi, A., 1980. Semantic Grammar of Hindi: A Study of Reduplication (New Delhi:
Bahri Publications Private Ltd.).
1984. “The Conjunctive Participle in Hindi-Urdu,” International Journal of Dravidian
Linguistics, 13.2, 252–263.
1990. “Experiential Constructions and the ‘Subjecthood’ of the Experiencer NPs in
South Asian Languages,” in Verma and Mohanan (1990: 253–268).
1992. Reduplication in South Asian Languages: An Areal, Typological and Historical
Study (New Delhi: Allied Publishers Ltd.).
1994. Semantic Universals in Indian Languages (Shimla: Indian Institute of Advanced
Study).
2001. A Manual of Linguistic Field Work and Indian Language Structures (Munich:
Lincom Europa).
2006. Endangered Languages of the Andaman Islands (Munich: Lincom Europa).
Achumi, R., 2000. “Sentential Complementation in Sema,” M.Phil. dissertation, Delhi
University, Delhi.
Achumi, R., and K. V. Subbarao, 2005. “Sentential Complementation in Sema,”
paper presented at the International Conference on South Asian Languages and
Literatures-6, Osmania University, Hyderabad, India.
Agnihotri, R., 1992. “India: Multilingual Perspectives,” in N. D. Crawhill (ed.), Demo-
cratically Speaking (Capetown: National Language Project), 46–55.
2001. Half the Battle and a Quarter, CIIL, Foundation Day Lectures (Mysore: Central
Institute of Indian Languages).
Aissen, J., 2003. “Differential Object Marking: Iconicity vs. Economy,” Natural Lan-
guage and Linguistic Theory, 21, 435–483.
Alexiadou, A., E. Anagnostopolou and F. Schafer, 2006. “The Properties of Anti-
causatives Crosslinguistically,” in M. Frascarelli (ed.), Phases of Interpretation
(Berlin: Mouton), 187–212.
Amritavalli, R., 2000. “Lexical Anaphors and Pronouns in Kannada,” in Lust et al.
(2000: 49–112).
2004. “Experiencer Datives in Kannada,” in Bhaskararao and Subbarao (2004: I,
1–24).
Anagnostopoulou, E., and M. Everaert, 1996. “How Exceptional are Nominative
Anaphors? A Case Study of Greek,” in L. Nash, G. Tsoulas and A. Zribi-Hertz
(eds.), Actes du deuxiēme colloque langue et grammaire (Paris: Department of
Language Sciences, Paris-8), 19–32.
336
Bibliography 337
Anderson, G. D. S. (ed.), The Munda Verb: Typological Perspectives (Berlin and New
York: Mouton de Gruyter, 2007).
(ed.) 2008. The Munda Languages (London: Routledge).
Anderson, G. D. S., and K. D. Harrison, 2008. “Sora,” in Anderson (2008: 299–380).
Annamalai, E. 1997. Adjectival Clauses in Tamil (ILCAA, Tokyo University of Foreign
Studies, Tokyo, Japan).
2000. “Lexical Anaphors and Pronouns in Tamil,” in Lust et al. (2000: 169–216).
Arora, H., 2004. Syntactic Convergence: The Case of Dakkhini Hindi-Urdu (Delhi:
Publication Division, Delhi University).
Arora, H., and K. V. Subbarao, 2004. “Syntactic Change and Convergence,” in
Bhaskararao and Subbarao (2004: II, 25–48).
Asher, R. E., 1985. Tamil. Lingua Descriptive Studies 7 (Amsterdam: North-Holland).
Asher, R. E., and T. C. Kumari, 1997. Malayalam (London and New York: Routledge).
Aze, F. R., 1973. “Clause Patterns in Parengi-Gorum,” in R. L. Trail (ed.), Patterns
in Clause, Sentence, Discourse in Selected Languages of India and Nepal, vol. I
(Norman, OK: Summer Institute of Linguistics, University of Oklahoma), 235–
313.
Baker, M. C., 1988. Incorporation: A Theory of Grammatical Function Changing
(Chicago: University of Chicago Press).
1996. The Polysynthesis Parameter (New York and Oxford: Oxford University Press).
2001. Atoms of Language (New York: Basic Books).
2010. “Formal Generative Typology,” in B. Heine and H. Narrog (eds.), The
Oxford Handbook of Linguistic Analysis (Oxford: Oxford University Press),
285–312.
Baker, M. C., and J. McLoskey, 2007. “On the Relationship of Typology to Theoretical
Syntax,” Linguistic Typology, 11.1, 285–296.
Bal, B. K., 1990. “COMP and Complementation in Oriya and English,” Ph.D. thesis,
Central Institute of English and Foreign Languages, Hyderabad.
Barbora, M., 2001. “COMPs in Assamese: Implications in Second Language Acquisi-
tion,” Ph.D. thesis, Tezpur University, Tezpur.
Bareh, C., 2007. “Descriptive Analysis of the Jowai and Rymbai Dialects of Khasi,”
Ph.D. thesis, North Eastern Hill University, Shillong.
Barua, S., 2003. “Typological Characteristics of Kokborok,” Masters’ degree fieldwork
dissertation, Delhi University.
Bashir, E., 1988. “Topics in Kalasha Syntax: An Areal and Typological Perspective,”
doctoral dissertation, The University of Michigan, Ann Arbor.
2003. “Dardic,” in G. Cardona and D. Jain (eds.), Indo-Aryan (London: Routledge),
818–894.
Bayer, J., 1996. Directionality and Logical Form: On the Scope of Focusing Particles
and wh-in-situ (Dordrecht, Boston and London: Kluwer).
2001. “Two Grammars in One: Sentential Complements and Complementizers in
Bangla and other South Asian Languages,” in Bhaskararao and Subbarao (2001:
11–36).
2004. “Non-nominative Subjects in Comparison,” in Bhaskararao and Subbarao
(2004: I, 48–76).
Bayer, J., T. Bhattacharya and M. T. Hany Babu, 2007. Linguistic Theory and South
Asian Languages: Essays in Honour of K. A. Jayaseelan (Amsterdam: John
Benjamins).
338 Bibliography
Beermann, D. A., and L. Hellan, 2002. “VP-Chaining in Oriya,” in M. Butt and

T. Holloway King (eds.), Proceedings of the LFG02 Conference, National Tech-
nical University of Athens, Athens (Stanford, CA: CSLI Publications, Stanford
University), 41–56.
Begum, R., 2004. “The Conjunctive Participle in Bodo,” Masters’ fieldwork dissertation,
Delhi University.
Benmamoun, A., 2006. “Licensing Configurations: The Puzzle of Head Negative Polar-
ity Items,” Linguistic Inquiry, 37, 141–149.
Bhaskararao, P., 1972. “On the Syntax of Telugu Existential and Copulative Predica-
tions,” in J. W. Verhar (ed.), The verb “be” and its synonyms. Foundations of
Language, supplementary series, 5 (Dordrecht: D. Reidel Publishing Company),
153–206.
Bhaskararao, P., and K. V. Subbarao (eds.), 2001. The Yearbook of South Asian Lan-
guages (New Delhi: Thousand Oaks; London: Sage Publications).
2004. Non-nominative Subjects, vols. I and II (Amsterdam and Philadelphia: John
Benjamins).
in preparation. A Grammar of Modern Telugu (Amsterdam and Philadelphia: John
Benjamins).
Bhat, D. N. S., 1998. “A Typology of Himalayan Languages,” keynote address delivered
at the 4th Himalayan Languages Conference, Pune, India.
Bhat, D. N. S., and M. S. Ningomba, 1995. Manipuri Grammar (Mysore, India: Central
Institute of Indian Languages).
Bhatia, T. K., 1993. Punjabi: A Cognitive–Typological Study (London: Routlege).
1995. Negation in South Asian Languages (New Delhi and Patiala: Indian Institute
of Language Studies).
2000. “Lexical Anaphors and Pronouns in Punjabi,” in Lust et al. (2000: 637–714).
Bhatt, R., 2003. “Locality in Correlatives,” Natural Language and Linguistic Theory,
21.3, 485–541.
2005. “Long-distance Agreement in Hindi-Urdu,” Natural Language and Linguistic
Theory, 23, 758–807.
Bhatt, R., and V. Dayal, 2007. “Rightward Scrambling as Rightward Remnant Move-
ment,” Linguistic Inquiry, 38.2, 287–301.
Bhatt, R. M., 1994. “The Case of Quirky Constructions,” in A. Davison, G. Silva and
W. Yan (eds.), Papers from the Fourth Annual Meeting of the Formal Linguistic
Society of Mid America, 23–34.
1999. Verb Movement and the Syntax of Kashmiri (London: Kluwer Academic
Publications).
Bhatt, R. M., and J. Yoon, 1991. “The Composition of COMP and Parameters of V2,”
WCCFL, 10, 41–52.
Bhattacharya, T., 2005. The Yearbook of South Asian Languages (Berlin: Mouton de
Gruyter).
Biberauer, T., A. Holmberg and I. Roberts, 2007. “Disharmonic Word-order Systems
and the Final-over-Final Constraint (FOFC),” in A. Bisetto and F. Barbieri (eds.),
Proceedings of XXXIII Incontro di Grammatica Generativa, 86–105.
2008. “Structure and Linearization in Disharmonic Word Orders,” in C. B. Chang
and H. J. Haynie (eds.), Proceedings of the West Coast Conference on Formal
Linguistics (Somerville, MA: Casadilla Proceedings Project), 96–104.
Bibliography 339
Biberauer, T., G. Newton and M. Sheehan, 2009. “The Final-over-Final Constraint

and Predictions for Diachronic Change,” Toronto Working Papers in Linguistics
(TWPL) , 31, 1–17.
Bickel, B., 1998. “Review Article: Converbs in Cross-linguistic Perspective,” Linguistic
Typology, 2, 381–397.
1999. “Nominalization and Focus in Some Kiranti Languages,” in Y. Yadava and
W. G. Glover (eds.), Topics in Nepalese Linguistics (Kathmandu: Royal Nepal
Academy), 271–296.
2004. “The Syntax of Experiencers in the Himalayas,” in Bhaskararao and Subbarao
(2004: I, 77–111).
Bickel, B., W. Bisang and Y. P. Yadava, 1999. “Face vs. Empathy: The Social Foundation
of Maithili Verb Agreement,” Linguistics, 37.3, 481–518.
Bloch, J., 1934. L’Indo-aryen du veda aux temps modern (Paris: Librairie Adrien
Maisonnneuve).
Burling, R., 2004. The Language of the Modhupur Mandi (Garo) (New Delhi: Biblio-
phile South Asia / Promilla & Co.).
Burrows, L., [1915] 1980. The Grammar of the Ho Language: An Eastern Himalayan
Dialect (New Delhi: Cosmos Publications).
Burzio, L., 1986. Italian Syntax (Dordrecht: Reidel).
Butt, M., 1993. “A Reanalysis of Long-distance Agreement in Urdu,” in Proceedings of
the Nineteenth Annual Meeting of the Berkeley Linguistic Society (Berkeley, CA).
1995. The Structure of Complex Predicates in Urdu (Stanford, CA: CSLI Publications,
Stanford University).
Butt, M., and W. Geuder, 1999. “On the Lexical Status of Light Verbs,” paper presented
at the Workshop on Semi-lexical Heads, Tilburg.
Butt, M., T. Holloway King and G. Ramchand, 1994. Theoretical Perspectives on
Word Order in South Asian Languages (Stanford, CA: CSLI Publications, Stanford
University).
Cardona, G., 1965. A Gujarati Reference Grammar (Philadelphia: University of Penn-
sylvania Press).
Cardona, G., and D. Jain, 2003. The Indo-Aryan Languages (London and New York:
Routledge).
Chandra, P., 2005. “Long-distance Agreement in Hindi,” paper presented at the South
Asian Languages Analysis conference, University of Illinois, Urbana-Champaign,
Illinois.
2007. “(Dis)Agree: Movement and Agreement Reconsidered,” doctoral dissertation,
University of Maryland.
Chatterji, S. K., 1926. The Origin and Development of the Bengali Language, 3 vols.
(Calcutta: Calcutta University Press; rpt. London: Allen and Unwin).
Chelliah, S., 1990. “Experiencer Subjects in Manipuri,” in Verma and Mohanan (1990:
195–212).
1997. A Grammar of Manipuri (Berlin and New York: Mouton de Gruyter).
Chomsky, N., 1965. Aspects of the Theory of Syntax (Cambridge, MA: MIT Press).
1973. “Conditions on Transformations,” in S. R. Anderson and P. Kiparsky (eds.),
A Festschrift for Morris Halle (Cambridge, MA: MIT Press).
1975. Reflections on Language (New York: Pantheon).
1981. Lectures on Government and Binding (Holland: Foris Publications).
340 Bibliography
1986. Knowledge of Language: Its Nature, Origins and Use (New York: Praeger).
1995a. The Minimalist Program (Cambridge, MA: MIT Press).
1995b. “A Minimalist Program for Linguistic Theory,” in Chomsky (1995a).
1999. “Derivation by Phase,” MIT Working Papers in Linguistics.
2000. “Minimalist Inquiries: The Framework,” in R. Martin, D. Michaels and
J. Uriagereka (eds.), Step by Step: Essays in Minimalist Syntax in Honour of
Howard Lasnik (Cambridge, MA: MIT Press), 89–155.
2001. “Derivation by Phase,” in M. Kenstowicz (ed.), Ken Hale: A Life in Language
(Cambridge, MA: MIT Press), 1–52.
Chomsky, N., and H. Lasnik, 1995. “The Theory of Principles and Parameters,” in
N. Chomsky, The Minimalist Program (Cambridge, MA: MIT Press).
Cinque, G., 2007. “A Note on Linguistic Theory and Typology,” Language Typology,
11.1, 93–106.
Clark, T. W., [1977] 1989. Introduction to Nepali (New Delhi: Heritage Book House).
Cole, P., 1987. “The Structure of Internally Headed Relative Clauses,” Natural Language
and Linguistic Theory, 5, 277–302.
Cole, P., and L. Sung, 1994. “Head Movement and Long-distance Reflexives,” Linguistic
Inquiry, 25, 355–385.
Cole, P., G. Hermon and C.-T. James Huang (eds.), 2001. Long-distance Reflexives.
Syntax and Semantics 33 (San Diego: Academic Press).
Comrie, B., 1978. “Ergativity,” in W. P. Lehmann (ed.), Syntactic Typology (Sussex:
The Harvester Press), 329–394.
1981. Language Universals and Linguistic Typology (Chicago: University of Chicago
Press).
Coopmans, P., 1983. “A Review of B. Comrie, Language Universals and Linguistic
Typology,” Journal of Linguistics, 19.1-2, 455–473.
Coupe, A. L., 2003. “The Mongsen Dialect of Ao: A Language of Nagaland,” Ph.D.
thesis, La Trobe University, Bundoora, Victoria.
Croft, W., 1990. Typology and Universals (Cambridge: Cambridge University
Press).
Cysouw, M., 2003. “Against Implicational Universals,” Language Typology, 7,
89–101.
Das, B., 2005. “PRO in Bangla,” term paper, M.Phil. course, Delhi University.
Das, P. K., 2006. Grammatical Agreement in Hindi-Urdu and its Major Varieties
(Munich: Lincom Europa).
Dasgupta, P., 1977. “The Internal Grammar of Compound Verbs in Bangla,” Indian
Linguistics, 38, 68–85.
1980. “Question and Relative and Complement Clauses in a Bangla Grammar,” Ph.D.
thesis, New York University, New York.
2003. “Bangla,” in Cardona and Jain (2003: 391–443).
2004. “Some Non-nominative Subjects in Bangla,” in Bhaskararao and Subbarao
(2004: I, 128–140).
2007. “The Ubiquitous Complementizer,” in Bayer et al. (2007: 163–174).
Dasgupta, P., A. Ford and R. Singh, 2000. After Etymology: Towards a Substantive
Linguistics (Munich: Lincom Europa).
Davison, A., 1981. “Syntactic and Semantic Indeterminacy Resolved,” in P. Cole (ed.),
Radical Pragmatics (New York: Academic Press), 101–128.
1988. “Operator Binding, Gaps and Pronouns,” Linguistics, 26, 181–214.
Bibliography 341
1991. “Feature Percolation and Agreement in Hindi-Urdu,” paper presented at the

Panel on Agreement in South Asian languages at the South Asian Languages
Analysis Conference, University of Wisconsin, Madison.
1992a. “Lexical Projection, Case and Clause Adjunction: Another View of ‘Case
Resistance,’” paper presented at University of Iowa.
1992b. “Control Experiencer Subjects: Implications for Phrase Structure,” paper pre-
sented at Delhi University.
1994. “Finite Clauses and the Structure of the Clausal Projection,” paper presented
at CIEFL, Hyderabad and at Delhi University.
1995. “Kashmiri V2, Hindi/Urdu V-final Order: The Role of C and its Features” (ms).
2000. “Lexical Anaphors and Pronouns in Hindi,” in Lust et al. (2000: 397–470).
2001. “Long-distance Anaphors in Hindi/Urdu: Syntactic and Semantic Issues,” in
Cole et al. (2001: 47–82).
2004a. “Non-nominative Subjects in Hindi-Urdu,” in Bhaskararao and Subbarao
(2004: 141–168).
2004b. “Structural Case, Lexical Case and Verbal Projections,” in V. Dayal and
A. Mahajan (eds.), Clause Structure in South Asian Languages. Studies in Natural
Language and Linguistic Theory (Dordrecht: Kluwer), 199–226.
2007a. “Word Order, Parameters and the Extended COMP Projection,” in Bayer
et al. (2007a: 174–198).
2007b. “Long-distance Agreement in Hindi/Urdu,” paper presented at University of
Iowa Colloquium.
Dayal, V., 1994. “Binding Facts in Hindi and the Scrambling Phenomenon,” in Butt
et al. (1994: 237–262).
1996. Locality in Wh Quantification (Dordrecht: Kluwer).
Deeney, J., 1979. Ho Grammar and Vocabulary (Chaibasa: Xavier Ho Publications).
Dilip, M., K. V. Subbarao and G. Uma Maheshwar Rao, 2011. “Quantification in
Telugu,” paper presented at All India Conference of Linguists, Punjab University.
Dixon, R. M. W., 1972. The Dyirbal Language of North Queensland. Cambridge Studies
in Linguistics 9 (Cambridge: Cambridge University Press).
Donegan, P., and D. Stampe, 2004. “Rhythm and the Synthetic Drift of Munda,” in
R. Singh (ed.), The Yearbook of South Asian Languages and Linguistics,
3–36.
Downing, B., 1978. “Some Universals of Relative Clause Structure,” in J. H. Greenberg,
C. A. Ferguson and E. Moravcsik (eds.), Universals of Human Language, vol. IV
(Stanford, CA: Stanford University Press), 375–418.
Dowty, D. R., 1989. “On the Semantic Content of the Notion ‘Thematic Role,’” in
G. Chierchia and B. H. Partee (eds.), Properties, Types and Meaning, vol. II,
Semantic Issues (Dordrecht: Kluwer), 69–129.
Dryer, M., 1992. “The Greenbergian Word Order Correlations,” Language, 69, 81–138.
2005. “Order of Subject, Object and Verb,” in M. Haspelmath, M. Dryer, D. Gill
and B. Comrie (eds.), The World Atlas of Language Structures (New York: Oxford
University Press), 386–389.
Dube, K. C., 1983. Ādibhāratı̄ Bhāg 2. Kur.ukh-Hindı̄ (Bhopāl: Ādimjāti Anusandhān
evam Vikās Saṁsthā).
Dwarikesh, P. P. S., 1971. “The Historical Study of the Conjunctive Participial Phrase in
the New Indo-Aryan Dialects of the Madhyadesa (‘Midland’) of Northern India,”
doctoral dissertation, University of Chicago.
342 Bibliography
Dwivedi, V., 1994. Syntactic Dependencies and Relative Phrases in Hindi (Amherst,
MA: GLSA Publications).
Emeneau, M. B., 1956. “India as a Linguistic Area,” Language, 32.3, 3–16.
1974. “India as a Linguistic Area Revisited,” in Southworth and Apte (1974).
Everaert, M., 1991. “Contextual Determination of the Anaphor/Pronominal Distinc-
tion,” in J. Koster and E. Reuland (eds.), Long-distance Anaphora (Cambridge:
Cambridge University Press), 77–118.
1999. “The Morphosyntax of SE(-Reflexivization): Some Preliminaries,” paper pre-
sented at the seminar on Grammatical Models, Tilburg.
2001. “Paradigmatic Restrictions on Anaphors,” in K. Megerdoomian and L. A. Bar-el
(eds.), WWCFL 20 Proceedings (Somerville, MA: Cascadilla Press), 178–191.
Everaert, M., and E. Anagnostopoulou, 1997. “Thematic Hierarchies and Binding
Theory: Evidences from Greek,” in F. Corblin, D. Godard and J.-M. Marandin
(eds.), Empirical Issues in Formal Syntax and Semantics (Bern: Peter Lang),
43–60.
Everaert, M., and K. V. Subbarao, 2011. “Verbal and Nominal Reciprocals in South
Asian Languages (SALs) and Obligatory Order Reversal in the Nominal Anaphor
in Dravidian,” paper presented at the South Asian Languages Analysis Conference,
CIIL, Mysore.
Fox, B., 1987. “The Noun Phrase Accessibility Hierarchy Reinterpreted: Subject Pri-
macy or the Absolutive Hypothesis?” Language, 63.4, 856–870.
Gair, J. W., 1998. Studies in South Asian Linguistics: Sinhala and Other South Asian
Languages (New York and Oxford: Oxford University Press).
2003. “Sinhala,” in Cardona and Jain (2003).
Gair, J., and W. S. Karunatillake, 2000. “Lexical Anaphors and Pronouns in Sinhala,”
in Lust et al. (2000: 715–773).
Gair, J., and J. Paolillo, 1997. Sinhala (Munich: Lincom Europa).
Gair, J., and K. Wali, 1988. “On Distinguishing AGR from Agr: Evidence from South
Asia,” in Papers from Parasession on Agreement in Linguistic Theory (Chicago:
Chicago Linguistics Society), 87–104.
Gambhir, V., 1981. “Syntactic Restrictions and Discourse Functions of Word Order in
Standard Hindi,” doctoral dissertation, University of Pennsylvania, Philadelphia.
Geeta Devi, T., 2000. “Relative Clauses in Manipuri,” M.Phil. dissertation, Delhi Uni-
versity.
Gill, D., 2000. Riau Indonesian: A VO Language with Internally Headed Relative
Clauses (Leipzig: Max Plank Institute of Evolutionary Anthropology).
Giridhar, P. P., 1980. Angami (Mysore, India: Central Institute of Indian Languages).
1994. Mao Grammar (Mysore, India: Central Institute of Indian Languages).
Greenberg, J. H., 1966. “Some Universals of Grammar with Particular Reference to the
Order of Meaningful Elements,” in J. H. Greenberg (ed.), Universals of Language,
vol. II (Cambridge, MA: MIT Press), 73–113.
Gumperz, J. J., and R. Wilson, 1971. “Convergence and Creolization: A Case from the
Indo-Aryan/Dravidian Border,” in D. Hymes (ed.), Pidginization and Creolization
of Language (Cambridge: Cambridge University Press).
Gurtu, M., 1992. Anaphoric Relations in Hindi and English (Delhi: Munshiram
Manoharlal).
Hany Babu, M. T., 1997. “Functional Categories,” Ph.D. thesis, Central Institute of
English and Foreign Languages, Hyderabad.
Bibliography 343
Hasan, M., 2005. “Backward Control in Subzapuri,” term paper, M.Phil. course, Delhi
University.
Haspelmath, M., 1993. “More on the Typology of Inchoative/Causative Verb Alterna-
tions,” in B. Comrie and M. Polinsky (eds.), Causatives and Transitivity (Amster-
dam: John Benjamins), 87–120.
Haspelmath, M., and E. König (eds.) 1995. Converbs in Cross-linguistic Perspective:
Structure and Meaning of Adverbial Verb Forms – adverbial participles, Gerunds.
Empirical Approaches to Language Typology 13 (Berlin: Mouton de Gruyter).
Herring, S. C., 1991. “Nominalization, Relativization, and Attribution in Lotha, Angami,
and Burmese,” Linguistics of the Tibeto-Burman Area, 14.1, 55–72.
Hock, H. H., 1975. “Substratum Influence on (Rig Vedic) Sanskrit?” Studies in the
Linguistic Sciences, 5.2, 76–125.
1982a. “The Sanskrit Quotative: A Historical and Comparative Study,” Studies in the
1982b. “Conjoined We Stand: Theoretical Implications of Sanskrit Relative Clauses,”
Studies in the Linguistic Sciences, 19.1, 93–126.
1991. Principles of Historical Linguistics (Berlin: Mouton de Gruyter).
2001. “Typology vs. Convergence: The Issue of Dravidian/Indo-Aryan Syntactic
Similarities Revisited,” in Bhaskararao and Subbarao (2001: 63–100).
2005. “How Strict is Strict OV?” in T. Bhattacharya (ed.), The Yearbook of South
Asian Languages and Linguistics (Berlin: Mouton de Gruyter), 145–164.
Holmberg, A., 2000. “Deriving OV Order in Finnish,” in P. Svenonius (ed.), The Deriva-
tion of VO and OV (Amsterdam: John Benjamins), 123–152.
Hook, P. E., 1973. Compound Verbs in Hindi (Ann Arbor, MI: University of Michigan).
1974. The Compound Verb in Hindi (Ann Arbor: Center for South and South-East
Asia Studies, The University of Michigan).
1976. “Is Kashmiri an SVO Language?” Indian Linguistics, 37, 133–142.
Hindi Structures: Intermediate Level. Michigan Papers on South and South-East Asia
16 (Ann Arbor, MI: Center for South and South-East Asia Studies, University of
Michigan).
1990. “A Note on Expressions of Involuntary Experience in the Shina of Skardu,”
BSOAS, 67, 77–82.
1991. “The Emergence of Perfective Aspect in Indo-Aryan Languages,” in E. Traugott
and B. Hein (eds.), Approaches to Grammaticalization, vol. II (Amsterdam: John
Benjamins), 59–89.
1996. “Kesar of Layul: A Central Asian Epic in the Shina of Gultari,” in W. Hanaway
and W. Heston (eds.), Studies in the Popular Cultures of Pakistan (Lahore: Sang-
e-Meel), 121–183.
1997. “Relative Clauses in Eastern Shina,” in P. Michalove, I. Hegedüs and
A. Manaster-Ramer (eds.), Indo-European, Nostratic, and Beyond: Festschrift
for Vitalij V. Shevoroshkin (Washington: Institute for the Study of Man),
140–154.
2001. “Where Do Compound Verbs Come From? (And Where Are They Going?)”
in Bhaskararao and Subbarao (2001: 101–130).
Hook, P., and O. N. Koul, 1996. “Kashmiri as a V-2 Language,” in V. S. Lakshmi
and A. Mukerjee (eds.), Word Order in Indian Languages (Hyderabad: Center of
Advanced Study in Linguistics, Osmania University), 95–105.
2004. “Case as Agreement,” in Bhaskararao and Subbarao (2004: I, 213–226).
344 Bibliography
Hopper, P., and S. Thompson, 1980. “Transitivity in Grammar and Discourse,” Lan-
guage, 56, 251–299.
Hornstein, N., 1999. “Movement and Control,” Linguistic Inquiry, 30.1, 69–96.
Huang, Y., 2000. Anaphora: A Cross-linguistic Approach (Oxford: Oxford University
Press).
Ichihashi-Nakayama, K., 1994. “On Dative Subject Construction in Nepali,” in
C. Genetti (ed.), Papers in Linguistics 6 (Santa Barbara: Department of Linguistics,
UC Santa Barbara), 41–76.
Jacquesson, F., ms. “A Dimasa Grammar.” Brahmaputra Studies database,
http://brahmaputra.vjf.cnrs.fr/bdd/MG/pdf/Dimasa Grammar-2.pdf.
Jaeggli, O., and K. J. Safir, 1989. The Null Subject Parameter. Studies in Natural
Language and Linguistic Theory (Dordrecht, Boston and London: Kluwer).
Jayaseelan, K. A., 1990. “The Dative Subject Construction and the Pro-Drop Parameter,”
in Verma and Mohanan (1990: 269–283).
1998. “Blocking Effects and the Syntax of Malayalam Taan,” in R. Singh (ed.),
Yearbook of South Asian Languages and Linguistics (New Delhi: Sage), 11–
27.
1999. Parametric Studies in Malayalam Syntax (New Delhi: Allied Publishers).
2000. “Lexical Anaphors and Pronouns in Malayalam,” in Lust et al. (2000: 113–
168).
2004. “The Possessor–experiencer Dative in Malayalam,” in. Bhaskararao and Sub-
barao (2004: I, 227–244).
2007. “The Argument Structure of the Dative Construction,” in E. Reuland,
T. Bhattacharya and G. Spathas (eds.), Argument Structure (Amsterdam and
Philadelphia: John Benjamins), 37–48.
Kachru, B., 1973. An Introduction to Spoken Kashmiri, part 1 (Urbana, IL: Department
of Linguistics, University of Illinois).
Kachru, Y., 1970. “The Syntax of ko-sentences in Hindi-Urdu,” Papers in Linguistics,
2.2, 299–316.
1979. “The Quotative in South Asian Languages,” South Asian Languages Analysis,
1, 63–77.
1980. Aspects of Hindi Grammar (New Delhi: Manohar Publications).
1981. “On the Syntax, Semantics and Pragmatics of the Conjunctive Participle in
Hindi-Urdu,” Studies in the Linguistic Sciences, 11.2, 35–49.
1990. “Experiencer and Other Oblique Subjects in Hindi,” in Verma and Mohanan
(1990: 59–75).
1993. “Verb Serialization in Syntax, Typology and Historical Change,” in M. K. Verma
(ed.), Complex Predicates in South Asian Languages (Delhi: Motilal Banarsidass),
115–133.
2006. Hindi (Amsterdam and Philadelphia: John Benjamins).
Kachru, Y., and T. K. Bhatia, 1977. “On Reflexivization in Hindi-Urdu and its Theoret-
ical Implications,” Indian Linguistics, 38.1, 21–38.
Kak, A., ms. “Complementation in Kashmiri,” University of Kashmir, Sri Nagar, India.
Kayne, R., 1994. The Antisymmetry of Syntax (Cambridge, MA: MIT Press).
Kazenin, K. I., 2001. “Verbal Reflexives and the Middle Voice,” in M. Haspelmath,
E. König, W. Oesterreicher and W. Raible (eds.), Language Typology and Language
Universals (Berlin: Mouton de Gruyter), 916–927.
Bibliography 345
Keenan, E. L., 1985. “Relative Clause,” in T. Shopen (ed.), Language Typology and
Syntactic Description, vol. II (Cambridge: Cambridge University Press), 141–
170.
Keenan, E. L., and B. Comrie, 1977. “Noun Phrase Accessibility and Universal Gram-
mar,” Linguistic Inquiry, 8.1, 63–99.
Kellogg, S. H., 1938. A Grammar of the Hindi Language (Delhi: Oriental Books Reprint
Corporation).
Kemmer, S., 1993. The Middle Voice. Typological Studies in Language 23 (Amsterdam
and Philadelphia: John Benjamins).
Kevichüsa, M., 1996. “Relative Clause Formation in Angami (Tenyidie),” M.Phil. dis-
sertation, Delhi University.
2007. “Apects of Angami (Tenyidie) Syntax,” Ph.D. thesis, Delhi University.
Kevichüsa, M., and K. V. Subbarao, 1998. “The Relative Clause in Tenyidie (Angami
Naga),” South Asian Language Review, 8.2, 40–64.
Kidwai, A., 2000. XP-Adjunction in Universal Grammar: Scrambling and Binding in
Hindi-Urdu. Oxford Studies in Comperative Syntax (Oxford: Oxford University
Press).
Klaiman, M. H., 1977. “Bengali Syntax: Possible Dravidian Influence,” International
Journal of Dravidian Linguistics, 4.2.
1979. “On the Status of Subjecthood Hierarchy in Hindi,” International Journal of
Dravidian Linguistics, 8, 17–31.
1980. “Bengali Dative Subjects,” Lingua, 51, 275–295.
1983. “Bengali Conjunctive Participle Constructions,” in Proceedings of the Berkeley
Linguistic Society, 19, 138–147.
Koh, T. J., 2003. “Agreement in Ho,” Masters’ field methods course dissertation, Delhi
University.
Koh, T. J., and K. V. Subbarao, ms. “A Grammar of Ho.”
Koul, O. N., and K. Wali, 2006. Modern Kashmiri Grammar (Springfield, VA:
Dunwoody Press).
Koshal, S., 1979. Ladakhi Grammar (Delhi: Motilal Banarasi Dass).
Krishnamurti, Bh., 1975. “Verbs of Cognition in Telugu,” Osmania Papers in Linguistics,
1.1, 1–15.
2003. The Dravidian Languages. Cambridge Language Survey (Cambridge:
Cambridge University Press).
Krishnamurti, Bh., and J. P. L. Gwynn, 1985. A Grammar of Modern Telugu (Delhi:
Oxford University Press).
Kuolie, D., 2006. A Structural Description of Tenyidie: A Tibeto-Burman Language of
Nagaland (Kohima: Ura Academy).
Kuiper, F. B. J., 1974. “The Genesis of a Linguistic Area,” in F. C. Southworth and M. L.
Apte (eds.), “Contact and Convergence in South Asian Languages,” International
Journal of Dravidian Languages, 3.1, 135–153.
Kumar, R., 1998. “Internal Relative Clauses in Hmar and Mizo,” M.Phil. dissertation,
Delhi University.
2006. Syntax of Negation and Licensing of Negative Polarity Items in Hindi (London
and New York: Routledge).
Kumar, R., and K. V. Subbarao, 2005. “Postposition Incorporation and Internally Headed
Relative Clauses in Hmar,” in Y. Yadava, G. Bhattariai, R. R. Lohani, B. Prasain
346 Bibliography
and K. Parajuli (eds.), Contemporary Issues in Nepalese Linguistics (Kathmandu:

Nepalese Linguistic Society), 129–138.
Kuroda, S. Y., 1965. “Generative Studies in the Japanese Language,” doctoral disserta-
tion, MIT, Cambridge, MA.
Lahaussois, A., 2003. “Nominalization in Thulung Rai,” Linguistics of the Tibeto-
Burman Area, 26.1, 37–62.
Lakshmi Bai, B., 1985. “Some Notes on Correlative Constructions in Dravidian,” in
V. Z. Acson and R. L. Leed (eds.), For Gordon H. Fairbanks (Honolulu: University
of Hawaii Press), 181–190.
Lalitha Murthy, B., 1994. “Participial Constructions: A Cross-linguistic Study,” Ph.D.
thesis, Delhi University.
Lalitha Murthy, B., and K. V. Subbarao, 2000. “Pronouns and Lexical Anaphors in
Mizo,” in Lust et al. (2000: 777–840).
Lambert, H. M., 1971. Gujarati Language Course (Cambridge: Cambridge University
Press).
Lasnik, H., 1999. Minimalist Analysis (London: Blackwell).
Lasnik, H., and J. Uriagereka with C. Boeckx, 2005. Minimalist Syntax (Malden: Black-
well).
Lee, H., 2003. “Prominence Mismatch and Markedness Reduction in Word Order,”
Natural Language and Linguistic Theory, 21, 617–680.
Lehmann, T., 1989. A Grammar of Modern Tamil (Pondicherry: Pondicherry Institute
of Linguistics and Culture).
Lehmann, W. P., 1973. “A Structural Principle of Language and its Implications,”
Language, 49, 47–66.
Lichtenberk, F. 1985. “Multiple Uses of Reciprocal Constructions,” Australian Journal
of Linguistics, 5, 19–41.
Lidz, J., 2001a. “Anti-locality,” in Cole et al. (2001: 227–254).
2001b. “Condition R,” Linguistic Inquiry, 32.1, 123–140.
2001c. “Argument Structure of Verbal Reflexives,” Natural Language and Linguistic
Theory, 19.2, 311–353.
2006. “The Grammar of Accusative Case in Kannada,” Language, 82.1, 10–32.
Lohse, B., J. A. Hawkins and T. Wasow, 2004. “Domain Minimization in English
Verb-particle Constructions,” Language, 80.2, 238–261.
Lord, C., 1976. “Evidence for Syntactic Reanalysis from Verb to Complementizer,” in
Papers from the Parasession on Diachronic Syntax (Chicago: Chicago Linguistic
Society), 179–191.
Lorrain, J. H. (a.k.a. Pu Buanga), [1940] 1982. Dictionary of the Lushai Language
(Culcutta: The Asiatic Society).
Lust, B., 1986. Studies in the Acquisition of Anaphora, vol. I, Defining the Constraints
(Dordrecht: Reidel).
Lust, B., K. Wali, J. W. Gair and K. V. Subbarao, 2000. Lexical Anaphors and Pronouns
in Selected South Asian Languages: A Principled Typology (Berlin and New York:
Mouton de Gruyter).
Lyngdoh, A., 2000. “Complementation in Khasi,” M.Phil. dissertation, Delhi University.
Macphail, R. M., 1983. An Introduction to Santali (Calcutta: Firma KLM Private Ltd.).
Magier, D., 1987. “The Transitivity Prototype: Evidence from Hindi,” Word, 38.3, 187–
199.
Bibliography 347
1990. “Dative/Accusative Subjects in Marwari,” in Verma and Mohanan (1990: 213–

220).
Mahajan, A., 1982. “Causatives in Indian languages,” M.Phil. dissertation, Delhi
University.
1987. “Notes on wh Questions in Hindi,” paper presented at “South Asian Languages
Analysis,” Cornell University, Ithaca, New York.
1989. “Agreement and Agreement Phrases,” in I. Laka and A. Mahajan (eds.),
Functional Heads and Clause Structure. MIT Working Papers in Linguistics 10
(Cambridge, MA: MIT Press), 217–252.
1990. “The A/A-bar Distinction and Movement Theory,” Ph.D. thesis, Cambridge,
MA: MIT.
1994. “Active Passives,” Proceedings of WCCFL, 13, 286–301.
1997a. “Rightward Scrambling,” in D. Beerman, D. LeBlanc and H. van Riemsdijk
(eds.), Rightward Movement (Amsterdam and Philadelphia: John Benjamins), 185–
213.
1997b. “Universal Grammar and the Typology of Ergative Languages,” in A.
Alexiadou and T. A. Hall (eds.), Studies in Universal Grammar and Typologi-
cal Variation (Amsterdam and Philadelphia: John Benjamins), 1335–1357.
2000. “Relative Asymmetries and Hindi Correlatives,” in A. Alexiadou, P. Law and
C. Wilder (eds.), The Syntax of Relative Clauses (Amsterdam and Philadelphia:
John Benjamins), 201–229.
2004. “The Origin of Non-nominative Subjects,” in Bhaskararao and Subbarao (2004:
I, 283–299).
Mahapatra, B. P., 1979. Malto: An Ethnosemantic Study (Mysore: Central Institute of
Indian Languages).
Mahto, P., 1989. “On the Nature of Empty Pronominals,” Ph.D. thesis, Central Institute
of English and Foreign Languages, Hyderabad.
Malhotra, S., 2002. “Sentential Word Order in Kokborok,” paper presented at the
Linguistic Society of India meeting, Shillong, North-Eastern Hill University.
Malhotra, V., 1982. “The structure of Kharia: A Study of Linguistic Typology and
Language Change,” Ph.D. thesis, Jawaharlal Nehru University, New Delhi.
Manzini, R. M., and A. Roussou, 1998. “A Minimalistic Theory of A-movement and
Control,” University College of London Working Papers, 11. Also in: Lingua, 10,
409–447 (2000).
Martin, R., 2001. “Null Case and the Distribution of PRO,” Linguistic Inquiry, 32.1,
141–166.
Martin, S. E., 1975. A Reference Grammar of Japanese (New Haven: Yale University
Press).
Masica, C. P., 1976. Defining a Linguistic Area: South Asia (Chicago: University of
Chicago Press).
1991. The Indo-Aryan Languages (Cambridge: Cambridge University Press).
1999. “Alternative Embedding Strategy in Gujarati,” in P. J. Mistry and B. Modi
(eds.), Vidyopasaanaa: Studies in Honour of Harivallabh C. Bhayani (Mumbai
and Ahmedabad: Image Publications Pvt. Ltd.), 135–156.
Matisoff, J., 1972. “Lahu Nominalization, Relativization and Genitivization,” in
J. Kimbal (ed.), Syntax and Semantics, vol. I (New York: Academic Press), 237–
257.
348 Bibliography
McAlpin, D., 1976. “Dative Subjects in Malayalam,” in Verma (1976: 183–194).

Minegishi, M., and G. Murmu, 2001. Santali Basic Lexicon with Grammatical Notes
(Tokyo: ILCAA, Tokyo University of Foreign Studies).
Mistry, P. J., 1998: “Objecthood and Specificity in Gujarati,” in J. H. Hill, P. J. Mistry
and L. Campbell (eds.), The Life of Language: Papers in Honour of William Bright
(Berlin: Mouton de Gruyter), 425–442.
2000. “Lexical Anaphors and Pronouns in Gujarati,” in Lust et al. (2000: 333–395).
2004. “Subjecthood of Non-nominatives in Gujarati,” in Bhaskararao and Subbarao
(2004: II, 1–32).
Mohanan, K. P., 1981. “Pronouns in Malayalam,” Studies in the Linguistic Sciences, 11,
67–75.
1984. “Operator Binding and the Path Containment Condition,” Natural Language
and Linguistic Theory, 2, 357–396.
Mohanan, K. P., and T. Mohanan, 1990. “Dative Subjects in Malayalam: Semantic
Information in Syntax,” in Verma and Mohanan (1990: 43–58).
1994. “Issues in Word Order in South Asian Languages: Enriched Phrase Structure
or Multidimensionality?” in Butt et al. (1994: 153–184).
Mohanan, T., 1994. Argument Structure in Hindi (Stanford, CA: CSLI Publications,
Stanford University).
Mohanty, R., 2002. “ECM Constructions in Oriya and English and Tree Adjoining Gram-
mar,” Ph.D. thesis, Central Institute of English and Foreign Languages, Hyderabad.
Monahan, P., 2003. “Backward Object Control in Korean,” in G. Garding and
M. Tsujimura (eds.), WCCFL 22 Proceedings (Somerville, MA: Casadilla Press),
356–369.
Montaut, A., 2004. A Grammar of Hindi (Munich: Lincom Europa).
Mukherjee, A., 1989. Language Change and Language Variation (Hyderabad: Centre
for Advanced Study in Linguistics, Osmania University).
Mukherjee, S. N., K. V. Subbarao and D. Walia, 2004. “Lexical Anaphors and Blocking
Effects in Hmar,” in T. R. Kansakar and M. Turin (eds.), Themes in Himalayan
Languages and Linguistics (Kathmandu, Nepal: Tribhuvan University; Heidelberg:
Heidelberg University), 103–114.
Nadkarni, M. V., 1970. “Noun Phrase Embedded Structures in Kannada and Konkani,”
doctoral disseration, UCLA.
1975. “Bilingualism and Syntactic Change in Konkani,” Language, 51, 672–683.
Nagaraja, K. S., 1984. “Relativization in Konyak,” Indian Linguistics, 45, 41–48.
1985. Khasi: A Descriptive Analysis (Pune: Deccan College).
1999. Korku Language (Tokyo: ILCAA, Tokyo University of Foreign Studies).
Nath, D. K., 2009. “Nominalization in Tiwa,” paper presented at the seminar on Nomi-
nalization at Tezpur University, Tezpur.
Neukom, L., 2001. Santali. Languages of the World / Materials 323 (Munich: Lincom
Europa). (= L. Neukom, ms “Santali Grammar”)
Neukom, L., and M. Patnaik, 2003. A Grammar of Oriya. Seminar für Allgemeine.
Sprachwissenschaft der Univerität Zürich.
Newmeyer, F. J., 2007. “Linguistic Typology Requires Cross-linguistic Formal Cate-
gories,” Linguistic Typology, 11.1, 133–158.
Noonan, M., 1985. “Complementation,” in T. Shopen (ed.), Language Typology and
Syntactic Description, vol. II (Cambridge: Cambridge University Press), 42–140.
Bibliography 349
n.d. handout. “Converbal Constructions in Chantyal,” http://pantherfile.uwm.edu/

noonan/www/converb.type1.pdf.
Osada, T., 1992. A Reference Grammar of Mundari (Tokyo: ILCAA, University of
Foreign Studies, Tokyo).
Pandharipande, R., 1981. “Syntax and Semantics of the Passive Construction in Selected
South Asian Languages,” doctoral dissertatiion, University of Illinois at Urbana-
Champaign, Urbana.
1990. “Experiencer (Dative) NPs in Marathi,” in Verma and Mohanan (1990: 161–
180).
1997. Marathi (London and New York: Routledge).
Pandharipande, R., and Y. Kachru, 1976. “Relational Grammar, Ergativity and Hindi-
Urdu,” Studies in the Linguistic Sciences, 6.1, 82–99.
Pandit, P. B., 1972. “India as a Sociolinguistic Area,” Gune Memorial Lectures, Pune
University, Poona.
Pangersenla, 2005. “Issues in the Interpretation of Anaphors in Ao (Mongsen),” M.Phil.
dissertation, Delhi University.
Paramasivam, K., 1979. Effectivity and Causativity in Tamil (Trivandrum: Dravidian
Linguistics Association).
Patnaik, M., 1998. “Quotatives Revisited,” South Asian Language Review, 8.2, 79–96.
1999. “Aspects of Juang Syntax,” Ph.D. thesis, Delhi University.
Patnaik, M., and K. V. Subbarao, 2000. “An Initial Note on Lexical Anaphors and
Pronouns in Juang,” in Lust et al. (2000: 841–859).
Peterson, J., 2006. “Kharia: A South Munda Language, vol. I, Grammatical Analy-
sis,” unpublished Habilitationsschrift, Universität Osnabrück. Also A Grammar of
Kharia (Leiden: Brill, 2010).
2008. “Kharia,” in G. D. S. Anderson (ed.), The Munda Languages (New York:
Routledge), 434–507.
Picallo, C., 1985. “Opaque Domains,” Ph.D. thesis, CUNY (City University of New
York), New York.
Pinker, S., and R. Jackendoff, 2005. “The Faculty of Language: What’s Special about
It?” Cognition, 95, 201–236.
Pinnow, H. J., 1960. Beiträge zur Kenntnis Juang-Sprache.
Piñón, C., 2001. “A Finer Look at the Causative–inchoative Alternation,” in R. Hastings,
B. Jackson and Z. Zvolensky (eds.), Proceedings of Semantics and Linguistic
Theory, 11 (Ithaca, NY: CLC Publications).
Polinsky, M., and E. Potsdam., 2002. “Backward Control in Tsez,” Linguistic Inquiry,
33.2, 245–285.
2003. Backward Control: Evidence from Malagasy. MIT Working Papers in Linguis-
tics 44 (Cambridge, MA: MIT Department of Linguistics and Philosophy).
Porizka, V., 1963. Hindi Language Course (Prague: Statni pedagogicke nakladatelstvi).
Poudel, T., 2007. Tense, Aspect and Modality in Nepali and Manipuri (Munich: Lincom
Europa).
Prakash, O., 2006. “Subordination in Mizo,” Ph.D. thesis, Delhi University.
Prakash, O., and K. V. Subbarao, 2003. “Internally Headed Relative Clauses in Mizo,”
paper presented at the Himalayan Languages Conference, CIIL, Mysore.
Prasad, B. R., 1991. Mising Grammar (Mysore: Central Institute of Indian Languages).
Premchand, 1966. Godān (Allahabad: Sarasvati Press).
350 Bibliography
Raina, A. M., 2002. “The Verb Second Phenomenon in Kashmiri,” in O. N. Koul and
K. Wali (eds.), Topics in Kashmiri Linguistics (New Delhi: Creative Books), 113–
128.
Ramakrishna Reddy, B., 1980. “Non-Dravidian Element in Manda Syntax: A Study in
Linguistic Convergence,” Osmania Papers in Linguistics, 6, 71–88.
1992a. “Syntactic Convergence in Central Indian Languages,” in M. K. Sastri (ed.),
Sri Korada Ramakrishnaiah Centenary Festschrift (Visakhapatnam: Sita Trust),
571–577.
1992b. “Predicate Agreement in Dravidian,” PILC Journal of Dravidic Studies, 1.1,
33–44.
Ramamurti, G. V., 1931. A Manual of the Sora (or Savara) Language (Madras: Govt.
Press).
Ramanujan, A. K., and C. Masica, 1969. “Toward a Phonological Typology of the
Indian Linguistic Area,” in T. E. Sebeok (ed.), Current Trends in Linguistics, 5:
Linguistics in South Asia (The Hague: Mouton), 543–577.
Ramarao, C., 1975. Telugu vākyam [“The Telugu Sentence” – in Telugu] (Hyderabad:
A. P. Sahitya Academy).
2003. Telugu velugulu [“Telugu Brightness” – in Telugu] (Hyderabad: Navodaya
Book House).
Ramasamy, K. 1981. “Correlative Relative Clauses in Tamil,” in S. Agesthialingom and
N. R. Nair (eds.), Dravidian Syntax. Annamalai University Publications in Lin-
guistics 73 (Annamalainagar: Annamalai University Department of Linguistics),
363–380.
Ray, T., 2000. “Lexical Anaphors and Pronouns in Oriya,” in Lust et al. (2000: 575–636).
Regmi, C. U., 2002. “Relativization in Nepali,” Ph.D. thesis, Tribhuvan University.
Reinhart, T., 2002. “The Theta System: An Overview,” Theoretical Linguistics, 28.3,
229–290.
Reinhart, T., and T. Siloni, 2005. “The Lexicon–Syntax Parameter: Reflexivization and
Other Arity Operations,” Linguistic Inquiry, 36.3, 389–436.
Riemsdijk, V. H., 2006. “Free Relatives,” in M. Everaert and V. H. Riemsdijk (eds.),
The Blackwell Companion to Syntax, vol. II (Oxford: Blackwell), 338–382.
Rizzi, L., 1997. “The Fine Structure of the Left Periphery,” in L. Haegeman (ed.),
Elements of Grammar (Dordrecht: Kluwer), 281–283.
Rosenbaum, P. S., 1967. The Grammar of Predicate Complement Constructions (Boston,
MA: MIT Press).
Ross, J. R., 1973. “Slifting,” in M. Gross and M. Schützenberger (eds.), The Formal
Analysis of Natural Languages (‘s Gravenhage, Holland: Mouton and Company),
133–172.
Sarju Devi, T., 2006. “Embedded Clauses in Manipuri,” doctoral thesis, Delhi University.
Sarju Devi, T., and K. V. Subbarao, 2002. “Reduplication and Case Copying: The Case of
Lexical Anaphors in Manipuri and Telugu,” Linguistics of the Tibeto-Burman Area,
25.2 (Fall), 47–72; also in R. Laury, G. McMenamin, S. Okamoto, V. Samiian and
K. V. Subbarao (eds.), Papers in Honor of P. J. Mistry (New Delhi: Indian Institute
of Language Studies, 2003).
Savio, D., 1991. “WH-questions in Tamil,” CIEFL Occasional Papers in Linguistics, 3,
55–67.
Saxena, A., 1985. “Reflexivization in Hindi: A Reconsideration,” International Journal
of Dravidian Linguistics, 14, 224–237.
Bibliography 351
Schiffman, H., 2000. A Reference Grammar of Spoken Tamil (Cambridge: Cambridge

University Press).
Schiffman, H., and M. Shapiro, 1981. Language and Society in South Asia (Delhi:
Motilal Banarsidass).
Sengupta, G., 2000. “Lexical Anaphors and Pronouns in Bangla” in B. Lust et al. (2000:
277–332).
Shibatani, M., and T. Bynon, 1995. Approaches to Language Typology (Oxford: Claren-
don Press).
Shibatani, M., and P. Pardeshi, 2001. “Dative Subject Constructions in South Asian
Languages,” in Bhaskararao and Subbarao (2001: 311–348).
Sigurdsson, H. A., 1989. Verbal Syntax and Case in Icelandic: A Comparative GB
Approach (Lund: Department of Scandinavian Linguistics, University of Lund).
1991. “Icelandic Case-marked PRO and the Licensing of Lexical Arguments,” NLLT,
9, 327–363.
2004a. “The Syntax of Person, Tense, and Speech Features,” Italian Journal of
Linguistics / Rivista di Linguistica, 16.1, 219–251.
2004b. “Icelandic Non-nominative Subjects,” in Bhaskararao and Subbarao (2004:
137–160).
Simpson, A., and T. Bhattacharya, 2003. “Obligatory Overt wh-movement in a wh-in-
situ Language,” Linguistic Inquiry, 34.1, 127–142.
Singh, J., 1993. “Case and Agreement in Hindi: A GB Approach,” Ph.D. thesis, York
University, UK.
Singh, U. N., 1979. “Some Aspects of Maithili Syntax: A Transformational-Generative
Approach,” Ph.D. thesis, Delhi University.
1980. “Bole: An Unresolved Problem in Bengali Syntax,” Indian Linguistics, 41,
188–195.
Southworth, F. C., and M. L. Apte (eds.), 1974. Contact and Convergence in South
Asian languages. Special issue of International Journal of Dravidian Linguistics,
3.1.
Speijer, J. S., 1886. Sanskrit Syntax (Delhi: Motilal Banarsidass; repr: 1980).
Sridhar, S. N., 1979. “Dative Subjects and the Notion of Subject,” Lingua, 49, 99–
125.
1981. “Linguistic Convergence: Indo-Aryanization of Dravidian,” Lingua, 53, 199–
220.
1990. Kannada (London: Routledge).
Srivastav, V., 1989. “Hindi wh and Pleonastic Operators,” NELS, 20.
1991a. “Subjacency Effects at LF: The Case of Hindi WH,” Linguistic Inquiry, 22,
762–769.
1991b. “The Syntax and Semantics of Correlatives,” Natural Language and Linguistic
Theory, 9, 637–686.
Steever, S. B., 1987. “Remarks on Dravidian Complementation,” Studies in the Linguis-
tic Sciences, 17.1, 103–119.
1988. The Serial Verb Formation in the Dravidian Languages (Delhi: Motilal
Banarsidass).
Stowell, T., 1981. “Origins of Phrase Structure,” Ph.D. thesis, MIT, Cambridge, MA.
Subbarao, K. V., 1971. “Notes on Reflexivization in Hindi Syntax,” Studies in the
1974. “Noun Phrase Complementation in Hindi,” Ph.D. thesis, University of Illinois.
352 Bibliography
1979. “Secondary Verbs in Telugu,” International Journal of Dravidian Linguistics,

8.2, 268–272.
1984a. Complementation in Hindi Syntax (Delhi: Academic Publications).
1984b. Syntactic Typology, Language Universals and Convergence in South Asian
Languages, UGC Project Report (Delhi: Delhi University).
1996. “The Conjunctive Participle in South Asian Languages,” paper presented at the
Indo-French Workshop on Computational Linguistics, Hyderabad: University of
Hyderabad.
1998a. “Linguistic Theory and Syntactic Typology: A Proposal for A Symbiotic
Relationship,” in Vāgbhārati: Proceedings of the International Conference on
South Asian Languages and Literatures (Moscow: Moscow State University,
5–23. Rpt. in Indian Linguistics, 60 (1999), 95–110).
1998b. “In Support of Agr as a Functional Category,” paper presented at the GLOW
Workshop, Central Institute of English and Foreign Languages, Hyderabad.
1999. “Agreement and Relative Clauses in Tibeto-Burman Languages,” Tokyo:
ILCAA, Tokyo University of Foreign Studies (ms).
2000. “Syntactic Typology and South Asian Languages,” in R. Singh (ed.), The
Yearbook of South Asian Languages and Linguistics (New Delhi: Thousand Oaks;
London: Sage Publications), 93–108.
2001. “Agreement in South Asian Languages: Minimalist Inquiries: The Framework,”
in Bhaskararao and Subbarao (2001: 457–492).
2003. “Language Types and Agreement in South Asian Languages,” in B.
Ramakrishna Reddy (ed.), Special Issue on Agreement, International Journal
of Dravidian Linguistics (Chennai: International Institute of Tamil Studies),
1–27.
2004. “Tense and Agreement in Backward Control,” paper presented at the Univer-
sities of Osnabrück and Lund and also at the conference on Control Verbs, ZAS,
Berlin.
2005a. “Backward Control,” plenary lecture delivered at the 25th South Asian Lan-
guages Analysis Conference, University of Illinois, Champaign-Urbana, Illinois.
2005b. “Case Alternations and Control Theory,” keynote address delivered at the
International Conference on South Asian Languages and Literatures, Osmania
University, Hyderabad, India.
2007a. “The Role of Particles and Clitics in Disambiguation in South Asian Lan-
guages,” in C. Masica (ed.), Old and New Perspectives on South Asian Languages:
Grammar and Semantics (Delhi: Motilal Banarsidass), 153–174.
2007b. “Typological Characteristics of South Asian Languages,” in B. B. Kachru and
S. N. Sridhar (eds.), Language in South Asia (Cambridge: Cambridge University
Press), 49–78.
2010. “Dravidian Languages: A Syntactic Typology,” special lecture presented at the
seminar on The Linguistic and Genetic Architecture of Dravidian at Dravidian
University, Kuppam, India, 22–24 February 2010.
Subbarao, K. V., and H. Arora, 1989. “Extreme Convergence: The Case of Dakhini
Hindi-Urdu,” in A. Mukherjee (ed.), Language Change and Language Variation
(Hyderabad: Centre of Advanced Study, Osmania University), 105–122.
2009. “The Conjunctive Participle in Dakkhini Hindi-Urdu: Making the Best of Both
Worlds,” Indian Linguistics, 70, 359–386.
Bibliography 353
Subbarao, K. V., and P. Bhaskararao, 2004. “Non-nominative Subjects in Telugu,” in

Subbarao, K.V., and M. Everaert, 2011. “Verbal and Nominal Anaphors in South Asian
Languages (SALs): Some Observations on Case Copying, Antecedent Restrictions
and Long-distance Binding,” paper presented at the workshop on “The World of
Reflexives” at University of Utrecht.
Subbarao, K. V., and P. Haokip, 2009. “Gap Relatives in Thadou with Special Reference
to the Comitative PP as Head,” paper presented at the seminar on Nominalization
at Tezpur University, Tezpur; also at South Asian Languages Analysis (SALA)
2011, Central Institute of Languages, Mysore.
2011. “EHRCs and IHRCs in Thadou (Tibeto-Burman),” paper presented at South
Asian Languages Analysis 29 (SALA 29) conference at the Central Institute of
Indian Languages, Mysore, India.
Subbarao, K. V., and M. Kevichüsa, 1999. “Internal Relative Clauses in Tenyidie
(Angami): A Case of Hierarchical Precedence vs. Linear Precedence,” Linguis-
tics of the Tibeto-Burman Area, 22.1, 149–181.
Subbarao, K. V., and M. Kevichüsa, 2005. “Relative Clauses in Sema,” in Bhattacharya
(2005: 255–272).
Subbarao, K. V., and B. Lalitha Murthy, 1994. “Ergativity in Mizo,” Gaveshana (Agra:
Central Hindi Institute), 257–266.
1997. “The Passive in Mizo,” in A. Abbi (ed.), Languages of Tribal and Indige-
nous Peoples of India: The Ethnic Space (Delhi: Motilal Banarsidass), 401–
414.
2000. “Lexical Anaphors and Pronouns in Telugu,” in Lust et al. (2000: 217–276).
ms. “Aspects of Mizo Grammar.”
Subbarao, K. V., and S. Munshi, 2000. “Agreement in Kashmiri: Minimalist Inquiries:
The Framework” (ms).
Subbarao, K. V., and T. Sarju Devi, ms. “A Grammar of Manipuri.”
Subbarao, K. V., and A. Saxena, 1987. “Language Universals: Inductive or Deduc-
tive?” in E. Bashir, M. Deshpande and P. E. Hook (eds.), Select Papers from
SALA 7 (Indiana: Indiana University Linguistic Club, Indiana University), 337–
342.
1989. “Reflexives and Reciprocals in Dravidian,” Studies in the Linguistic Sciences
(University of Illinois), 17.1, 121–136.
Subbarao, K. V., and G. Temsen, in press. “Comitative PP as Head in Externally Headed
Relative Clauses in Khasi,” Mon-Khmer Studies: A Journal of Southeast Asian
Languages (Bangkok, Thailand: Mahidol University), 187–205.
Subbarao, K. V., R. K. Agnihotri and A. Mukherjee, 1991. “Syntactic Strategies
and Politeness Phenomena,” International Journal of the Sociology of Language,
35–53.
Subbarao, K.V., Th. Geeta Devi and Th. Sarju Devi, 2003. “Embedded Clauses in
Manipuri,” in T. R. Kansakar and M. Turin (eds.), Themes in Himalayan Linguistics
(Heidelberg: South Asian Institute; Kathmandu: Tribhuvan University), 173–196.
Subbarao, K. V., U. R. Hakacham and M. Gogoi-Dutta, ms. “A Typological Grammar
of Rabha.”
Subbarao, K. V., U. R. Hakacham and T. Sarju Devi, 2007. “Case-marked PRO: Evidence
from Rabha, Manipuri, Hindi-Urdu and Telugu,” in R. Bielmeier and F. Haller
354 Bibliography
(eds.), Linguistics of the Himalayas and Beyond (Berlin: Mouton de Gruyter),

291–322.
Subbarao, K. V., S. Malhotra and S. Barua, 2010. “Aspects of Kokborok Syntax,”
Interdiciplinary Journal of Linguistics, 3, 1–43.
Subbarao, K. V., C. Viswanath Rao and A. Saxena, “Agreement in Telugu,” in
B. Lakshmi Bai and B. Ramakrishna Reddy (eds.), Studies in Dravidian and
General Linguistics: A Festschrift for Bh. Krishnamurti (Hyderabad: Center for
Advanced Studies in Linguistics, Osmania University, 1991), 215–232.
Subbarao, K. V., C. Viswanath Rao, N. Rao and A. Saxena, 1989. “Verb Say in
South Asian Languages,” in A. Mukherjee (ed.), Language Change and Lan-
guage Variation (Hyderabad: Centre of Advanced Study, Osmania University), 89–
104.
Tang, C. C. J., 1989. “Chinese Reflexives,” Natural Language and Linguistic Theory,
7, 93–121.
Temsen, G., 2006. “Embedded Clauses in Khasi,” Ph.D. thesis, Delhi University.
Temsen, G., 2007. “What is the Order?” paper presented at the National Seminar on the
Emerging Linguistic Scene in North-East India.
Temsen, G., and K. V. Subbarao, 2004. “Control Structures in Khasi,” paper presented
at the Fourth Himalayan Languages Conference, Central Institute of Indian Lan-
guages, Mysore, India.
in preparation. “A Grammar of Khasi.”
Tikkanen, B., 1987. The Sanskrit Gerund: A Syntactic Diachronic and Typological
Analysis. Studia. Orientalia 62.
1995. “Burushaski Converbs in their South and Central Asian Areal Context,” in
Haspelmath and König: 487–528.
2001. “Converbs,” in M. Haspelmath, E. König, W. Oesterreicher and W. Raible (eds.),
Language Typology and Language Universals (Berlin and New York: Walter de
Gruyter), 1112–1123.
Ura, H., 2000. Checking Theory and Grammatical Functions in Universal Grammar
(Oxford: Oxford University Press).
Ura, H., 2003. “Case,” in M. Baltin and C. Collins (eds.), The Handbook of Syntactic
Contemporary Theory (Malden, MA: Blackwell), 334–373.
van der Wurff, W., 1989. “The Syntax of Participial Adjuncts in Eastern Bengali,”
Journal of Linguistics, 25, 373–416.
van Driem, G., 1993. A Grammar of Dumi (Berlin and New York: Walter de Gruyter).
Varija, N., 2005. “Syntactic Convergence in Bhalavali Marathi,” paper presented at
the Prof. M. B. Emeneau Centenary International Conference on South Asian
Linguistics, Central Institute of Indian Languages, Mysore, India.
Verma, M. K., 1971. The Structure of Noun Phrase in English and Hindi (Delhi: Motilal
Banarsidass).
1976. The Notion of Subject in Indian Languages (Madison: University of Wisconsin).
1991. “Exploring the Parameters of Agreement: The Case of Magahi,” Language
Sciences, 13.2, 125–143.
Verma, M. K., and K. P. Mohanan, 1990. Experiencer Subjects in South Asian Languages
(Stanford, CA: CSLI Publications, Stanford University).
Wali, K., 2000. “Lexical Anaphors and Pronouns in Marathi,” in B. Lust et al. (2000:
513–574).
Bibliography 355
2004. “Non-nominative Subjects in Marathi,” in Bhaskararao and Subbarao (2004:

II, 223–252).
2006. Marathi (Delhi: Indian Institute of Language Studies).
Wali, K., and O. N. Koul, 1997. Kashmiri: A Cognitive Descriptive Grammar (London
and New York: Routledge).
Wali, K., and K. V. Subbarao, 1991. “On Pronominal Classification: Evidence from
Marathi and Telugu,” Linguistics, 29, 1093–1110.
Wali, K., O. N. Koul, P. E. Hook and A. K. Koul, 2000. “Lexical Anaphors and Pronouns
in Kashmiri,” in B. Lust et al. (2000: 471–512).
Wasow, T., 1986. “Reflections on Anaphora,” in B. Lust (ed.), Studies in the Acqui-
sition of Anaphora, vol. I, Defining the Constraints (Dordrecht: Reidel), 107–
122.
Wasow, T., A. Perfors and D. Beaver, 2005. “The Puzzle of Ambiguity,” in Morphology
and the Web Grammar: Essays in Memory of Steven H. Lapointe (Stanford, CA:
CSLI Publications, Stanford University).
Watters, D. E., 2002. A Grammar of Kham, Cambridge Grammatical Descriptions
(Cambridge: Cambridge University Press).
Weinreich, U., 1958. Languages in Contact: Findings and Problems, Linguistic Circle
of New York 1 (The Hague: Mouton).
Woolford, E., 1999. “More on the Anaphor Agreement Effects,” Linguistic Inquiry, 30,
257–287.
2003. “Burzio’s Generalization, Markedness, and Locality Constraints on Nominative
Objects,” in E. Brandner and H. Zinsmeister (eds.), New Perspectives on Case
Theory (Stanford: CSLI Publications), 301–329.
Wurmbrandt, S., 2003. Infinitives and Clause Structures (Berlin and New York: Mouton
de Gruyter).
Yadava, Y. P., 1997. “Maithili Agreement” (ms).
1999. “The Complexity of Maithili Verb Agreement,” in R. Singh (ed.), The Yearbook
of South Asian Language and Linguistics (Delhi: Sage), 139–152.
2004. “Non-nominative Subjects in Maithili,” in Bhaskararao and Subbarao (2004:
II, 253–264).
2007. “Raising from a Tensed Clause and Linguistic Theory: Evidence from Maithili,”
in Bayer et al. (2007: 143–162).
Yamabe, J., 1990. “Dative Subject Constructions in Indic Languages,” unpublished
Masters’ thesis, Department of Linguistics, University of Tokyo.
Yoon, J., 2004. “Non-nominative (Major) Subjects and Case Stacking in Korean,” in
Zaenen, A., J. Maling and H. Thráinsson, 1985. “Case and Grammatical Functions: The
Icelandic Passive,” Natural Language and Linguistic Theory, 3, 441–483.
Author index
Abbi, 24 Cardona, 194

Achumi, 20, 25, 205, 216, 234–235, 237 Chandra, 114–116
Agnihotri, 19, 149 Chatterji, 18, 22, 42
Aissen, 178 Chelliah, 137–138, 323
Alexiadou, 179 Chomsky, 1, 4–5, 8, 12, 44–45, 75–76, 84,
Amritavalli, 51, 60–62, 65, 75, 83, 88, 137, 93–94, 112, 132, 135, 196, 213–214, 257,
171 259
Anagnostopoulou, 72 Cinque, 3
Anderson, 39, 104, 113 Clark, 284, 287
Annamalai, 52, 301 Cole, 76, 87–88, 295
Arora, 32, 74, 153, 197, 205, 284 Comrie, 4, 15, 264, 271, 276, 278, 288, 294,
Asher, 50, 144, 152, 274, 276, 289–293, 300
301 Coopmans, 4
Aze, 104 Coupe, 55
Croft, 3
Baker, 3, 12, 102–103, 106, 110–114,
310 Dasgupta, 40, 139, 140, 161, 164, 168, 173,
Bal, 207 174, 175, 238, 315, 324, 327n.9, 328n.18
Barua, 15 Davison, 26–27, 46, 49, 68, 70, 72–73, 76–79,
Bashir, 248 82, 85, 88, 96, 114–116, 135, 154, 156–157,
Bayer, 6, 27, 161, 168, 206–209, 220, 159–160, 162–163, 172, 177, 181–182, 192,
222–223, 226–227, 229, 238–239 207, 209, 220, 224, 226, 231, 233–235, 239,
Beermann, 160–161 241–244, 260, 279, 313, 319n.23, 320n.27,
Benmamoun, 94, 106 324n.15, 325, 326n.2, 328n.11, 329n.23
Bhaskararao, 3, 35, 79–80, 91, 137, 140, 146, Dayal, 40, 85–87, 206, 220–221, 223–226,
151, 153, 161–162, 166, 170, 172, 178–179, 230–231
181, 189, 191–192, 229, 254 Deeney, 25, 58, 202
Bhat, 210–211, 216, 226, 319n.17 Dixon, 60
Bhatia, 73–74, 101, 168, 200, 285, 319n.25 Donegan, 12, 48
Bhatt, Rajesh, 40, 114–116, 119–120, Downing, 264
123–124, 141–142, 144–145, 160, 179, 199, Dowty, 7
266, 268–269, 277, 316n.17, 325n.11, Dwivedi, 207
329n.6, 330n.17
Bhatt, Rakesh, 16, 20, 199, 207, 272, 274, Emeneau, 1, 18, 27, 42, 134, 194, 201
326n.3 Everaert, 43, 59, 72, 78, 92, 313, 317–318n.2,
Biberauer, 19, 27, 210, 240, 243–244, 319n.19
329n.26
Bickel, 37, 160, 169, 260 Gair, 3, 41, 49, 62, 67, 140–141, 202–203,
Bloch, 42 216, 279, 284, 286
Burling, 151, 248 Gambhir, 29
Burrows, 63, 104 Gill, 295
Butt, 23, 114–116, 221 Giridhar, 248
356
Author index 357
Greenberg, 19, 21, 314–315 Macphail, 39, 105

Gumperz, 20 Magier, 174–175, 177–178
Gurtu, 155, 231, 235 Mahajan, 85–87, 96, 114–116, 124, 149,
174–175, 177–178, 188, 226, 231, 235,
Hany Babu, 198, 202 239–240, 323
Haspelmath, 60 Mahapatra, 127, 198
Herring, 37 Mahto, 100, 114
Hock, 19–20, 27, 194, 197, 201, 203 Malhotra, 15, 64
Hook, 22–23, 50, 101, 115, 135, 149, Masica, 1, 11, 18, 22–24, 26–27, 30, 42, 134,
162 140, 146, 188, 194–195, 202, 209, 216,
Hornstein, 247–248, 257 270–271
Matisoff, 37
Ichihashi-Nakayama, 160, 169, 260 McAlpin, 146, 167
Minegeshi, 58
Jaeggli, 27 Mistry, 49, 63, 97–98, 194, 201, 211–212, 218,
Jayaseelan, 50–51, 74, 80, 82, 84, 87, 90, 137, 248, 280
149, 162, 170, 175, 181, 184–185, 188, 192, Mohanan, 140, 146, 155, 168–169, 188, 229,
229 235
Mohanty, 207
Kachru, 20, 26–27, 96, 135, 146, 170, 194, Montaut, 152
202, 279 Mukherjee, 88
Kak, 141, 160, 200–201, 205, 274, 281 Murmu, 58
Kayne, 13, 112–113, 133, 239
Kazenin, 59–61, 63, 78 Nadkarni, 290
Keenan, 4, 15, 264, 269, 271, 276, 278, 288, Nagaraja, 44
294, 300 Neukom, 123, 203
Kevichüsa, 20, 22, 37, 55, 59, 216, 268, Newmeyer, 7
298–299, 302 Noonan, 194
Kimenyi, 110
Klaiman, 167 Osada, 207
Koh, 9, 57, 63, 78, 105, 125, 197, 206, 248
Koshal, 248 Pandharipande, 24, 96, 135, 137–138, 142,
Koul, 20, 50, 101, 114, 135–136, 150, 236, 167, 170, 194, 198, 200, 204, 209, 215,
243, 272–273, 281, 285 241, 243, 273–274, 285–287, 328–329,
Krishnamurti, 26, 34, 36, 47, 60, 121, 139, 332n.26
167, 170, 275, 291 Pandit, 19–20
Kuiper, 27, 194, 201 Paramasivam, 176
Kumar, 270, 297, 308 Patnaik, 44, 48, 123, 202, 207, 327
Kumari, 50, 289, 291–292, 301 Peterson, 25, 38–39, 48, 58, 64, 142, 159, 200
Kuolie, 22 Picallo, 72
Kuroda, 247 Piñón, 60
Polinsky, 246–248, 256–257
Lahaussois, 37 Porizka, 96
Lalitha Murthy, 48, 51–53, 60, 62, 65, 68–69, Poudel, 145
74, 80–81, 83, 85, 107, 111, 126, 128–130, Prakash, 129
135–136, 141, 145, 159, 202, 219, 247,
249–250, 254, 259 Raina, 20
Lambert, 202 Ramakrishna Reddy, 127, 165
Lasnik, 5, 8, 29, 257, 259 Ramamurti, 39, 64, 102–104
Lehmann, 52, 97, 176 Ramanujan, 18
Lichtenberk, 56 Ramarao, 301
Lidz, 178 Riemsdijk, 263
Lohse, 14 Rizzi, 209
Lord, 194 Rosenbaum, 193
Lust, 3, 5–6, 43, 45, 83, 87, 156 Ross, 223
358 Author index
Safir, 27 Tang, 28, 60, 88, 108

Sarju Devi, 37, 54–55, 90, 116, 211 Temsen, 55, 61, 125, 142, 213, 217, 219, 222,
Saxena, 5, 27, 51, 70, 155, 254 277, 308
Schiffman, 18, 176 Tikkanen, 202
Shibatani, 146, 154, 170, 307–308
Sigurdsson, 13, 154, 171, 247, 259
van Driem, 123, 146
Singh, 98–99, 114–115
Varija, 35, 189
Sridhar, 97, 146, 150, 152, 170, 188, 216, 274,
Verma, 100, 140, 168
290
Srivastav, 236–237, 239, 277
Steever, 275 Wali, 3, 20, 49–50, 67–71, 73, 76–79, 88, 101,
Stowell, 195, 220, 226, 238 114, 135–136, 150, 165, 194, 212, 214,
Subbarao, 3, 5, 15, 18, 20, 22, 27–28, 32, 35, 226–228, 235–236, 238, 272–273, 278,
37–38, 40, 42, 44, 48, 51–55, 57, 60–63, 65, 280–282, 285
68–70, 74, 76, 78–81, 83, 85, 88, 90, 105, Wasow, 14, 43
107–108, 111, 114–117, 119–123, 125–126, Woolford, 72
128–130, 131, 135–137, 140–142, 145–146, Wurmbrandt, 116
151, 153, 161–162, 165–166, 170, 172,
178–179, 181, 189, 191–192, 194, 197,
Yadava, 69, 99, 114, 120, 215
202–206, 213–214, 216, 219–223, 237–240,
Yamabe, 70
247–250, 254, 257, 259, 261, 268, 277, 279,
Yoon, 157, 199
288, 290, 298–299, 302, 304, 308, 313,
315n.8, 316, 318, 319n.17, 320, 322n.16,
323, 325–327, 332n.26, 334 Zaenen, 168
Language index
Austro-Asiatic, 1, 9, 11, 15, 18–19, 38, 48, 55, Kharia, 10, 25, 38–39, 44–45, 48, 57–59,
81, 94, 114, 125, 134, 300, 313 63–64, 81, 142, 144, 158–159, 193,
Mon-Khmer, 1–2, 7–11, 15, 18, 21, 33, 42, 195, 200–201, 294, 346, 348
44–45, 48, 56, 59–61, 64, 81, 84, 91, Mundari, 10, 45, 48, 57, 61, 126–127,
102, 106–108, 122, 125, 132, 137, 142, 193, 195–196, 202, 207, 347
144, 149, 159, 193, 196–197, 212–213, Santali, 10, 39, 45, 48, 57–58, 61,
217–220, 222, 249, 263, 265, 271, 276, 105, 132, 193, 195, 202–203,
279, 294, 299–302, 304–305, 313–314, 345–347
316–317, 320–321 Savara, 10, 39, 107, 348
Khasi, 7–9, 12, 17–22, 29, 33, 42, 44–45, Sora, 10–13, 39, 48, 63–64, 94, 102–104,
48, 55–56, 59–61, 64, 81, 84, 91, 102, 107–108, 111–114, 132–133, 313, 336,
106–108, 113, 122, 125, 132, 137, 348
141–142, 144, 149, 159, 193, 195–197, Dravidian, 1–2, 9, 11–12, 15–16, 18, 21,
212–214, 217–220, 222–223, 249, 265, 23–24, 26–28, 31, 33–37, 40–41,
276–277, 279, 294, 301–302, 304–306, 44–47, 50, 54–55, 60–62, 65–66, 74,
308, 313–318, 320–321, 325, 329, 333, 79, 90–91, 94–95, 103, 108, 114, 121,
336, 345, 347, 352–353 124, 126–128, 133–134, 136–137, 140,
Pnar, 7, 9, 17, 21, 107, 314, 321 142, 149, 151–154, 158–159, 180,
Rymbai, 9, 17, 21, 107, 314, 318, 321, 187–188, 193, 195–197, 201–203, 206,
336 218, 241, 244, 248, 257, 259–260,
Munda, 1, 10–12, 15, 18, 25, 27–28, 33, 262–264, 266–271, 274–279, 282, 284,
38–40, 42, 44–45, 48, 57–64, 66, 78, 286–289, 291–294, 299–300, 311,
81, 91, 94, 102–105, 107–108, 316–318, 325–327, 330, 332
111–114, 122, 125–128, 132–134, Kannada, 9, 34–36, 51, 55, 60–62, 65,
141–142, 144, 158–159, 188, 193–197, 74–75, 83, 88, 97, 121, 135, 137,
200–203, 205–207, 245, 249, 262–263, 150–154, 156, 178, 188–190, 216, 248,
265, 271, 274, 276–279, 294, 300–301, 257, 259, 274, 276, 290–291, 311, 319,
312, 317 332, 335, 345, 347, 350
Ho, 1–2, 5, 10–11, 13, 19–20, 22–23, 25, Malayalam, 8–9, 27, 34, 36, 44–45, 50–51,
27, 30–31, 33, 44–45, 48, 50, 53–54, 70–72, 74, 80, 82, 84–85, 87, 90–91,
57–58, 60–63, 67, 69, 71, 75–76, 78, 95, 134, 137, 139, 144, 152–153, 162,
81, 86, 91, 93, 95, 101, 103–105, 115, 171, 173, 175–178, 184–185, 187,
119–120, 125, 135, 139, 143, 145, 189–190, 198, 202, 229, 248, 257, 274,
149–151, 161–162, 165–166, 173, 179, 276, 289, 291–293, 301, 305, 317, 336,
193–197, 201–203, 205, 209–211, 213, 343, 346–347
217, 234, 237, 244, 247–248, 257, 262, Manda, 9, 126–128, 165, 348
276, 291, 299, 303–305, 309–310, 313, Tamil, 8–9, 23, 36, 52, 70–71, 97, 121,
316–317, 319, 326, 329, 331–332, 123, 134, 139, 171, 173, 175–178, 198,
335–340, 342, 344, 346, 349, 352–353 229, 248, 270, 275, 284, 290, 301, 305,
Juang, 10, 44–45, 48, 195–196, 207, 327, 311, 324–325, 336, 345, 348–349,
348 351
359
360 Language index
Dravidian (cont.) 257, 259–260, 262, 266–268, 274,

Telugu, 9, 14, 23–24, 26, 29, 31–32, 34–36, 277–286, 288, 293–294, 301, 316, 320,
44, 46, 48, 50–52, 54–55, 60–62, 322, 324–325, 327–333, 335–337,
65–67, 69–72, 74–75, 77, 79–80, 339–340, 343–344, 347, 351–352
82–86, 89–90, 108, 114, 121, 123, Japanese, 11, 75, 154, 307, 308, 344
135–137, 139, 143–149, 151–156, Kashmiri, 9, 15–18, 20, 26–27, 30, 40–42,
158–164, 166–173, 176, 178–183, 45, 50, 69–73, 91, 95, 101, 114,
186–191, 196–198, 204, 206, 225, 118–119, 123, 135–136, 141–142,
228–230, 232–237, 244–249, 253–254, 144–146, 149–150, 159–162, 179–180,
257, 259–260, 275–276, 280, 284, 287, 193, 195–196, 199–201, 205, 209, 228,
289, 291–293, 301, 305, 311, 316, 319, 235–236, 238–239, 241, 243, 257,
322, 324–325, 327–328, 332, 337, 344, 272–274, 280–281, 285–286, 288, 294,
349–353 301, 314, 324, 326, 337, 340, 342–344,
Toda, 9, 44–45, 50, 91, 141, 289, 318 348, 352–353
Indo-Aryan (IA), 1–2, 5–6, 9, 11–12, 15, 18, Konkani, 9, 23–25, 27, 35, 40–41, 196, 241,
21, 23–24, 26–28, 30–33, 35, 39–41, 258–259, 276, 282, 301, 311–312, 317,
44–46, 48–49, 58, 62–63, 73, 77, 79, 325–326, 332–333, 347
94–95, 100–101, 103, 105, 108, Korean, 11, 76, 88, 157, 196, 307, 308, 354
114–115, 124, 126, 133–134, 136, 142, Maithili, 9, 69, 95, 97–101, 114–115, 120,
151, 159, 174, 193–197, 201–202, 209, 165, 215, 338, 350, 354
218, 222, 238, 241, 243–244, 248, 257, Mangalore Konkani, 9, 23–24, 35, 258–259,
259, 263–264, 267–272, 274–282, 286, 311, 317, 325–326, 332–333
288, 292–294, 299–301, 311–312, Marathi, 5–6, 9, 13, 23–24, 27, 30, 35,
316–317, 320, 330, 332 40–41, 43–45, 49, 66–69, 72–73,
Assamese, 9, 23–24, 27, 32, 45, 136, 138, 76–77, 79, 83, 91–92, 101, 135–139,
141, 188, 193, 196, 203, 212–214, 217, 142, 154, 164–165, 193–194, 196–198,
238, 241, 248, 257, 281, 315, 328, 331, 200, 203–204, 209–210, 212, 214–215,
334, 336 218, 226–228, 238, 241–243, 248,
Bangla (a.k.a. Bengali), 8–9, 23–24, 26–27, 272–274, 278, 280, 282–286, 288, 294,
30, 32–33, 40–41, 45, 57, 70–71, 79, 301, 311, 319, 326–327, 330, 332, 347,
82, 134, 136, 139–141, 144, 148–149, 353
152, 156–158, 161–162, 164, 168–169, Nepali, 9, 27, 45, 136, 142, 145, 159–160,
171, 173–175, 177–178, 188, 193, 196, 162, 169, 186, 193, 196, 241, 248, 257,
206–209, 212–214, 217, 220–222, 260, 280, 284–285, 287–288, 301, 332,
226–227, 237–239, 241–242, 246, 248, 339, 342, 348–349
257, 260, 280–283, 288, 301, 315–316, Oriya, 9, 15, 23–24, 27, 30, 41, 45, 48, 57,
320, 327–332, 336, 339 79, 123, 126, 136, 160–161, 193,
Bhalavali Bhasha, 9, 246, 248, 276, 196–198, 207, 238, 241, 248, 280,
311–312 282–284, 287–288, 294, 301, 311, 315,
Dakkhini, 9, 31–32, 193, 196–197, 331–332, 336–337, 347, 349
204–206, 240–241, 246, 248, 257, 271, Punjabi, 9, 13, 26–27, 32, 40–41, 45, 70–74,
280, 284, 294, 311–312, 325–327, 332, 95, 98, 101, 114–115, 124, 135–136,
336, 351 139, 142, 149, 151–152, 160–161, 168,
Eastern Bangla, 9, 248, 257 176, 186, 193, 195–196, 199–201, 209,
Gujarati, 9, 40, 44–45, 49, 63, 91, 97–98, 238–239, 241, 259–260, 262, 278,
101, 136, 193–194, 196, 200–202, 281–283, 285–286, 288, 294, 301, 316,
210–212, 218, 241, 248, 278, 280, 294, 329–330, 332, 337
317, 327, 338, 345–346 Shina, Eastern, 9, 330, 342
Hindi-Urdu, 9, 13, 20–24, 26–32, 40–41, Shina of Gultari, 9, 342
45–46, 48–50, 68–73, 76, 78–79, 82, Shina of Skardu, 9, 101, 342
88, 91, 95–96, 98, 101, 108, 113–119, Sinhala, 9, 41, 44–45, 49, 62, 67, 91, 136,
124, 133, 135–144, 147–163, 167–177, 140–141, 193, 196, 202–203, 216, 241,
180, 185–187, 193–197, 199–201, 270, 276, 279–280, 284–286, 294, 311,
205–206, 209, 216–217, 220–226, 327, 330, 332, 341
230–231, 233, 236–241, 243, 246, 248, Subzapuri, 9, 246, 257, 341
Language index 361
Swat-Dir Kohistani, 9, 248 Manipuri, 7, 10, 28, 37, 44–45, 54–55, 59,
Torwali, 9, 248 81–82, 89–90, 93, 107–108, 137–138,
Tibeto-Burman, 1, 7, 10–11, 13–16, 18, 21, 140, 151, 159, 202, 210–212, 216, 218,
23, 25, 27–28, 31–33, 37–41, 44–45, 259, 261–262, 278, 295, 302–306, 308,
47–48, 52, 59–61, 65–66, 72, 80, 91, 319, 323, 333–334, 337–338, 341,
94–95, 102–106, 108–109, 111, 114, 348–349, 352
120, 122, 125, 132–135, 137, 140, 149, Mao, 10, 248, 341
151, 159, 177, 188, 193, 195–196, 201, Mising, 10, 21, 334n.43, 348
203, 206, 218, 241, 244, 248, 263–265, Mizo, 7, 10, 15–16, 27, 45, 47–48, 52–53,
268–269, 271, 274, 276, 278–279, 290, 59–60, 62, 65, 68, 72, 80–83, 86, 91,
294–295, 298–300, 302–304, 306–307, 95, 103, 106–107, 109, 111, 114, 120,
312, 315–317, 319, 321, 323, 325, 122, 125–126, 128–133, 135–137, 140,
329 145, 149, 202, 218, 246, 248–252, 278,
Ao, 10, 44–45, 55, 59, 248, 302, 339, 348 294–297, 302–303, 305–306, 308,
Bodo, 10, 21, 23, 45, 70–71, 107–108, 137, 318–319, 322, 329, 344–345, 348, 352
139–140, 147, 151, 159, 173, 177, 248, Paite, 10, 27, 48, 95, 102–103, 106, 109,
276, 334n.43, 337 111, 122, 132, 149, 278, 297, 303, 322,
Dimasa, 10, 303, 333n.43, 334, 343 333–334
Dumi, 10, 123, 145–146, 324, 353 Rabha, 10, 23, 28, 45, 55, 108, 140, 159,
Garo, 10, 151, 248, 338 218–219, 223, 259, 276, 323, 334, 352
Hmar, 10, 16, 21, 27, 45, 49, 60, 62, 64–66, Sangtam, 7, 10, 269–270, 295–296, 298,
68, 72, 78, 88, 91, 93, 95, 102–103, 302–307
105–107, 109–111, 114, 122, 125, Sema, 7, 10, 20–21, 25, 204–205, 216,
128–130, 132, 135, 137, 141, 149, 202, 234–235, 237, 276, 295, 298–299, 308,
264–265, 270, 278, 294–297, 302–303, 335, 339–341, 343, 346–351
305–306, 308–312, 315, 319, 321–322, Tenyidie (a.k.a. Angami), 7, 10, 20–22, 28,
333, 334n.48, 344, 347 37, 44–45, 55, 59, 265, 267–268, 276,
Kham, 10, 111, 139, 151, 177, 353 295, 302, 304–308, 316, 319, 343–344,
Kokborok, 10, 13, 15, 33, 42, 45, 108, 351
137–138, 148–150, 159, 193–194, Thadou, 10, 16, 38, 45, 48, 103, 128,
202–203, 248, 315, 319, 336, 346, 352 131–132, 135, 137, 244–245, 278, 294,
Konyak, 7, 10, 295, 298, 302, 306, 347 297, 303–306, 308, 313, 333n.40, 351
Ladakhi, 10, 248, 344 Tiwa, 10, 102, 145–146, 323, 347
General index
A-bar movement, 85–86, 92 Backward Control, 12, 14, 31–32, 158,

ablative, 99–100, 287, 291, 298–300, 305 246–260, 262, 317, 329, 341, 348,
abstract case-marking, 264 351
accompaniment, 7, 265, 291, 300, 308 movement analysis, 257
accusative case-marked subject, 140 Binding Theory, 44–45, 76, 83–84, 214, 247,
accusative subject(s), 134, 140, 345 254
acquisition of knowledge or a skill, see dative blocking effects, 43–44, 87–88, 92, 343,
subjects – domains of occurrence of 347
adposition incorporation, 94, 102, 109–110,
132, 264–265, 309–311 canonical position of the internal head in
in IHRCs, 264–265 IHRCs, 264
against Rightward Extraposition, 196, capabilitative, 96, 142–143, 166
239–240 Capabilitative Passives, 96
agent-oriented adverbs, 180 case alternations, 246–247, 351
agreement, 12–13, 15–16, 23, 27–28, 39–40, case and agreement in SALs, 94, 128–129
43, 57, 63, 72, 77–79, 87–88, 91–112, case checking properties of NNS, 135
114–135, 137–138, 141–142, 146–147, Case Copying, 44, 51, 54, 89–91, 349
165–166, 176, 181–183, 187–188, 192, case-marked PRO, 259, 261–262
195–196, 206–207, 212, 215, 240, Case Resistance Principle (CRP), 195–196,
249–251, 253, 259, 262, 277, 294–297, 218–221, 226, 238, 240, 328
311, 320–325, 330, 334, 337–342, c-command, 43, 45, 68–69, 74, 83, 85, 88, 94,
344–345, 348, 350–354 106–108, 154, 156, 195–196, 214, 240,
in Malto, 127 247, 254
in Manda, 126 clefts, 16, 41, 94, 128–131, 193, 317
agreement typology of SALs, 95, 97, 99, 101, in Mizo, 128–129
103, 105, 107 clitic(s), 7, 14, 34, 39, 44, 54, 64, 74, 77–78,
ambiguity resolution, 264 92, 95, 102–104, 108–111, 122–126,
ambiguous interpretation of EHRCs and 132, 145–146, 191, 207, 224, 234–235,
IHRCs, 278 239, 312, 318, 320, 322–324, 351
A-movement, 85–86, 92, 346 cognitive gerunds, 36–37, 317
Anaphor Agreement Principle, 72, 79, coindexing, 43, 49, 59, 61, 66, 71, 76–78,
91 81–82, 85, 103, 105, 132, 156–157,
animacy, 8, 174, 176–178, 183, 325 183, 214, 217, 231, 233, 247, 253, 255,
antecedent–anaphor binding, 115 264–266, 269, 272, 290, 295, 297, 319,
anticausative, 48, 59–61, 75, 91, 179, 335 327
applicative constructions, 94, 309–10 comitative PP, 7, 14, 38, 109–110, 132, 265,
asymmetries 273, 278, 286–288, 291–292, 294,
between the pre-nominal and post-nominal 298–308, 312–313, 318, 322, 331–334,
relative clauses, 265 351–352
between the root and embedded clause in as head, 7, 292, 299–308, 312–313, 318,
Kashmiri (IA) complements, 199 332, 351–352
in relative clauses in Indo-Aryan, 277 sociative PP, 291
362
General index 363
COMP, 16, 165, 179, 188, 194–201, 203–207, as a controllee, 179–180

209, 211, 215, 222, 224–225, 227–228, natural phenomena pertaining to body,
232, 234, 237–241, 327, 336–337, 340 146–7
COMP agreement, 195–196, 206, 240 part–whole relationship (single and
comparative, 21, 113, 319, 342, 349 double dative marking), 147, 153
complementation, 16, 120–121, 193–196, 198, perceiver of visual and auditory actions/
200, 202, 204, 206, 208, 210, 212, 214, perceptions, 146, 149
216, 218, 220, 222, 224, 226, 228, 230, physiological and mental ailments, 146–7
232, 234, 236, 238, 240, 242, 244–245, psychological states and emotions, 146–7
326, 335–336, 343, 345, 347, 350 subject of predicates expressing
complement clauses, 6, 13, 15–16, 27, 30, 37, obligation and necessity
116–117, 177, 195, 198–199, 217–218, (desideratives), 147, 151
220–223, 239–241, 273, 319, 327–328, temporary possession (to indicate /
339 involving), 139, 149, 151
clause-final position, 193, 219, 315 to denote a recipient, 147, 152
clause-initial position, 193, 221–222 to express possession and kinship, 147,
clause-medial position, 218, 220–221 149
complementizer(s), 6–7, 12, 14, 16, 27, 30, 40, [−transitive] nature of the host verb, 94,
81, 154, 164–165, 193–196, 199, 109
201–202, 204, 206–207, 209–213, modification of, 280
217–219, 224–225, 227, 232, 237–244, default agreement, 95–97, 108, 124, 141–142,
277, 313, 326–328, 336, 339, 345 147, 181–182, 187–188, 322, 324
complex anaphor, 6, 14, 45, 50, 70–72, 79, 83, detransitivizer, 16, 38, 59–61, 75
89–91, 171, 179–180, 214 diachronic change, 196, 240, 243–244, 337
complex predicates and verb agreement, 182 diachronic explanation, 13, 114, 133
compound verbs, 22, 162, 339, 342 Differential Object Marking, 8, 178, 335
conjunctive participle (CP), 22–23, 26, 32, DOM, see Differential Object Marking
127–128, 157, 159, 246, 257, 259, 262, direction of c-command, 195–196, 240
335, 337, 343–344, 350–351 and the occurrence of the Negative Polarity
CP-adjoined relative clause, 266 Items, 196
conjunct verbs, 22–23 direct object, 15–16, 19–21, 27–29, 39–40, 43,
control structures, 31, 246–247, 353 50, 52–54, 59, 63, 73, 77, 93, 97–105,
convergence, 1–2, 11, 14, 20, 31, 33, 41, 108–110, 112, 122, 125, 128, 132–133,
126–127, 195, 197, 313–314, 336, 135, 171, 218, 233–234, 254, 267–274,
341–342, 344, 348, 350–351, 353 278, 280–281, 283–285, 288–290, 294,
coordinate reduction, 168 296–298, 325, 329
covert operator, 263–264, 267, 271, 283–286 modification, 267, 280, 285, 289–290, 296
creative errors, 173, 217, 246 direct vs. indirect speech, 215
domain, 14, 45, 49, 58, 67–68, 75, 83–84, 134,
dative subject(s), 15, 34–35, 44, 51, 61–62, 146–147, 149, 151, 153, 192, 320, 345,
69–71, 74, 89, 97–101, 108, 126–127, 348
134–136, 146, 148–149, 151–155, double complement clauses, 199
157–162, 164–170, 177, 179–192, double/multiple dative subject constructions,
253–256, 260, 272, 278, 280, 292–294, 34, 135, 154, 188–192, 325–326
319, 325–326, 330, 342–344, 346–347, Dravidian languages, 12, 23–24, 27, 33–36,
349–350, 354 44, 50, 54–55, 62, 91, 95, 103, 121,
construction, 15, 34, 51, 61–62, 70, 74, 89, 126–127, 136–137, 140, 142, 149,
97, 134, 154, 166–167, 183, 188–190, 151–153, 158–159, 180, 187, 193, 195,
192, 254, 293, 325–326, 330, 342–343, 201–203, 206, 257, 259–260, 266,
see also DSC 269–270, 275–276, 278, 282, 284,
default agreement and inherent case 286–289, 291, 293, 300, 311, 316,
marking, 181 325–326, 330, 332, 344, 350–351
domains of occurrence of DSC, 8, 15, 89, 134, 155–156, 168, 170–171,
acquisition of knowledge or a skill, 147, 173, 175–181, 183–185, 187, 192, 319,
152 see also dative subject(s), construction
364 General index
echo word(s), 9, 20, 22, 26, 199, 314, 315–316 two FCs occurring in a row, 198
ECM, 79, 81–82, 95, 123–125, 133, 135, Final Complement, 6, 154, 164, 193–195, 207,
144–146, 178–179, 320, 347, see also 241–244
Exceptional Case Marking final complementizer, 6, 154, 164, 193–195,
cases in Dravidian, 124 207, 241–243, see also FC
cases in Munda, 125 Final over Final Constraint / FOFC, 19, 27,
cases (Small Clauses) and long-distance 196, 210, 240, 243–245, 329, 337
agreement, 123 Finiteness of the IC-clause vs. post-verbal
cases (Small Clauses) in Khasi, 125 position in blocking Wide Scope, 226
subjects Forward Control, 14, 31–32, 158, 246–250,
in Mizo and Hmar, 125 252–253, 258–259
accusative case-marked, 144 functional category, 14, 93–95, 106, 126, 128,
nominative case-marked, 145 185, 194, 240, 350
EHRCs, 263–265, 269, 271, 277–278, 283,
289, 292, 294–296, 299–300, 305, genitive subject, 7–8, 33, 44, 68, 70–71, 134,
307–309, 311–313, 318, 322, 332, 351, 136, 147–148, 156–157, 161, 164,
see also Externally Headed Relative 167–168, 178, 187–188, 259–260, 319
Clause grammatical functions (GF), 106, 110, 168,
embedded relative clause, 128, 263, 266–267, 271, 273, 275, 277–278, 293, 297, 336,
269, 276, 295, 299, 309, 326 353–354
embedded subject accessible to relativization in sentential
accusative case-marked, 135 relative clauses, 271
ergative case-marked, 135, 283 grammaticalization, 65, 91, 342
ergative grammaticalized form, 194, 198, 201, 218
case marking, 281, 283, 330 grammatical relation, 264, 270
subject modification, 279
subjects, 154 Head Direction Parameter, 6, 12, 194
Ergative–Absolutive, 41, 135, 187 Head to Head Movement, 6, 12
Evidence for Backward Control, 249, 251,
253, 255 IC-clause in clause-initial position, 221
Exceptional Case Marking, 79, 123, 135, IC functioning as
144–145, 171–173, 179, 323, see also a disjunctive marker, 200, 201, 307
ECM an expletive IC in non-restrictive clauses,
existential possession, 137 201
experiencer (dative) as a yes/no question marker in Kharia
as null PRO, 180 (Munda), 201
modified, 180 a subordinating linker with the
expletive as an additional question word interpretation of then, 201
(wh-expression), 236–237 IHRC(s), 38, 263–266, 269–270, 277–278,
expletive expressions in SALs – absence of, 290, 295–300, 302–313, 329–330, 351,
30–31, 316 see also internally Headed Relative
expletive light verb, 35, 37 Clause(s)
Externally Headed Relative Clause, 263–265, Pattern I, 297
267–268, 277, 279, 281, 283, 285, 287, Pattern II, 297
289, 291, 293, 299, 313, 352, see also imperfect participle, 172, 278
EHRC implications of postposition incorporation, 308
inalienable possession, 35, 137–138, 140, 153,
factive, 59, 62–63, 67–68, 195, 208, 210–212, 189, 192
215, 227–228, 232, 238 incorporation, 7, 11–12, 14, 38–40, 93–95,
factors responsible for ambiguity and 102–114, 126, 132, 264–265, 304,
disambiguation in IHRCs, 278 306–313, 317, 320–322, 333–334, 336,
FC, 6, 19, 27, 154, 165, 191, 193–219, 344
223–225, 227–230, 234–236, 238–240, in polysynthetic languages, 111, 113
243–245, 326–329, 337, see also final into the predicate, 39
complementizer nouns, 11, 107, 112
General index 365
of a direct object, 102, 104, 320 299, 308, 313, 329, 339, 341, 344, 348,
of a DP, 321 see also IHRC(s)
of adverbs, 7, 107, 308 internal head of an IHRC with an overt case
of inclusive particles, 106, 108 marker, 269
of intensifiers, 106 nature of, 264
of locative and instrumental PP, 39, 102 with no overt postposition, 270
of locative postposition in Paite, 111 intransitive nature of the dative predicate, 192
of Negative Polarity Items / NPIs, 111, 106,
108 Kayne’s generalization, 13, 112–113, 133,
of quantifiers, 106, 108 239
of the comitative postposition, 14, 110, 132,
322 language contact, 14, 18–20, 31, 33, 41, 95,
possessor of object and direct object, 104 126–127, 196, 240, 243, 246, 312, 314
possessor of subject, 105 and agreement, 126–127
postpositions, 38, 95, 106, 111, 126, 304, language–dialect dichotomy, 246
306–311, 313, 317, 321, 323, 336 LDA, see long-distance agreement
pronominal suffixes, 105 left peripheral, 193
verbal reciprocal, 308 COMP, 193, 195, 199, 241
indirect object(s), 15, 20–21, 27, 29, 39–40, lexical case, 90, 94–95, 97, 99, 108, 181, 260,
43, 53–54, 93, 98, 100, 102–104, 295, 340
108–109, 132–133, 135, 267, 269–271, marker (postpositional), 94
273, 278, 285, 290, 294, 296–298 lexical case marking, 12, 100, 264
modification, 285, 290 lexical/inherent (case), 134
indirect speech, 30, 195, 215, 240, 316 lexical passive, 74, 96, 140, 143, 167
infinitival clauses as NPs, 116 locative, 6, 12, 39, 44, 46, 73, 82–83, 95, 97,
infinitival form of the embedded verb, 265 102, 108, 111, 126–127, 132, 134–135,
infinitival (nominalized), 263 139–140, 150–151, 153, 157, 163–164,
inherent Case, 12, 15, 94, 134–135, 181–185, 168, 181–182, 188–191, 254, 261–262,
187–188, 191–192, 246, 248, 325, 329 271, 273, 286–287, 290, 292–293, 298,
assignment in DSCs, 181, 183, 185, 187 300, 305, 310–311, 325, 331
complex predicates and verb agreement, subject(s), 73, 108, 157, 164, 168, 292–293
182 long-distance agreement, 13, 94–95, 100,
default agreement and inherent Case, 181 114–124, 126, 133, 259, 262, 321, 324,
role of aspect, 185 337–338, 340
role of tense, 184 in Hindi-Urdu, 115–118, 259, 262, 337
role of the verb “to come”, 186–187 in Kashmiri, 119
marker, 12, 15, 94, 108, 134–135, 181–185, in Mizo, 120
187–188, 192, 246, 248, 325, 329 in Telugu, 121
marking of the dative subject, 135 long-distance anaphoric relationships, 115
of the embedded subject, 246 long-distance antecedent, 43, 76, 83, 87, 92,
inherently case-marked, 15, 94, 134, 181, 135–136, 214
191–192, 261 long-distance binding, 5–6, 14, 43–45, 66,
Initial Complement, 6–7, 12, 27, 193, 195, 75–78, 82, 87, 92, 319–320
201, 209, 213, 220, 224–225, 241–242,
313 Minimalist Program, 43, 93–94, 157, 320,
Initial Complementizer (IC), 6, 27, 188, 338–339
193–201, 204–226, 228–231, 236, modals, the case of, 166
238–241, 243, 313, 317, 321, 327–329 modification
instrumental, 12, 18, 39, 46, 54, 73, 78, 102, of possessor, 278, 288
132, 134–135, 137, 141, 176, 188, of the comitative, 286, 288, 294, 299–301
286–287, 291, 298, 300, 305 of the locative phrase, 286
(by-passive), 134 of the oblique object, 287
intensifiers, 13, 93, 106, 109, 132 monomorphemic anaphor, 5, 38, 43–45, 53,
Internally Headed Relative Clause(s), 7, 15, 55, 75–77, 91
38, 263–266, 269–270, 277, 295, 297, movement of clauses, 193
366 General index
multi-head relative clauses, 268–269 146, 148, 150–152, 154, 156, 158, 160,
multiple sets of agreement features, 95 162, 164–166, 168, 170, 172, 174, 176,
178, 180, 182, 184, 186, 188, 190, 192,
Narrow and Wide Scope, 195, 223–225, 259, 278, 292, 322, 326, 336–337,
227–229, 231, 233–235, 237–240 339–340, 346, 350–351, 353–354, see
factors affecting Wide and Narrow Scope, also accusative subject(s); dative
224 subject(s); genitive subject; subject,
finite and non-finite nature of the Initial genitive case-marked; locative,
COMP (IC) clause, 225 subject(s); NNSs
finiteness of the IC-clause vs. post-verbal as head, 292
position in blocking Wide Scope, 226 subject in by-passive (instrumental subject),
nature of the matrix predicate and COMP, 140
227 non-subject properties of NNSs, 165
occurrence of clitics/particles as scope Noun Phrase Accessibility Hierarchy, 15, 264,
inhibitors, 234–5 271, 275, 276, 278, 285, 288, 294, 300,
position of occurrence of the 332, 341
complementizer: initial or final, 224 NP-adjoined relative clauses, 266
position of question expressions and Scope NPAH, see Noun Phrase Accessibility
possibilities, 230 Hierarchy
position of the FC clause, 229 null (covert) operator, 263, 277, 279–281
role of intonation, pause and
morphophonemic criteria, 228 object complementation, 193
topicalization and Wide Scope, 233 Object of the Genitive, 267, 288, 332
use of pleonastic question expressions as obligative constructions, 101, 139
scope markers, 235 Oblique Object(s) (OO), 27, 40, 43, 59, 93,
Wide Scope reading, 224, 227–230, 103, 108–109, 126, 267, 271–273, 278,
232–236, 238–240 286–288, 290, 294, 297–298, 300–301,
natural phenomena pertaining to the body – 305, 332
dative subjects, 146, 148 modification, 286, 294, 297
nature of the predicate in DSCs, 170–171, 173, operator, see null (covert) operator
175, 179 operator scope, 193
Negative Polarity Items, 13, 93, 106, 108–109,
132, 195–196, 325, 337, 344 parameter(s), 1, 5, 6–8, 12, 19, 27–28,
NNSs, see also non-nominative subjects 111–112, 178, 194, 224, 239, 257, 313,
and subject-oriented verbs, 154, 162 325
as antecedents to anaphors, 154 negatively specified, 6–7, 208
as controllers of PRO, 154, 157 positively specified, 6
domains of occurrence, 134, 146–147, 149, parametric, 1–3, 10, 12, 19, 27, 29, 42, 94–95,
151, 153, 192 195, 209, 224, 246, 257, 343
formal representation, 192 choice, 246, 257
nominal anaphor, 5–6, 16, 38–42, 44–45, variation, 1–3, 10, 12, 19, 27, 29, 195, 224
48–54, 56–59, 62, 67–69, 72, 76, 78, participial, 22–23, 26, 41, 67, 127, 158,
81, 91, 155–156, 254, 341 160–161, 172, 177, 202–203, 212, 246,
nominalization, 37–38, 265, 269, 338, 342, 259–260, 263–268, 271, 277, 279–293,
344, 346–347, 351 311–312, 330, 332, 340, 344, 353
nominalizer, 37, 65, 319 infinitival relatives, 264
nominal reflexive, 15, 48, 53, 68, 254 participial-type, 288, 312
nominative–accusative, 7, 41, 126, 135, part–whole relationship, 35, 147, 153, 189
166–167, 175, 178, 183 single and double dative case marking, 147,
construction, 135, 167, 178, 183 153
nominalized-type, 312 passive, 16, 38, 48, 50, 57–59, 62–63, 68,
non-factive, 195, 210–212, 215, 227–228 72–74, 91, 96, 98–100, 106, 110,
non-finite type, 263–264 128–135, 140–144, 159, 166–167, 171,
non-nominative subjects, 3, 15, 32, 68–69, 71, 179, 281–282, 323, 346–347, 352,
73, 94, 134, 136, 138, 140, 142, 144, 354
General index 367
in Mizo, 129, 352 pronominal clitics, 39, 103–104, 110, 122,

in Thadou, 131–132 125, 132
patient modified in a dative subject Pronominal Strength Hierarchy, 13, 95,
construction, 293 122–123, 126, 133
perceiver of visual and auditory actions – in agreement, 13, 95, 122
dative subject, 146 psychological states and emotions, 146–147
perfect participle, 26, 32, 172, 259, 278–279, purpose marker, 27, 203
281
person harmony, 27, 30, 215–218, 240 quantifiers, 13, 93, 106, 108–109, 114, 132,
Phase Impenetrability Constraint, 195, 325
213–214, 321 question markers, 193
phi-features, 72, 93, 116, 121, 150, 327 quotative, 6, 22, 27, 30, 40, 194–197, 201–204,
phonological evidence from FC in Marathi, 206–207, 209, 214, 216, 218–219,
154 228–229, 241, 244, 327, 342–343, 348
physiological and mental ailments, 146–147 as an identifier, 203
polarity question, 27, 209, 241–244, 325
polymorphemic, 51, 59, 76–77, 81–83, 92, raising infinitives, 124
217 reanalysis of the Initial Complementizer, 195
polysynthesis, 91, 94, 95, 102–103, 108, reciprocals, 43, 45, 51, 59, 76, 89, 306, 320,
111–114, 132–133, 336 341, 352
polysynthetic-like, 95 reduplication, 14, 22, 24–26, 42, 44, 46,
polysynthetic languages, 102–103, 111, 50–52, 54, 89–91, 264, 299, 312, 315,
113–114 335, 349
position of occurrence of the head in IHRCs, 264
of an FC-clause, 195 reduplicated form of the anaphor, 254
of an IC-clause, 195, 219 relative clause, 7, 14–16, 38, 40, 94, 128–133,
possession, 35, 137–140, 147, 149–151, 153, 193, 201, 240, 263–300, 302, 304, 306,
186, 188–189, 192, 298–299 308–313, 317, 326, 329–330, 332, 334,
possessive reflexive, 46, 73, 86, 156, 190 339–344, 348–349, 351–352
possessor in Mizo, 128, 130, 348
agreement, 13, 15, 39–40, 100, 105, 322 types of, 40, 263–265, 267, 269–270, 277
incorporation, 104 relative–correlative clause, 40, 263–264, 266,
modification, 271, 294 268, 270–271, 273–274, 276–277, 288,
raising, 104, 106 294, 311
post-nominal, 263–268, 275, 277, 327 relativizable positions, 265, 271, 276
relative clauses, 263–266, 275, 277 in EHRCs, 265
post-sentential, 14, 193 reported speech, 146, 195
COMP/complementizer, 27, 165, 195, 207 restructuring, 116, 124, 133
potential ambiguity in relative clauses, 264 infinitive, 116
pre-nominal, 263–268 right-adjoined relative clause, 267, 272
correlative structures, 264 right peripheral, 40, 193, 195, 201
preposition incorporation, 103, 110 COMP, 40, 195, 241
pre-sentential, 14, 193 Rightward Extraposition, 195–196, 239–240
COMP, 224
Principles A, B, C, 83 sandhi, 34, 37, 321
Principles and Parameters approach, 5, 8, 257 self-benefactive, 59, 62–63, 67–68
PRO, 12–13, 32, 76–78, 115–118, 133, semantic case, 94
135–136, 154, 157–162, 165, 180, semi-polysynthetic (languages), 102–103, 132
247–249, 253, 257, 259–262, 317, 327, sentential relative clauses, 264, 271
339, 346, 349, 352 sentential-type, 263–264
in dative case-marked position, 260 Small Clauses, 44, 78–79, 81, 123, 125, 177
triggering agreement, 262 sociative, 291, 332
pro, 27–28, 106, 123, 187, 270, 297, 316, specificity, 8, 70–71, 96–98, 171, 174–178,
329–330 188, 192, 325, 346
pro-drop, 27–28, 53, 101, 103, 187, 343 split ergative languages, 135
368 General index
split ergativity in person, 103, 122 Tensed-S Condition, 213–214

[+/−stative] nature of the verb in NNS tense harmony, 30, 195, 215, 217–218, 316
constructions, 170 Thematic Eligibility Condition, 7, 288, 300,
Strict OV Constraint, 264 308, 312
structure, 12, 18–19, 51, 53, 69, 94, 108, 134, thematic relationship, 7, 265, 300, 302,
168, 183, 192, 195, 204–205, 207, 209, 304–308, 311–312, 329
211–212, 221, 240, 246, 310, 314, 340, theta criterion, 112
344–345 time adverbial, 22, 29–30, 274, 286–287,
structural case, 12, 69, 94, 221 291
structural case marker, 12, 69, 94, 221 transitivity, 8, 68, 178, 341–342, 345
structural case marking, 53, 94, 183 transitivizing effect, 94, 309, 334
structural differences between an IC and an triple agreement markers, 99–100
FC, 195, 204–205, 207, 209, 211–212 typology, 1–6, 8,10–11, 17–19, 43, 94–95,
structural nominative case, 192, 246 97, 99, 101, 103, 105, 107, 246, 314,
subcategorization, 6, 156, 312–313 317
subcommand, 13, 43–44, 88, 92, 322 of agreement patterns, 94–95
subject, 3, 7–8, 12–16, 19–20, 27–35, 39–44,
49, 51–55, 57, 59, 61–63, 66, 68–82, UG, see Universal Grammar
85–91, 93–105, 108–111, 115–193, unaccusative, 15, 134, 171–172, 288
195–196, 202, 206–207, 212–217, 222, uninterpretable features of goal and probe,
229–231, 233, 240, 246–256, 258–262, 132
267–268, 271–274, 278–285, 288–289, Universal Grammar (UG), 1–2, 4, 10, 12, 246,
292–297, 304, 306, 308, 310–311, 257, 343–344, 346, 350, 353
316–327, 329–347, 349–351, 353–354
and non-subject properties of the NNS V2
construction, 154 language, 17–18, 40, 199, 272
complementation, 120–121, 193 order, 15, 199, 272, 274
ergative case-marked, 135, 283 position, 9, 20, 199, 314
genitive case-marked, 118, 281, 283 vector, 22–23, 56, 162–164, 177, 199, 315, 318
in by-passive, 140 verbal anaphor, 5–6, 16, 38–39, 41, 43–45,
modification, 267, 279–280, 289, 294 49–50, 52, 56–59, 61–62, 68, 76–77,
of predicates expressing obligation and 80–82, 91–92, 254–255, 306, 317–318,
necessity (desideratives), 147 320, 333
properties of NNSs in SALs, 134, 154, 165 verbal reciprocal, 16, 38, 43, 48–49, 52,
subordinate clause, 15, 26, 30, 199, 206, 263, 55–60, 65, 67, 89, 302–308, 318, 333
272 as a colloborative marker, 57
subordinate relative clause markers, 193 as a group activity marker, 56
subordinating linker with the interpretation of as a hortative marker, 56
then, IC as a, 201 used in the formation of nouns, 58
subordination linker, 194 verbal reflexive, 5–6, 16, 38, 43–45, 48,
suffix, 16, 48, 56–57, 61, 63, 103, 105–106, 52–53, 55, 57–59, 61–68, 75, 77–78,
112, 121–123, 125, 141, 182, 211, 80–81, 89, 91, 141–142, 154, 254–255,
232–233, 290, 305–306, 309–310, 312, 315, 319, 321, 334
319 verbal reflexive / verbal reciprocal – their
superlative, 21, 319 multi-functionality, 58
syntactic changes, 195, 264, 311 argument suppression in activities
syntactic convergence, 1, 11, 41, 126, 336, pertaining to (self’s) body, 62
348, 353 as a detransitivizer, 59
syntactic passive, 74, 140–141, 143 as a locality instigator, 66
syntactic reanalysis in language contact as a marker of passive, 63
situations, 196 as a self-benefactive, 62
as a verbal emphatic, 64
TEC, 7, 300, 308, 312 direct and indirect reflexives, 59
temporary possession, 139, 149, 151 in the formation of special lexical items,
tense and person harmony, 27, 30, 240 65
General index 369
obligatory verbal reflexive with (some) disharmonic, 337

verbs, 66 Greenberg’s universal 22, 315
used in taboo expressions, 66 in relative clauses in Kashmiri, 273
verbal reflexive with adjectives, 64 in SALS, 20, 21, 299
free, 27, 28
wh-relative clause-type, 265 SOV in the V2 language Kashmiri,
wh scope, 196, 223, 239 272
Wide and Narrow Scope, see Narrow and typology, 6, 19, 314
Wide Scope universals, 239, 314
word order unmarked, of Polarity Question marker and
basic, 40, 329 the IC-clause in Hindi-Urdu, 240
in IHRCs, 295
change from VSO to SVO in Khasi, yes/no question marker, 200–201, 209, 212,
222 234–235

(Cambridge Books Online) Subbārāo, Kārumūri V - South Asian Languages - A Syntactic Typology-Cambridge University Press (2012)

Uploaded by

Copyright:

Available Formats

You might also like

(Cambridge Books Online) Subbārāo, Kārumūri V - South Asian Languages - A Syntactic Typology-Cambridge University Press (2012)

Uploaded by

Document Information

Original Title

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

(Cambridge Books Online) Subbārāo, Kārumūri V - South Asian Languages - A Syntactic Typology-Cambridge University Press (2012)

Uploaded by

Copyright:

Available Formats

South Asian Languages

k ā r u m ū r i v . s u b b ā r ā o is Radhakrishnan Chair Professor in Humani-

This publication is in copyright. Subject to statutory exception

First published 2012

Printed in the United Kingdom at the University Press, Cambridge

Library of Congress Cataloguing in Publication data

ISBN 978-0-521-86148-9 Hardback

Additional resources for this publication at www.cambridge.org/subbarao

Cambridge University Press has no responsibility for the persistence or

2 South Asian languages: a preview 18

3 Lexical anaphors and pronouns in South Asian languages 43

3.9 The reflexive in a locative PP 82

4 Case and agreement 93

5 Non-nominative subjects 134

7 Backward Control 246

8 Noun modification: relative clauses 263

(Some relevant material concerning the syntactic typology of SALs is placed

Balthasar Bickel, Gisbert Fanselow, Yogendra Yadava, Udaya Narayana Singh,

neg pple negative participle

1.1 Languages of South Asia

1.2 Aim of the book

1.3 Linguistic theory, language universals and language typology

counterexamples to the theoretical claims made, enabling linguists to revise and

1.4 Inductive and deductive approaches to language analysis

linguistic inquiry, we argued elsewhere that the deductivist is a microinduc-

Limited Induction Base → Deduction → Induction

(See Subbarao and Saxena 1987 for details.)

1.5 Relevance of linguistic theory: some illustrations

That is, the occurrence of a nominal anaphor in a subcategorized locative PP

This can be formulated in terms of the following parameter.3

That is, the occurrence of a comitative PP is underspecified for the occurrence

as specificity and animacy. Thus, with regard to differential object marking

We present further arguments in this work to demonstrate that typology without

1.6 The framework

1.6.1 The Principles and Parameters approach

1.7 The data

1.8 Relevance of the present work

1.10 Contributions of cross-linguistic studies to our understanding

chapter 4) and the non-verb-final structures in Kokborok (Tibeto-Burman

We demonstrate that when an adposition is incorporated into the verb,

Kokborok (TB), a phenomenon unparalleled in any South Asian language,

clause, while in complement constructions there is verb-final order in the

1.11 The binding link for all chapters

2.2 Typology of South Asian languages

2.3 Typological characteristics of SALs

Some Tibeto-Burman languages, such as Tenyidie, Bodo, Hmar, Mising and

2.4 Evidence in support of “South Asia as a linguistic area”

it, “modifies the meaning of the main verb in terms of ‘manner-specification’

2.4.2 Conjunct verbs

In all the Dravidian languages and in many Indo-Aryan languages (Assamese,

Repetition of the verb stem

2.4.4 Echo words

2.4.5 Conjunctive participles

2.4.6 The quotative

2.5 Parametric variation and SALs

2.5.1 Null pronominals

Tibeto-Burman languages such as Manipuri, Rabha, Tangkhul do not exhibit

The absence of the embedded subject is indicated by pro in (20).