Saartje Verbeke

Alignment and Ergativity in New Indo-Aryan Languages

Empirical Approaches
to Language Typology

Editors
Georg Bossong
Bernard Comrie
Kristine Hildebrandt
Yaron Matras

Volume 51
Saartje Verbeke
Alignment
and Ergativity
in New Indo-Aryan
Languages
ISBN 978-3-11-029253-4
e-ISBN 978-3-11-029267-1
ISSN 0933-761X

Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche
Nationalbibliografie; detailed bibliographic data are available in the Internet
at http://dnb.dnb.de.

© 2013 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin
Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen
♾ Printed on acid-free paper
Printed in Germany

www.degruyter.com
Contents
Acknowledgments ix
List of Abbreviations x
Introduction 1

1 Theoretical preliminaries 8
1.1 Ergativity from a functional and typological perspective 8
1.1.1 Ergativity: the standard functional-typological definition 9
1.1.2 Core arguments and grammatical relations 13
1.1.2.1 The subject issue 14
1.1.2.2 Grammatical relations 21
1.1.2.3 Case 28
1.1.2.4 Verb agreement 32
1.1.3 Alignment splits based on referential hierarchies 34
1.1.3.1 Silverstein’s hierarchy 35
1.1.3.2 Differential object marking: the markedness discussion 36
1.1.3.3 Differential subject marking 39
1.1.4 Tense/Aspect/Mood and head- vs. dependent-marking 41
1.1.4.1 Tense/Aspect/Mood split 41
1.1.4.2 Case marking and verb agreement 42
1.2 Direct and indirect motivations of ergativity 44
1.2.1 Ergativity in discourse-functional and cognitive linguistics 46
1.2.1.1 Du Bois: “The discourse base of ergativity” 46
1.2.1.2 Cognitive accounts 48
1.2.2 Historical motivations 54
1.2.3 Case and transitivity 56
1.2.3.1 An alternative view of ergativity 56
1.2.3.2 Transitivity: from Sapir to Hopper and Thompson 60
1.3 Conclusions 63

2 Indo-Aryan 65
2.1 Geographical distribution of the Indo-Aryan languages 65
2.2 Alignment in Hindi 68
2.3 Origin of the ergative pattern in Indo-Aryan 75
2.3.1 Historical overview 75
2.3.1.1 Old Indo-Aryan 75
2.3.1.2 Middle Indo-Aryan 78
2.3.1.3 New Indo-Aryan: Early Hindi 80
2.3.2 Passive reanalysis 80
vi Contents

2.3.3 Ergative remains ergative? 83

2.4 Some key concepts of the Indo-Aryan languages 89
2.4.1 Masica’s layer system 89
2.4.2 The verb system in Indo-Aryan 91
2.4.2.1 The participial base of the verb system in Indo-Aryan 91
2.4.2.2 Light verbs and lexical compound predicates 92
2.4.2.3 Perfect, perfective, and past 94
2.4.2.4 Passives 96
2.4.2.5 Causatives 97
2.4.3 Experiencer constructions 98
2.4.4 Unergatives 108

3 Eastern Indo-Aryan: Asamiya 111

3.1 Asamiya: description 113
3.1.1 Case marking in Asamiya 113
3.1.2 Case marking of the subject 115
3.1.3 Differential object marking and the dual marking of IO 119
3.1.3.1 O-marking 119
3.1.3.2 Dual marking of IO 121
3.1.3.3 Experiencer constructions 124
3.1.4 Verb agreement in Asamiya 126
3.2 The transitive verb 127
3.2.1 The transitive verb in Eastern Indo-Aryan 128
3.2.1.1 The transitive verb in the Bihari languages 131
3.2.1.2 Honorific agreement in Maithili 134
3.2.1.3 Honorific agreement in Magahi 137
3.2.2 The transitive verb in an Eastern Hindi variety 139
3.3 Ergative marking in Eastern Indo-Aryan 142
3.4 Conclusion 144

4 Northern Indo-Aryan: Nepali 146

4.1 Nepali: description 147
4.1.1 Case marking in Nepali 147
4.1.2 Case marking of the core arguments 148
4.1.3 Verb agreement in Nepali 152
4.2 Reanalysis and extension of the postposition le 156
4.2.1 Previous accounts 159
4.2.1.1 Focalization 159
4.2.1.2 Syntactic restrictions 161
 Contents vii

4.2.1.3 Disambiguation hypothesis 161

4.2.1.4 Stage- vs. individual-level predicates 163
4.2.2 Perfective aspect expressed by le 165
4.3 Language contact 170
4.4 Conclusion: alignment in Nepali 173

5 Western Indo-Aryan: Kashmiri 175

5.1 Kashmiri: description 176
5.1.1 Case marking in Kashmiri 176
5.1.2 Verb agreement in Kashmiri 180
5.1.2.1 The verb system 180
5.1.2.2 Imperfective constructions 182
5.1.2.3 Ergative case marking in the perfective 185
5.2 Pronominal suffixes in Western Indo-Aryan 188
5.2.1 Bound pronouns or agreement markers? 188
5.2.2 Pronominal suffixes in Kashmiri 191
5.2.3 Pronominal suffixes in Poguli 199
5.2.4 Pronominal suffixes in Sindhi 201
5.2.5 Pronominal suffixes in Siraiki 204
5.3 Second person importance:
hierarchical relations and Marathi 206
5.4 Conclusion 208

6 Central Indo-Aryan: Rajasthani 212

6.1 Description of Marwari 213
6.1.1 Case marking 213
6.1.2 Verb system and agreement in Marwari 214
6.2 Differential marking 218
6.2.1 Marwari subject marking: a historical perspective 218
6.2.2 Differential marking in Harauti 223
6.2.3 Object marking in Marwari and Gujarati 230
6.2.4 The marking of subjects and experiencers
in Gujarati and Punjabi 234
6.2.5 Differential marking: summary 239
6.3 Central Indo-Aryan and Hindi: variation on a micro-scale 240
6.3.1 Hindi and Braj: O-marking and agreement 240
6.3.2 Bundeli and Bagheli 244
6.3.3 Varieties of Hindi: conclusion 246
6.4 Central Indo-Aryan: Conclusion 246
viii Contents

7 Conclusion 248
7.1 The definition of ergativity: summary 248
7.2 Splits 251
7.2.1 The influence of referential hierarchies 252
7.2.1.1 Referentiality of O 252
7.2.1.2 Referentiality of A 254
7.2.2 Tense/Aspect/Mood splits 256
7.2.3 Classifying the Indo-Aryan languages 259
7.2.4 Case marking versus verb agreement 260
7.3 Core arguments and grammatical relations 261
7.4 The motivation of the ergative pattern in Indo-Aryan 263
7.5 Final observations 266

8 Appendix 268
8.1 Appendix 1 268
8.2 Appendix 2 274

9 References 296

Index of subjects 316

Index of languages 320
Acknowledgments
This book is a revision of my 2011 doctoral dissertation, defended at Ghent Uni-
versity and made possible by a grant from the Research Foundation Flanders
(FWO). I would like to thank my thesis supervisor Klaas Willems and Peter Hook.
I am also thankful for the insightful comments given by Eva De Clercq, Annie
Montaut, Omkar N. Koul, Seppo Kittilä and Yaron Matras.
This book would not have been possible without the help of my native speak-
ers: Rudhira Mitra, Jayeeta Mazumder, Alpha Maheshwari, Sarit Ray, Pratik
Udeshi, Rohit Goyal, Jeanne Vaz, Prof. Ramesh Chandra Sharma, Durgesh,
Khemchand and Suman Sharma, Ria De, Meenaxi Barkati-Ruscheweyh, Dr. Jyoti-
prakash Tamuli, Vishma Raj Thapaiya, Chand Mati Pokharel, Saraswati Chaud-
hary and Satish.
Most importantly, for their love and support, I wish to thank my parents,
friends and family. And my handsome husband, Bram Vincent.
List of Abbreviations
The glosses of the examples follow the “Leipzig glossing rules” (2008). The
transcription of the examples follows the traditional conventions in Indological
studies (I do not provide a phonetic transcription, but an orthographic translit-
eration).

ABL ablative
ABS absolutive
ACC accusative
AOR aorist
AUX auxiliary
CAUS causative
CL class
CONT continuous
CVB converb
DAT dative
DEF definite
DES desiderative
DET determiner
DOM differential object marking
DSM differential subject marking
EDOM extended differential object marking
ERG ergative
EMPH emphasizing particle
F feminine
FUT future
GEN genitive
H honorific
INCL inclusive
INJ injunctive
INS instrumental
INT interrogative
INTS intensivum
IPRF imperfect
ITR intransitive
LNK linker
LOC locative
MH middle honorific
N neuter
 List of Abbreviations xi

NEG negation
NFUT non-future
NH non-honorific
NOM nominative
OBJ objective
OBL oblique
PASS passive
PL plural
PRF perfect
PROG progressive
PST past/perfective
PTCP participle
PURP purposive
Q question marker
QUOT quotation marker
SAP Speech Act Participant
SBJV subjunctive
SG singular
TAM tense/aspect/mood
TR transitive
VOC vocative
WITN witnessed
Introduction
A certain construction in the Indo-Aryan language Hindi-Urdu has long attracted
the attention of linguists, due to its rarity among the languages of the Indo-Euro-
pean family. In this construction, the particle ne is used, which does not appear
in any other construction in the language. The following example may serve as an
illustration (the example is taken from Harris and Sharma 1968: 81):

(1) maiṃ=ne us-e tumhār-ī gand-ī kamiz-eṃ

I=erg (s)he.obl-obj your-f dirty-f shirt[f]-nom.pl
de+dīṃ
give+give.pst.f.pl
‘I gave him/her your dirty shirts back.’

To language learners, the ne-construction is something of a mystery, because the

subject in ex. (1), maiṃ, is followed by ne, and because the transitive verb, dīṃ,
does not agree with the subject of the sentence but agrees with the direct object
tumhārī gandī kamizeṃ. This type of construction occurs only with transitive
verbs in the perfect aspect/past tense in Hindi-Urdu. Imperfective constructions
follow the traditional Indo-European pattern of syntactic constructions; that is,
the verb agrees with the subject, and the subject takes the nominative case.
In older accounts such as Kellogg’s Grammar of the Hindi language (1938:
239–240), the ne-construction is presented as a passive; with dīṃ being a passive
verb, and ne the postposition of the passive agent. However, there is no active
construction that could be regarded as the counterpart of this “passive” in Hindi-
Urdu. Quite on the contrary, if one seeks a corresponding construction in another
voice in the language, then one finds a construction that resembles a periphras-
tic passive of the “passive” construction. Moreover, other Indo-Aryan languages,
such as Punjabi and Gujarati, display the same unusual pattern in their grammar;
however, importantly, pronominal arguments pattern differently in these lan-
guages. Upon closer investigation, we may find a number of additional dissimi-
larities, but the important point is that several Indo-Aryan languages have a con-
struction that is very similar to the ne-construction encountered in Hindi-Urdu.
Recent accounts have indicated that the type of Indo-Aryan construction
exemplified by the Hindi-Urdu ne-construction has the same pattern as the con-
struction that is typically called the “ergative construction” in descriptions of
other language families, particularly Australian and Caucasian languages. Before
long, it was suggested that the Hindi-Urdu ne-construction is also an ergative con-
struction, and the ne particle was labeled an ergative case marker accordingly.
However, this conclusion, which is primarily based on a number of cross-lin-
2 Introduction

guistic similarities, raises more questions than it answers. After all, Hindi-Urdu
is an Indo-European language, whereas the ergative construction has tradition-
ally been associated with “exotic” languages, such as Australian and Caucasian
languages, or language isolates, such as Basque. Moreover, Hindi-Urdu derives
from Sanskrit, which is a highly “accusative” language with at least eight cases,
but none of these cases remotely resembles the ergative case. In other words, the
belief that Hindi-Urdu possesses an ergative construction, requires a very con-
vincing explanation.
Most accounts of ergativity in Hindi-Urdu have traditionally been diachronic.
These accounts focus on the developments from Old Indo-Aryan to New Indo-
Aryan in search of a possible foundation from which the ergative pattern may
have emerged in the course of time. However, the discussion of the historical
origins of the Hindi-Urdu ergative construction encounters considerable difficul-
ties. The analysis and interpretation of the historical material are all but straight-
forward, and there exists the more fundamental problem of finding sufficient
and reliable historical data. Furthermore, the evolution from Old Indo-Aryan to
New Indo-Aryan has sometimes been reduced to a simplified evolution from San-
skrit to Hindi-Urdu. Frequently, evidence of other related languages and dialects
spoken at a certain point of time in the long history of the Indo-Aryan languages
(which covers a period ranging from ca. 1500 BCE to the present) is ignored.
The present study approaches the matter of ergativity in Indo-Aryan lan-
guages from a different angle. This study aims to elucidate the structure and form
of ergative constructions in Indo-Aryan languages from a synchronic point of
view, based on a broad cross-linguistic comparison of a range of different Indo-
Aryan languages. To this end, in addition to Hindi-Urdu, less studied Indo-Aryan
languages are duly taken into account. The research reported in the study is con-
ducted within a typological framework. Thus, the study’s objectives are consist-
ent with the goal of modern typological research on ergativity; that is, to describe,
compare and analyze the “alignment” patterns of languages across the world.
Below, I will briefly explain the concept of “alignment” patterns and then clarify
my stance vis-à-vis typological research.
My use of the term “alignment” refers to the ways in which transitive and
intransitive sentences are constructed and their often complex morphosyntactic
properties. The concept of ergativity has been extensively discussed in the typo-
logical literature. As I noted above, the ergative construction in Hindi-Urdu has
primarily been identified on the basis of cross-linguistic comparison. However,
I believe that a full-fledged typological framework must also pay due attention
to less frequent or even exceptional patterns in different languages. The present
study aims to contribute to this evolving line of research by comparing the align-
ment patterns in a number of modern Indo-Aryan languages in a comprehensive
 Introduction 3

manner. This study focuses on the distribution of ergative and not-so-ergative

features occurring in the Indo-Aryan languages currently spoken on the South
Asian subcontinent. It should be noted from the outset that the focus of the study
is not restricted to Hindi-Urdu. Undoubtedly, most of the currently available lit-
erature on ergativity in Indo-Aryan languages is strongly directed at Hindi-Urdu.
However, an important secondary aim of this study is to question the central posi-
tion of Hindi-Urdu in the research on ergativity in Indo-Aryan languages. Other
languages (including varieties) are often equally as relevant as Hindi-Urdu, and
occasionally even more revealing in view of a comprehensive study of ergativity
in the Indo-Aryan languages.
The present study is entirely devoted to comparing different languages within
a language family. It is emphatically not a cross-linguistic study that discusses a
representative sample of the languages of the world. The “sample” of languages
that will be analyzed in the subsequent chapters is explicitly biased because it
is derived from a single language family. Languages that are spoken in the area
on which the study focuses but that belong to other language families, such as a
number of Tibeto-Burman and Iranian languages (Iranian of course being more
closely related to Indo-Aryan), are briefly taken into account when this is con-
sidered relevant, but they are not discussed extensively. This qualification is
important in view of the scope and typological aspirations of the present study. It
should be borne in mind that any statement regarding the linguistic phenomena
discussed in the study only applies to this phenomenon as it is attested in the lan-
guage family of Indo-Aryan languages, and does not apply beyond the confines
imposed by this focus. This study does not intend to derive language “universals”
from the observations pertaining to the Indo-Aryan languages to be discussed.
Nevertheless, a study of one language family is far from irrelevant. Compared
with a typological study based on a large number of sample languages from
across the world, a contrastive comparison of the languages of one (admittedly
diverse) language family may have several advantages. For instance, this type of
study avoids the vexing problem of selecting an adequate sample of languages. A
sample should be representative of all languages of the world, but it is common
knowledge even among typologists that samples are often biased in a particular
direction simply because the languages of a particular family, a specific area or
a particular sociolinguistic layer are more extensively described than others, and
thus more likely to be included in the sample. Another complicating factor is that
many historical situations of linguistic contact and genetic relations between lan-
guages are not yet fully understood and described. Furthermore, by extracting one
particular construction from a sample of languages, one derives abstraction from
the language systems (or “grammars”) in which different constructions show spe-
cific relations and contrasts with one another. Therefore, typological research is
4 Introduction

constantly at risk of tearing apart what language systems join together. In short,
it is extremely difficult to compose a representative sample that simultaneously
includes a satisfactory quantity of the variation to be examined and depicts the
individual languages with sufficient accuracy to prevent false generalizations. By
focusing on the languages within a single language family, we are able to care-
fully observe the limits and restrictions of the sample. However, the choice of one
language family allows us to study each language of the sample in detail, and it is
also an invitation to pay close attention to the systematic relationships between
different constructions within the particular languages and the way they inter-
act and influence each other. This level of research extends beyond the simple
comparison of one specific construction in different languages. Because the
Indo-Aryan languages have the same historical ancestor, it is possible to define
a number of concepts that are valid for all the languages to be analyzed in the
present study; however, without losing sight of some of the fundamental differ-
ences between them.
The material used in this study primarily consists of textbooks, a corpus of
texts that have been sampled particularly for this purpose, and various existent
grammars. The choice of written sources rather than spoken languages is deliber-
ate and is based on the conviction that one should take advantage of such sources
whenever the opportunity presents itself. Furthermore, textual evidence has the
not inconsiderable advantage of providing a clear and, importantly, fairly con-
sistent picture of the grammars of particular languages (or varieties). This advan-
tage is especially relevant to the present study, which is based on a comparison
of a considerable number of different Indo-Aryan languages (and varieties). The
choice of written, often literary, material is arguably not entirely uncontroversial,
and relying on printed texts has unquestionably certain drawbacks, particularly
with respect to the questions of how many texts one must consult and how many
examples one must find in the corpus before one can claim to offer a representa-
tive picture of a language, variety, or construction. However, when these types
of problems occurred, I was able to obtain the knowledge and judgments of at
least two native speakers for every main language analyzed in the study. These
informants were consulted in all cases of doubt, particularly when the textual
evidence was insufficient to produce a convincing interpretation of certain exam-
ples. Native speakers were also asked to formulate their opinions about the gram-
maticality of certain variations in constructions that did not regularly occur in the
texts of our corpus.
The present study is organized as follows. In Chapter 1, I address the theoreti-
cal preliminaries of ergativity from a typological perspective. The basic assump-
tions expounded in this Chapter are derived from cross-linguistic comparative
studies, and some of these assumptions are claimed to be universally valid.
 Introduction 5

Therefore, any reference to Indo-Aryan languages is avoided when possible in

this chapter, in order to ensure an unprejudiced view of the general literature
on ergativity. The research questions to be addressed in the subsequent chapters
are formulated on the basis of this first chapter. Broadly speaking, the questions
guiding Chapters 2 through 6 are the following:
– What are the syntactic, morphological, and semantic properties of case
marking and verb agreement in the Indo-Aryan languages spoken on the
South Asian subcontinent? To provide a balanced overview of these proper-
ties across the Indo-Aryan languages, Chapters 3 through 6 each focus on
the languages of one particular region, viz. Eastern Indo-Aryan, Northern
Indo-Aryan, Western Indo-Aryan, and Central Indo-Aryan, respectively.
Chapter 2 is conceived as an introductory Chapter to Indo-Aryan in general
and contains illustrations from Hindi-Urdu.
– To what extent do the properties of case marking and verb agreement that
we encounter in Indo-Aryan languages conform to the generally accepted
typological definition of ergativity, and in what respects can the proper-
ties of case marking and verb agreement attested in Indo-Aryan languages
modify the general typological view of ergativity and alignment? These
questions further develop the theoretical background provided in Chapter 1
and relate the concept of ergativity to the empirical attestation of ergative
constructions in the Indo-Aryan languages.
– What are the alignment types that are shared by all Indo-Aryan languages,
and in what respects do these alignment types vary in our sample lan-
guages? These questions draw on the findings based on the previous ques-
tions and allow us to make general observations regarding alignment and
ergativity in Indo-Aryan.
– A final (admittedly more speculative) question concerns the conceptual
status of ergativity. If ergativity occurs in Indo-Aryan, does this occurrence
imply that the Indo-Aryan speakers have a different conceptualization of
the state of affairs expressed in their language as compared with speakers
of languages in which ergativity is absent? Although I will not tackle this
question head on in the present study, this issue will run as an undercurrent
through several chapters.

Chapters 2 through 6 each contribute to obtaining a comprehensive answer to the

questions listed above. Chapter 2 is a transitional Chapter between the theoreti-
cal Chapter 1 and the empirical Chapters 3 through 6. The Chapter begins with
a brief geographical overview of the spread of the Indo-Aryan languages across
the South Asian subcontinent. Additionally, relevant findings of earlier scholarly
research on Hindi-Urdu are presented, particularly with respect to the diachronic
6 Introduction

developments in Indo-Aryan that are assumed to have led to the emergence of

the ergative construction. Subsequently, various key concepts that are important
in the grammars of the Indo-Aryan languages are illustrated by a large number
of examples that are predominantly drawn from Hindi-Urdu. These “key con-
cepts”, such as postpositional case marking, the formation of the tenses, unerga-
tive verbs and converbial constructions, refer to properties that are recurrent in
the grammars of Indo-Aryan languages. These concepts are introduced in this
chapter, but their language-particular instantiations are discussed in detail in the
ensuing chapters that examine the Indo-Aryan languages of particular regions.
The chapters following the theoretical Chapter 1 and the presentation of
the Indo-Aryan languages and some of their basic properties in Chapter 2 each
focus on languages and varieties in one specific area in the Indo-Aryan region.
In Chapter 3, I explore Eastern Indo-Aryan, particularly the language Asamiya.
In contrast with Hindi-Urdu, modern Eastern Indo-Aryan is assumed to possess
no ergative constructions, although ergative constructions are attested in early
sources. However, I defend the view that modern Asamiya shows clear evidence
of ergative case marking. The second issue discussed in this Chapter is the dif-
ference in conjugation between the transitive and intransitive verbs in Asamiya.
By comparing Asamiya with the Eastern Indo-Aryan Bihari languages, I show
that this difference in conjugation can be traced to a particular type of object
agreement.
Chapter 4 examines the North Indo-Aryan languages, particularly Nepali.
Nepali is an unusual language among the Indo-Aryan languages. Associated with
Nepal and substantially influenced by – and, in turn, influencing – the Tibeto-
Burman languages spoken in this country, Nepali is usually only briefly men-
tioned in older comparative grammars of Indo-Aryan. In typological linguistics,
Nepali is never considered in its identity of a branch of Indo-Aryan but is simply
considered as a cross-linguistically interesting language with respect to its align-
ment pattern. I will adduce evidence for the view that the fascinating alignment
pattern found in Nepali occupies an intermediary position between Eastern Indo-
Aryan and Central Indo-Aryan, by simultaneously showing ergative case marking
and subject agreement.
After Nepali, Chapter 5 proceeds to examine Western Indo-Aryan, particu-
larly the Northwestern Indo-Aryan languages, which are represented in this study
by Kashmiri. Kashmiri and the Northwestern Indo-Aryan languages are interest-
ing because they display “pronominal suffixes” on the verbs, which indicate the
person of core arguments. In this respect, the alignment in Kashmiri is very dif-
ferent from the alignment in Nepali. However, both languages have in common
that they are spoken in mountainous regions in which there is frequent language
contact with Tibeto-Burman languages. Indeed, if we look beyond the differences
 Introduction 7

in formal marking of the arguments and the different types of verb agreement,
there is evidence of influences due to language contact in Western and Northern
Indo-Aryan. These influences are particularly evident in the case marking pattern
and occur in different degrees.
In Chapter 6, I discuss Central Indo-Aryan. Although Hindi-Urdu is arguably
the best known and most important of all the Central Indo-Aryan languages, I
choose to treat Hindi-Urdu only as a frame of reference in this chapter. Chapter 5
focuses on the Rajasthani languages and, albeit to a lesser extent, a number of
varieties of Hindi-Urdu. The Rajasthani languages provide an extremely complex
picture of alignment in Indo-Aryan. The variation in alignment patterns, the
local preferences for one or another case marker, and the dialectal divergences in
general prove to be more complex in the Rajasthani languages than in any other
Indo-Aryan language.
In Chapter 5, a substantial part of the discussion is devoted to agreement
features, particularly the phenomenon of pronominal suffixes in Western Indo-
Aryan. In contrast, Chapter 6, on Central Indo-Aryan, focuses on a number of
fundamental issues concerning the nature of case marking in Indo-Aryan lan-
guages. Central Indo-Aryan differentiates between postpositional case marking
and inflectional case marking. However, postpositions and inflectional case
morphemes also have different functions on the semantic and the syntactic
level. These functional differences, which run parallel to the formal differences,
have not yet been given the attention they deserve. In addition, the relationship
between verb agreement and postpositional/inflectional case marking must be
specified.
Chapter 7 concludes the study and summarizes the evidence for ergativity
in the different Indo-Aryan languages. This Chapter retreats from the areal focus
that guided the previous chapters and focuses on structural similarities between
the Indo-Aryan languages.
1 Theoretical preliminaries

1.1 Ergativity from a functional and typological perspective

Contemporary views on ergativity in functional typology depart from two articles

written in the 1970s by Comrie (1978) and Dixon (1979). These authors described
ergative features in different languages and analyzed the ergative pattern as
a system that exists for reasons beyond those of other argument structuring
systems. As a result of these thorough analyses of the phenomenon, ergativity
has been discussed primarily in the functional-typological literature.¹
Comrie (1978) and Dixon (1979) demonstrated that non-ergative constructions
occur in every language that has been termed an “ergative language”. Therefore,
no language is entirely ergative. Deviating constructions often display the accu-
sative pattern, but other constructions have also been reported. Evidence from
the description of an increasing number of languages demonstrates the extent
of the variation. Since the 1990s, linguistic typology has more and more become
a distinct branch of linguistics in its own right (cf. Song 2011). As a result of the
ever increasing body of descriptive data, every linguistic category that had been
quite generally accepted as universally valid has been questioned. For instance,
the traditional category of “subject” became problematic in view of ergativity,
because the ergative pattern displays a different morphosyntactic coding of the
transitive “subject” compared with the accusative pattern. The definitions of
categories were subsequently broadened, and categories began to be defined as
either cross-linguistic prototypes or clusters of features. Dryer (1997) has main-
tained that the effort to obtain a universal definition of “subject” is beside the
point as long as “subject” remains a heuristically useful category that is defined
in terms of one language. In addition to linguistic categories, Greenbergian “uni-
versals” have been questioned (Greenberg 1966). In contemporary linguistic
typology, every concept has been subjected to scrutiny, and among all of Green-
berg’s famous language universals only one or two universals remain undisputed
(cf. Haspelmath 2010).

1 In addition to functional-typological comparisons and descriptions, ergativity has been

treated in functionally inspired theoretical frameworks, such as Relational Grammar (Perlmutter
1983), Functional Grammar (Dik 1978, 1980) and later Role and Reference Grammar (Foley and
Van Valin 1980, 1984). For a critical overview of the conception of ergativity in these frame-
works, see Dixon (1994: 232–236).
 Ergativity from a functional and typological perspective 9

In the following Sections 1.1.1–1.1.4, the functional-typological perspective of erg-

ativity will be presented. Section 1.1.1 outlines the standard functional-typologi-
cal definition of ergativity. The consequences of this definition for grammatical
relations, and particularly for the “subject”, are discussed in Section 1.1.2. Split
patterns and their motivations are examined in Sections 1.1.3 and 1.1.4.

1.1.1 Ergativity: the standard functional-typological definition

The articles by Comrie (1978) and Dixon (1979) are considered fundamental in
establishing the definition of ergativity. The two articles appeared during a period
of general interest in universal patterns in syntax. For instance, a volume edited
by Plank (1979) discussing ergativity was published around the same period
(also, Tchekhoff 1972; Li 1976, 1977). Plank (1979: 4) used the term “alignment”
to indicate the argument structure of a language, i.e. the way the arguments are
coded.² Both Comrie and Dixon preferred the terms “system” or “pattern”, but
“alignment” became a common term in later works on ergativity and argument
structure in general (e.g., Harris 1990, Comrie 2008, Wichmann 2008, Bickel and
Nichols 2009: 305).
According to Comrie (1978: 329), ergativity is “a system of nominal case
marking where the subject of an intransitive verb has the same morphological
marker as a direct object, and a different morphological marker from the subject
of a transitive verb”. Comrie distinguishes five possible types of systems in the
languages of the world. To define these systems, he uses the terms S, A and P.
In Comrie (1978), the definitions of these terms are quite rudimentary, and are
primarily based on the subject/object distinction from the accusative system. S is

2 “Argument coding” is a general term that refers to the various ways an argument can be
identified in an argument structure; such ways may include the argument’s case, postposition,
or cross-reference on a verb. In this study, “marking” will be used only to refer to the formal
marking on the argument itself; the term “verb agreement” will be used for cross-referencing
on a verb. The term “marked” refers here exclusively to formal markedness (syntagmatic com-
plexity, Moracvcsik and Wirth 1980); for example, an unmarked argument refers to an argument
with a nominative (zero) case marking, as opposed, for instance, to an accusative-marked argu-
ment. Whenever “marked” is used differently in other theoretical approaches discussed in the
present study, the meaning of the term will be specified. I am aware that formal markedness
often coincides with “semantic markedness”; for instance, a formally unmarked term occurs
more frequently and is semantically more “expected”. However, to allow for exceptions to this
correlation (cf. Haspelmath 2006) and to avoid any confusion, “markedness” is restricted to
form in this study.
10 Theoretical preliminaries

“the single argument of an intransitive verb” and is “clearly reminiscent of the

word subject”. A is “the argument of a transitive verb which would be its subject
in a non-ergative language like English”, and “P refers to the argument that
would be the direct object” (Dixon 1979 uses O rather than P). Comrie (1989: 125)
describes the five possible systems of a language as follows:

The nominative-accusative system groups S and A (nominative) together against P (accusa-

tive). The ergative-absolutive system groups S and P (absolutive) together against A (erga-
tive). Both of these systems are widespread across the languages of the world. The neutral
system would have the same form for all three primitives, but since this is tantamount to
lack of case marking for these relations, it is not directly relevant to our considerations: as
a system, it is, of course, widespread in the languages of the world, but most languages
with this system have other means, such as verb agreement or word order, to indicate which
noun phrase is A and which is P in the transitive construction. The fourth possible type,
tripartite, would have distinct cases for each of the three primitives. The fifth type would
group A and P together as against S.

He considers the fourth and fifth types to be extremely rare; these types are either
absent from the languages of the world or occur as exceptional patterns in lan-
guages with otherwise ergative or accusative systems.³ In defining the different
patterns, Comrie focuses on the “nominal” morphological marking of the argu-
ments. The case markings that are relevant to determining the type of system are
the nominative and accusative cases for the accusative system, and the ergative
and absolutive cases for the ergative system. Of these pairs, the accusative and
ergative are typically the formally marked cases. Therefore, it is customary to
refer to an “accusative system” rather than a “nominative-accusative system” and
to an “ergative system” rather than an “ergative-absolutive system”. The nomi-
native and absolutive are normally the unmarked cases. Although the distinc-
tion between the terms “nominative” and “absolutive” may be convenient to
emphasize the differences between the two systems, the nominative and abso-
lutive are not necessarily formally different in so-called split alignment systems
(cf. Section 1.1.2.3). As a rule, nominal marking constitutes the grounds for the
division in different alignment systems. Comrie also mentions verb agreement
as a parameter that can be used to distinguish between the five possible systems
(Comrie 1978: 338, cf. also Mallinson and Blake 1981: 55–58). However, Comrie
seems to assign secondary importance to verb agreement. He points out that verb

3 For example, certain Iranian languages display some constructions that group A and P vs. S
as a stage in the evolution from a predominantly ergative system to a predominantly accusative
system (Comrie 1989: 125), and the Australian language Ritharngu displays the tripartite pat-
tern but only with arguments referring to humans and intelligent animals (Comrie 1989: 131).
 Ergativity from a functional and typological perspective 11

agreement may determine neutral case marking systems “as another mechanism
besides case marking”. Thus, whereas case marking is of primary importance in
determining different alignment types, verb agreement has a role inasmuch as it
modifies the alignment pattern that is determined by the case marking.
The next fundamental study of ergativity was conducted by Dixon (1979).
Dixon begins by extending the concept of alignment-determining coding from
case marking to pre- and postpositional marking, verbal coding in the form of
cross-referencing of arguments on the verb (agreement), and eventually word
order. For instance, in an ergative word order pattern, the O-argument is consid-
ered to be placed in the same position with regard to the verb as the S-argument,
whereas the A-argument is considered to be placed in a different position. Word
order is considered the least efficient strategy for determining the alignment type
because one can only compare the word order pattern of an intransitive sen-
tence with that of a transitive sentence when the verb is in a medial position. For
example, one could conjecture that a language in which O precedes the verb and
A follows it (OVA) is ergative if S precedes the verb (SV). Conversely, if S follows the
verb (VS) in a verb-medial language, the OVA construction would be accusative.
However, in verb-initial and verb-final languages, it is impossible to determine
the alignment pattern on the same basis. In languages with verb-initial and verb-
final constituent orders, the order of the arguments may vary but not with respect
to the verb position. Therefore, one cannot compare the position of one of the two
arguments in a transitive construction with the position of a single argument with
respect to the verb in an intransitive construction (all arguments are positioned
either before the verb in verb-final constructions or after the verb in verb-initial
constructions). Because only verb-medial languages can distinguish ergative pat-
terns from accusative patterns through word order and because it is generally
acknowledged that word order may be determined by a host of different factors,
word order is commonly considered an unreliable factor in establishing align-
ment patterns (Dixon 1979: 65–68, Givón 1995: 255–256). In some recent accounts,
inflectional cases and pre- and postpositions generally have been united under
the single heading of “case marking” in contradistinction to “verb agreement”
(see, e.g., Nichols 1986, Blake 2001).
Dixon (1979: 60) defines ergativity as follows: “a language is said to show
ergative characteristics if intransitive subject is treated in the same manner
as transitive object, and differently from transitive subject”. The “treatment”
includes every possible type of argument coding, and in this sense, Dixon’s defi-
nition is more general than the definition provided by Comrie’s, who focused on
nominal marking.
Although both Comrie and Dixon emphasized that different patterns may occur
in one language, they also suggested that each language has a “predominant”
12 Theoretical preliminaries

system, upon which variations may occur, yielding so-called “split systems”.⁴ In
practice, any alignment is subject to various conditions and occurs only in cir-
cumstances that are often not easily defined (cf. Creissels 2008, Section 1.1.3 and
1.1.4). Although no language in which all constructions are ergative has yet been
discovered, the term “ergative language” frequently occurs in the literature (e.g.,
Dik 1978; Cooreman et al. 1984: 24; Farrell 2005) and surprisingly in Dixon (1994)
(cf. Harris 1997: 362). Such a language is considered to have a dominant erga-
tive pattern, but it is difficult to determine exactly how many constructions in a
language must be ergative to consider the language ergative. Though Dixon and
Comrie have acknowledged that ergative alignment always occurs in combina-
tion with other alignment patterns, their objective is to find properties that might
elevate ergativity above the status of “a superficial morphological phenomenon”
(Comrie 1978: 334). The formal coding of ergative features is regularly denoted as
“morphological ergativity”. Dixon’s description of syntactic ergativity (1979: 64,
1994: 143, Section 1.1.2.1) is one such attempt to establish that ergativity is more
than simply a morphological phenomenon. However, morphological ergativity is
undoubtedly more common than syntactic ergativity (Comrie 1978: 350).
A number of linguists have considered the idea of sorting languages accord-
ing to alignment types. For instance, the division of all languages into discrete
types of alignment is an explicit aim of the typological branch of Russian lin-
guists; indeed, Russian linguists contributed extensively to early ergative studies
(cf. Bokarev 1950; Mescaninov 1967; Kibrik 1979, 1997 among others). Klimov
(1973: 5, 1979; Nichols 1990) is the founder of the “content-oriented typologi-
cal framework”, also known as “contentive typology”, in which three language
systems/types are distinguished: the ergative type, the active type, and the nomi-
native type. These types are diachronically related to each other; in particular,
a language can change from the accusative type through the active type to the
ergative type. The distinction between language types is based on syntactic fea-
tures relating to verb agreement and case marking, but also on factors relating to
“content”. In this context, “content” refers to various other grammatical features
with semantic bases. For instance, Klimov (1973: 5) associates languages with an
“active” alignment with the distinction between agent and patient in argument
coding, yet also with grammatical distinctions between constructions of alien-
able and inalienable possession, the distinction between inclusive and exclusive
first person pronouns, and the formal distinction between animate and inanimate

4 Note that according to Comrie and Dixon, “system” means a method of argument coding
rather than referring to the Saussurian “language system”. However, in this respect, “align-
ment” is arguably a more suitable term than “system”.
 Ergativity from a functional and typological perspective 13

nouns etc. (Nichols 1990: 95). In contrast, he argues that ergative systems are
based on “the opposition of so-called agentive and factitive functions” (Klimov
1979: 331; Kibrik 1979); the agentive function is exemplified by the ergative case
and possessive verbs, amongst other structural features, and the factitive is
exemplified by the absolutive case and affective verbs, etc. (Tournadre 1996: 20).
Nichols (1990) claims that her own typology, which associates a coding pattern
with certain alignment types, is more applicable to the languages of the world
than Klimov’s essentially content-based typology. Nichols distinguishes head-
marking (marking on the head of a clause, e.g., verb agreement) from dependent-
marking (marking on the dependent element in a clause, e.g., case marking). She
argues that head-marking often occurs in combination with ergative or accusa-
tive alignment; this argument is not surprising considering Comrie’s definition,
in which the identification of ergative and accusative alignment patterns is pri-
marily dependent on the overt case markings of the arguments. Nichols further
argues that Klimov’s association between certain “content” features (e.g., posses-
sive constructions, grammatical gender etc.) and ergative/accusative alignment
should be attributed to a tendency of these features to occur in either a head-
marking or dependent-marking language rather than to the ergativity or accusa-
tivity of the language.
In contemporary linguistics, the content-based typology has largely been
abandoned, at least in the form proposed by Klimov (cf. Wichmann 2008: 6).⁵
However, Comrie’s and Dixon’s definitions have been adopted in modern func-
tional typology, and their definitions are still widely accepted as standard defini-
tions of ergativity (cf. Primus 2011: 304).

1.1.2 Core arguments and grammatical relations

Core arguments are the arguments in a clause that stand in a syntactic and seman-
tic relationship with the verb, and the number and nature of the arguments are
determined by the valency of the verb (cf. Van Valin and LaPolla 1997: 26–30). In
the traditional linguistics of Standard European languages, the two core argu-
ments of a transitive verb have been labeled as “subject” and “object” on the
basis of their morphosyntactic behavior. The ergative pattern jeopardizes this tra-
ditional labeling because the ergative coding properties of the core arguments do
not allow one to identify the subject in a transitive clause with the subject of an

5 Even in current Russian linguistics, the tradition of content-based typology has not been
further pursued (Yakov Testelets, p.c.).
14 Theoretical preliminaries

intransitive clause. Given that it is impossible to identify the subject and object
cross-linguistically on the basis of identical formal coding properties in different
alignment patterns, difficulties arise because the categories of subject and object
are no longer efficient for comparing argument structures across different lan-
guages. The subject of a construction in one language has coding properties that
differ from the subject of a construction in another language. Furthermore, even
in a single language, “subject” can become a problematic category if this particu-
lar language possesses more than one type of alignment pattern. Therefore, other
criteria beyond the formal properties must be employed in order to identify the
subject and the object. For instance, the identical syntactic behaviors of certain
arguments or their indistinguishable semantic roles are also instrumental in defin-
ing the subject cross-linguistically. The discussion regarding the “subject” in lan-
guages primarily pertains to the question of how morphological, syntactic and
semantic conditions treat arguments in terms of their alignment with each other.

1.1.2.1 The subject issue

In the traditional account of core arguments, the subject is distinguished from
the (direct) object (Farrell 2005: 4–6) in a straightforward grammatical way, as
illustrated by the following German sentences:

(2) Der Mann läuft (‘The man walks’)

(3) Der Mann sieht den Jungen (‘The man sees the boy’).

Formally, the verb forms läuf-t and sieh-t agree with der Mann. Der Mann is in the
unmarked nominative case in German. Due to these morphosyntactic factors of
case marking and agreement, der Mann is traditionally called the “subject” of the
clause. Conversely, the accusative case is considered to be the case of the “(direct)
object”, and in the German example we find the accusative de-n Junge-n (nomina-
tive: der Junge). The subject and the direct object are the “grammatical relations”
that are distinguished in most Standard European languages. The subject, direct
object and verb constitute the building blocks of a sentence. In languages such
as German and English, the arguments are also ordered in a well-defined way; as
a rule, this ordering also contributes to their identification (e.g., subject before
direct object, albeit with a considerable amount of variation across and within
languages).
In a language with an ergative pattern, the ostensible subject of a transitive
verb does not have the same morphosyntactic properties as the sole argument of
an intransitive clause, nor does it control the verb agreement. In the following
 Ergativity from a functional and typological perspective 15

examples from Tsez (Polinsky and Comrie 1999: 112), there is agreement between
bikori ‘snake’ and the verb b-exu-s; the morpheme b- indicates the specific noun
class of the argument (viz., class 3):

(4) bikori b-exu-s

snake.cl3 cl3-die-pst.witn
‘The snake died.’

(5) už-ā bikori b-exu-r-si

boy-erg snake.cl3 cl3-die-caus-pst.witn
‘The boy killed the snake.’

If the morphosyntactic features illustrated above by the German examples (2)

and (3) are applied to examples (4) and (5), then bikori is the “subject” in both
sentences, whereas už-ā ‘boy’ is the direct object in example (4). Several older
accounts have indeed identified the ergative argument in constructions such as
(4) as the direct object and the nominative argument as the subject (e.g., Tchek-
hoff 1978). However, the translations of ex. (4) and (5) suggest another analysis.
It seems plausible to identify the subject ‘the boy’ in the English translation with
the Tsez so-called direct object užā and the direct object ‘the snake’ with the sup-
posed subject bikori. It would seem counterintuitive to consider den Jungen in
ex. (3) as the direct object but bikori in ex. (5) as the subject. This discordance
partially explains why most scholars have abandoned the idea that the ergative
argument is the direct object. Identifying the subject and direct object in an erga-
tive construction on the sole basis of morphosyntactic features is not a cross-lin-
guistically valid method.
In addition to the above-mentioned morphosyntactic properties, purely syn-
tactic factors can be used to identify core arguments (Donohue and Wichmann
2008). For instance, Dixon (1979: 113–115) observes that the subjects of transi-
tive verbs (A) and intransitive verbs (S) in many languages behave identically in
certain constructions, whether or not they display the same morphosyntactic fea-
tures of verb agreement and nominative case marking. He argues, for instance,
that A and S are both potential addressees of imperative and jussive construc-
tions. For instance, compare the construction, I ordered him to go, in which go is
intransitive, with the construction I ordered him to bring the water, in which the
verb bring is transitive. Both the S of go and the A of bring are coreferential with
the O of the main clause. According to Dixon (1979: 114), this observation is uni-
versally true in every language, regardless of each language’s alignment pattern.
Similarly, in coreferential NP’s in constructions with main verbs meaning ‘can’,
‘begin’, ‘finish’ etc. – e.g., I began to laugh, I began to paint the wall – Dixon notes
16 Theoretical preliminaries

that the S of laugh and the A of paint are treated identically because each can be
raised to be arguments of the main verb (but see Bhat 1991: 101).
The above constructions illustrate that S and A are syntactically treated the
same way, regardless of the dominating alignment pattern. This finding suggests
that there is a syntactic basis for identifying a cross-linguistically valid category of
subject in almost every language, irregarding of whether the languages’ prevail-
ing morphological patterns are ergative or accusative. The linking of O and S on
a syntactic level is much rarer. Morphological ergativity might still be explained
as a “surface” phenomenon that is caused, for instance, by purely formal rules
along the lines of “agreement with the closest unmarked argument” (cf. Saxena
1984; Anderson 1976). However, if syntactic ergativity were common, this ten-
dency would strongly indicate that the ergative grouping of S and O profoundly
influences every aspect of a language. It would also corroborate the claim that
there are two inherently different language systems, i.e., ergative and accusative
languages, which have nothing in common as far as case marking and alignment
is concerned.
Syntactic ergativity is especially problematic in defending the subject cat-
egory because languages with syntactically ergative patterns show that S and
A pattern differently in both morphology and syntax. The Australian language
Dyirbal is one of the few languages in the world that display syntactic ergativ-
ity. Ex. (6) (Dixon 1995; Bickel 2011: 423) is an example of conjunction reduction
with a control verb. The O-argument of the verb that means ‘to see’ – simultane-
ously with the S of the verb that means ‘to want’ – is omitted. Note that walŋgarra
is intransitive in Dyirbal. Ex. (7) (Dixon 1994) is an example of coordination in
which the O of the coordinated verb that means ‘to see’ is the same as the S-argu-
ment of the first verb meaning ‘return’. Thus, S and O are syntactically treated the
same way in both examples. In sum, syntactic ergativity transgresses the merely
formal level of coding properties and, according to some authors (Dixon 1979;
Anderson 1976), pertains to the “deeper” level of syntactic structures rather than
the more “shallow” level of morphological structures.

(6) bayi yara walŋgarra-nyu baŋgun

det.m.sg man want-nfut det.f.sg.erg
yibi-ŋgu bura-li
woman-erg see-purp
‘The man wanted the woman to see him.’

(7) ŋuma banaga-nyu yabu-ŋgu bura-n

father return-nfut mother-erg see-nfut
‘Father returned and mother saw him.’
 Ergativity from a functional and typological perspective 17

Dixon (1979) proposes to use the concept of “pivot” to refer to the grouping of
arguments with identical syntactic behavior. In syntactically ergative construc-
tions, the pivot is {S, O}. In ex. (7), the pivot is ŋuma, ‘father’, which is the S of
the verb banaga-nyu ‘return’ and the O of the verb bura-n ‘see’. In accusative
constructions and morphologically ergative constructions, the pivot is {S, A}. For
instance, in the English sentence Father returned and saw mother, the NP father is
the S of return and the A of saw. Thus, the ‘pivot’ is essentially a construction- and
language-particular category; for example, in Dyirbal, {S, O} is the pivot, and in
English the pivot is {S, A}.
Syntactic ergativity is, to say the least, rare in the languages of the world. Even
in Dyirbal, it is not the only pattern attested. According to Heath (1979, 1980) syn-
tactically ergative constructions do not often occur in Dyirbal and most of these
constructions can be explained by other mechanisms such as the “indefinite-NP
deletion” rule (Heath 1979: 415). In his view, all indefinite NP’s can be deleted in
Dyirbal, irrespective of whether they are A, S, or O. Dixon’s interpretation that S
and O are treated in the same way in coordination patterns is nothing but a mis-
understanding of this “indefinite-NP deletion” rule. Heath further argues that the
apparent syntactic ergative pattern is in fact caused by mechanisms that primar-
ily function to distinguish nominal from pronominal coreferential arguments. In
the following examples from Dyirbal, according to Heath, there is no {S, O} pivot
but only coreferential deletion of A. In the first example, the “deleted” A of the
coordinated clause is coreferential with the A baŋgul+yaṛaŋgu of the preceding
sentence. In the second example, the “deleted” A of the verb buŋal-ma-n is coref-
erential with the S bayi (Heath 1979: 428–429, examples from Dixon 1972: 134, 376;
but with Heath’s glossing):

(8) bala+yugu baŋgul+yaṛaŋgu nudi-n bayi+ñalŋga bunju-n

the+stick the+man cut-pst the+child spank-pst
‘The man cut the tree and (the man) spanked the child.’

(9) añja bayi bili-ñu midi+ban bilŋgiŋga buŋal-ma-n

as for he climb-pst small+the on thighs go down-ins-pst
‘He climbed up and brought the small one down across his thighs.’

Not surprisingly, Heath’s (1979) account has been strongly criticized by Dixon
(1979b, later also by Thomsen 1994). It should be borne in mind, however, that
the aim of Heath (1979) is not to show that Dyirbal is an “accusative language”;
his aim is to demonstrate that the evidence in support of the view that there is
syntactic ergativity in Dyirbal is scant, and that other functional motivations
for the patterns in which S and O are treated alike must not be ignored. Dixon
18 Theoretical preliminaries

(1979b) reiterates his own thesis that there is syntactic ergativity in Dyirbal and
that Heath misinterprets the evidence (for instance, ex. (9) is not a coordinated
construction, but simply two separate sentences, according to Dixon). However,
Dixon (1979b: 1003) acknowledges, in a quite contradictory fashion, that “there
is not too much evidence that Dyirbal is ergative”. In more recent linguistic lit-
erature on the subject, Dixon’s approach to Dyirbal has generally been followed,
although it is widely accepted that the same extreme pattern of syntactic ergativ-
ity has never been attested in any other language. Insight in Dixon’s data could
give the opportunity to revaluate the theories on syntactic ergativity; however,
thus far the data on Dyirbal has not been made public.
The core arguments in a clause also have semantic functions. For instance,
der Mann in ex. (2) and (3) fulfils the semantic function of the agent of the action
expressed in the clause, whereas den Jungen in the second sentence is the patient.
Agent and patient are “semantic roles”, which are typically assumed by core argu-
ments. Because they rely on semantics, semantic roles can be cross-linguistically
distinguished in a fairly straightforward manner. For instance, every language
can express an agent that performs an action. Fillmore (1968) has introduced
a number of semantic roles (also called “thematic” or “notional” roles, Palmer
1994), including “Agent”, “Patient”, “Theme”, and “Recipient”. In subsequent
studies, the number of semantic roles has varied considerably and these roles
have not always been defined consistently or exhaustively (cf. Dowty 1991: 548;
Palmer 1994: 5). The naturalness of semantic roles partially explains why the
identification of the subject on the basis of morphosyntactic properties comes
across as counterintuitive. As discussed above, the direct object den Jungen in ex.
(3) fulfils the patient role, but in the ergative construction in ex. (5), the patient
bikori is morphosyntactically the “subject”. Based on our knowledge of “Stand-
ard European” languages, the subject has been identified with the agent (the
argument performing the action) and the direct object has been identified with
the patient (the argument undergoing the action). Because the subject is also the
argument that displays the morphosyntactic features of nominative case and the
control of agreement, the observation that the patient in an ergative construction
displays the morphosyntactic features of a subject is, to say the least, surprising.
Semantic roles do not have a direct bearing on the argument coding of ergative
patterns; for instance, in ex. (10) from Tsez (Polinsky and Comrie 1999: 110), žek’u
is the agent but remains unmarked, in contrast with the agent už-ā in ex. (5).

(10) žek’u Ø-ik’i-x

man.cl1 cl1-go-prs
‘The man goes.’
 Ergativity from a functional and typological perspective 19

Additionally, it is worth pointing out that in accusative constructions, the agent is

not always the subject of the sentence. For instance, Mary in Mary falls from the
table is not a volitional agent, she rather undergoes the action.
In one type of alignment the semantic roles do determine the morphosyn-
tactic coding of the arguments. Wichmann (2008: 3–4) has labelled this type of
alignment “semantic alignment”. Semantic alignment includes various types of
constructions in which the only argument of an intransitive sentence is coded on
a semantic rather than a syntactic basis. As a result, the argument coding of S
can differ according to the circumstances. Whereas ergative alignment jeopard-
izes the distinction between subjects and direct objects, semantic alignment led
Wichmann to question the existence of an intransitive subject, because this “S”
can be formally coded in various ways. Prior to the proposal of semantic align-
ment, linguists commonly referred to “active” or “agentive” languages to indicate
languages with predominantly semantically-based argument coding (cf. Klimov
1974; Tuite et al. 1985). One type of semantic alignment is indeed based on an
active-stative and agent-patient distinction (cf. Nichols 1990; Mithun 1991). In an
active-stative construction, the semantics of the verb determines the argument
coding. For example, if the verb is active (e.g., ‘dance’, ‘jump’ and most transitive
verbs), S is coded in the same way as A. If the verb is stative (e.g., ‘fall’, ‘sleep’), S
is coded in the same way as O. Conversely, in an agent-patient construction, the
semantics of the arguments (i.e., whether S is agentive or patientive) determine
the coding; for instance, in John falls, John may fall either intentionally, to draw
attention, or John may accidentally stumble and fall (cf. DeLancey 1981; McGregor
2009: 487). Dixon (1979) uses the terms “split S-marking” and “fluid S-marking”
to indicate the difference between the active and agentive constructions respec-
tively; Merlan (1985, Creissels 2008) opts for the term “split intransitivity” for the
same phenomenon.⁶
In summary, due to the attestation of ergative and semantic alignment, the
terms “subject” and “object” can no longer be used to indicate core arguments
of a clause. Semantic roles were previously considered an alternative means, but
although semantic roles are convenient universal descriptive tools, they are not
always reflected by the coding properties of arguments, particularly not in accu-

6 Donohue and Wichmann (2008) find both Dixon’s and Merlan’s terms ill-chosen because the
terms “split/fluid S-marking” and “split intransitivity” are based on the notion of “S”, although
S is exactly the argument whose existence is being criticized in these accounts. In both Dixon’s
and Merlan’s terminology, the idea of a “split” S is involved. However, according to Donohue
and Wichmann, when “S” is split, S no longer exists as a homogeneous, unitary category in the
language. Therefore, they consider any reference to S in, e.g., “split-S” marking, as misguided.
20 Theoretical preliminaries

sative or ergative alignment patterns (cf. Comrie 1973). A further complication is

that Fillmore’s roles could be endlessly extended because properties of events are
conceptually without limit (for instance, the “killer” role with the verb ‘to kill’,
the “sitter” role with the verb ‘to sit’, etc.).
For these reasons, Dowty (1991) abandons semantic roles and the traditional
subject/object distinction altogether and argues in favor of “protoroles”. Pro-
toroles are prototypical semantic roles based on clusters of semantic and gram-
matical features that have irregular distributions in different languages. This
way, Dowty avoids the vexing problem of endlessly extending the list of possible
semantic roles and creates the possibility for cross-linguistic comparison on a
basis that differs from the controversial subject/object distinction. As semantic
prototypes, “Proto-agent” and “Proto-patient” correlate with syntactic features
for argument selection. Foley and Van Valin (1984, Van Valin and LaPolla 1991)
opt for “macroroles”, i.e., “Actor” and “Undergoer”. The concepts of protoroles
and macroroles are similar but not identical. For Dowty, protoroles are them-
selves prototypes of certain semantic roles, whereas for Van Valin macroroles can
have prototypes: “The prototypical Actor is an agent and the prototypical Under-
goer a patient” (Van Valin 1990: 226). In a sense, both accounts attempt to replace
the subject and object categories, which are cross-linguistically considered as
purely syntactic relations, with concepts that are based on semantics. However,
although macroroles are semantically motivated, Van Valin (1990: 226) remarks
that they are primarily “the two primary arguments of a transitive predication,
either one of which may be the single argument of an intransitive verb”. In his
view, therefore, proto- or macroroles are similar to what has been proposed by
Comrie (1978) and Dixon (1979) with S, A and O/P. Comrie (1978) defines S as “the
single argument of an intransitive verb”, and A and P in a distinctly more seman-
tic way, i.e. A and P are “reminiscent of the semantic terms agent and patient”.
However, despite the correlation between A/P and the semantic roles of agent/
patient, A and P are in “syntactic opposition”, rather than “semantic opposi-
tion” like agent/patient (Comrie 1978: 330–331). The agent and the patient are
the participants in a “prototypical transitive situation”, irrespective of the syntax
(Comrie 1989: 111). The grammatical roles A and P have the advantage over the
semantic roles of agent and patient because grammatical roles may be identified
even in non-prototypical transitive situations. They are “syntactic terms, whose
prototypes are defined in semantic terms”, and Comrie considers them as “primi-
tives” of every language. Similar to the subject and object before them, A, S and
P have been interpreted as purely syntactic categories (cf. Farrell 2005: 8), but
Comrie’s definition indicates that they are not and that they do rely on semantics
in their definitions.
 Ergativity from a functional and typological perspective 21

Dixon too defines S, A and O as “semantico-syntactic” categories (i.e., as a mix of

both semantics and syntax). Dixon (1979: 108) distinguishes A, S and O as “primi-
tives”. His definition of S is identical to Comrie’s definition; A is “the NP in a
TRANSITIVE clause which CAN BE AGENT”, and O is “the OTHER OBLIGATORY
NP in a TRANSITIVE clause”. Dixon continues to rely on the value of a “universal
category” of subject, which groups {A, S} because they can both be the “initiat-
ing/controlling agents”. This grouping is visible in constructions in which S and A
are treated the same way, such as jussives, imperatives and ‘can’-verbs (cf. Dixon
1979: 112). Furthermore, in most morphologically ergative constructions, {S, A}
functions as the pivot. Hence, although S and A are morphologically different,
their syntactic behaviour is identical in these constructions. Comrie (1973) also
notes that in the final analysis there are more arguments to retain the concept of
subject than to reject it. Furthermore, recent psycholinguistic research has dem-
onstrated that A and S are cognitively processed in the same way; this observa-
tion seems to indicate that “subject” is actually a cognitive reality (cf. Bornkessel-
Schlesewsky et al. 2008).
Although S, A and O (or P) have become common in typological linguistics
(whereas protoroles and macroroles are limited to more formal frameworks, e.g.,
Primus 1999 and studies conducted in the spirit of Role-and-Reference grammar),
these terms continue to be criticized. A common criticism is that the definition
of S differs completely from the definitions of A and O. While A and O refer to
semantic-syntactic relations, S is simply defined as the only argument present in
the clause, as Chafe and Mithun (1999) have rightly indicated (cf. also Andrews
2001, Creissels 2009). Moreover, following Dixon’s definition, O is the “other
obligatory NP in a clause”. Hence, if A cannot be identified (e.g., because of ellip-
sis) O cannot be identified either because the identification of O is dependent on
that of A. Note that Comrie’s definition of P as the “most patient-like argument”
is clearer in that respect. A second criticism is related to how different types of S
are coded in semantic alignment patterns; thus the category of S must be split.
Dixon (1979) uses Sa and So, to refer to the S coded as A and the S coded as O,
respectively, in languages with semantic alignment. However, he also remarks
that S functions as the “syntactic prime”, whereas a grouping of {A, Sa} would
overemphasize the semantic agency and disregard the different coding of A and S
in an ergative pattern (Dixon 1979: 108–109).

1.1.2.2 Grammatical relations

Grammatical relations are exclusively identified by morphosyntactic features, and
they relate core arguments with the verb (Palmer 1994: 10; Cole and Sadock 1977;
Plank 1979; Comrie 1989: 65–66; Bickel 2011: 399). Other names for these rela-
22 Theoretical preliminaries

tions are “syntactic relations” or “syntactic roles”. The traditional grammatical

relations are the subject and object, the latter including both direct and indirect
object. Palmer (1994) differentiates between grammatical relations and “gram-
matical roles”, which are A, O and S.⁷ The grammatical relations determine the
alignment. For instance, the grammatical relation subject is S=A when expressed
in terms of grammatical roles, and the grammatical relation direct object is O. If
the grammatical relations are S=A vs. O (subject vs. direct object), the alignment
is accusative. Palmer calls this the “Subject-Object relation”. For lack of a better
term that parallels the Subject-Object relation in accusative alignment, Palmer
uses the term “Ergative-Absolutive relation” to refer to the grammatical relations
(S=O vs. A) determining an ergative pattern. He draws on the case terminology
(ergative/absolutive) rather than using separate names referring to grammatical
relations (subject/object). According to Palmer, every language has one dominat-
ing type of grammatical relation, either the Subject-Object relation or the Erga-
tive-Absolutive relation and these relations distinguish accusative and ergative
“languages” (Palmer 1994: 14).
Comrie (1989: 66) notes that many scholars take the universality of gram-
matical relations cross-linguistically for granted; that is, they hypothesize that
subjects, direct objects and indirect objects can be identified across languages.
However, Comrie argues that one must find criteria that establish the grammati-
cal relations in a language independently. These criteria may be identified, for
instance, by the cross-linguistic comparison of subject properties. Comrie refers
to Keenan (1976), who proposed various properties of how the subject in the lan-
guages of the world can be described. These “subject properties” include formal
“coding properties” such as case marking, word order and verb agreement;
“behaviour-and-control properties” such as deletion, coordination properties
and reflexivization; and “semantic properties” such as agency and autonomous
existence. However, Keenan does not indicate the importance of each property,
which seems to vary for each language, nor does he identifies a property that is
indispensable for proving that an argument is a subject (cf. Dixon 1979: 110–112).
Givón (2001: 195–196; 1995: 230) takes issue with the properties advanced by
Keenan. He argues that the semantic properties are in fact “functional reference-
and-topicality properties” and suggests that topicality is the main factor that
determines the subject of a sentence. Givón excludes “agency” from the subject-
identifying properties and argues that, if agency were a determining property
of subjects, arguments of non-agentive (often intransitive) verbs would not be

7 Palmer (1994) himself uses the abbreviations A, P and S, A and P standing for Agent and
Patient respectively, S for “Single argument”.
 Ergativity from a functional and typological perspective 23

considered subjects. This result would be undesirable, however, because these

arguments generally do not display fewer subject properties than, for instance,
ergative subjects of transitive clauses (Givón 1995: 301). Givón orders the prop-
erties hierarchically from most universal to least universal. The most universal
properties are also the most strongly related to topicality:

functional reference-and-topicality properties < behaviour-and-control properties < word

order < grammatical agreement < nominal case-marking

The category of subject can be universally determined if one observes the left-
most properties within the hierarchy (Givón 1995: 251–252, Cole et al. 1980). Case
marking, the rightmost property, is considered to be highly language-particular
and hence much less helpful in determining the subject. Givón’s take on the
matter showed that not all of Keenan’s subject properties are equally basic to the
definition of the subject category. Keenan’s subject properties are arguably so
diverse and general that they fail to meet the goal of establishing independent
criteria that are necessary and sufficient to define the category of “subject” across
languages (cf. Barđdal and Eythórsson 2012). The aim of describing the properties
of arguments that are ostensibly subjects in particular languages is precisely to
identify an argument as a subject in a language. According to Comrie (1989: 66), if
arguments in different languages show “a reasonable degree of overlap” in their
properties, then these arguments may be assigned the subject relation. However,
the extent to which the overlap is “reasonable”, seems to be determined by the
language specialist, and is not cross-linguistically settled.
In contrast with the above attempts to find a cross-linguistic basis to compare
grammatical relations, Dryer (1997) claims that the question “Are grammatical
relations universal?” is a non-issue because, in his view, grammatical relations
are not even “cross-linguistic notions” (Dryer 1997: 115). According to Dryer, if a
category cannot be applied cross-linguistically without changing its definition to
a cluster definition of different features (clearly, the category of subject is such
a case), then it is irrelevant whether the category of subject occurs in every lan-
guage of the world. No subject in one language has exactly the same properties as
any subject in another language.⁸ Subjects and similar concepts such as “object”
or even “case” can be found and defined language-internally, but in a compari-
son between languages, an identical subject or case category is impossible.
The grammatical relation of the indirect object is also questioned. Dryer (1986)

8 Though one can find pairs of languages in which the subject properties are nearly identical;
for instance, because of a historical or dialectal relationship between the two languages.
24 Theoretical preliminaries

observes that languages can code the arguments in ditransitive constructions

differently. He distinguishes indirective alignment, in which the O-argument of
a monotransitive construction patterns as the O-argument of a ditransitive con-
struction, from secundative alignment, which differentiates the primary object (O
of a monotransitive construction and the recipient-object from a ditransitive one)
from the secondary object (i.e. the patient-object of a ditransitive construction)
(cf. Haspelmath 2001; Croft 2003: 152; Siewierska 2004: 57; Kittilä 2006, 2007b).
The following examples from German illustrate the in directive alignment: the O
den Apfel remains in the same case, whether the construction is monotransitive
or ditransitive (Malchukov et al. 2010: 4):

(11) ich aß den Apfel

I.nom ate the.acc apple
‘I ate the apple.’

(12) ich gab dem Kind den Apfel

I.nom gave the.dat child the.acc apple
‘I gave the child the apple.’

The following examples from West Greenlandic illustrate secundative alignment

(Fortescue 1984: 193, 88; Malchukov, Haspelmath and Comrie 2010: 4). The O,
takurnata.q in the monotransitive construction in ex. (13) is absolutive, whereas
in the ditransitive construction in ex. (14), the O aningaasa-nik is instrumental.
The recipient in ex. (14) is in the absolutive case.⁹

(13) piita-p takurnarta.q tuqup-paa?

peter-erg.sg stranger.sg kill-int.3sg->3sg
‘Did Peter kill the stranger?’

(14) (uuma) niisi aningaasa-nik tuni-vaa

(that.erg) Nisi.sg money-ins.pl give-ind.3sg->3sg
‘He gave Nisi money.’

9 Secundative and indirective alignment can occur in different constructions in the same lan-
guage. For instance, in West Greenlandic, O can be expressed by means of an absolutive case
instead of an instrumental as in ex. (14), and the recipient by means of an allative case or an
absolutive case (Fortescue 1984: 89).
 Ergativity from a functional and typological perspective 25

In some languages, the recipient and the patient are marked in the same way; an
example of such marking is the double object construction in English (Malchu-
kov, Haspelmath and Comrie 2010: 4). The arguments of tritransitive construc-
tions (e.g., double causatives) are also coded differently in the languages of the
world (Kittilä 2007). Hence, even the indirect object is not a universally identifi-
able category.
In typology, Dryer’s successful argumentation that a cross-linguistic defini-
tion of a category is worthless if one cannot define the categories conclusively has
led to the opinion that every notion of universal categories must be revaluated.
For instance, Haspelmath (2010) follows Dryer and suggests a distinction between
comparative concepts and descriptive categories. According to Haspelmath,
descriptive categories are used to describe one particular language but are not a
basis for cross-linguistic comparison. Conversely, comparative concepts are tools
of the linguists constructed for cross-linguistic comparison. They must be defined
in very general terms that may be universally applied. However, this extreme view
of typology has received due criticism, e.g., from Newmeyer (2010), who claims
that “descriptive categories” inevitably form the basis for the definition of “com-
parative concepts” because any definition of comparative concepts necessarily
draws on language-particular definitions of categories. Therefore, Newmeyer
concludes that the relation between comparative concepts and descriptive cate-
gories is to a certain extent circular, particularly because determining the descrip-
tive categories in a non-described language, one has to resort to definitions that
are based on comparative concepts. In this view, interaction between compara-
tive concepts and descriptive categories cannot be avoided. Goddard’s (1982: 169)
observation regarding case may be applicable in this respect: “[b]ehind the case
terminology is a history of usage that has established a certain core or criterial
meaning for each case label”. Case labels and terms such as “subject” etc. have a
long history in the linguistic research of many languages; thus it is impossible to
consider them as purely descriptive categories (i.e., as categories that are defined
on the basis of data from one particular language).
Another view of grammatical relations has been proposed by Bickel (2011).
Following Van Valin (1990: 226), Bickel (2011: 402) claims that the best way to
define grammatical roles (or proto- or macroroles) is to define them as “the
minimal set distinguished by numerical valence, i.e. by the distinction between
intransitive (one-place), transitive (two-place) and ditransitive (three-place)
verbs”. He identifies S, A and O in the same way as Comrie (1978) and Dixon
(1979). As many scholars before him, Bickel (2011) observes that the identification
of grammatical relations is a complicated matter both cross-linguistically and
language-internally. Even within one language, one can find different types of
alignment patterns. For instance, S and O can be marked identically, as in an erga-
26 Theoretical preliminaries

tive pattern, although the verb agrees with S and A, as in the accusative pattern.
Due to these inconsistencies in alignment within one language, Bickel (2011: 401)
advances the view that grammatical relations are not determined by a language
but by particular constructions. Cross-linguistically, a transitive construction in
one language can have different grammatical relations compared with a transi-
tive construction in another language. Language-internally, Bickel also espouses
a clear differentiation between the “agreement construction” and the “case con-
struction”. The “agreement construction” (i.e. the morphosyntactic rule of verb
agreement) has a particular set of grammatical relations (for instance, the verb
only agrees with S and A). The “case construction” (i.e. the rule of case marking)
possesses its own set of grammatical relations, which are not necessarily similar
to the set of the agreement construction (for instance, S and O are marked in the
same way). Bickel explains as follows:

In general, then, a GR [grammatical relation] is defined as the set of arguments that is

selected by a construction for a particular syntactic purpose, for example, for agreement
rules or case government. This means that an argument can bear as many GRs as it enters
constructions in a given syntactic context, and these GRs need not be the same across
constructions. How GRs are selected, how they are defined for each construction, to what
degree their distribution overlaps across constructions, how types of GRs correlate with
each other, and how they are distributed in the languages of the world – these are the core
issues that define research in GR typology. (Bickel 2011: 401)

Bickel’s approach resolves a number of contradictions with respect to argu-

ment coding patterns. For instance, the combination of an ergative case marking
pattern with an accusative agreement pattern within a language is due to differ-
ent selections of grammatical relations. In languages in which all other construc-
tions are ergative, the syntactic grouping of A and S, for instance as addressees of
imperatives (cf. Dixon 1979), is associated with a particular choice of grammatical
relations in a particular construction. The dissociation of “grammatical relations”
from “languages” is advantageous because it is no longer necessary to maintain
that there is one dominant alignment pattern in a language and that the deviant
patterns are idiosyncratic “exceptions”. On the contrary, all types of alignment
can in principle be considered equally “eligible” in any construction. However,
Bickel’s definition of “construction” is not completely clear in his approach; he
seems to equate the term with “rule” (cf. Dryer 1986). He maintains that “[w]hile
the difference among rules, constructions and ordered constraint sets is of criti-
cal importance for the architecture of formal grammar models, it is irrelevant to
defining typological variables” (Bickel 2011: 401). However, if one considers a
construction to be a form-meaning unit, then referring to “agreement construc-
 Ergativity from a functional and typological perspective 27

tion” seems unfortunate because “agreement construction” refers to only the syn-
tactic rule of agreement and not to any semantic property of the construction.
According to Bickel, the term that I have used thus far, “ergative construc-
tion” consists of a number of “constructions” that differ with respect to verb
agreement, case marking, and a number of other syntactic patterns. Although
Bickel’s approach may have some drawbacks with respect to the theoretical
issue of what a “construction” entails, the view that grammatical relations are
construction-specific allows us to give an unbiased overview of every possible
alignment pattern in any one language without treating certain constructions as
exceptional and requiring an additional explanation. The question of primary
interest in this kind of typological research is no longer how one can group lan-
guages according to their alignment; the important questions ask what combina-
tions of alignment types occur in languages and what alignment patterns show
cross-linguistic restrictions.
As indicated above, Bickel’s view entails that one language is no longer
restricted to one set of grammatical relations. Bickel (2011: 404) provides an over-
view of possible grammatical relations and their commonly used names, pre-
sented here in Table 1.1.

Grammatical relation Commonly used name

{S} intransitive subject, nominative

{S, A} subject, nominative; accusative alignment
{A} transitive subject, ergative
{O, T} direct object, accusative; indirective alignment
{O, G} primary object, dative; secundative alignment
{T} secondary object
{G} indirect object, dative
{S, O, T} absolutive; nominative; ergative alignment
{S, O, G} absolutive; nominative; ergative alignment

Table 1.1 Bickel (2011: 404) on grammatical relations

Table 1.1 shows that the terminology is far from consistent in the literature, which
refers to alignment types, cases, and grammatical relations without any clear dif-
ferentiations. The advantage of using abbreviations and curly brackets such as
“{S, A}” is that this type of nomenclature is consistent and treats every conceiv-
able grammatical relation in the same, unbiased way. Nevertheless, Bickel con-
tinues to use “subject” for {S, A} and “object” for any relation with O. The indica-
tion of the grammatical relations by curly brackets used by Bickel resembles the
“pivot” term used by Dixon (1979). Bickel restricts the notion of ‘pivot’ to a type of
28 Theoretical preliminaries

grammatical relation found in certain “biclausal constructions”, such as in sub-

ordinate constructions with non-overtly expressed arguments (ex. (7)).
The terminology used for the grammatical roles associated with recipients
varies considerably from one linguist to another. As illustrated in Table 1.1, Bickel
uses G and T (following Croft 2003: 143). G is reminiscent of the semantic role
‘goal’, and T is related to ‘theme’. Other accounts have employed the abbreviation
R, based on the recipient, rather than G. The names of the grammatical relations
in ditransitive constructions differ from the name of the grammatical relations
in transitive constructions. A and O can be identified in transitive construc-
tions, and A, T and G can be identified in ditransitive constructions. However,
Dixon (2010: 116) uses the same term to refer to the most patient-like argument in
monotransitive and ditransitive constructions, viz. O. He also introduces E to refer
to any third argument in a construction that already contains an A and an O. E
refers to “extension to core”. In addition to E, IO refers to this third argument but
is restricted to arguments with the semantic role of recipient or beneficiary that
occur in ditransitive constructions.
In the present study, I will use the abbreviations S, A, O, and IO. S and A
are well-established labels in the literature, S refers to the single argument of an
intransitive clause, and A refers to the most agent-like argument of a transitive
clause. I prefer O to P for the other argument of a transitive clause because P is
generally associated with the semantic role of patient, whereas O does not carry
this association and also includes other semantic roles, such as theme, and refers
more generally to object entities or domains (cf. Bossong 2006: 237). For the same
reasons, I prefer the abbreviation O to refer to the theme/patient argument in
ditransitive constructions. IO is preferred for specific reasons of argument coding
in Indo-Aryan languages, to which I turn in Chapter 2 in more detail. Whereas E
would be a satisfactory alternative, the “extension to core” definition seems too
broad and requires that obligatory locative adjuncts must also be included (for
instance put + on). These extensions are coded in various ways in different lan-
guages (e.g., by means of cases such as the locative, allative, dative, comitative
etc.), yet IO consistently refers to one particular recipient/beneficient/goal argu-
ment in ditransitive constructions and is coded in a particular way.

1.1.2.3 Case
The case of the core arguments has traditionally been considered as an instru-
mental coding property in determining the grammatical relations of construc-
tions. A case, in its original meaning, is defined as belonging to a paradigm that
expresses “the systemic differences in the form of all declined nominals, adjec-
tives and participles, corresponding to differences in their functional and cate-
 Ergativity from a functional and typological perspective 29

gorial status” (Dik 1978: 158). Goddard’s definition (1982: 167) is similar: a case
system is “a set of morphosyntactic categories inferred from formal differences
between nominals in different syntactic or semantic contexts”. Whereas these
authors refer to form-function units manifested in a paradigmatic way, Blake
(2001: 1, and Croft 1988; Haspelmath 2009; Siewierska and Bakker 2009: 291,
among many others) emphasizes the relations that cases express and claims that
“case marks a relationship of a noun to a verb at the clause level, or of a noun to
a preposition, postposition or another noun at the phrase level”. The relationship
is one of semantic dependency in which case marks the dependent partner in
the relation. According to Siewierska and Bakker (2009), unpredictable seman-
tic relations are more likely to be associated with overt case marking. For these
authors, core arguments are less prone to taking case than adjuncts because the
dependency relation between the verb and the core argument is often already
identified by means of verb agreement.
It is customary to differentiate between syntactic cases, which code core
arguments, and semantic cases. Syntactic cases are known as “structural” or
even “functional” cases, because a syntactic case “is assigned on the basis of the
structure or position in the sentence” (Barđdal 2011: 624). Semantic cases have
been called “inherent”, “thematic”, “lexical”, or “idiosyncratic” cases; each
term suggests slightly different meanings (Chomsky 1981: 171; Zaenen, Maling
and Thraínsson 1985; Yip, Jackendoff and Maling 1987; Haspelmath 2009;
Barđdal 2011). Semantic cases mark all types of adjuncts. The terms “lexical”
and “idiosyncratic” are not used for cases that mark pure adjuncts, rather, these
terms are used for cases that mark non-canonical core arguments, including the
experiencer argument in some languages. Thus, the terms “lexical” and “idi-
osyncratic” cases contrast with the “structural” case (Barđdal 2011: 622–623;
Corbett 2006: 148).
The names of structural cases have been used to indicate grammatical rela-
tions; in particular, the relation between the ergative and absolutive cases has
been proposed as parallel to the relation between the subject and object (Palmer
1994: 15; Croft 2003: 144; Bickel 2011: 404). However, subjects and objects are also
distinguished in certain ergative constructions; which can obscure the parallel
organization of the grammatical relations in alignment patterns. Moreover, there
are many other grammatical relations in addition to those implied by the names
Subject/Object and Absolutive/Ergative (cf. Bickel 2011, Table 1.1). Case names are
also used to indicate alignment patterns (e.g., ergative or accusative alignment)
including other coding properties determining alignment. For instance, ergative
verb agreement normally refers to agreement with an O-argument (Haspelmath
2005, Section 1.1.4.2).
30 Theoretical preliminaries

In current linguistics, a case is considered a form/meaning pairing (cf. Willems

1997 for some historiographical background information). In the traditional defi-
nition, formal case marking refers to inflectional marking (cf. Blake 2001: 1; Dik
1978: 158). However, the concept of case has been expanded to other forms of
coding, including adpositional marking (cf. Zwicky 1985; Blake 2001: 9; Haspel-
math 2008) and marking by means of clitics (Butt 2006: 11). A case meaning/
function is not necessarily associated with a single form. As Goddard (1982: 168)
indicates, a case can have different forms; for instance, a case may have different
forms for masculine and for feminine nouns (see also Spencer 2006). Conversely,
one case form can be used for different case functions, particularly as a conse-
quence of case syncretism (cf. Blake 2001: 122). Goddard (1982: 169), following
Wierzbicka (1981), argues that every case has a cross-linguistically established
core meaning. For instance, the accusative case is commonly associated with
patient marking, and the instrumental case is associated with expressing tools,
etc. However, in this view the nominative and absolutive cases are problematic
because they cannot be straightforwardly associated with core meanings. Tradi-
tionally, the nominative is considered the case of A and S and the case for quota-
tion forms (Creissels 2009: 448). With regard to ergative constructions, the desig-
nation “absolute” or “absolutive” refers to the formally unmarked counterpart of
the ergative case (Creissels 2009; Haspelmath 2009).
The nominative and absolutive cases are both unmarked. The only appar-
ent difference is that the nominative contrasts with the accusative in an accu-
sative pattern, whereas the absolutive is in opposition to the ergative. However,
the terminology of nominative and/or absolutive is confusing in languages in
which both accusative and ergative patterns occur because two different names
can designate an identical form-function unit (Creissels 2009: 448), namely, the
S-argument. Moreover, in semantic alignment, only non-agentive S-arguments
are in the nominative case. Therefore, Creissels (2009: 455) and Goddard (1982:
183) suggest that the name “absolutive” should be abandoned, because it is too
intertwined with the ergative alignment type to be a suitable alignment-neutral
name for a case. As outlined above, ergative alignment is associated not with one
particular language, but with constructions. If the term “absolutive” is used to
refer to a case, then the language in which it occurs is considered to be strictly
ergative.¹⁰ However, cross-linguistic comparisons have revealed that linguists

10 One might object that the ergative case is also associated with the ergative case alignment.
However, the ergative case is defined as the case indicating the transitive subject A, irrespec-
tive of the case marking of O or the agreement pattern of the verb. The absolutive case can only
be defined by referring to the “ergativity” of the construction; that is, a particular argument is
 Ergativity from a functional and typological perspective 31

have been unable to find a strictly ergative language among the languages of the
world.
Whereas Goddard defines the nominative as the case of S and the quotation
form, Creissels, in view of the constructions that formally differentiate between
an active and a patientive S, argues that only the quotation form is the “extra-
syntactic” core meaning of the nominative.¹¹ However, Haspelmath (2009) argues
that all cases must be defined with respect to particular languages. Case labels
are primarily “mnemonic devices” that can reflect only some of the semantic
properties they express. Therefore, the name given to a particular case is not a
particularly important matter for Haspelmath. Moreover, the properties attrib-
uted to a specific case in a particular language, rather than the case names, are
the basis of cross-linguistic comparison.
Attributing a semantic value to every case, including the syntactic cases, is
known as the “indexing” or “characterizing” approach to case (Mallinson and
Blake 1981: 91–93). Semantic cases are commonly defined by the function of the
adjuncts they determine; for instance, the locative refers to a place, and the geni-
tive refers to a possessor. In the characterizing approach, the value of a syntactic
case is established in the same way as that of a semantic case. The syntactic case
is determined by the semantic role the case prototypically marks or by the syn-
tactic function the marked argument assumes in the construction (Primus 2009).
Another approach to syntactic case is the “distinguishing” or “discrimina-
tory” approach, according to which cases function as discriminatory mechanisms
to distinguish between the two arguments of a transitive construction (Næss
2006: 309–310).¹² In the distinguishing approach, semantic distinctions only aid

absolutive only if A is ergatively marked, and if the verb agrees with this absolutive argument.
If the latter conditions are not fulfilled, the label “absolutive” is generally not used in the litera-
ture. I will adhere to this convention in the remainder of this study.
11 The view that the nominative is the case of a word’s “quotation (or, citation) form” has
a long tradition in Western case theory, this view has been held since the Greek and Roman
period of language studies (see Willems 1997: 183–186). Creissels (2009: 457) proposes to
use the name “absolute” for the case of citation forms, in contrast with the “integrative” case
for “syntactically licensed” cases. However, this terminology has thus far not been commonly
used.
12 Although less explicitly addressed in the literature, the same distinction between a dis-
criminatory and characterizing approach can be found regarding the case marking of ditransi-
tive constructions. The characterizing approach is often preferred because the typical marking
of the IO is related to the semantic role of recipient. Nevertheless, the attestation of secunda-
tive and indirective alignment in ditransitive constructions has shown that a discriminatory ap-
proach can also be applied to ditransitive alignment. In this view, the IO is distinguished from
O by treating the O of a monotransitive construction in the same way as the O of a ditransitive
32 Theoretical preliminaries

in distinguishing the arguments from each other. For example, in a hypothetical

situation, one may find a formally marked A in a certain language, although the A
in that language is normally unmarked. In both the discriminatory and indexing
approaches, the markedness of A is considered to be related to the lower agency
of A (for instance, it may be a non-human A). In the indexing approach, the lower
agency of A is sufficient to explain its formal markedness. However, in the dis-
tinguishing approach, A is marked because it cannot be properly distinguished
from the O-argument, considering that the O, which is the most patient-like argu-
ment in the clause, is typically low in agency, not the A-argument. In many recent
accounts, the indexing and discriminatory approaches have been combined and
both approaches are considered necessary to identify the arguments by the case
marking (Næss 2007; Kittilä 2011).

1.1.2.4 Verb agreement

Agreement, particularly verb agreement,¹³ is often referred to as a second mor-
phosyntactic mechanism that distinguishes the core arguments in addition to
the mechanism of case marking. Other designations for agreement are “cross-ref-
erencing” and “concord”; many linguists have their own preferences. However,
“agreement” is the most common designation in the literature, and this term will
be used throughout the present study. I will occasionally resort to the term “cross-
reference”, particularly when I wish to focus on agreement from the point of view
of the argument controlling agreement rather than the verb agreeing with the
argument (cf. Corbett 2006: 5–7). Haspelmath (2005) uses the term “indexing”
to refer to agreement as opposed to “flagging”, which is Haspelmath’s term for
case marking. However, considering that “indexing” is used as a synonym for the
“characterizing” function of case marking (cf. Section 1.1.2.3), I prefer to avoid
this term when discussing agreement.
Verb agreement is often understood as person agreement (i.e., the indication
of the person of an argument on the verb). Person agreement can take the form of
an affix (prefix and/or suffix) attached to the verb stem. In some languages, the
addition of such a suffix may cause phonological changes in the verb stem. In

construction (indirective alignment) or by treating the monotransitive O in the same way as the
IO (secundative alignment).
13 Agreement can refer to any marking on a head that cross-references to a dependent ele-
ment. For instance, agreement features such as the case, gender and number of a noun, can
be marked on an accompanying adjective, possessed noun or adposition (cf. Siewierska 2004:
138–148). However, I am only concerned with verb agreement, and an analysis of nominal
agreement is outside the scope of the study.
 Ergativity from a functional and typological perspective 33

traditional Graeco-Roman linguistics, the addition of person suffixes to the verb

form was called the “inflection” of the verb. However, affixes are not the only
way for denoting person agreement. In general, one assumes that person agree-
ment originally derived from bound pronouns, pronominal forms that became
attached to the verb stem. The process is as follows: free pronouns occurring
close to the verb become bound to the verb form, become clitics (which have a
relatively free position but can no longer be used independently) and ultimately
become inflections (Corbett 2006: 75). Cross-linguistically, one may note the dif-
ferent stages in this development in different languages. There are languages that
display free pronouns, person clitics or agreement suffixes (inflection), all indi-
cating the person of an argument on the verb. Clitics are freer than inflectional
agreement, because they can occur in any position in a clause, depending on the
language (Corbett 2006: 13). According to Corbett, there is a functional parallel to
the formal difference between inflectional agreement and clitics, which he calls
“pronominal affixes”. Inflectional agreement is obligatory in most languages;
agreement in the form of pronominal clitics is sometimes optional. Corbett (2006:
103) argues that pronominal affixes may or may not be considered as agreement,
depending on certain criteria. For instance, agreement can only refer to an argu-
ment in one particular case, whereas pronominal clitics and free pronouns can
refer to more than one argument with different case markings. A second criterion
is referentiality: free pronouns are referential, whereas agreement suffixes are
“non-committal with respect to anaphoric or referential status” (Corbett 2006:
103; Evans 1999: 256). Clitics occupy a middle position. However, in many lan-
guages there are exceptions to these tendencies; therefore, a separate discussion
of whether a certain verb suffix conveys agreement or is instead an anaphoric
bound pronoun is often required for each particular language (Corbett 2006: 14,
Chapter 5). Therefore, Corbett (2006: 112) concludes that “a rigid classification
into languages with agreement or with pronominal affixes would limit rather
than enhance future research”.
Although person agreement is the most remarkable feature of agreement,
verbs can also agree in gender and number with an argument. This type of agree-
ment does not necessarily derive from free pronouns becoming attached to a
stem, but may have originated in a different way. For instance, if the verb was a
participial (adjectival) form in a prior stage of the language, gender and number
agreement are likely to be indicated on the verb.
The function of agreement is less easily defined than that of case marking.
Levin (2001: 21–27) summarizes three functions of agreement. Essentially, agree-
ment on the verb repeats a certain feature already given by the overt argument,
including features such as gender, number or case. Therefore, it is generally
assumed that agreement expresses some type of redundancy. However, Corbett
34 Theoretical preliminaries

(2006: 275) believes that redundancy is not the main function of agreement,
because in many languages, the controller of agreement is not overt; thus the
agreement is the only indicator of the presence of an argument. Levin’s second
function is reference tracking, defined by Corbett (2006: 275) as “allowing the
hearer to keep track of the different referents in a discourse”. The function of
reference tracking is similar to the distinguishing function of case marking; just a
case marking, agreement allows the hearer to identify and distinguish the argu-
ments in discourse (Siewierska and Bakker 2009: 293). The third function refers to
the syntax of a construction, and entails that agreement marks the constituency
of a construction. Although all three functions apply to a certain extent, Corbett
(2006: 275) argues that none of them sufficiently accounts for all instances of
agreement. Alternatively, he argues for an amalgam of “different combinations
of functions in different languages”. In many accounts, agreement has been con-
sidered as a mechanism that primarily indicates syntactic relations between a
verb and its arguments; however, the attestation of bound pronouns that do not
necessarily agree with an overt argument in a clause has introduced a semantic
dimension to the discussion, in the realization that these forms are more than
simple indications of cross-reference.
Case marking and agreement can be considered as complementary (of course,
agreement is only comparable to the case marking of the core arguments, not to
the case marking of adjuncts). Siewierska and Bakker (2009: 293) argue that the
difference between case marking and agreement is that case marking denotes
“the relation holding between two entities”, whereas agreement is considered to
be more specific. Agreement also indicates a relation, but codes only one particu-
lar “property” of an argument, such as its person, gender or number. In all other
respects, case marking and agreement demonstrate the same distinguishing and
indexing functions.

1.1.3 Alignment splits based on referential hierarchies

Pairings of grammatical relations constitute alignment patterns. For instance

{S, A} as opposed to O is the pattern of accusative alignment. As Bickel (2011)
shows, there are many possible pairings of grammatical relations. A single lan-
guage normally does not display only one alignment type. Though the possibili-
ties of alignment are numerous, in general there are certain recurring regularities
in the distribution of alignment patterns. For instance, languages with predomi-
nantly ergative alignment may show accusative alignment under particular cir-
cumstances. These regularities have led to the assumption of so-called alignment
“splits”, i.e., regular occurrences of different alignment patterns within one and
 Ergativity from a functional and typological perspective 35

the same language. Splits may have various motivations, including the semantic
factors entailed by the so-called “referential hierarchies”.

1.1.3.1 Silverstein’s hierarchy

Split alignment has been a common term in the literature since Dixon (1979). For
instance, languages with split alignment may display a predominantly ergative
pattern, but they will have constructions diverting from this alignment, often in
favor of an accusative pattern. The reasons for these deviating constructions are
manifold, but certain alignment patterns appear to be recurrent in the splits. On
the basis of these recurrent constructions, one can hypothesize possible motiva-
tions for the splits.
Since Silverstein (1976), the concept of “referential hierarchy” has been
related to case marking and “splits” in alignment patterns. Silverstein noted
that referential features of core arguments influence their case marking. He con-
structed a feature inventory of these referential features and related the inventory
with the case marking patterns in different languages. On the basis of a compari-
son between the feature inventory and the case marking in different languages,
Silverstein proposed the following hierarchy of features:

first/second person > third person > proper nouns > human > animate > inanimate

In this “universal” hierarchy, there are three different types of features distin-
guished, viz. features referring to person, nominality, and animacy. Silverstein’s
hierarchy is known as the “animacy hierarchy”, “nominal hierarchy” or “person/
indexicality hierarchy”. More recent research has shown that other features can
also be arranged in a hierarchy. Siewierska (2004: 149–161) distinguishes a defi-
niteness hierarchy, which ranks the definiteness of the arguments from definite/
specific to indefinite/non-specific, and a focus hierarchy, which ranks topicalized
arguments above non-topical arguments. Bickel (2008, 2011: 410) uses “referen-
tial hierarchies” (RH) as an overarching term to include all the different hierar-
chies thus far proposed in the literature, and he notes that politeness and other
notions can also be ordered in a referential hierarchy.
The hierarchies decisively influence the marking of the core arguments
because they predict that A-arguments with features from the left side of the hier-
archy will be unmarked, whereas O-arguments with features from the right side
of the hierarchy will take no marking (Comrie 1989: 129–137; Bickel 2008: 191 “the
Marking-based RH hypothesis”). Silverstein (1976) also claims that the hierarchies
are related to the alignment pattern. More specifically, Silverstein contends that
the core arguments with features toward the left of the hierarchy pattern accusa-
36 Theoretical preliminaries

tively, while the arguments more to the right pattern ergatively (Bickel 2008: 192
“the Alignment-based RH hypothesis”). The latter hypothesis is stronger than the
Marking-based hypothesis because according to the Alignment-based RH hypoth-
esis, other factors of alignment, such as verb agreement, are believed to be influ-
enced by the hierarchies.
Silverstein’s feature hierarchy (1976) described and predicted the linguistic
regularities he encountered in a range of languages. Later, functional explana-
tions were sought for the hierarchies and for their correlations with case marking
(McGregor 2009: 490). Dixon (1979) suggests that the “potentiality of agency” (or
“natural agentivity”, DeLancey 1981: 645) is the prime motivating factor behind
the hierarchies. Agency is “the degree of control which the referent of a noun
phrase has over the situation described by the verb with which it is associated”
(Comrie 1978: 356). Dixon claims that the arguments at the left of the hierarchy
are more agentive-like (i.e., they occur more naturally as agents of an action
expressed in a clause). These arguments do not require a special case marking
and remain in the nominative case when they are in the agent role because they
are naturally interpreted as agents. In other words, they do not need to distin-
guish or discriminate themselves from the other argument(s). Similarly, when the
arguments with features from the left of the hierarchy are assigned patient roles,
they are likely marked because these roles are not usually associated with these
arguments in a clause. However, arguments at the right side of the hierarchy are
rather “unexpected” in an agentive role. Therefore, it is more likely that they are
marked, for instance with an ergative case. Their occurrence as patients, though,
is considered to be entirely natural, and they are expected to remain unmarked
in this function.
Nevertheless, recent typological studies have criticized and modified the
“potentiality of agency” hypothesis in several respects. Two particular cross-
linguistic phenomena were instrumental for the criticism, differential object
marking and differential subject marking.

1.1.3.2 Differential object marking: the markedness discussion

Differential object marking (DOM) has been explicitly explored in the literature
since Payne (1980) and Bossong (1985), who described DOM in Iranian languages.
DOM refers to the phenomenon in languages in which O can be marked in differ-
ent ways. In most languages with DOM, O is either unmarked or marked. This
varying markedness of O has been attributed to the influence of the features of
the referential hierarchies. The semantic criteria determining the marking of O
are usually definiteness and/or animacy, but the topicality level of O can also
have an influence (Bossong 1983; Lazard 2001).
 Ergativity from a functional and typological perspective 37

An O-argument with features from the left side of the referential hierarchy should
be marked because these arguments are the least likely to function as patients. In
this view, O’s should be inanimate, indefinite and non-topical, and generally not
salient to human perception and interests, in contrast with A. The core arguments
A and O must be maximally distinguished from each other to be clearly identified
in the clause (cf. Aissen 1999, 2003). However, the prototypicality of an inanimate/
indefinite O has been contested since Hopper and Thompson (1980). These authors
have argued that an affected and individuated O contributes to a higher level of tran-
sitivity and is therefore “more natural” than a non-affected or non-individuated O.
Individuation refers to a high level of nominality, animacy, concreteness, singular-
ity and definiteness (Hopper and Thompson 1980: 253). Affectedness refers to the
extent to which O is affected by the action. The concept is illustrated in Hopper and
Thompson (1980: 253) by underlining the difference between, e.g., drinking up the
milk compared to drinking some milk, in which some milk is said to be a less affected
O than the milk. According to Hopper and Thompson, sentences with an agentive A
and an individuated O constitute the prototypical transitive sentence, in which O is
definite, concrete, and animate – all of which are properties that are usually attrib-
uted to the agent in the discriminatory account, rather than to the O-argument.
Drawing on Hopper and Thompson (1980), Comrie (1989: 136) argues that O need
not be inanimate or indefinite (i.e. the opposite of A); O is simply less animate/
definite than A is. This argument has been rephrased by de Swart (2006: 253) as
the Minimal Semantic Distinctness principle¹⁴ (contra Aissen 2003): “the agent
must outrank the patient on the relevant semantic scales”. However, this statement
still implies that O must be semantically distinguished in features from A, and this
implication contrasts with the indexing approach, according to which the individu-
ation of the patient is on the same level as the individuation of the agent.
Næss (2004) acknowledges the contradiction between the view of Hopper
and Thompson (1980) and the discriminatory approach and attempts to find a
middle way by arguing that the distinguishing feature of O is not individuation,
but affectedness. In this view, a prototypical patient is more affected by the action
than the agent is. However, Næss (2004: 1203) argues that the affectedness of O
is instantiated by a high level of animacy and definiteness, because living beings
and definite entities (e.g., the milk instead of some milk) will naturally be more
affected. It is not entirely clear whether Næss’ reasoning amounts to more than
a terminological discussion. She attributes definiteness/animacy to the property
of affectedness rather than individuation, contrary to Hopper and Thompson

14 Not to be confused with Næss’ principle of “maximal semantic distinctness” of the argu-
ments, that is presented in the same volume (Næss 2006: 310, see Section 1.2.3.2).
38 Theoretical preliminaries

(1980); but ultimately, affectedness and individuation are similar, if not identi-
cal, properties of a transitive construction because both are based on the same
diagnostic features of definiteness and animacy.
Cross-linguistically, definiteness and animacy are clearly the main factors
determining DOM. However, definiteness and animacy are different types of fea-
tures. Animacy is an inherent property of an argument, whereas definiteness is
contingent (cf. Siewierska 2004: 156; Bossong 1983), in the sense that definite-
ness is a property that can be attributed to an argument and is not inherent to the
argument.¹⁵ Moreover, animacy and definiteness are not equally important for
case marking in all languages. For instance, Klein and de Swart (2011) maintain
that in a language in which animacy and definiteness influence case marking,
animacy can have a stronger impact than definiteness. If this claim is true, a defi-
nite but inanimate O should not be marked, whereas an animate but indefinite O
should be marked. Definiteness is only a factor when the O is inanimate because
an animate O is always marked.¹⁶ Klein and de Swart (2011) also distinguish
between “trigger” and “result” factors. Animacy is a trigger of overt case marking
because the markedness of O is a result of its animacy. Definiteness, though, is an
interpretation of O resulting from the marking. Clearly, the distinction between
“trigger” and “result” creates an inequality among the features found in the hier-
archies. Therefore, it is not surprising that Klein and de Swart question the rel-
evance of the referential hierarchies for DOM (cf. also Durie 1988; Du Bois 2003:
31). They opt for a “multidimensional” account of binary features rather than fea-
tures ranked in a hierarchy. These binary features include other non-referential
but grammatical principles such as word order and tense.
The question of the validity of referential hierarchies for the case marking
of O is related to the marking of A. In particular, a phenomenon called “mark-
edness reversal” complicates the perspective related to referential hierarchies.
The concept of “markedness reversal” has an immediate bearing on differential
subject marking.

15 However, there is one exception that is often ignored in typological literature. Proper names
are per definitionem definite, therefore definiteness is an inherent property of a proper name
(cf., Willems 2000: 98).
16 This conclusion is largely based on evidence from Hindi-Urdu, which Klein and de Swart
(2011) however slightly misrepresent (cf. Chapter 2). The authors adduce a better example from
Kannada (Lidz 2006), but in this language, definiteness is split into specific/non-specific; thus
it is unclear whether the semantic properties that interact (and that are subsequently ranked
against each other) are restricted to animacy and definiteness, or whether a more fine-grained
view of definiteness and animacy that incorporates specificity, humanness, and givenness
should be developed.
 Ergativity from a functional and typological perspective 39

1.1.3.3 Differential subject marking

Differential subject marking is the phenomenon in which the subject can be either
marked or unmarked. The term originated in the context of ergative constructions
and semantic alignment. In DSM, the influence of referential hierarchies is rel-
evant only for the marking of A. In ergative constructions, A is normally marked.
However, a certain type of A does not take the ergative case marking but remains
unmarked. Two hierarchies in particular are said to determine typical examples
of DSM, viz. the person and the nominal hierarchy. Silverstein (1976) claims that
in such splits, pronominal A’s tend to be unmarked whereas nominal A’s take
the ergative case. In other languages, first/second person pronominal A’s are
unmarked, whereas third person pronominal and non-pronominal A-arguments
are in the ergative case. It is also possible that the difference lies between the first
and the second person, one of which is unmarked and the other marked.
Other hierarchies do not seem to apply to A-marking. For instance, Næss
(2004) indicates that the level of definiteness of A has no influence on A-marking.
Malchukov (2008: 207–208) notes an influence of referential hierarchies in the
other direction; that is, in certain languages, only an animate or volitional A is
ergatively marked and not an inanimate or non-volitional A. This observation
contradicts the expectation that only a non-typical A, with features on the right
side of the referential hierarchies, will be marked. In fact, the opposite is true: an
A that is highly volitional tends to be marked.
According to Wierzbicka (1981), there is no evidence that the speaker is more
typically an agent than any other person. In the same vain, Mallinson and Blake
(1981: 82–86) conclude that there is only a difference in agency between humans
and non-humans, not between pronouns and nouns designating humans. Conse-
quently, Wierzbicka (1981: 76) replaces agency with topicality and argues that the
more topical arguments tend to be marked less. Malchukov (2008: 216) confirms
that topical A-arguments tend to be unmarked, whereas a newly introduced A is
rather marked.
Silverstein (1976) and Dixon (1979) claim that the referential hierarchies are
parallel in both directions. The properties that are typical of A are thought of as
expressing features that are opposites of the properties typical of O. Numerous
scholars have considered this situation as an instance of “markedness reversal”
(Aissen 2003). However, as demonstrated above, cross-linguistic evidence has
shown that DOM and DSM are essentially determined by different hierarchies (de
Swart 2006: 251). Animacy and definiteness determine the marking of O, but the
differential marking of A can be explained by topicality (Malchukov 2008: 216).
Conversely, the marking of O is not directly determined by a nominal or person
hierarchy, though animacy and definiteness of course interact with these hierar-
chies. A second, quantitative, argument against the notion of markedness rever-
40 Theoretical preliminaries

sal is that there are far fewer languages with DSM than there are with DOM. If
both A and O were determined by the need to be distinguished from the respective
other argument, many more parallel splits in argument marking would need to be
observed (Næss 2004).
Bickel (2008) concludes that the markedness-based RH hypothesis is correct
in that there is a cross-linguistic tendency for leftmost A-arguments to be less
marked than rightmost A-arguments. This tendency holds exclusively for case
marking. However, on the basis of cross-linguistic data, Bickel also maintains
that there is no relation between referential hierarchies and alignment patterns.
Alignment concerns both case marking and verb agreement. Because verb agree-
ment is rarely influenced by referential hierarchies, establishing a relationship
between the hierarchies and alignment patterns is extremely difficult. However,
Bickel and Witzlack-Makarevich (2008) have found substantial evidence of lan-
guages in which a pronominal A is marked and a non-pronominal A is unmarked.
Therefore, Bickel concludes that differential subject marking is often caused by
a convergence of historical conditions rather than by influences of features from
the referential hierarchies (cf. Filiminova 2005, Chapter 6).
There are still other instances in which the subject can be marked in different
ways, which may be grouped under “semantic alignment” (cf. Wichmann 2008,
Section 1.1.2.1). The pattern found in the so-called “active” languages is a well-
known case in point: the coding of the arguments is based on the semantics of
the verb or the patientive/agentive nature of S (cf. Section 1.1.2.1). Another type of
semantic alignment occurs in combination with a particular lexical set of verbs.
These verbs are transitive in that they take two core arguments, but they lack
other semantic features of transitivity (for instance, an agentive A, an affected O,
telicity of the verb, etc., cf. Hopper and Thompson 1980). As a result, the subject
takes an oblique case, often the dative or a comparable case (Nichols 2008; Kittilä
2011, cf. Chapter 2).
A final type of DSM occurs primarily in the Australian languages and has been
called “optional ergative marking” (cf. McGregor 2010). In this type, the marking
of A with an ergative is entirely optional and is not restricted by any grammatical
feature. The functional motivations behind this type of DSM extend beyond pure
referential hierarchy; the marking seems to be caused by discourse motivations or
other reasons (cf. McGregor 2010; Meakins and O’Shannessy 2010).¹⁷

17 Nepali illustrates this type of DSM. See Section 4.2 for an elaborate discussion.
 Ergativity from a functional and typological perspective 41

1.1.4 Tense/Aspect/Mood and head- vs. dependent-marking

Many splits in alignment are not determined by grammatical factors; rather, they
are determined by semantic factors that influence grammatical coding. In splits
that are determined by referential hierarchies the semantics of the core argu-
ments have a central role. However, the semantics of the verb may also be a deci-
sive factor in the alignment split.

1.1.4.1 Tense/Aspect/Mood split

Dixon (1979) draws attention to splits that are conditioned by the tense/aspect/
mood (TAM) of verbs. There is a cross-linguistic tendency in which a TAM split in
alignment entails that constructions with a perfective verb take ergative align-
ment and constructions with an imperfective verb take accusative alignment.¹⁸
Consequently, it has been argued that there is a semantic similarity between an
ergative coding and a perfective tense/aspect; that is, both point to the result of
an action. The perfective is considered to emphasize the completion of an action
and can be interpreted as either resultative or bounded (cf. Comrie 1973; Peter-
son 1998; Malchukov and de Hoop 2011). The ergative construction has been con-
sidered as a “patient-oriented” construction, as opposed to the “agent-oriented”
accusative construction (cf. Regamey 1954; Hock 1986). Whereas the patient is
considered to be central to the construction, the agent is considered peripheral
in an ergative construction (Mallinson and Blake 1981). In this view, every event
expressed in a transitive construction is considered to have a conceptual “starting
point” and “endpoint”. The starting point is the agent, the argument that insti-
gates the action, and the endpoint is the patient, the argument that represents
the result of the action. The agent is the focus in accusative constructions, but the
patient is the focus in ergative constructions (cf. DeLancey 1981, Section 1.2.1.2).
This relationship is considered to be similar to the relation between perfective
and imperfective constructions. A perfective construction refers to a completed
action, whereas an imperfective construction expresses an action that has begun
but has not yet come to an end (cf. DeLancey 1981; Dixon 1994). Although this
reasoning has been proposed as a functional explanation for the split, many
tense/aspect splits in languages seem to be the consequences of certain historical
developments that have led to an ergative pattern (cf. Kachru and Pandharipande

18 There is one profound exception; in some languages of the Cariban language family the
ergative pattern is found only with imperfective verb forms (Gildea 2003).
42 Theoretical preliminaries

1979; Abraham 2000). In this view, the TAM split is not necessarily functionally
motivated, but rather an effect of an indirect motivation (cf. Creissels 2008).¹⁹
Finally, Dixon (1979) has also drawn attention to splits in the alignment
between main clauses and subordinate clauses. The motivation for this type of
split is similar to that of the TAM split because the mood in main clauses (e.g.,
realis) tends to differ from that of subordinate clauses (e.g., irrealis, future)
(Dixon 1994).

1.1.4.2 Case marking and verb agreement

There are two main morphosyntactic mechanisms coding the arguments in a
clause, the coding of arguments on the verb and the marking of the arguments
themselves (recall that in this study “coding” is used as an overall term irre-
spective of whether the coding occurs on the verb or on the arguments, whereas
“marking” exclusively refers to coding on the arguments.) Nichols (1986, cf. also
Helmbrecht 2001) distinguishes languages that prefer one of these mechanisms
and calls them “head-marking” and “dependent-marking” languages respec-
tively. Head-marking refers to the coding on the verb that is the head of the clause
and is known as verb agreement (Croft 1988; Dixon 1994; Haspelmath 2005:
2). Dependent-marking refers to the morphological marking on the arguments
themselves and can range from inflectional case marking to the use of adposi-
tional marking and clitics. Although most languages show only one of the two
coding types (Nichols 1992), some languages combine both types. However, in
these languages, head- and dependent-marking do not necessarily display the
same alignment pattern. This peculiarity has been investigated by Klimov (1973)
who distinguishes “verbal ergativity”, “nominal ergativity”, and “mixed ergativ-
ity”. With these terms, he refers to three ways in which ergative coding presents
itself in languages; in the form of an ergative case (nominal ergativity), in the
form of verb agreement with O (verbal ergativity), or in the occurrence of both
forms. According to Comrie (1978), the five systems of morphological marking
(viz., the accusative, ergative, tripartite, neutral, and S≠A=O pattern), are also
attested for verb agreement (cf. Section 1.1.1). Moreover, relatively rare morpho-
logical systems such as the tripartite and S≠A=O systems occur more frequently
as agreement patterns than as case marking patterns. Comrie draws attention to
the Iranian language Danesfani as showing the latter type of verb agreement: the
verb agrees only with the S-argument. However, this peculiar agreement has a

19 For example, consider the development from a passive construction to an ergative construc-
tion, cf. Section 1.2.2.
 Ergativity from a functional and typological perspective 43

diachronic explanation in Comrie’s view: Danesfani is in a transitional stage from

an ergative-absolutive system that is apparently “breaking down” and “being
replaced by a nominative-accusative or neutral system” (1978: 342). Because of
the clash between the new accusative agreement with A and the old ergative
agreement with O, the language “opts” to delete all ambiguous agreement suf-
fixes and retains only the agreement with S. In Comrie’s opinion, this system will
further develop into a system in which S-agreement is also lost or in which new
ways of coding A and O are created.
Accusative verb agreement regularly occurs in combination with an erga-
tive case marking pattern; however, “the inverse is rare or non-existent” (Comrie
1978: 340; Woolford 2006). Dixon (1994: 94; also Mallinson and Blake 1981: 104)
provides a possible reason for this asymmetry. Dixon chooses the term ‘bound’
versus ‘free’ split to designate the split between case marking and verb agree-
ment. The terminology ‘bound’ versus ‘free’ refers to the split in alignment
between the ‘bound’ cross-references (agreement markers) on the verb and the
‘free’ case marking on the pronouns. Such a situation is attested, for instance,
in the New Guinean language Enga in which the case marking is ergative and
the verb agreement accusative (cf. Bossong 1984). Dixon (1994: 94) reduces the
‘bound’ versus ‘free’ split to the fairly common split between the accusative align-
ment of sentences with a pronominal A and the ergative alignment of sentences
with a non-pronominal A. Dixon (1994: 86) argues that pronominal A-arguments
are unmarked due to their “verb controlling nature”. However, as pointed out in
Section 1.1.3.3, the correlation between verb agreement and the referential hier-
archy of “agency” is not confirmed cross-linguistically. The accusative pattern of
pronominal A-arguments is often not a result of their more agentive roles, but
particular historical developments may have had a decisive role. The second
reason Dixon provides for the link between pronouns and cross-referencing is a
historical argument and may be more convincing: in many languages, the affixes
on the verb that refer to its arguments evolved along a cline from earlier pronomi-
nal forms through bound pronouns to affixes. Dixon concludes that agreement in
languages with a ‘bound’ versus ‘free’ split should be accusative, whereas case
marking, which applies to both pronouns and nouns, should be ergative (cf. also
Bossong 1984: 383).
If an alignment split between case marking and agreement is possible, then
the question arise as to whether the two coding properties have the same function
and whether (and how) they influence each other. According to Givón (2001: 231),
verb agreement and case marking have complementary functions. He argues that
agreement can fill in a lacuna created by the case marking. For instance, if the
case marking is nominative and distinguishes between the grammatical relations
of subject/object, then verb agreement may be a means of indicating the seman-
44 Theoretical preliminaries

tic roles of agent/patient. This idea is informed by Nichols (1990), who argues
that verb agreement (head-marking) is associated with one particular alignment
type (i.e., the semantic alignment in the so-called stative languages). Accord-
ing to Croft (1988, 2003: 147), agreement is different from case marking because
agreement directly refers to properties of the arguments, whereas case marking is
primarily a means to code a semantic relation between the argument(s) and the
event. In this view, agreement has the discourse function of keeping track of the
arguments and is therefore akin to the way anaphoric relations are expressed by
anaphoric pronouns.
The above-mentioned different functions of case marking and verb agree-
ment have led to the assumption that case marking and verb agreement are two
different mechanisms that exist independently of each other and do not necessar-
ily influence each other (cf. Jelinek 1984; Subbarao 2000 and Butt and King 2004;
but see Woolford 2006 and Chapter 6).

1.2 Direct and indirect motivations of ergativity

According to Anderson (1976), most languages display only a morphological type

of ergativity; moreover, this type only occurs in a limited number of constructions.
As a result, Anderson has claimed that ergativity is merely a “trivial fact about
morphology” (Anderson 1976: 16–17). Other scholars have sought to answer the
question why this morphological variation exists. Some scholars have attributed
a (cognitive) function to ergativity (cf. Du Bois 1987; Langacker 1991), whereas
others have concluded that Anderson’s approach (1976) is essentially correct and
that the ergative pattern is merely a result of other grammatical developments (cf.
DeLancey 2004; Gildea 2004; Creissels 2008).
The most straightforward functional explanation to date has been provided
by Comrie (1978; Tchekhoff 1972: 47). According to Comrie, the motivation behind
the occurrence of different alignment types is simply to distinguish core argu-
ments from each other. In a transitive sentence, two arguments must be iden-
tified. In an intransitive sentence, only one argument must be identified. The
factors that are associated with this unique S-argument may be identical to one
of the two arguments of a transitive sentence. If the argument identical to S is A,
the pattern is accusative, if the identical argument is O, the pattern is ergative.
Other possible alignment patterns are considered uneconomical by Comrie (tri-
partite type) or provide no information with respect to the identification of the
arguments (neutral pattern). Therefore, the accusative and ergative patterns are
preferred as a result of economy and iconicity (by avoiding ambiguity).
 Direct and indirect motivations of ergativity 45

However, one does not find an even distribution of the accusative and erga-
tive patterns across the languages of the world. Even in a single language, there is
usually a preference for one of the two types of constructions. Therefore, in cog-
nitive and functional accounts, Comrie’s typological explanation is considered
insufficient; rather, these accounts focus on finding cognitive and/or functional
motivations for the pairing of {S, A} and {S, O}. Garcia-Miguel (2011: 772) aptly
summarizes the approach towards ergative and accusative languages in cogni-
tive linguistics as follows: “languages grammaticalize either one of two possible
orientations in the conceptualization of events with two participants: starting
from subject and eventually extending to an object or starting from the nuclear
relation with an absolutive and eventually extending to an ergative”. The goal of
cognitive accounts is to find the rationale of {S, A} vs. {S, O}; that is, the cogni-
tive motivation(s) for the tendencies of certain languages to code S and A in the
same way and for other languages to code S and O in the same manner. However,
because they seek a motivation, cognitive accounts seem to underestimate the
diversity of alignment patterns. Many cognitive and functional accounts of erga-
tivity use the term “ergative language” as opposed to “accusative languages”, and
the class of “active languages” constitutes the third pattern that they distinguish
(cf. Cooreman et al. 1984; Dik 1978). Indeed, if one language were associated with
a single alignment type, one could argue that the conceptualization of events in
one language or alignment is fundamentally different from that in another lan-
guage or alignment. However, the proof of the viability of a theory is in the splits.
If ergativity is considered to be essentially determined by a cognitive motivation
that demands the identical coding of S and O, then the same cognitive motivation
should also be able to account for the many non-ergative patterns in languages
with a predominantly ergative pattern. Unfortunately, in the empirical investiga-
tion of languages, these motivations have not been supported.
Similarly, in cognitive linguistics, scholars endorse the view that the abso-
lutive case differs from the nominative case and argue that the absolutive is the
case of {S, O}, whereas the nominative is the case of {S, A} (e.g., Du Bois 1987:
808; Langacker 1991: 380, Section 1.1.2.3). The distinction between the nomina-
tive and the absolutive emphasizes the difference between the two patterns to the
extent that they seem to have nothing in common. Only DeLancey (1981) hesi-
tates to assume a fundamental distinction between the nominative and absolu-
tive. DeLancey refers to Trubetzkoy (1939), who suggested, according to DeLancey
(1981: 630), that “‘nominative’ and ‘absolutive’ are in some sense the same cat-
egory.” (cf. Section 1.2.3).
In this Section 1.2, I will show that scholars have adduced both “direct” and
“indirect” motivations to explain ergativity in languages. Direct motivations are
at the core of functional approaches which attempt to account for the ergative pat-
46 Theoretical preliminaries

terning in languages from a discourse perspective, and of the approaches from a

cognitive perspective which refer to cognitive conceptualizations to explain erga-
tive alignment. The idea of ‘competing motivations’ is incorporated into the func-
tional approaches to account for the splits discussed in Section 1.1.3 and 1.1.4. In
Creissels’ approach (2008) ergative patterns are caused by “indirect” motivations.
Creissels favors a historical account in which the ergative pattern is described as
an “effect” of other diachronic developments in the language related to align-
ment patterns. This approach adeptly handles the splits in alignment, because
the ergative pattern is considered to be a coincidental epiphenomenon rather
than an independent cognitively motivated system.

1.2.1 Ergativity in discourse-functional and cognitive linguistics

1.2.1.1 Du Bois: “The discourse base of ergativity”

Functional accounts have sought functional similarities between O and S to
account for the similar coding of these two core arguments. Compared with the
well-known subject similarities between A and S, O and S seem to have little in
common.
Since Dixon (1979), it has been demonstrated that A and S can be grouped
together as prototypical agents and as syntactic subjects of particular construc-
tions (cf. Section 1.1.2.1). On the basis of Dixon, the functionalist Givón (2001)
claims that {S, A} is the result of a grammaticalization process of the “primary
clausal topic”. The topic of a clause is what is being talked about, the accessi-
ble information, and the implication of A and S expressing the discourse topic is
that both are coded in the same way (cf. Givón 1983; Cooreman et al. 1984; Chafe
1994). In this view, the grouping of S and A is determined by discourse motiva-
tions rather than by semantic roles or syntactic functions.
In his article “The discourse basis of ergativity”, Du Bois (1987) argues in
favor of a similar discourse-based account for {S, O} in ergative constructions.
He emphasizes the validity of describing {S, A} as the most agentive, topical and
human-like arguments, but argues that another, “competing”, motivation may
cause a deviation from this accusative pattern that groups S and A. The main
thrust of his account is that S and O are associated with lexical and new informa-
tion; and in this respect, they differ from A. According to Du Bois, the main dis-
course function of A is to maintain topic continuity in narrative discourse; there-
fore, A is less likely to contain new information (cf. Givón 1983). All languages
display a “preferred argument structure” which dictates that only one new lexical
core argument should be introduced in a clause and that the argument represent-
ing this new information should not be A (Du Bois 2003: 34). From this perspec-
 Direct and indirect motivations of ergativity 47

tive, S and O are grouped on the basis of their common discourse function of
introducing new information, and this grouping is arguably a functional motiva-
tion for their identical coding in ergative constructions. However, according to Du
Bois, the motivations to group {S, A} as agentive topics and {S, O} as conveyers of
new information compete with each other, and this competition causes the split
patterns. The split along Silverstein’s hierarchy of pronouns that are coded dif-
ferently from nouns is also accounted for by this idea of competing motivations.
Notably, pronominal arguments tend to pattern accusatively because these argu-
ments are natural agents. The semantic nature of pronouns is said to carry more
weight than their possible, but unlikely, discourse role of conveying new infor-
mation. Pronouns are necessarily already “given”; that is, they are either men-
tioned in the discourse or referential (Du Bois 1987: 844). Pronouns are not likely
to be grouped as {S, O} because they do not, as a rule, convey new information
(although in some languages they do pattern ergatively, cf. Newmeyer 1998: 149;
Bickel and Witzlack-Makarevich 2008, see Section 1.1.3.3 and Chapter 6).
The account given by Du Bois (1987) is based on an analysis of quantitative
data from the Mayan languages to determine the distribution of new and old
information across S, A and O. The results obtained seem to be confirmed by data
from other languages (e.g., Scancarelli 1985; Du Bois et al. 2003; references in
Du Bois 2003: 38–40 and McGregor 2009: 486). Du Bois’ account has also been
incorporated into grammaticalization theory, because his account presents a neat
example of how discourse becomes grammaticalized into morphology and syntax
(cf. Hopper and Traugott 1993). Nevertheless, Du Bois’ account does not explain
all splits described in the literature; for instance, it does not explain the alignment
split based on TAM. There is no obvious explanation for why a certain discourse
motivation would “win” in a perfective construction but “lose” in an imperfective
construction. The variation between ergative and accusative patterns cannot be
ignored and raises the questions of why and under what circumstances one moti-
vation takes precedence over another one. Clearly, the preference for one moti-
vation over another cannot be determined for all languages, because different
patterns can occur in one particular language. This view has also been adopted
by Harris and Campbell (1995, repeated in Gildea 2004), who have added that
historical developments may lead to different alignment patterns. For instance,
in the transition process from a passive to an ergative pattern, the coding pattern
seems to contradict the preferred argument structure because an oblique-marked
argument (i.e., the passive agent) that introduces new information is reanalyzed
as an ergatively marked A-argument, which typically conveys old information.
Thus, from a historical point of view, the claim that there is a preferred argument
structure based on discourse preferences is not valid because the A-argument,
which supposedly conveys old information, derives from an argument that typi-
48 Theoretical preliminaries

cally introduces new information into the clause. Interestingly, Newmeyer (1998:
149–150) has presented an even more fundamental criticism from a generative
perspective: “we have no way of placing limits on what the relevant set of pos-
sible motivating factors are for any particular phenomenon”. Thus, other factors
in addition to the semantic and discourse motivations mentioned by Du Bois may
also have roles in determining the argument coding, and the number of possible
motivations is unlimited. This claim seems to be confirmed by the wide variation
of different coding patterns and grammatical relations in languages, which draw
on more than a simple discourse grouping of {S, O} or semantic grouping of {S, A}.

1.2.1.2 Cognitive accounts

To explain splits in alignment, DeLancey (1981) introduces two cognitive con-
cepts: viewpoint and attention flow. He defines viewpoint as “the perspective
from which the speaker describes the event” and attention flow as “the order in
which the speaker expects the hearer to attend” to core arguments (DeLancey
1981: 626). According to DeLancey, all events have a natural attention flow, which
runs from the source to the goal. The source of the attention flow forms the “start-
ing point” and the goal is the “terminal point”. In the previously mentioned
accounts, {S, A} represents the agent or the topic. DeLancey considers {S, A} as
representing the starting point and O as the terminal point. He argues that the
source and goal of the natural attention flow can conflict with the viewpoint of
the speaker. In normal circumstances, the viewpoint of the speaker is the same
as the starting point of the attention flow. However, the speaker may consciously
choose another viewpoint, for instance by using a passive construction. In a
passive construction, the end-point of the event is presented first, in the form
of the patient-subject. Because the hearer is unlikely to expect this viewpoint, it
must be indicated by formal measures such as passive syntax.
DeLancey subsequently claims that the viewpoint chosen by the speaker in
ergative constructions is invariably O, viz. the terminal point of the action. As a
result, viewpoint and attention flow conflict. Consequently, the natural starting
point, which is A, must be marked by means of a genuine ergative case to indi-
cate the reversal of the viewpoint. Gildea (2004) considers the ergative case as a
marker of “the peripheralization of the agent”, in the sense that the agent does
not seem to be included in the verb-terminal point unity but rather stands apart.
This observation is linked to the view that {S, O} in an ergative construction rep-
resents the topic, i.e., the “central entity” of the construction, and that A is the
“peripheral” argument of the construction because A is located outside of the
verb-patient unit (cf. Wierzbicka 1980; Mallinson and Blake 1981: 103).
 Direct and indirect motivations of ergativity 49

DeLancey relates the viewpoint versus attention flow idea to two types of
regular splits occurring in alignment patterns; i.e., the splits based on the referen-
tial hierarchies of the arguments (in particular the different case markings of pro-
nouns and nouns) and the TAM splits connected with the semantics of the verb.
With regard to the first split, DeLancey argues that the accusative patterning
of pronominal A-arguments is logical. Pronouns are deictic centres; hence, they
always represent the viewpoint of the speaker. Whereas Du Bois (1987) focuses
on the linguistic “givenness” of the pronouns, DeLancey argues that the “deictic
character” of the pronouns groups them as {S, A}. Due to their deictic value and
their highly topic-worthy nature, they are the most typical nominative subjects
(cf. also Mallinson and Blake 1981: 104).
DeLancey (1981: 646–647) accounts for the second type of splits, the TAM
splits, on the basis of the same cognitive conceptualization of attention flow
and viewpoint. In imperfective (accusative) constructions, the natural starting
point of attention flow is usually identical to the viewpoint of the speaker, given
that the action starts from the agent and has not yet been terminated. However,
a perfective construction emphasizes the goal of an action (i.e., the natural ter-
minal point of an event). Languages with a TAM split apparently “don’t permit a
conflict between aspect and viewpoint assignment” (DeLancey 1981: 647, cf. also
Wierzbicka 1980; Regamey 1954; Mallinson and Blake 1981: 106), because result-
oriented perfective constructions cannot occur in combination with a “starting
point-oriented” accusative alignment pattern. Conversely, imperfective, starting
point-oriented constructions cannot display an ergative result-oriented align-
ment pattern. However, this connection between the semantics of perfective con-
structions and the ergative pattern has been discarded by Anderson (1992: 355) as
“elusive and essentially stipulative”. Anderson’s criticism well-taken, DeLancey’s
approach demands a willingness to interpret the meaning of a perfective as
similar to the meaning of an ergative construction, considering that the perfec-
tive construction conveys a particular tense/aspect meaning that is not conveyed
strictu senso by the ergative construction.
DeLancey’s approach to ergative alignment faces the same objections as Du
Bois’ approach. The attention flow/viewpoint perspective cannot account for the
variation in alignment patterns within a language. DeLancey attempts to account
for some regularly occurring splits, but he considers them as exceptions to the
ergative pattern and assumes different (exceptional) motivations for all of these
splits. DeLancey’s explanation is essentially a cognitive approach in which case
is defined as a locational relation in a conceptual space. Langacker’s cognitivist
view of ergativity shows the same desire to define the ergative case on a seman-
tic, localistic base. Langacker (1991: 380) argues against the assumption that the
“ergative/absolutive” pattern marks subject/object relations because the ergative
50 Theoretical preliminaries

case does not mark all subjects, and the “absolutive” is not restricted to only one
grammatical relation. The obvious rationale behind the accusative pattern is pre-
cisely its correlation with the grammatical relations of subject/object. However,
in ergative patterns, there is no correlation between the coding properties of the
arguments and the subject/object distinction. A and S do not pattern in the same
way; thus, they do not constitute the category of “subject”. The motivation behind
the existence of the ergative pattern must therefore be sought elsewhere.
Langacker, in contradistinction to DeLancey (1987), applies a broad defini-
tion of ergativity. In contemporary linguistics, one differentiates between “typo-
logical ergativity” (cf. Dixon 1979), in which the ergative pattern is morphologi-
cally or syntactically indicated, and “lexical-semantic ergativity”, in which an
ergative pattern is observed at the lexical or semantic level (cf. Halliday [1966]
2003: 61–67, 1967; Davidse 1992; Lemmens 1998; McGregor 2009).²⁰ Although
originally introduced by Halliday, the idea of lexical-semantic ergativity has been
influenced by the generative interpretation of ergativity. In this interpretation,
ergativity is essentially a property of the verb rather than a property of the con-
struction. According to Peeters (2000: 688) “among (a decreasing number of)
generativists, verbs are called ergative if their subject at surface structure level
does not derive from an underlying subject, but instead from a direct object”; this
idea harkens back to Burzio (1986), Perlmutter (1983) and Wunderlich (1985: 194).
They consider the subject of an “ergative” verb not as a true agentive subject but
an underlying direct object.
Langacker adopts the definition of Halliday, contra Dixon (1994: 20), arguing
that ergative patterns can be observed in all languages, including English. For
instance, Langacker claims that nominalizations in English pattern ergatively. In
the following examples, “of” indicates S and O, whereas “by” is the marker of A
(Langacker 1991: 380):

(15) a. the chanting of the demonstrators

b. the chanting of the slogans

c. the chanting of the slogans by the demonstrators

The difference between the constructions in (15) and morphologically ergative

constructions is the following: the verb to chant is divalent; that is, it can take
(no more than) two arguments. However, the construction in which to chant can

20 Thus far, I have considered only “typological ergativity” in this study. I will continue to refer
to typologically ergative constructions when using the simple term “ergativity”.
 Direct and indirect motivations of ergativity 51

occur, may be transitive or intransitive: (15a) is an intransitive construction and

(15b) and (15c) are transitive constructions. Much earlier, Tchekhoff (1978: 42) had
already considered this type of nominalization in French to be an illustration of
what he called “neutralité … pour la voix” (i.e., the voice of the verb is neutralized
and is neither passive nor active). However, in morphologically ergative construc-
tions, divalent verbs that can occur in transitive and intransitive constructions
are rare. Even if a divalent verb occurs without O, the construction is still consid-
ered transitive, in the assumption that one argument is not overtly realized.
Langacker’s broad definition of ergativity also implies that English verbs
such as break and melt are considered ergative because they can occur in both
one-argument and two-argument constructions. In the following examples, the
ice is the subject in (16a) but the direct object in (16b), yet semantically, it remains
the theme in both occurrences.

(16) a. The ice broke.

b. John broke the ice.

Note that in languages with morphological ergativity, verbs of the ‘break’ type
are expected to be exceptional, because the case marking of the core arguments
depends on the transitivity of the verb. Dixon and Aikhenvald (2000: 4) labelled
them “ambitransitive” because they can be transitive or intransitive.²¹
Morphologically ergative marking is inherently related to transitivity because
the verbs in ergative constructions must be transitive; if not, their subjects will
not take ergative case (cf. Genetti 2007: 107 on Dolakha Newar). Dixon (1994) has
argued against the use of the term “ergative” to designate lexical-semantic ergativ-
ity in languages such as English, on the basis of the observation that it is confus-
ing to treat a pattern determined by morphosyntactic features in the same way as
a lexical-semantic pattern. However, McGregor (2009: 486) sees no reason not to
extend the term from typological ergativity to lexical-semantic ergativity because
other phenomena in addition to case marking are included in typological erga-
tivity already (for instance patterns of syntactic ergativity). Therefore, McGregor
considers ergativity not a purely morphosyntactic phenomenon, but liable to an
extension to the lexical-semantic level. In a lexical-semantic account of ergativity,
a verb is either “ergative” (e.g., ‘break’) or “transitive” (e.g., ‘kill’) because erga-

21 Note that break differs from a verb such as eat. When eat is used in an intransitive construc-
tion, then this is merely a case of ellipsis, rendering O automatically indefinite, as in he is
eating (cf. Næss 2007: 125).
52 Theoretical preliminaries

tivity is considered a property of the verb. In contrast, constructions with erga-

tive morphology or syntax are necessarily transitive. From Dixon’s point of view,
ergativity is a property of a construction rather than a property of a verb – if the
verb were not transitive, there would be no ergative morphology/syntax. Hence,
there is a fundamental difference between the two views on ergativity. Extending
the term from morphologically ergative constructions in Dyirbal to English ‘break’
constructions is clearly not the same as extending the term from case marking
to other coding patterns. Furthermore, Abraham (2000: 133) points out that if we
wish to analyze the English ‘break’ examples, without the observation of formal
coding properties, “all we are thrown back to are the syntactic diagnostics – which
turn out to be of a dramatically different nature [than the formal coding proper-
ties used to identify typologically ergative constructions]”. Abraham applies three
syntactic analyses (coordination, passivization and relative clause constructing)
to typologically ergative constructions from Dyirbal and to the English ‘break’ type
of examples and concludes that the two constructions have nothing in common.
Remarkably, although Langacker (1991) considers ergativity to be a lexical-
semantic property of verbs, which is clear from his choice of examples, he also
applies his approach to morphological ergativity by attempting to account for
the ergative and absolutive cases. Like DeLancey, Langacker argues that each
event follows a natural path. The path runs along a prominence hierarchy from
the “primary figure” to the “secondary figure” in the event, and these figures are
projected on a “ground”. In spatial terms, the primary figure is the “trajectory”
or the starting point of an action, and the secondary figure is the “landmark”
or the goal. Langacker’s notion of primary and secondary figures is parallel to
Givón’s idea of primary and secondary topics. The primary figure/topic is the
subject/nominative, and the secondary figure/topic is the object/accusative in
accusative constructions. The ergative pattern presents the event differently.
Langacker argues that the nominative and absolutive cases, which are both for-
mally unmarked, are the “more basic cognitive” terms in an alignment theory.
Clearly, the nominative is the starting point as the primary figure, but the absolu-
tive is also a starting point, albeit “of a natural path defined in terms of concep-
tual autonomy/dependence” (Langacker 1991: 382). According to Langacker, this
“autonomy/dependence” layering defines a path “running counter to the flow
of energy along an action chain”. This approach implies that the starting point
of an action is the “theme”. Langacker illustrates this concept with the ‘break’
example: “(glass break)” is the autonomous proposition (the “event conception’s
conceptually autonomous ‘core’”, as Langacker 1999: 84 puts it) to which one can
add an element such as ‘hammer’. Thus, ‘hammer’ is the dependent element or
the new influx of “energy” in the event structure (Langacker 1991: 382, 1999: 85).
In this view, the absolutive case is part of the “thematic relationship”. In short,
 Direct and indirect motivations of ergativity 53

the absolutive is the “theme”, and the ergative case is the “agent”. “Theme” and
“agent” are defined in terms of autonomy/dependency; “theme” expresses the
autonomous, thematic relation, and “agent” expresses the input of “energy”.
Mithun and Chafe (1999: 583) argue similarly that the absolutive case is the “par-
ticipant most immediately or directly involved in the event or state”. According
to Mithun and Chafe, the “immediacy of involvement” determines the ergative
pattern, and according to Garcia-Miguel (2011: 768), this concept is identical to
Langacker’s autonomy/dependency distinction. “[T]he theme is more intrinsic to
the overall event conception than other participants” (Langacker 1999: 86), and
this constitutes the motivation for the ergative construction.
The account of ergativity by Langacker (1999) is subsumed under the
Chapter “The meaning of of”; thus, Langacker provides essentially an account
of the “ergative” construction in English in this chapter. However, his account
has been applied to languages with morphological ergative constructions (e.g.,
Mithun and Chafe 1999; Montaut 2004, 2009b, Chapter 2). With regard to the
extension to typologically ergative languages, Langacker assumes that the theme
and the verb are central to the proposition, whereas the agent in an ergative con-
struction is a peripheral argument, not obligatory and easily omitted. However,
this assumption runs counter to what is found in most ergative constructions, for
the expression of the agent in a transitive clause is as obligatory as the expression
of an agent in accusative constructions.
In addition to the focus on lexical-semantic ergativity, Langacker’s account
is flawed in the same way as DeLancey’s and Du Bois’ accounts: it cannot explain
the extensive variation among the alignment patterns in languages. Why would
certain constructions prefer to indicate “the immediacy of involvement” of the
participants, whereas the same constructions with a different verb tense would
then indicate “the starting point” of an action instead? Another unsettled point is
how the “theme”-based approach could account for languages in which A and O
are both overtly marked (cf. Chapter 2).
Split patterns are the greatest problem for functional and other motiva-
tions of ergativity. Cognitive accounts of ergativity assume a different underly-
ing conceptualization of ergative and accusative constructions, but they overlook
the fact that the same event can be expressed in the same language by means
of different alignment patterns, without any indication that there is a different
conceptualization of the event. For example, another alignment pattern may be
used if a verb tense changes or if the first person changes to the third person.
Such findings caution against a purely “conceptual”, cognitive-based framework
explaining ergativity. Moreover, Gildea (2004) claims that from a historical per-
spective, “neither the putative semantic similarity between O and S, nor the dis-
course affinity between O and S, has ever been observed to lead to the creation of
54 Theoretical preliminaries

an absolutive grammatical category”. In a later article, DeLancey (2004) refrains

from considering ergativity as a taxonomic, independent phenomenon. He refers
to ergativity as a “blue bird”, implying that ergativity is as much a determining
feature to classify languages as being blue is a taxonomic feature to classify birds.
The observation is interesting, but it does not provide any information about the
cognitive foundation of language. An attempt to find functional or conceptual
motivations for a bird’s being blue or a language’s being ergative, may turn out to
be a vain enterprise.

1.2.2 Historical motivations

Explaining ergativity from a purely historical point of view is almost tantamount

to denying any functional importance of the ergative system from a synchronic
point of view and arguing that it is merely an epiphenomenon of linguistic change.
Historical motivations assume that the ergative construction emerged at a certain
time in a language that had previously displayed only accusative constructions.
Specifically, historical motivations are based on research on languages with erga-
tive constructions for which historical material is available, such as Indo-Aryan
and Indo-Iranian. To study other languages, scholars must rely on reconstruc-
tions that often depend on available material from neighboring languages.
Creissels (2008) argues that the ergativity of a construction is often caused
by indirect motivations that are likely to be neglected in the literature. He dis-
tinguishes three types of historical indirect motivations that may explain why
languages may display a split alignment rather than a single alignment pattern.
According to Creissels, splits in alignment patterns are not necessarily function-
ally motivated, alignment may also have changed in a mechanical way, being in
fact epiphenomenal to changes in other domains of the grammar (Creissels 2008:
3). He suggests three possible historical pathways leading to variable alignment
patterns.
The first way of transforming alignment is caused by so-called “light verbs”.
Creissels (2008: 10, following Jespersen 1909–1949) defines a light verb construc-
tion as a construction in which “the verb is semantically “light” in the sense that
its contribution to the conceptualization of an event is relatively small in com-
parison with that of one of its complements, very commonly, but not exclusively,
a noun in P role”. For instance, vefat etmek in Turkish literally translates as ‘death
do’, meaning ‘to die’. Light verb constructions are transitive, taking an A and an
O (vefat ‘death’ being O in the Turkish example). However, due to their lexicalized
meaning, such verbal constructions can be semantically reanalyzed as intransi-
tive. The O-argument is then considered part of the verb phrase rather than a free
 Direct and indirect motivations of ergativity 55

patient-argument. If the predominant alignment pattern in the language is accu-

sative, no problem of case marking arises because A and S are coded alike and the
transitivity of the verb has no bearing on their coding. However, if the predomi-
nant alignment pattern is ergative, then the A of a light verb construction should
normally be ergative, given that the verb is transitive and takes an O. After the
reanalysis of the light verb construction, the verb and the patient form a single
complex unit and the verb is considered intransitive; hence, the marking of A may
change to the nominative marking of an S (Creissels 2008: 10–12, Section 1.2.2).
A second possible way of changing alignment is arrived at through ellipsis
(Creissels 2008: 13–14). If the O-argument in an ergative construction is often
absent (for instance, when it is indefinite as in the English clause he is eating), the
construction may be reinterpreted as intransitive. However, the coding remains
ergative, and again there is a deviant alignment pattern because intransitive con-
structions are now taking ergatively marked A’s. Similarly, accusative transitive
constructions without an overt A can display deviating alignment because they
only have an overt argument coded as O.
The third pathway leading to a different alignment pattern is also the best
studied, particularly with respect to Indo-Aryan and Indo-Iranian languages.
Creissels (2008: 20–28) calls this pathway the grammaticalization of TAM periph-
rases. The best known among this type of pathway is the passive-to-ergative
development, and many early conceptions of ergativity have been based on this
concept. According to this approach (e.g., Dik 1978), although the ergative is not
passive, it nevertheless originated from a passive construction. Creissels rejects
this view and maintains that the constructions that have generally been described
as passive at the origin of ergative constructions are actually resultative construc-
tions; thus he argues that the grammaticalization of periphrastic constructions
was the source of the ergative. I return to an extensive discussion of this transition
with regard to Indo-Aryan in Section 2.3. Note that Creissels does not consider the
ergative construction as the goal of the reanalysis of the periphrases but rather
as an effect of a change in the grammatical TAM paradigm of the languages. As a
result, the ergative construction is not functionally motivated but rather consid-
ered epiphenomenal of other diachronic changes in Creissels’ account. This view
is shared by Gildea (2004: 49) who describes ergativity as merely a “historical
accident”. However, Gildea also notes that because ergative morphology is effec-
tive in distinguishing the A from the O-argument and thus has the function of
distinguishing the arguments, it remains fairly consistent in the languages of the
world and does not easily disappear.
56 Theoretical preliminaries

1.2.3 Case and transitivity

The following two sections focus on the treatment of case and transitivity in
earlier approaches to ergativity. In contrast with functional/cognitive concepts
such as agentivity, topicality and result-orientedness, case and transitivity have
largely remained in the background of the debate about the phenomenon of erga-
tivity. In the ensuing Chapters 2–6 on Indo-Aryan, the role of case and the tran-
sitivity of the verb in the form of agreement will be analyzed. I will argue that
ergative case was the original motivation for claiming the existence of “ergative
languages”, whereas O-agreement is rather a consequence of a historical tran-
sition in the tense paradigm of Indo-Aryan. Moreover, case and transitivity are
complementary: case is a matter of paradigmatic choice, transitivity acts on a
syntagmatic level. The ergative case and the transitivity of a verb are intrinsically
related, to the extent that previous linguists even used the term “transitivity”
from verbal morphology to indicate case morphology.

1.2.3.1 An alternative view of ergativity

The early accounts of ergativity primarily considered ergativity as the ergative
case member of the case paradigm of a given language. According to Seely (1979:
191), Albert Dirr, in 1912, was the first researcher who used the term ‘ergative’ for
a distinct case marking.²² Dirr related the term to the Greek word έ̓ργον, meaning
‘work’ or ‘deed’, implying that the ergative case primarily referred to an active
“agent”. Dirr researched Caucasian languages, and the ergative case remained
associated with these languages for a long time before the term was incorporated
into general linguistics (cf. Vaillant 1936). Prior to Dirr’s research, scholars also
distinguished the ergative case in languages such as Basque and Inuit languages;
however, they coined other terms (e.g., “narrative”) or extended well-known
terms, such as instrumental, to include the ergative. Some of the earliest scholars
who explored ergativity are Gabelentz (1861), Pott (1873), Schuchardt (1896), and
Uhlenbeck (1901). They shared the view that the ergative construction must be
interpreted as a special type of passive (cf. Trask 2002: 274). Uhlenbeck conducted
studies of American Indian languages. He distinguished a case called “Aktivus”

22 According to Manaster-Ramer (1994), Dirr borrowed the term from the Caucasianist Trom-
betti (1903), who attributed the term to Fr. Wilhelm Schmidt (1902). Schmidt had used the term
in a description of languages from New Guinea. It is possible that Schmidt in turn borrowed the
term from Ray and Haddon (1893), but these authors used the term to refer to a type of locative
(Butt 2006; McGregor 2009: 481).
 Direct and indirect motivations of ergativity 57

from a case called “Passivus”. The Aktivus is the “Kasus der handelnden Person”,
and the Passivus is the “Kasus der leidenden Person/Sache” (Uhlenbeck 1916).
Whereas the Aktivus could be used only for a subject, the Passivus could be
used for a subject or a direct object. According to Uhlenbeck, languages with
an Aktivus/Passivus case lack the nominative and accusative cases. Uhlenbeck
extended the terms Aktivus/Passivus from verb morphology to case morphology.
Vaillant later remarked as follows : “son [Aktivus and Passivus case] existence
implique une distinction tranchée et régulière de l’ “actif” et du “passif”, mais
qui s’exprime dans le sujet, tandis que le verbe continue et peut ne plus distinguer
deux constructions également “passives” par leur origine” (Vaillant 1936: 95).²³
The hypothesis that ergative is passive remained popular for quite some time
and was even incorporated into the first functional definitions of ergativity (e.g.,
Estival and Myhill 1988). Hock (1986: 19) points out that Comrie “characterizes
ergativity as “passive” morphology and/or syntax without the existence of a cor-
responding and more basic “active” ”. Because of this lack of a corresponding
“active” form and because languages with ergative constructions can also display
passive forms, the passive approach to ergativity was finally abandoned (Tchek-
hoff 1978).²⁴ Nevertheless, the similarity between passive and ergative construc-
tions has been noted in more recent studies. As pointed out previously, several
accounts suggest that the resemblance between the ergative and passive con-
structions is historically motivated and argue that ergative constructions origi-
nated from passive constructions (cf. Dik 1978, Section 1.2.2).

23 According to Uhlenbeck, and others who followed him, Proto-Indo-European also had an
Aktivus and a Passivus (Uhlenbeck 1901: 170; for an overview of authors discussing ergativity
in early Indo-European see Villar 1984: 168; Rumsey 1987). This hypothesis about PIE was in-
formed by the fact that the neuter gender in old Indo-European languages has only one marking
for a nominative and accusative case, whereas masculine and feminine nouns have a separate
marking for both cases. Uhlenbeck among others attributed this fact to a former PIE ergative
case that was used only for animate, “agentive” subjects, whereas the inanimate or neuter
subjects did not receive the PIE ergative. Later, this hypothesis was debunked, by scholars
harkening back to Silverstein. Silverstein (1976) argued that unexpected, inanimate agents
needed to be marked ergatively, not typical agents (cf. Villar 1984: 167). These early scholars
had all knowledge of Basque, and Dakota was also a frequently quoted language in their work
(not coincidentally a language with semantic alignment).
24 In some predominantly ergative languages, there are “antipassive” constructions that have
essentially the same function as passives; that is, downgrading the agent and promoting the
patient (Comrie 1978). In this study, I will not be concerned with the antipassive constructions
because, although they are an important topic in many ergative studies, they are absent in
Indo-Aryan.
58 Theoretical preliminaries

In the wake of these first accounts of ergativity, the European structuralist Tru-
betzkoy (1939) notes that every transitive construction consists of a verb and (at
least) two arguments. One argument is in “le cas sujet”, and the other argument
is in “le cas déterminant immédiat d’un verbe”. Ergative and accusative construc-
tions pattern differently depending on which argument is the “cas sujet” and
which is the “cas déterminant”. In accusative constructions, O (“l’objet attaint”
for Trubetzkoy) is “déterminant”, whereas in ergative constructions, the A-agent
is the case determining the verb. Below, den Jungen ‘the boy’ in ex. (17) is the
determining case in the German accusative construction, whereas in ex. (18)
from the Kabardian language (Kumakhov et al. 1996: 94), the ergative fɘz-χe-m
‘the women’ is the determining case. In both examples, the nominative is “le cas
sujet” according to Trubetzkoy:

(17) der Mann sieh-t den Junge-n

the.nom man see-prs.3sg the.acc boy-acc
‘The man sees the boy.’

(18) fɘz-χe-m bostej-χe-r ø-ja-dɘ-r

woman-pl-erg dress-pl-nom 3-3pl-sew-prs
‘The women are sewing the dresses.’

Thus, the ergative and the accusative form “le cas déterminant immédiat d’un
verbe” and they are in opposition to the nominative case, or “le cas sujet”. Tru-
betzkoy assumes the existence of only one unmarked nominative case in either
ergative or accusative constructions (this view contrast with the conjecture of an
“absolutive” case). “Le cas déterminant immédiat d’un verbe” determines the
valency of the verb; in particular, it causes the construction to be transitive. In
this sense the accusative/ergative stands in opposition to the nominative, which
is always present in the clause and is the first argument of the verb. The ergative/
accusative is also in opposition to the genitive or any other “semantic” case. Tru-
betzkoy calls the latter cases “le cas déterminant adnominal”, whereas accusa-
tive and ergative are “des cas adverbals”, intrinsically related to the verb.
Tchekhoff (1978) claims (following Martinet 1962) that the subject-argument
is the obligatory argument in any verb construction. According to Tchekhoff, the
subject constitutes the argument that is semantically and syntactically unmarked.
The subject, as the “first” argument, occurs in transitive and intransitive con-
structions and has no well-described semantic role except that it is a non-agent.
Tchekhoff proposes that the “subject” can fulfil the semantic role of “auteur”
and of “patient sémantique”. However, the ergative argument is invariably the
agent, both syntactically and semantically (Tchekhoff 1978: 34). Tchekhoff does
 Direct and indirect motivations of ergativity 59

not search for a cognitive or functional motivation for the {S, O} pairing; he simply
assumes that they are the unmarked cases that can take any remaining semantic
role (cf. the distinguishing approach by Comrie). S and O are defined in purely
negative terms as non-agents.
The difference between Trubetzkoy, Martinet, and Tchekhoff is that Trubetz-
koy has explored case and paradigmatic oppositions in the system of case that
lead to a different syntagmatic construction. In particular, Trubetzkoy (1939)
argues that the accusative and the ergative are opposites although they have the
same function; that is, they are the cases that determine the verb. In contrast,
Martinet and Tchekhoff focus on grammatical relations in the clause.
Whereas Martinet and Tchekhoff seem to assume that categories such as
subject and agent are universally applicable, Coseriu (1987) argues, from a struc-
tural-functionalist perspective, that there are different linguistic-descriptive
roles in ergative and accusative constructions (“languages”). He distinguishes
between two linguistic-descriptive roles, which are said to be language-specific
categories, viz. “Inaktivus” and “Aktivus”. The “Inaktivus” can be found in erga-
tive systems, and the “Aktivus” can be found in accusative systems. According to
Coseriu, the linguistic pairing of the agent with the “Intransitivaktant” character-
izes nominative languages. Together, the agent and the “Intransitivaktant” form
the linguistic-descriptive role of “Aktivus”, which is the grammatical relation {S,
A}. Whereas Uhlenbeck opposed “Passivus” to “Aktivus”, Coseriu chooses to refer
to the pairing of O and the “Intransitivaktant” S as “Inaktivus”. The “Inaktivus” is
the linguistic-descriptive role {S, O}. According to Coseriu, the “Intransitivaktant”
(i.e. S) is the “erweiterte Agens” in an accusative construction; thus he empha-
sizes the similarity in agency between S and A, whereas the “Intransitivaktant”
is an “erweitertes Objekt” in ergative constructions. Coseriu’s main argument is
that the presentation of the events expressed in clauses is “grammaticized” in dif-
ferent ways according to accusative and ergative alignment types. In nominative
languages, the event is presented from the vantage point of a “Quelle”, or source,
which Coseriu has not defined in spatial terms; rather, he has defined “Quelle” as
a mental vantage point from which the event originates. Conversely, in ergative
languages, the event is presented from the vantage point of an “Ort”, or place,
which is defined as the abstract place where the event is mentally situated.
All the above approaches share the view that {S, O} should be grouped not
on the basis of cognitive concepts, but on the basis of the language-particular
similarity between the two core arguments. S and O are the unmarked arguments,
and thus it is rather impossible to find criteria to define them as the nominative
case – they are in the nominative case. However, the ergative case is a formally
marked case in the paradigm and has a semantic basis. Each approach refers to
60 Theoretical preliminaries

“agentivity” as the meaning of the ergative, but the empirical evidence suggests
that the ergative case is related to the transitivity of the verb (cf. Trubetzkoy 1939).

1.2.3.2 Transitivity: from Sapir to Hopper and Thompson

Sapir (1917) reacted to Uhlenbeck’s hypothesis (1916) that the ergative construc-
tion is a type of passive. In this 1917 review article, Sapir focuses on verb agree-
ment, because this coding property is dominant in the Amerindian languages. He
observes that number agreement in Native North American languages often pat-
terns ergatively in the sense that the verb stem agrees with the O in number (Sapir
1917: 84); cf. the following examples from the language Koasati (Mithun 1999: 86,
examples taken from Kimball 1991: 447):

(19) ittó-n attiní:-li-lit

wood-obj burn.sg-1sg.agent-pst
‘I burnt a log’

(20) ittó-n attíni:-li-lit

wood-obj burn.pl-1sg.agent-pst
‘I burnt some logs’

In these examples, the verb meaning ‘to burn’ takes the stem attiní: when O is
singular and the stem attíni: when O is plural.²⁵ Sapir subsequently argues that
for ergative number agreement, the link between S and O lies in the causative
origins of the verb form. For instance, if a verb with the meaning ‘to kill’ has a
form that originally meant or can be interpreted as the causative ‘to cause to die’,
then it stands to reason that the O-argument influences the meaning of the verb
much more than in constructions not derived from a causativization (Sapir com-
pares ‘one man dying’ with ‘several men dying’). He introduces the notion of a
“casus inertiae”, following Uhlenbeck’s “Passivus”, but he attributes a different
meaning to this notion. According to Sapir, a “casus inertiae” is a “caseless form
which takes on all the functions not specifically covered by the transitive or active
case” (Sapir 1917: 85). Note that Sapir uses the term “transitive case” to refer to

25 In these examples, there is also A-agreement, which is indicated by the suffix li. In con-
temporary Amerindian linguistics, one no longer considers the alternating verb stems of these
verbs as true agreement but rather indicates this phenomenon as “verbal number” in the sense
that the verb itself can be plural or singular and this “verbal number” does not necessarily
coincide with the number of its arguments (cf. Mithun 1999: 83).
 Direct and indirect motivations of ergativity 61

the subject of a transitive verb. He continues as follows: “in other words, the I of
I sleep, and the me of he kills me may be identical in form, not because of any
identity of verb-morphology, but merely by way of contrast to the distinctively
transitive form of the I of I kill him” (Sapir 1917: 85). This short quotation clearly
indicates that Sapir countenances a distinguishing approach to case marking,
and essentially claims that the nominative case simply expresses an opposition
to another case. He considers the ergative as a transitive case (cf. Hjelmslev 1972:
154, casus transitivus).²⁶
In the range of contemporary motivations for the ergative pattern, the factor
of transitivity has rarely been explicitly treated but rather has been assumed as
obvious from the beginning. Gildea (2004) refers to Givón (2001) and Hopper
and Thompson (1980) as being among the few scholars who relate ergativity
explicitly with transitivity. According to Givón, the motivation of accusative
constructions is the topicality of the subject, {S, A}. In active constructions, the
semantic roles of agent and patient are morphosyntactically distinguished from
each other. However, in ergative constructions, {A} is opposed to {S, O} because it
marks the transitivity of the construction (Givón 2001: 208). According to Coore-
man, Fox and Givón (1984: 2–3), ergative “morphology is oriented toward the
transitivity properties of the clause, and transitivity is itself a complex epiphe-
nomenon that correlates directly neither with propositional-semantic roles nor
with subject/topic status, although both clearly play important parts in defining
transitivity”.
Ever since Hopper and Thompson (1980) published their seminal paper on
transitivity in language, the concept of transitivity has developed in linguistics
from a purely formal indication of the valency of a verb to a semantic bundle of
properties at the level of the clause (Kittilä 2011: 347; Næss 2007). In the formal
approach, the traditional argument is that a verb is transitive when it has (at
least) two core arguments (cf. Lazard 1998: 160). However, in recent approaches
to transitivity, the presence of two or more participants is only one of the possible
properties of transitivity, and transitivity is defined with respect to a continuum.
This implies that the perspective has shifted from verbs that were analyzed as
either transitive or intransitive to constructions that are analyzed as being either

26 Sapir is also interested in constructions with an active kind of semantic alignment, which
constitute a common pattern in the Amerindian languages. With respect to these constructions,
Sapir contends that it is possible that they are related to, or even originate from, transitive
impersonal or experiencer constructions: this claim was much later adopted and reinterpreted
by Malchukov (2008: 77). However, Sapir also clearly states that ergative constructions are not
related to these impersonal constructions.
62 Theoretical preliminaries

more or less transitive depending on the number of transitive properties they

display. The parameters that Hopper and Thompson (1980: 252) distinguish to
determine high or low transitivity are the following:

HIGH LOW

A. Participants 2 or more participants 1 participant

B. Kinesis action non-action
C. Aspect telic atelic
D. Punctuality punctual non-punctual
E. Volitionality volitional non-volitional
F. Affirmation affirmative negative
G. Mode realis irrealis
H. Agency A high in potency A low in potency
I. Affectedness of O O totally affected O not affected
J. Individuation of O O highly individuated O non-individuated

Table 1.2 Hopper and Thompson’s properties of transitivity (1980:252)

These parameters can be divided into properties pertaining to A, O, and the verb.
The first property, the number of participants, is dependent on both the argu-
ments of the verb. A common criticism of Hopper and Thompson’s continuum is
that the properties are not presented as equal; some properties are valued more
highly than others. For instance, Bhat (1991) argues against the presence of two
participants as the primary criterion to determine A, S and O. He thinks that
the property “volitionality” might equally determine the grammatical relations
between A, S and O. According to Næss (2006, 2007), the properties of affected-
ness of O and volitionality/agency of A are fundamentally important for a proto-
typical transitive construction. In her opinion, the first condition of transitivity is
that the two core arguments are “maximally semantically distinct”; this implies
that O is affected and A is volitional. In other words, transitivity depends on the
semantic interaction between the two arguments rather than simply the pres-
ence of two arguments. According to Næss , the other properties in Hopper and
Thompson’s list, particularly those properties pertaining to verbs, are of second-
ary importance; however, she believes that transitivity should be defined with
respect to a continuum, that may involve additional factors (Kittilä 2011: 349). Her
account is highly reminiscent of Cooreman, Fox and Givón (1984), who argue that
the conditions for semantic transitivity are the presence in the clause of an agent-
cause (visible, salient, intentional, controlling), a patient-result (visible, salient,
non-intentional, non-controlling), and a “compact” (i.e., non-periphrastic) per-
fective or realis verb.
 Conclusions 63

Transitivity is said to determine ergative constructions. Formally, the coding of A

is dependent on the (possible) presence of O. But semantic transitivity also influ-
ences the coding. Givón (1984: 152) argues as follows:
The choice of ergative marking of clausal subjects in ergative languages is
thus sensitive, in one manner or another, not simply to “absolute” semantic ele-
ments such as “prototypical patient object” and “prototypical agent subject”, but
rather to a composite degree of transitivity of the clause. However, individual lan-
guages are not sensitive to exactly the same sub-components of transitivity, nor
do they divide transitive from intransitive at exactly the same point on the various
scales of properties which, as a cluster, determine transitivity.
Transitivity has also been related with alignment patterns in other ways.
Kittilä (2008: 356), following Tsunoda (1985: 391), assumes that accusative and
ergative alignment patterns indicate the coding of prototypical events, whereas
other case frames, such as nominative/dative or nominative/instrumental frames,
are indicative of a lesser degree of transitivity. For instance, some experiencer
verb constructions do not show accusative or ergative patterning; rather these
constructions show a dative case marking for the experiencer role. According to
Hopper and Thompson (1980: 254) these constructions are lower in transitivity.
Kittilä (2011: 364) explains the TAM split in terms of transitivity. In his opinion,
imperfective events lack an affected patient. In addition to the agency of the
agent, the affectedness of the patient is the criterion to obtain a highly transitive
construction. Therefore, in languages with a different alignment pattern based
on the TAM of the verb, imperfective constructions do not pattern ergatively. Con-
versely, perfective constructions display a high degree of transitivity in argument
coding, by means of the ergative case and in semantics, by virtue of the affected
patient and agentive agent.
The advantage of a transitivity-based approach to ergativity over a cogni-
tive approach is that a different conceptual background need not be assumed for
ergative constructions vis-à-vis the accusative construction. As a result, the focus
is less conceptually biased and is entirely based on the linguistic possibilities of
indicating different grammatical relations and semantic roles.

1.3 Conclusions

Functional typology has established a firm basis in the research on ergativity.

However, many questions concerning ergativity have not yet been answered.
The ever-increasing number of languages described and analyzed from a typo-
logical perspective has caused typologists to doubt the existence of ergativity as a
pattern that models the linguistic system of a language in a strictly defined way,
64 Theoretical preliminaries

motivated by a unique rationale. Thus, among typologists ergativity is no longer

a unitary pattern that stands in a cognitive contrast with other alignment system,
“ergativity” has been increasingly used as a descriptive term to refer to one of
the many possible alignment patterns of a construction. A particular construc-
tion may display the ergative pattern in case marking or agreement, for a variety
of reasons that are often language-specific. Historical developments may have
contributed to the motivations for ergativity due to processes such as the merging
of case markers, changes in tense systems and reanalyses of light verb construc-
tions. Synchronic motivations may also be a factor for ergativity; for example, the
ergative pattern is as economical in distinguishing the arguments as the accusa-
tive pattern.
In the next Chapters 2–6, the alignment in the Indo-Aryan languages will be
investigated on the basis of the conceptual and terminological distinctions I have
presented and discussed in the present Chapter. In each chapter, the focus will be
on constructions in these languages that comply with the definition of ergativity
given by Comrie (1978) and Dixon (1979). At the same time, alignment and the
grammatical relations determining this alignment in the Indo-Aryan languages
will be investigated. Thirdly, the attested splits will be scrutinized. Of course, con-
structions that code ergatively are the main subject matter of the ensuing chap-
ters. However, when other patterns seems to be relevant in order to understand
the alignment system of the language, they will also receive due attention as we
go along.
2 Indo-Aryan

2.1 Geographical distribution of the Indo-Aryan languages

The Indo-Aryan (henceforth IA) and the Iranian language families together include
most Eastern representatives of the Indo-European language family.¹ Modern IA
languages are mainly spoken on the South Asian subcontinent including India,
Pakistan, Bangladesh, Nepal, Sri Lanka and parts of Afghanistan. Altogether,
there are almost one billion speakers of IA languages. As a result of the Indian
diaspora, the languages have spread to Europe (in particular Great Britain), the
USA, Australia, and the Middle East. These developments have taken place over
roughly the last 150 years. Sinhala, spoken in Sri Lanka, and the Romani dialects
spoken all over Europe are Indo-Aryan languages that have left the subcontinent
much earlier. In total, there are about 221 IA languages (www.ethnologue.com).
The IA language with the most speakers is Hindi-Urdu (ca. 400 million speakers),
followed by Bangla (ca. 181 million speakers). The differentiation between Hindi
and Urdu is a socio-political matter, founded in the partition of the British Indian
Empire into India and Pakistan in 1947. In Pakistan, the name Urdu is common,
whereas in most parts of India, Hindi is more frequent (except for areas with a
predominantly Muslim population, such as Kashmir, Lucknow, and Hyderabad).
Although Hindi and Urdu are structurally more or less identical, differences
exist that are related to vocabulary and register (style). Hindi displays a greater
Sanskrit influence in its vocabulary and in its poetic register than does Urdu,
whereas Urdu literature shows stronger influences of Perso-Arabic vocabulary
and of the Persian and Arabic register. Notably, Hindi is written in the Devana-
gari script, whereas Urdu is written in the Arabic Nasta’liq script. The differences
between spoken Hindi and Urdu are socially and, to a certain extent, politically
determined. The speaker is free to over-emphasize or to de-emphasize them. For
instance, it is well known that the famous Indian film industry (known as Bolly-
wood) often prefers Urdu to Hindi for its lyrical songs because Urdu is considered
to be the more poetic language, thanks, for example, to the many Urdu ghazals (or
poetic songs) composed at the courts from the 12th century onwards.
Hindi is the primary official language of India, followed by English, which is
a subsidiary official language. Hindi is a language of instruction in primary and
secondary schools throughout North India (often in addition to English) and it is
also taught at schools in South India. It is a language with grammatical studies,

1 Except for the IE language Tocharian, which has been extinct for over a millennium.
66 Indo-Aryan

dictionaries and extensive literary and other cultural works, including novels,
poems, and diaries. Television programs are mainly broadcast in Hindi through-
out the subcontinent. It is also the language of the bureaucracy in India (again
in addition to English). Likely because of this dominant position in India, Hindi-
Urdu is also the best studied IA language in contemporary linguistics and has
been analyzed in functional as well as generative linguistic frameworks.
Twenty-two regional languages are included in the “Eighth Schedule to
the Indian Constitution”, meaning that they have a certain “official” status. Of
these languages, the following belong to the IA language family (besides Hindi
and Urdu): Asamiya (Assamese), Bangla (Bengali), Dogri, Gujarati, Kashmiri,
Konkani, Maithili, Marathi, Nepali, Oriya, Punjabi, Sanskrit, and Sindhi. In this
study, the Asamiya, Nepali, Kashmiri, and Rajasthani languages will be investi-
gated in detail. In contrast to the first three languages, Rajasthani is not included
in the “Eighth Schedule” because it is not a single language but a group of closely
related languages that are often considered dialects. In the present study, the four
aforementioned languages were chosen to represent particular regions of the IA
language area. They are not necessarily the dominant languages in their specific
regions, but they all present interesting features with respect to alignment, and
most of them are generally underrepresented in linguistic studies of the IA lan-
guage family. The choice for the four regions, i.e. the Eastern, Northern, North-
western and Central region, is arguably an artificial division, geographically as
well as linguistically. Because people move, the geographic borders between the
regions are not strict, and other authors will include languages in other regions
than in the division followed here. Linguistically, languages per region share a
number of features, but so do languages from across borders, depending which
feature is investigated. The overview in Masica (1991: 446–463, see Section 7.2.3)
shows that there is no consensus on a historical classification of the Indo-Aryan
languages (as Masica 1991: 454 notices to the point: “It may seem that just about
every conceivable way of carving up the NIA pile has been advocated by one
scholar or another”). Since this is a synchronic study, I opted for a synchronic
and rather pragmatic division of the languages, based on their commonest
accepted regional position and their linguistic similarities in terms of alignment
patterns. For the Northwestern and the Eastern group, the geographic division is
in line with the linguistic features of the languages; the languages pertaining to
these groups are always classified together. The main problem is with the Central
group, for which it is unclear if Gujarati, Punjabi and the Bihari languages pertain
to it or rather form separate groups. I preferred to include Gujarati and Punjabi in
the Central area, and the Bihari languages with the Eastern group. This classifica-
tion served my purposes in view of the alignment features I discuss. Though it is
a quite common classification, I do not intend this generalization to encompass
 Geographical distribution of the Indo-Aryan languages 67

all linguistic features; quite clearly, languages such as Awadhi display phono-
logically a much greater resemblance to Hindi than to Bangla. Similarly, I have
included Nepali into a separate chapter, again on the basis of its alignment fea-
tures, which are specifically different from Hindi and the Rajasthani languages,
pertaining to the Central group. In other respects, Nepali displays similarities to
Rajasthani and Northern Hindi varieties, and again, for other linguistic features,
it may be included with the Central group.
Asamiya, Bangla² and Oriya are the languages that are traditionally consid-
ered to constitute the Eastern IA branch. Asamiya is the easternmost IA language
and is spoken in the Indian state of Assam. Bangla is spoken in the Indian state
of Bengal and in Bangladesh. Bangla has various dialects; the standard variety
is spoken in and around Kolkata. Oriya is the language of the state of Orissa and
shows many similarities to Bangla. While the area in which these three languages
are primarily spoken forms the eastern border of the area where the Indo-Aryan
languages are spoken, another group of languages represents central Eastern
Indo-Aryan. This group includes Bhojpuri, Maithili, Bihari, and Magahi, as well
as smaller dialects such as Sadri.
In the present study, Northwestern IA is represented by Kashmiri. Kashmiri is
spoken in the state of Jammu and Kashmir in India. Sindhi, Siraiki, and Shina are
other languages spoken in this northern border area between India and Pakistan.
Further south, Marathi is a relatively important language spoken in Southwest
India, and its little sibling Konkani is spoken in and around Goa.
Central India is known as the “Hindi belt” because this is the area where Hindi
and its “varieties” are the most common languages. The western side of the Hindi
belt is the region where the Rajasthani languages are spoken, of which Marwari
has the most speakers. Harauti, Bagri, Mewari and other languages (or variants)
are spoken by fewer people. The language Gujarati is spoken in the southwest,
north of Marathi, but it is historically more closely related to Rajasthani than to
Marathi. Awadhi, Braj, and Bundeli are considered to be closely related to Hindi.
Awadhi belongs to the Eastern Hindi branch, whereas Braj and Bundeli belong to
the Western Hindi branch. Awadhi and Braj were once very important in medieval
literature, but have now been replaced by Hindi and only remain in use in their
colloquial forms.
Because Hindi-Urdu is the predominant language spoken throughout India
and has emerged as the main language in linguistic research in general, the

2 For the names of these languages I have followed Cardona and Jain (2003), who prefer to
use the forms of the names that are used in the language communities themselves and not the
Anglicized versions “Assamese” and “Bengali”.
68 Indo-Aryan

remainder of this Chapter will elaborate on a number of linguistic issues con-

cerning Hindi-Urdu. In Section 2.2 the alignment pattern in Hindi-Urdu is briefly
illustrated. Section 2.3 gives an overview of the historical developments that are
relevant to the alignment patterns of Hindi-Urdu and the other IA languages.
Section 2.4 introduces key concepts in the description of IA (all of which are
illustrated with examples from Hindi-Urdu) such as the layer system of the case
markers (2.4.1), the verb system (2.4.2), experiencer constructions (2.4.3), and
unergative constructions (2.4.4). The focus on Hindi-Urdu in this Chapter is inevi-
table because much of our current knowledge of linguistic patterns in IA comes
from this language.

2.2 Alignment in Hindi³

Hindi has been described as a split ergative language in the literature; more spe-
cifically, it has been described as a language with ergative features only in perfec-
tive verb constructions (cf. Pandharipande and Kachru 1977; Mahajan 1990; Das
2006; Ura 2006; Keine 2007). The cases in Hindi are indicated by postpositions.
The postposition ne is the marking of the ergative. In most grammars or descrip-
tions of Hindi, ko is said to be the accusative or dative postposition because it
marks either O or IO in a clause. It can also mark an adjunct, such as a place or
time indication. Because of this, and because all of its uses are “objectival” (as
opposed to the subject) I prefer the neutral designation “objective” for the ko-
marking and its counterparts in other IA languages, thereby following Beames on
Hindi (1872: 252) and Iggesen (2005: 92) on the cross-linguistic category of objec-
tives. Note that pronouns in Hindi can display an objective form with ko as well
as a synthetic, shortened form (e.g., mujh-e besides mujh=ko). Table 2.1 below
contains all forms of the pronominal paradigm in Hindi.
Hindi is a verb-final language. Except for the subjunctive, imperative and the
future tense, all finite verb forms are periphrastic, consisting of a participle and
a copula. Imperfective constructions include an imperfect (present) participle. A
and S are in the nominative and O is marked with ko only when it is definite and/
or animate, e.g., ex. (21). If the construction is ditransitive, the IO is obligatorily
marked with the objective ko, as in ex. (23). The finite verb agrees with {A, S}.

3 From now onwards, the designation ‘Hindi’ will be used to refer to ‘Hindi-Urdu’ for the sake
of brevity and because most of the examples are derived from Devanagari texts. It should,
however, be understood that every remark is valid for Urdu as well.
 Alignment in Hindi 69

1sg 2sg 3sg 1pl 2pl 3pl

nom maiṃ tū yah, wah ham tum ye, we

obl mujh tujh is, us ham tum in, un
erg maiṃ=ne tū=ne is=ne, us=ne ham=ne tum=ne inhoṃ=ne,
unhoṃ=ne
obj mujh-e, tujh-e, is=ko, us=ko ham-eṃ, tuṃ-heṃ, in-heṃ,
mujh=ko tujh=ko ham=ko tuṃ=ko un-heṃ,
in=ko, un=ko

Table 2.1 Pronominal paradigm in Hindi

(21) rām sitā=ko dekh+rahā+h-ai

Ram sita=obj see+prog+aux-prs.3sg
‘Ram is watching Sita.’

(22) maiṃ so jā-ūṃ-g-ī

I.f sleep go-1sg-fut-f
‘I will go to sleep.’

(23) tum kitāb rām=ko kyoṃ de-tā+h-o?

you.m book[f] ram=obj why give-prs+aux-prs.2pl
‘Why are you giving Ram the book?’⁴

Perfective constructions, which can be recognized by the presence of the perfect

participle, have a different alignment than imperfective constructions. In the per-
fective, A is formally marked by the postposition ne, and the verb agrees with O.
However, this is the case only when that constituent is unmarked. This construc-
tion is illustrated in ex. (24):

(24) maiṃ=ne kītāb paṛh-ī

I=erg book[f] read-pst.f.sg
‘I read a book.’

4 Throughout this study, the present tense in Hindi is either translated with an English simple
present or with an English present continuous, considering that the meaning of the Hindi verb
form hovers between a simple present and a present continuous. However, as this has more to
do with the English translations than with the Indo-Aryan languages per se, I will always opt for
what I consider the most natural translation with respect to the context.
70 Indo-Aryan

(25) rām=ne sitā=ko pahle dekh-ā+th-ā

Ram=erg Sita=obj before see-pst.m.sg+aux.pst-m.sg
‘Ram had seen Sita before.’

(26) maiṃ so ga-yā+h-ūṃ

I.m sleep go-pst.m.sg+aux.prs-1sg
‘I have gone to sleep.’

The perfective pattern illustrated in the above examples is generally described

as being “ergative” in nature. However, only ex. (24) represents an ergative con-
struction that complies with Dixon’s definition of an ergative construction (cf.
Section 1.1.1). In ex. (25), the O sitā=ko is not treated in the same way as the S
maiṃ in ex. (26). Quite on the contrary, sitā=ko is marked, whereas maiṃ is
unmarked. Furthermore, the finite verb dekhā thā in ex. (25) does not show agree-
ment with the feminine O. The only true ergative construction is therefore ex. (24),
in which the conditions are correct: the verb form paṛhī is perfective; the O kitāb
is inanimate/indefinite and hence unmarked; O controls verb agreement; and the
A maiṃ=ne is marked as ergative.
The marking of O in Hindi is often quoted as a typical example of Differential
Object Marking (DOM) (cf. Lazard 2001; Aissen 2003). As discussed in Section 1.1.3,
animacy is an inherent feature of an argument, whereas definiteness is not inher-
ent but contingent (cf. Bossong 1983: 8; Siewierska & Bakker 2009: 292). In Hindi,
the features of animacy and definiteness are of varying importance. It has been
claimed that all animate O’s in Hindi have to take a marker: for instance, accord-
ing to Mohanan (1994: 79), the following example is ungrammatical. In her
opinion (which is the opinion of a native speaker), baccā needs to be replaced by
the marked form bacce=ko (recall that ko marks the objective):

(27) ?ilā=ne baccā uṭhā-yā

Ila=erg child[m] lift-pst.m.sg
‘Ila lifted a child.’

Mohanan’s example (27) has been quoted and repeated many times to illustrate
that the criterion of animacy is stronger than the feature definiteness (Malchu-
kov 2006: 331; 2008: 205; Keine 2007; Klein and de Swart 2011). Nevertheless,
sentences with an unmarked but animate O such as ex. (28) and (29), do occur in
Hindi. Ex. (28) is based on Montaut (2004: 171) who argues that the sentence is
 Alignment in Hindi 71

grammatical because of the indefinite nature of laṛkā. Ex. (29) has in turn been
considered grammatical by different informants.⁵

(28) maiṃ laṛkā dekh+rahā+h-ūṃ

I.m boy[m] see+prog+aux-prs.1sg
‘I look for a boy [to marry my daughter].’

(29) maiṃ=ne kaī aurat-eṃ dekh-īṃ

I.m=erg some.pl woman[f]-nom.pl see-pst.f.pl
‘I saw some women.’

Mohanan (1994: 80–81) herself observes that the situation is in fact much more
complicated than a simple distinction between animacy and definiteness. In par-
ticular, animate O’s can be divided into human and non-human, and non-human
indefinite O’s are often left unmarked, e.g., maiṃ=ne ek gāy dekhī ‘I saw a cow’.
Definiteness, on the other hand, is accompanied by specificity. Indefinite O’s may
be marked if they are specific (i.e., if they have been mentioned before in the
discourse) or else if they are referential, (i.e., if they refer to a particular referent
that “the speaker has in mind”, cf. Heusinger 2011: 2). According to Masica (1986:
125), in many examples of inanimate O’s, the O is considered “non-generic” and
“non-specific” (e.g. kām karnā, lit. ‘work do’, meaning ‘to work’).⁶ Hence, rather
than being a “stronger” factor than animacy, definiteness interacts with animacy.
As Mohanan (1994: 79) points out, the “canonical” marking of an animate O is ko,
and the “canonical” marking of an inanimate O is zero-marking. Therefore, the
majority of transitive constructions with an animate O display overt marking of O,
as is illustrated in the well-known ex. (30) (Mohanan 1994: 104):

5 Mohanan (1994: 108) would consider ex. (28) as an instance of incorporation; because the
O, laṛkā, is generic and indefinite, the example yields the interpretation ‘I am boy-searching’.
However, this is theoretically complicated. Incorporated arguments are not supposed to control
agreement, because they are “verb internal” arguments. In a perfective construction, how-
ever, the verb would agree with laṛkā. Mohanan (1994: 110) therefore argues that the noun
constitutes a lexical compound together with the predicate (cf. Section 2.4.2). For a historical
overview of an assumed noun incorporation process in Hindi, see Klaiman (1990).
6 Note that Masica (1986) prefers the term “Identified object marking” to refer to the mark-
ing of an animate/definite (identified) O in Hindi. He distinguishes four features of objects:
specific/generic and identified/unidentified. However, he does not specify how each of these
categories influence the marking of O in Hindi.
72 Indo-Aryan

(30) ilā=ne bacch-e=ko uṭā-yā

Ila=erg child-obl=obj lift-pst.m.sg
‘Ila lifted the child.’

The fact that a certain construction is canonical does not imply that there are no
exceptions to it. In particular, if an exceptional marking pattern occurs, seman-
tic modifications of O follow. If an animate O in a construction is unmarked, it
normally has a connotation of indefiniteness. Conversely, if an inanimate O is
marked, it must be interpreted as definite. However, to complicate matters, a
definite O without ko is also possible and occurs quite often (Mohanan 1994: 83);
consider the following example from one of my informants:

(31) rām=ne wah ghar banā-yā

Ram=erg that house[m] make-pst.m.sg
‘Ram constructed that house.’

To explain examples such as these, Klein and de Swart’s (2011) distinction

between trigger and effect features may be useful (cf. Section 1.1.3.2). The authors
consider animacy as a trigger of DOM; because animacy is an inherent property of
the argument, every animate O-argument needs to be marked. However, definite-
ness is instead an effect; as a result of the marking of O, O is interpreted as defi-
nite. It turns out that Hindi follows this pattern of trigger and effect. An animate
O is most often marked with ko and only exceptionally unmarked (for instance,
when O is animate but not human, or when it is used in a non-specific, generic
sense). A definite but inanimate O, on the other hand, is only marked when it
explicitly needs to be interpreted as definite. The definite interpretation of the
argument is hence a result of the marking.
Some individual verbs restrict the marking of O to the canonical marking such
as the above-mentioned verb banānā ‘to make’. For instance, the O’s of the verbs
likhnā ‘to write’ and paṛhnā ‘to read’ must be unmarked, whereas the animate O
of pīṭnā ‘to beat’ must take ko (Mohanan 1994: 81)⁷. It is not unthinkable that the
verb uṭānā ‘to lift’, which can occur with an animate or inanimate O (cf. ex. (30)),

7 According to one native speaker the combination of paṛhnā and a marked O is not impos-
sible. He gives the example maiṃ kitāb=ko paṛh-tā hūṃ. In his opinion, this sentence cannot
mean ‘I am reading the book’ but only ‘I read only books [and nothing else]’. Therefore,
although a marked O combined with paṛhnā may be technically possible, there is a certain con-
notation associated with the use of ko that seems to evoke restriction.
 Alignment in Hindi 73

always needs its O to be canonical; that is, an animate O needs to be marked, and
an inanimate O cannot be marked.
I have labeled ko as the marker of the objective case. Mohanan (1994: 91), on
the other hand, argues that ko is a syncretised form of two case functions, viz.
accusative and dative. She rejects the view that ko is the marker of one single
case. She adduces several reasons for her point of view. First of all, most Hindi
speakers apparently prefer to have only one ko-marked argument in a sentence,
preferably the IO, as in the following example from Mohanan (1994: 85):⁸

(32) ilā=ne māṃ=ko baccā/ (?)bacc-e=ko di-yā

Ila=erg mother=obj child[m] child-obl=obj give-pst.m.sg
‘Ila gave mother the child.’

Secondly, a ko-marked O can alternate with a nominative O, whereas ko is obliga-

tory when marking an IO. However, what Mohanan actually shows is that the
grammatical relations O and IO are clearly distinguished in Hindi, not that there
exist separate accusative and dative cases that are expressed by means of a single
syncretic marker. Furthermore, whereas she assumes that there is a separate
dative and accusative case, she maintains rather inconsistently that there is only
one case for an unmarked O and S, which is the nominative (cf. Woolford 1997).
She considers the nominative case of S and A in an imperfective construction
to be the same case as the case of an unmarked O, primarily because the verb
agreement is such that the verb agrees with a nominative argument -irrespec-
tive of whether that argument is S, A or O. However, in imperfective transitive
constructions, the verb only agrees with A and not with an unmarked O. In these
constructions, the unmarked A and O would then have a different case. There are
further objections to the view defended by Mohanan that result from a compari-
son with agreement patterns in other IA languages. Other IA languages do not
show the same preference as Hindi for agreement with an unmarked argument
(see Comrie 1984 and Chapter 6). In order to avoid such inconsistencies, I prefer
to distinguish clearly between cases and grammatical relations and to maintain

8 Note that Mohanan here draws on a strictly formal preference for not using ko twice in a
single clause. If a sentence similar to ex. (32) is constructed with a pronominal O, the pronoun
will take the pronominal objective form ending on -e, as, e.g., in the admittedly somewhat odd
sounding but not unconceivable example: ilā=ne māṃ=ko mujhe diyā ‘Ila gave me to mother’
(or, ‘Ila gave mother to me’). Moreover, the single use of ko is also a matter of preference of
the speaker, at least to a considerable extent. Some of my informants tell me that they have
no problems whatsoever with sentences with a double ko marking such as ilā=ne bacce=ko
māṃ=ko diyā from ex. (32).
74 Indo-Aryan

one label for one case; viz. ko is the objective case which may mark O as well as
IO.⁹ In this respect, I follow Lazard (2002), who maintains the asymmetric view
that one form can have different meanings, but that different forms can express
different meanings.
In fact, cross-linguistically, it is not unusual that in languages with DOM (such
as Hindi) the IO is marked in the same way as a definite/animate O. The occur-
rence of ko marking a definite/animate O and the IO is accounted for by Kittilä
(2006, cf. also Malchukov 2008). He suggests a typology of object-, animacy- or
role-based strategies for object marking. A language following an object-based
strategy displays a pattern in which the O and the IO of a ditransitive clause are
marked in the same way (this pattern is also known as “neutral” alignment, cf.
Haspelmath 2005). In a role-based strategy, the language marks IO differently
from O. This is the “indirective” alignment known from many Standard Euro-
pean languages and it runs parallel to the distinction between the accusative
and dative case. The third strategy of marking objects is animacy-based. In this
pattern, the marking is determined according to the semantic features found in
DOM. Interestingly, according to Kittilä (2006: 14), languages with DOM are open
to what he calls EDOM (“Extended Differential Object Marking”). In EDOM, the
use of the marker of O has been “extended” to mark the function of the IO as well,
on the basis of the same criterion of animacy. Because an IO expresses the seman-
tic role of a recipient or beneficiary, it is almost always animate and marked by
the marking of O. The marking becomes generalized to each and every IO in a
language with EDOM. However, although Kittilä’s hypothesis is very convincing
from a synchronic point of view, it is at odds with the attested historical develop-
ments, which all indicate that the marking of an IO occurred first and the marking
of a definite/animate O followed only later. According to Lehmann (2002: 97), a
dative case marker may further develop into an accusative case marker, but the
reverse development is unlikely. Nevertheless, the synchronic principle of EDOM
makes it clear that the marking of a definite/animate O and the marking of an IO
(i.e., the objective marking) are based on the semantic properties of animacy and
definiteness. The function of this marking is the “indexing” of the arguments, not
merely “distinguishing” the arguments from each other (cf. Chapter 1).

9 Of course, Mohanan’s distinction is determined by the generative framework she adopts, in

which universal case features are subject to strict rules.
 Origin of the ergative pattern in Indo-Aryan 75

2.3 Origin of the ergative pattern in Indo-Aryan

The origin of the ergative pattern in Hindi has been extensively discussed in the
literature, mainly because IA is one of the few language families in the world for
which there are literary sources in practically every language stage.¹⁰ The Sec-
tions 2.3.1–2.3.3 give an overview of the various stages in the development of IA
that have ultimately led to ergativity in Hindi, and of their treatment in the litera-
ture. Note that most diachronic accounts focus on Hindi as the “end point” of the
transition process (but see Khokhlova 1992 and Montaut 2004, 2009, who also
consider other Central Indo-Aryan languages).

2.3.1 Historical overview

By convention the evolution of the Indo-Aryan languages is divided into three

stages, based on the grammatical differences one can discern in each of the
periods (Masica 1991: 50):
1. Old Indo-Aryan (OIA): 1500 BCE – 600 BCE
Vedic, Classical Sanskrit
2. Middle Indo-Aryan (MIA): 600 BCE – 1000 CE
Prakrit, Apabhramśa
3. New Indo-Aryan (NIA): 1000 CE –
Hindi, Bangla, Nepali etc.

2.3.1.1 Old Indo-Aryan

Old Indo-Aryan (1500–600 BCE) had a rich system to express tenses and case
functions (Whitney 1896). The past tense only was divided into a perfective and
imperfective past, an aorist, and an analytic construction with a perfect partici-
ple. In ex. (33), the past tense is a synthetic verb form expressing perfect aspect
which consists of a conjugated form of the reduplicated stem ca-kar, i.e. ca-kār-a.

(33) devadatta-ḥ kaṭa-ṃ ca-kār-a

Devadatta-nom mat-acc prf-make-3sg
‘Devadatta (has) made a mat.’

10 Parts of this Section are based on a revision of Verbeke and De Cuypere (2009).
76 Indo-Aryan

However, another construction becomes an increasingly popular way to express

a perfective from the Classical Sanskrit period onwards. This is the periphrastic
construction with a perfect participle. It is generally acknowledged that the erga-
tive construction in Hindi originated from this construction (cf. Pray 1976; Ander-
son 1977; Comrie 1978: 371; Dik 1978; Estival and Myhill 1988; and Dixon 1994:
190). An example is offered in (34).

(34) devadatt-ena kaṭa-ḥ kṛ-ta-ḥ

Devadatta-ins mat-nom.sg make-prf.ptcp-nom.sg
‘The mat is made by Devadatta.’

The participle in this construction has the ending -ta. The agent, if overt, takes an
instrumental case (devadatt-ena), and the patient is in the nominative (kaṭaḥ).
It is also the patient argument which controls the agreement. The patterning of
the arguments of the participle is the same as in a passive construction, which
explains why the participle is often considered to have a passive meaning. In
transitive periphrastic constructions, the participle is mostly interpreted as a
passive, although an active interpretation is not excluded. A common verb that
allows for the two interpretations is pitā, the participle of ‘drink’, which can mean
‘having drunk’ as well as ‘being drunk’ (cf. Stronski 2009: 78; Speijer [1886] 1998:
280–281; Peterson 1998: 189). The ta-participle has an active counterpart which
is formed by adding the suffix -vat to the participle on -ta, yielding, e.g., kṛ-ta-vat
‘having made’. This active participle is only occasionally used, but it is far less
frequent than the standard ta-participle, which is amenable both to an active and
passive interpretation (Klaiman 1978; Hock 1986).
The ta-participle does not occur exclusively in periphrastic constructions. In
OIA, it can also be used as an adjective and in an attributive function, as in the
following example (35):

(35) ha-ta-ḥ vyagra-ḥ

kill-prf.ptcp-nom tiger-nom
‘the killed tiger’

Note that if an instrumental agent is added to the construction in ex. (35), then
one has the same perfective construction as in ex. (34).
Formally, the ta-participle is based on the bare root of the verb (Whitney 1896:
203, 340) to which the suffix -ta is immediately added. Although the participle
does not have any formal marker of tense or mood, it refers to a perfective event.
This means that the participle is most often interpreted as designating something
that happened in the past, although it can have a “present value” (Hock 1986: 16);
 Origin of the ergative pattern in Indo-Aryan 77

that is, it can convey the meaning of a past action that still has a bearing on the
present. Intransitive participles are active, in the sense that they take a nomina-
tive agent; for instance, in ex. (36) devadattaḥ is in the nominative and the action
expressed by the participle is actively carried out by the nominative argument:

(36) devadatta-ḥ ga-ta-ḥ

Devadatta-nom go-prf.ptcp-nom
‘Devadatta has gone.’

This participial construction, which is considered to be passive if the participle

is transitive, is not the only “passive” construction in OIA, and certainly not the
most exemplary passive. The “true” passive is a synthetic construction in OIA.
Every verb has an active as well as a passive synthetic conjugation. Ex. (37) is an
example of such a synthetic passive construction in OIA:

(37) devadatt-ena kata-ḥ kri-ya-te

Devadatta-ins mat-nom make-pass-prs.3sg.pass
‘The mat is being made by Devadatta.’

The synthetic passive is formed on the basis of a present or past tense root and
functions as the passive counterpart to the active present and past tense construc-
tions. Note that passive constructions are not limited to transitive verbs in Old
Indo-Aryan, intransitive verbs can also be passivized, such as ex. (38) (Klaiman
1978: 205):

(38) t-ena śāy-ya-te

he-ins lie-pass-prs.3sg.pass
‘There is being laid down by him.’ > ‘he lies down.’

The argument structures of ex. (34) and ex. (37) are similar. In both construc-
tions, there is an agent in an instrumental case and a patient in the nominative,
and the latter is the argument that controls the agreement. However, the perfect
participle only agrees with the nominative in gender and number, which is also
what one expects from a nominal participle, whereas the synthetic passive verb
not only agrees in gender and number with the nominative patient-subject, but
also in person.
In later texts in Classical Sanskrit, there is a tendency for perfectives to
become more and more expressed by means of the ta-participle construction
(Gonda 1951: 107–108). One can also note that the frequently used formal pattern
of a ta-participle transitive construction is extended to an intransitive construc-
78 Indo-Aryan

tion, as in ex. (39). In this example, the agent of the action is expressed with an
instrumental case (mayā), just like the agent-object of a passive:

(39) mayā ga-ta-ṃ

I.ins go-prf.ptcp-nom.n.sg
‘It is gone by me.’ > I went

The normally expected form would be ahaṃ gataḥ, in which ahaṃ is the nomina-
tive form of the pronoun, and with the verb gataḥ agreeing with ahaṃ in gender
and number. However, constructions like ex. (39) seem to have remained rather
exceptional in OIA (cf. Klaiman 1978: 213).

2.3.1.2 Middle Indo-Aryan

All Middle Indo-Aryan (MIA) languages share features on all levels, from phonol-
ogy to morphology and syntax. In general, there is phonological and morpho-
logical simplification in MIA as compared to OIA (Pischel 1900). On the other
hand, there are also some new forms added to MIA which were not yet present
in OIA; for instance, certain case-endings were restricted to pronouns in Sanskrit
but are being used with common nouns in Middle Indo-Aryan. One of the “sim-
plifying” measures is the decrease in the number of different verbal forms in MIA.
It is interesting to note that the means to express a perfective aspect are getting
restricted to one form only, viz. the construction with the ta-participle:

(40) iyaṃ dhaṃmalipi lāj-inā

this.nom true inscription[n]nom.sg king-ins
likhā-pi-tā
write-caus-prf.ptcp.nom.m.sg
‘This true inscription was caused to be written by the king.’

In this example from the 2nd century BCE, the perfect participle agrees in gender
and number with the patient and the agent takes an instrumental case. This is the
common way to express past tense in MIA. However, an active interpretation of
the sentence is also possible. In that case, the interpretation would be: “The king
caused this true inscription to be written”.
Apabhramsha, or Late Middle Indo-Aryan, displays an even more radical
case syncretism. This is a fascinating period to investigate the development of
the ergative construction, in particular because it is a period in which the massive
case syncretism in the history of MIA culminates. The system of seven cases found
in Sanskrit is reduced to a system of three syncretic cases, due to the merging of
 Origin of the ergative pattern in Indo-Aryan 79

i) the nominative and accusative, ii) the instrumental and locative, and iii) the
ablative, genitive, and dative (Bubenik 1998: 66; De Clercq 2003: 2110, 2010). The
following examples illustrate the alignment in Late MIA (Bubenik 1998: 148).

(41) laddh-a tuhuṃ maiṃ im-aṃmi van-aṃmi

find-prf.ptcp.nom you.nom I.ins this-loc wood-loc
‘You are found in this forest by me.’ > ‘I have found you in this forest.’

(42) tā keumai-eṃ hauṃ ghar-aho nī-ya

then Ketumaki-ins I.nom home-obl take-prf.ptcp.nom
‘Then I was brought home by Ketumaki.’ > ‘Then Ketumaki brought me
home.’

In each of these examples of Late MIA the participle agrees with the patient in
gender and number. The first person pronoun maiṃ is the instrumental agent
in ex. (41), in ex. (42) hauṃ is the nominative patient. Intransitive sentences (for
instance ex. (43)) are constructed with a perfect participle and an agent in a nomi-
native case (De Clercq 2003).

(43) salil-u pavaḍḍhi-u sīẏa-hĕ ṇām-eṃ

water[m]nom.sg rise-prf.ptcp.nom .m.sg Sita-gen fame-ins
‘Because of Sita’s fame, the water rose.’

The participle and the nominative salilu agree in number and gender. Note
however that the nominative in the paradigm has merged with the accusative.
Bubenik (1998: 142) points out that in Late Middle Indo-Aryan intransitive verbs
take an agent in the nominative case; however, the nominative case is also used
as the case of the patient, in a transitive construction with the perfect participle.
The verb then agrees with this nominative patient, and the agent is expressed by
an instrumental. Because of the case syncretism and the fact that nominative and
accusative have merged, it is impossible to analyze the agent as either a nomina-
tive subject or an accusative direct object. Bubenik interprets this finding as the
“appearance of the absolutive case”, given that the absolutive is the case that
can mark the object as well as the subject of an intransitive sentence. Therefore,
Bubenik considers the period of Late MIA as crucial in the evolution of the erga-
tive construction in the history of IA, and he calls the transitive participial con-
structions from MIA “semi-ergative” or “semi-passive” (1989: 389).
80 Indo-Aryan

2.3.1.3 New Indo-Aryan: Early Hindi

In the period of Early New Indo-Aryan the changes that began to occur in Late
Middle Indo-Aryan continue. The case syncretism goes one step further, until
only two cases are eventually left over: the nominative and the oblique.
In the following two examples of Early New Indo-Aryan taken from Drocco
(2008) the verb agrees with the patient.

(44) maiṃ itane dina t-eṃ āgaran-a ki-y-o

I so much day that-obl waking[m]-nom.sg do-prf.ptcp-m.sg
‘I stayed awake for so many days.’

(45) rājā bibeka dvārapāla=kauṃ ājñ-ā kar-ī

king bibeka guard=dat order[f]-nom.sg. do-pst.f.sg
‘King Bibeka ordered the guards.’

In ex. (44) āgarana agrees in number and gender with the verb kiyo. The personal
pronoun maiṃ is the only form left of the original, much more elaborate para-
digm of personal pronouns. In this example it takes the agent role, but formally it
is unclear what the case of the pronoun is, given that it can either be an oblique
or a nominative. Ex. (45) presents the same pattern: ājñā agrees in gender and
number with the verb karī, whereas rājā bibeka is standing on its own, unaffected
by the verb and not capable to control verb agreement. Hence, in Early New IA, a
rather ambivalent situation seems to arise, in that one cannot determine whether
a perfective construction instantiates an accusative passive pattern or an ergative
pattern.

2.3.2 Passive reanalysis

The development described in the previous Section 2.3.1 has been interpreted
in various ways. One of the most common explanations is based on the view
that ergative constructions are to be conceived of as (some kind of) passives.
The hypothesis that an ergative construction is in fact a special type of passive
construction originated in Uhlenbeck’s time, and it was picked up in several
early grammars of Hindi (e.g., Kellogg 1938; Beames 1872). In Kellogg’s analy-
sis (1938), the ergative case is labeled as the case of the ‘Agent’. He recognizes
that the patient of the action is put in the nominative case when it is used with
a “Perfect Participle of Transitive verbs”. He calls this construction a “Passive
construction”, which proves that Kellogg endorses a passive interpretation of the
construction in question.
 Origin of the ergative pattern in Indo-Aryan 81

In more recent research, the idea that an ergative construction is passive in

nature has been abandoned, as was discussed in Chapter 1 (cf. Comrie 1978;
Dixon 1979). However, because of the IA and Indo-Iranian historical descrip-
tions, in which the passive is assumed to be the origin of the ergative construc-
tion, the view that ergative constructions are essentially derived from a passive
construction still has its defenders. Anderson (1976), for instance, interprets the
rise of ergative marking in Hindi as the most important example of the so-called
“passive to ergative” evolution (see also Pray 1976; Estival and Myhill 1988). The
well-known functional approach to ergative languages by Simon Dik (1978: 157)
is another offshoot of this passive-as-ergative view. Dik assumes that “ergative
languages” develop from nominative languages. He distinguishes three stages in
the transition of a nominative to an ergative language.¹¹ In the first stage, there
is a functional markedness opposition between the active and passive construc-
tions. According to Dik (1989: 38), the semantic markedness of a construction is
determined by the “expectedness” and the “frequency” of the construction. In
the opposition between passive and active, the passive is the marked construc-
tion because it occurs less frequently than the active, and because it is safe to
say that its occurrence is not generally expected. However, markedness is, in his
view, not “a fixed, immutable property […] It may vary with the environment in
which it is used, and with the frequency with which it recurs” (Dik 1989: 41).
This explains why, in the second stage of the development from nominative to
ergative, the passive construction eventually becomes the functionally unmarked
construction in the language (i.e., the most frequent construction). At this stage,
the opposition between the active construction and the passive construction is
neutralized by the passive construction; that is, the functionally unmarked con-
struction. Finally, in the third stage, the former passive construction has acquired
the status of an active construction that is called “ergative” by Dik. The language
is at this moment considered to be ergative and no longer possesses a passive con-
struction. A possible fourth stage begins when a new passive construction arises,
which is then predicted to have a periphrastic form.
It is remarkable, and perhaps not coincidental, that the evolution path Dik
assumes strongly resembles the traditional idea of the origin of ergativity in IA.
Dik’s approach has been applied by Bubenik (1998: 133) to the evolution of the
different language stages in Indo-Aryan (Figure 2.1). Stage 1 can be identified
with OIA, viz. an accusative language with a marked passive construction (the
ta-construction). Stage 2 is exemplified by MIA, viz. an accusative language with

11 Note that Dik refers to “ergative languages”. As explained in Chapter 1, it is more consistent
to speak of “ergative constructions” rather than “ergative languages”.
82 Indo-Aryan

an unmarked passive (the ta-construction has evolved to the standard means for
expressing the perfective). Finally, stage 3 is NIA, when the ergative has taken
over. Hindi seems to have reached stage 4, given that it possesses a new analytic
passive that is constructed with the verb jānā ‘to go’ (see Section 2.4.2).

obsolete unmarked marked

OIA akārṣat ‘he did’ (Active Aorist) tena kṛtam ‘it was done by him’
↙ ↙ (Passive)
MIA akāsi ‘he did’ tena kaṭa ‘it was done by
him’/’he did’ (Passive/Ergative) ↙
NIA us=ne kiyā ‘he did’ (Ergative) us=se/us=ke dvārā kiyā gayā ‘it
was done by him’ (Passive)

Figure 2.1 The pathway of the reanalysis according to Bubenik (1989: 379)

Although superficially the different language stages in the history of IA seem to

correspond to Dik’s hypothetical stages, Bubenik makes some important remarks
after applying Dik’s theory on Indo-Aryan. First of all, it is important to note that,
contrary to what Dik predicts, the passive construction is attested in every stage
of the development of IA. Moreover, Bubenik correctly points out that the analytic
jānā-passive started to appear in MIA at a time when the language was still clas-
sified as being in the second stage (much earlier than one would expect on Dik’s
view) (Bubenik 1989: 380). The new analytic passive gradually replaced the older
synthetic one. Stage three, when there is an ergative construction but no longer
a passive, is therefore not attested in the history of the IA languages. Further-
more, the synthetic passive of OIA was also still productive in MIA, and to date a
form of a synthetic passive appears in modern IA languages, such as Rajasthani
(Khokhlova 1995).
Second, the rigid opposition between active and passive that Dik assumes
with his semantic markedness-based approach¹² is not attested in the IA lan-
guages. From OIA onwards, the ta-participle appears in a passive construction;
however, as was already noticed by Whitney (1896: 203, 340), it does not have a
tense marker and is not in opposition to any other tense or voice. For this reason,
Whitney places the participle outside of the tense system of the OIA verbs.
Burrow (1965: 354, cf. Bynon 2005: 7) classified the ta-construction as “Prakrit

12 Recall that Dik explains markedness primarily in terms of frequency and as determined by
the “environment” in which the (un)marked element occurs.
 Origin of the ergative pattern in Indo-Aryan 83

in disguise”, implying that it is more adequate to regard the ta-construction as

a construction resulting from a sociolinguistic preference than on the basis of a
grammatical “gap” in the voice paradigm. The passive counterpart of every active
tense form is the synthetic passive in OIA. If there had been a markedness shift
between passive and active, it would have taken place between the participle on
-ta and the participle on -tavat. However, the tavat-participle was never as fre-
quent as the ta-participle in OIA. Thus, if it is true that markedness is based on
frequency, it seems that the ta-participle, rather than the tavat-participle, was the
unmarked member of the opposition.¹³
Although Dik’s hypothesis is essentially a diachronic one, he also recognizes
a similarity between passive and ergative from a synchronic perspective. For
instance, he argues that, cross-linguistically, the instrumental case of the passive
agent is often formally identical to the ergative case in a language. He maintains
that the ergative A is regularly omitted and that the same happens to the agent of
a passive construction. However, neither observation holds true for Hindi. Firstly,
the ergative postposition ne is formally different from the ablative postposition
se, which marks the agent of a passive construction. Secondly, it is true that the
ergative A is occasionally omitted in Hindi, but this is true for every core argu-
ment in Hindi; for example, (46).¹⁴

(46) tum=ne khānā khā-yā? – haṃ, khā-yā+h-ai

you=erg food[m] eat-pst.m.sg yes eat-pst.m.sg+aux-prs.3sg
‘Have you eaten? – Yes, I ate.

2.3.3 Ergative remains ergative?

The central thrust of the “ergative” approach to the development of alignment in

IA is that the perfect participle construction has no active counterpart. A conse-
quence of this viewpoint is that the perfect participle construction is not consid-
ered to be a passive construction. Such an approach also keeps in mind that the
instrumental case is lost in Late MIA and that, as an agent marker, it is only much
later replaced by a postposition in Hindi.

13 In general, cross-linguistically, passives are supposed to occur less frequently than their
active counterparts (Haspelmath 1990).
14 As a rule, the fact that an ergative A is not overtly expressed is no reason to assume that
the ergative construction is passive. As Haspelmath (1990: 27) points out, this finding only
indicates that the ergative A lacks certain “subject” properties.
84 Indo-Aryan

Although scholars agree that the OIA ta-participle construction is not passive,
specific accounts vary with respect to the question of what is the exact point in
time at which the perfect participle construction can be analyzed as an ergative
construction. Various authors claim that the ta-construction is not entirely erga-
tive but rather “semi-ergative” or “ergative-like” (Bubenik 1998; Peterson 1998;
Bynon 2005). On the other hand, Klaiman (1978), Hock (1986) and Hook (1991)
claimed that the OIA perfect participle construction must have been ergative from
the beginning; and, that the ergative argument must be the subject of the con-
struction.¹⁵ These authors explicitly raised the question as to the “subject proper-
ties” of the instrumental agent of the perfect participle construction. However,
determining subject properties is a delicate issue (cf. Section 1.1.2), and it is there-
fore not surprising that different conclusions have been drawn.
Hook (1991: 178) draws attention to the fact that the instrumental agent is
often in the subject position, i.e. the first position, in constructions with the par-
ticiple. He gives the following example from a text from the 2nd or 3rd century BCE
(OIA) (cf. Hook 1991: 178):

(47) na hi pāṇin-īna śabd-āḥ prok-tāḥ

not for Panini-ins word-nom.pl teach-prf.ptcp.nom.pl
‘For it was not words that were taught by Panini.’

However, Sanskrit word order is relatively free, so the claim that the instrumen-
tal agent is considered the subject is controversial, as Hook himself points out.
His second argument in favour of the hypothesis that the instrumental agent is a
subject is that the instrumental case often also serves as the case of the agent of
a converb (or “conjunctive participle”) in a clause. This is illustrated in ex. (48)
(Hook 1991):

(48) tataḥ śabd-ād abhijñā-ya sa vyaghr-eṇa ha-ta-ḥ

then word-abl recognize-cvb he.nom tiger-ins kill-prf.ptcp-nom
‘The tiger recognized him by his voice and killed him.’

The instrumental argument vyaghreṇa is considered as the agent of the converb

abhijñāya, and indirectly also as the agent of the main verb hataḥ. However, in
Whitney’s Sanskrit grammar we find that in constructions with a converb, the

15 Note that these latter authors do not necessarily endorse the view that the ergative
construction in OIA is identical to the one in NIA; they simply assume that OIA already had an
ergative construction.
 Origin of the ergative pattern in Indo-Aryan 85

participle is always “qualifying the actor whose action it describes” (Whitney

1896: 355). In other words, this “actor” can be the nominative patient or the
instrumental agent, but also the dative indirect object, viz. whatever argument in
the clause that is semantically amenable to be the actor of the action expressed
by the converb.
Hook’s third and main argument draws on the behaviour of the reflexive
pronoun, which at a certain point in time turns to take the phrase in the instru-
mental case as antecedent instead of a nominative phrase. This is considered
proof that the subject property of being an antecedent of a reflexive pronoun has
been transferred from the nominative patient to the instrumental agent.
Hook concludes that the ta-construction in Sanskrit turned into an ergative
construction very early in its evolution, if it was not ergative from the very start.
Formally, the instrumental case does not change in any way when it is used as A
or in a purely instrumental sense. Hence, the observation that there is no proper
form of an ergative case in OIA does not contradict the claim that the construction
is of an ergative nature.
Hock (1986) regards the ta-construction in Sanskrit as “patient-oriented”. In
Hock’s view, the patient is grammatically in focus in Sanskrit, in the sense that it
is the object which takes the nominative case and agrees with the verb. However,
Hock assumes that the oldest forms of Sanskrit were “agent-oriented”. Rather
than a change from passive to ergative, Hock suggests that a transition took place
from a patient-oriented to an agent-oriented construction. This hypothesis takes
into account the difficulty that one and the same structure can be interpreted as
passive or as ergative. Hock emphasizes that the gerundival construction in OIA,
which consists of a gerundive on -tavyā and an agent in a genitive, is ergative.
Moreover, in a later phase of OIA, the instrumental case can as well be used for
the agent of a gerundival construction. According to Hock, the gerundival con-
struction plays a crucial role in the development towards an ergative construc-
tion in OIA, and he considers the gerundival construction itself as already being
ergative in nature. Peterson (1998) however rejects Hock’s analysis and points out
that the gerundival construction does not accord to the definition of an ergative
construction because in the gerundival construction it is S and A that are in the
same case, not S and O. Intransitive gerundives are also combined with a geni-
tive/instrumental agent.
The view that ergative constructions already existed in Sanskrit is most force-
fully advocated by Klaiman (1978). According to Klaiman (1978: 205), there can be
no doubt that the Sanskrit examples offered in (49) and (50) display an ergative
patterning:
86 Indo-Aryan

(49) nar-eṇa ved-āḥ paṭhi-tāḥ

man-ins veda-nom.pl read-prf.ptcp.nom.pl
‘The Vedas were read by the man.’ > The man read the Vedas.’

(50) sa śayi-ta-ḥ
he.nom.sg lie-prf.ptcp.-nom.m.sg
‘He lay down.’

Klaiman’s interpretation draws on Dixon’s definition of ergativity: O, ved-āḥ, and

S, sa, are in the same case and in both sentences the verb agrees with the nomina-
tive argument. The A, nar-eṇa, however, takes another case marking (instrumen-
tal) and does not agree with the verb. O and S are thus treated differently from A.
Klaiman’s argument is based on the agreement between the participle and O, this
type of agreement being a determining feature of ergative languages. However, a
second feature of ergative languages, viz. the ergative case marking of the subject,
is less prominent than O-agreement in OIA. In the above examples there is no
marking of the subject that can be analyzed as being exclusively ergative. Stron-
ski (2009) points out that Klaiman (like Hock) does not take into account the use
of an instrumental agent with an intransitive verb in her account of the ta-con-
structions in OIA (for instance mayā suptaṃ ‘by me it is slept’, mayā praviṣtaṃ
‘by me it is entered’). The intransitivity of the latter verb praviṣtam may be subject
to some discussion (after all, something can be entered), but it is nonetheless
clear that Stronski’s argument is directed against an ergative interpretation of the
ta-construction in OIA, in which S is not treated in the same way as O but is more
similar to A. Other accounts, such as Bynon (2005), follow Benveniste (1952), who
argued for a possessive interpretation of the participle construction. Benveniste
(1952) studied Old Persian and focused on constructions such as Old Persian ima
tya manā kartam, lit., ‘this is done of me’ (meaning ‘what I have done’). Accord-
ing to Benveniste, this is not a passive construction but a possessive one in which
manā is the genitive agent, so that the entire expression is semantically similar
to Latin quod mihi factum (est), lit. ‘what is done with respect to me’ (meaning
‘what I have done’) (Debrunner 1954: 937), in which mihi is traditionally regarded
as a “dativus auctoris” (Montaut 2009, 1996). Considering the close relationship
between Old Persian and OIA, a similar evolution from a possessive to an ergative
similar could be assumed for OIA. However, the main problem with considering
the OIA ta-participle construction as possessive is that the agent is normally not
expressed in the genitive case but in the instrumental case. As pointed out by
Cardona (1970), the instrumental is the typical case of the passive agent-object
in OIA. If one were to argue that the ta-participle construction is possessive, one
would have to assume that the original agent was a genitive and that the instru-
 Origin of the ergative pattern in Indo-Aryan 87

mental is a later innovation. This is what Bynon (2005) suggests, yet the historical
evidence does not support such an approach: the instrumental is attested much
earlier than the genitive as the case marking of the agent in constructions with the
ta-participle (Cardona 1970; Stronski 2009). It seems that Cooreman et al.’s (1984:
29) remark for Tagalog ergative constructions also applies to the ta-construction
in OIA, viz. that “[t]here is nothing structurally distinct in the […] ergative clause
that could tell us whether it is functionally a passive or ergative clause type.”
From the above discussion, it may be concluded that the ergative nature of
the OIA ta-construction is to a large extent a matter of interpretation. However,
still another analysis has proven to be cross-linguistically fruitful. Peterson (1998)
argues that the participial construction should not be considered as passive or
“ergative” but as primarily expressing a resultative meaning. Peterson (1998)
analyzes the perfect participle construction in MIA Pali. He concedes that the
perfect participle construction in Pali has passive features but he argues that
this is because the construction is derived from an OIA resultative construc-
tion that “describes the state which results from the action denoted by the verb
stem, and not merely the action itself” (Peterson 1998: 24). Peterson considers
the development as a transition from a resultative, adjectival construction in OIA
into an (ergative?) verbal construction in MIA. The transition was complete from
the moment that the participial verb form was no longer seen as an “adjective
describing a state” but as a perfect verb form expressing an action (Peterson 1998:
194–195). Based on this view, ex. (36) from OIA, repeated below as (51), should be
translated by means of the English copula ‘to be’, because the participle gataḥ
only describes a state.

(51) devadatta-ḥ ga-ta-ḥ

Devadatta-nom go-prf.ptcp-nom
‘Devadatta has/is gone.’

However, Peterson (1998: 23) interprets the following example from Pali as a
verbal construction expressing perfect aspect. He analyzes the agent so as a core
argument of the verb and not as an argument determined by an adjectival (parti-
cipial) form such as devadattaḥ in ex. (51):

(52) so an-āga-to
he.nom not-come-prf.ptcp.nom
‘He has not come.’

Creissels (2008) argues that it is problematic to interpret a historical periphrastic

construction as a passive or a resultative (compare the participle in OIA (51) and
88 Indo-Aryan

MIA Pali in (52)). Not surprisingly, in the various accounts of the ta-participle
in IA, we find different interpretations of this participle. According to Creissels,
nominal verb forms can be the source of a passive and also be “involved in TAM
periphrases”:

[T]he mere fact that a resultative form can be identified in the source of an ergative construc-
tion is not sufficient to conclude that the source construction was passive. The confusion
between resultative (forms) and passive (constructions) is probably the reason why so many
authors have overestimated the importance of passive constructions as the source of erga-
tivity, and neglected the possibility of changes in alignment patterns induced by the gram-
maticalization of aspectual, temporal or modal periphrases. (Creissels 2008: 20)

Peterson (1998: 198) embraces the idea that the resultative construction led to a
periphrastic tense in MIA and draws parallels with developments in Roman and
Germanic tense systems. The Germanic perfect has been explained by Bybee and
Dahl (1989, Creissels 2008) as originating either from a resultative construction
(e.g., German er ist gefahren ‘he has driven [a vehicle]’) or from a possessive con-
struction (e.g., German er hat das Auto gefahren ‘he has driven the car’). Peterson
argues that the ergative construction in IA has a similar origin; that is, the MIA
intransitive ta-participle construction derives from a resultative construction (cf.
ex. (52), Trask 1979), whereas the transitive ta-participle construction comes from
a possessive construction (as argued much earlier by Benveniste 1952). Hence,
Peterson still needs to account for the instrumental case of the agent, which is
not a case typically associated with a resultative or a possessive construction (cf.
Cardona 1970). In Peterson’s view, the first ta-participles used to convey perfect
aspect in OIA must have been intransitive. At the time when transitive verbs took
the participle form, the instrumental agent was an instrumental of “concomi-
tance” that resulted in constructions such as ‘with me is a written book’ (Peterson
1998: 211). Now, if it is conceded that the concomitant function of the instrumen-
tal is close to the possessor construction of genitive agents (e.g., ‘I have a book
written’), then this could explain why the genitive can also be used as an agent of
a ta-participle construction. Although Peterson’s analysis has the disadvantage
that he only applies it to Pali (cf. Stronski 2009), the analysis provides a possible
and plausible pathway of change, in particular in its view that the ta-participle
expresses a resultative and not a passive meaning. Moreover, Peterson also shows
that the resultative which first developed into a perfect, will ultimately evolve into
a general past or perfective as was the case in Hindi (see ex. (24)). The same devel-
opment has been demonstrated by Bybee and Dahl for English (1989), adding a
cross-linguistic precedent for this kind of transition.
 Some key concepts of the Indo-Aryan languages 89

2.4 Some key concepts of the Indo-Aryan languages

Indo-Aryan languages form one language family, which means that certain prop-
erties recur in every language of the family. This Section introduces a number of
these basic syntactic features. Obviously, the properties do not recur in exactly
the same way in every language. The properties relating to alignment will be dis-
cussed for each particular language separately, but in this Section 2.4 they are
first briefly explained using Hindi as an example. Some of the matters are as yet
unsettled in linguistic debate, but it is not the goal of this study to discuss these
debates in detail.

2.4.1 Masica’s layer system

The historical development of the IA languages led to a system of case marking

that is built on postpositions. The case system in IA was described by Masica
(1991), Butt and King (2005), and Schmidt and Kohistani (2008), among others.
Masica (1991: 231) distinguishes three layers of markers in IA. Layer I markers are
the inflectional cases. The nominative is the primary inflectional case, with a zero
marking. In Hindi, the non-nominative, or the oblique, is the second case inflec-
tion of layer I. For instance, one finds in Hindi the nominative laṛk-ā ‘boy’ and
the oblique laṛk-e. A lexical item in the oblique almost never occurs on its own
in Hindi; as a rule, the oblique form is followed by a postposition, e.g., laṛk-e=ke
sāth ‘with the boy’.¹⁶ Postpositions form layer II or III.¹⁷ Layer II are simple post-

16 Occasionally, one finds an oblique form without postposition in locative expressions, for
instance ḍakṣan-e ‘to the post office’ (Butt and King 2005: 12).
17 All postpositions are considered to be clitical forms in Hindi and not affixes like the layer I
inflectional markings. Butt and King (2005, following Mohanan 1994) make a different distinc-
tion between layer II markers (which they call “clitics”) and layer III markers (which they label
“postpositions”). They argue that layer II markers are case clitics on the basis of the observa-
tion that they have scope over more than one noun, for instance, an adjective combined with
a noun followed by a “clitic” will take the oblique case. Moreover, the focus clitic hī can be
inserted between the noun and the “clitic”. While it is true that these are all features of clitics,
layer III markers share the same features. Butt and King’s main reason for considering layer
III as postpositions, not as clitics, is that they are formally more complex and cannot mark a
subject. However, this opposition between “clitics” and “postpositions” with respect to layer II
and layer III markers seems to be farfetched, considering that there are more similarities than
differences between the two sets of markers. For instance, layer III postpositions display all the
properties of free clitics. Moreover, Butt and King (2005) admit that layer II locative markers
as meṃ ‘in’ and par ‘on’ have several similarities to layer III markers, although the former are
90 Indo-Aryan

positions that consist of one simple form and that mark both core arguments and
adjuncts (for instance the ergative ne, the objective ko, the locative meṃ in Hindi).
The postposition ko is used not only to mark O and IO but also to mark an adjunct
designating a direction, (e.g. ghar=ko ‘to the house’), or a time, (e.g. rāt=ko ‘at
night’). Finally, layer III postpositions are complex postpositions used to indicate
adjuncts. They consist of a genitive layer II postposition combined with a lexical
element (for instance us=ke=niche, ‘this=layer II=under’, meaning ‘under this’).
Postpositional marking in IA differs from inflectional marking in that post-
positions are invariable forms that are added only to the head of the phrase and
not to every part of it. For instance, in Hindi, the correct form of a determiner
phrase is is laṛkī=ko ‘this girl=to’, and not *is=ko laṛkī=ko ‘this=to girl=to’. For-
mally, layer II postpositions do not behave as case morphemes in the strict sense
of case inflections (layer I), but it has been argued that given that they also indi-
cate grammatical relations, layer II postpositions are functionally on a par with
case inflections (cf. for Hindi, Mohanan 1994: 59; Butt & King 2004: 170). Layer
III postpositions are invariably adjunct markers, and do not occur in combination
with core arguments.
The division into layers proposed by Masica (1991) is historically motivated.
The layer I suffixes are the remains of the Old and Middle IA inflectional case
system, whereas layer II and III markers derive from later developments. In Hindi,
the three layers can be clearly distinguished; in other IA languages, however,
postpositional markers have merged with the verb stem into a kind of inflection,
e.g., Asamiya rāma-k, with the ending -k, formerly a postposition from which the
end vowel has disappeared (see Section 3.1.1). It is therefore impossible to distin-
guish all three aforementioned layers of case marking in every IA language (as
Masica 1991: 233 remarks, it is in a way a conventional division), although all IA
languages reflect in some way or another the loss of inflectional cases and the rise
of a new kind of case marking.
In Hindi, the first postpositions surfaced in the Late MIA stage or in Early
Hindi (cf. Bubenik 1998: 67). What constitutes the etymological roots of the post-
positions remains a contentious issue. For many postpositions, a lexical origin
has been suggested. For instance, the Hindi objective postposition ko has been

said to be “case clitics”. Finally, the layer II locative marker meṃ as well as the layer III locative
marker ke pās are argued to mark “subjects”, which indicates that the argument that only layer
II clitics mark subjects is non-exclusive (though Butt and King have a rather broad conception
of “subject”, cf. Section 2.4.3). Therefore, I prefer to follow Montaut (2004: 60–68) and Masica
(1991) in using the straightforward and unbiased terms “simple” and “complex” postpositions
to differentiate between layer II and layer III markers.
 Some key concepts of the Indo-Aryan languages 91

related to the Sanskrit past participle kṛta- ‘done’ (cf. Montaut 2004: 65) but also
to kakṣe ‘side, armpit’ (cf. Beames 1872: 257). The origin of the Hindi ergative
postposition ne is equally uncertain, and in general three different hypotheses
are found in the literature (cf. Drocco 2008; Verbeke and De Cuypere 2009). A
grammaticalization of a lexical source is one hypothesis, e.g. lāgī ‘attached to’ (cf.
Beames 1872: 267; Kellogg 1938: 131; Chatterji 1926: 968; Butt 2001: 116; Montaut
2009). Another, early hypothesis, suggested that ne is derived from the Sanskrit
instrumental case -ena or -īna (cf. Monier-Williams 1858: 104–105; Trumpp 1872:
113; Lehmann 2002: 99), but this proposal has proven to be invalid (Beames 1872:
267 and Kellogg 1938: 131). The third hypothesis argues for the borrowing of ne in
Hindi from Old Rajasthani (Hoernle 1880: 224–225), in which ne (or nai) was used
as the objective postposition from the 15th century onward (cf. Kellogg 1938: 132;
Khokhlova 2001: 163). This is a tentative explanation; to confirm this view, more
historical data need to be analyzed.

2.4.2 The verb system in Indo-Aryan

2.4.2.1 The participial base of the verb system in Indo-Aryan

The verb system in contemporary IA is largely based on participles.¹⁸ There are
two participles, the imperfect (present) and perfect participle. Depending on the
individual language, the participles are either used as autonomous lexical items
that agree with a core argument in gender or number, or they are combined with
a form of the copula, which displays person agreement. Depending on the lan-
guage, the copula can be attached to a particular form of the verb, thus consti-
tuting one lexical unit, or it can simply follow a participial form. For instance,
a perfect participle followed by a past auxiliary designates a past perfect (e.g.,
Hindi yah gayā thā ‘he had gone’), and a present participle followed by a future
form of the auxiliary is an imperfect presumptive (e.g., Hindi maiṃ jātā hūṃgā
‘I will probably go’). In Hindi, only the future and the subjunctive tenses display
person marking and do not have a participial base (e.g., future vah karegī ‘she will
do’, subjunctive maiṃ karūṃ ‘I would do’). The most frequently used verb forms
in Hindi are thus periphrastic.
All IA languages have converbs, which are also called “absolutes” or “con-
junctive participles” in grammars of the IA languages. Converbs are non-finite
verb forms that designate an action that happened before or during the action
expressed by the main verb of the clause. In Hindi, these converbs can take the

18 For a full overview of all possible verb forms in Hindi, see Montaut (1996, 1999).
92 Indo-Aryan

form of the stem of a verb, the verb stem + kar, or the verb stem + ke. The argu-
ments of the converb may or may not be overtly expressed. In Hindi, arguments
of a converb cannot take the ergative case, but if they differ from the argument of
the main verb, they are expressed with a genitive case, cf. the following example
from Bickel and Yadav (2000: 351):

(53) tumhār-e kal nahīṃ ākar kuch kām nahīṃ

you.gen-obl.sg tomorrow not come.cvb some work[m] not
hogā
be-fut.3sg.m
‘If you don’t come tomorrow, there won’t be any work.’

Other non-finite verbal forms are participles and infinitives. The agents of par-
ticiples and infinitives are expressed by genitive constructions in Hindi; e.g.,
the infinitive construction rām=ke baiṭhne=par ‘on [the moment] of the sitting
of Ram’ > ‘when Ram was sitting’, or the participial construction rām=ke baiṭhte
hī ‘during the sitting of Ram’ > ‘while Ram was sitting’ (examples from Mohanan
1994: 78).

2.4.2.2 Light verbs and lexical compound predicates

Typical of IA is a phenomenon called “verb stapling”; that is, one verb may be
combined with another verb to express a complex lexical meaning or (more often)
a grammaticalized meaning. The main verb may take any form (participle, root,
or infinitive), and is either invariable or in agreement in gender/number with
the O in the clause (depending on the auxiliary with which it is combined). The
auxiliaries, also known as “light verbs” (or “conjunct verbs”, “compound verbs”,
“serial verbs” or “vector verbs”) are conjugated and display person marking and
gender/number agreement. In Hindi, rahnā is one of the frequently used light
verbs. Its lexical meaning is ‘to stay’, but when it is used as a light verb in com-
bination with the stem of another verb, rahnā expresses the progressive tense
of that other verb, which is then the main verb in the clause. Variants of rahnā
are also found in other IA languages, in which the verb rah can be more or less
grammaticalized. In Nepali, for instance, rah is incorporated in the verb form (as
in gar-e-rah-e-chu ‘I was doing’) (cf. Section 4.1.3). Other frequent light verbs in
Hindi are ānā ‘to come’, jānā ‘to go’, karnā ‘to do’, lagnā ‘to attach’, denā ‘to give’,
and lenā ‘to take’, etc. Whereas light verbs are quite common in languages such
as Hindi and Punjabi, the number of light verbs varies considerably across the
IA languages. In Kashmiri and Marathi, for instance, light verbs are much less
frequent (cf. Hook 1988: 293).
 Some key concepts of the Indo-Aryan languages 93

In IA languages, it is also possible to combine a verb with a noun, resulting in

what are called “lexical compound verbs” (Kellogg 1938: 272–279; Mohanan 1994:
110). The nominal part of a lexical compound verb is semantically incorporated
into the verbal construction (e.g., Hindi śuru karnā ‘to begin’, pasand ānā ‘to like’,
yād karnā ‘to remember’, etc.). The noun in such a construction can never be
marked for animacy or definiteness (cf. Klaiman 1990: 337). However, although
lexical compound constructions are instances of incorporation, it is important
to point out that the incorporation only takes place at a semantic level. Syntacti-
cally, there are a number of reasons why the nouns cannot be said to be com-
pletely incorporated into the verb construction but should instead be regarded as
independent O-arguments. For instance, in contrast to the traditional definition
of incorporation, lexical compounds can control verb agreement when the verb
is transitive and perfective. Furthermore, a lexical compound verb is still seen as
transitive (it takes, for instance, an ergative marked A in the perfective) (Klaiman
1990).¹⁹ In sum, it is reasonable to assume that there is incorporation in Hindi
on a semantic level, but that has no bearing on the syntax of the construction.
The meaning of the combination of a verb with a noun is often to a certain extent
idiosyncratic. For instance, in Hindi, gāliyā khānā literally means ‘to eat abuses’,
but its idiomatic lexical meaning is ‘to be scolded’. In Asamiya bhai means ‘fear’
and khā means ‘to eat’, but the lexical compound bhai khā means ‘to be scared’.
In Hindi, the transitivity of the light verb determines the transitivity of the
entire construction in which it is used. For instance, verb constructions with
rahnā are always intransitive whether the main verb is transitive or intransitive.
Constructions with khānā ‘to eat’, however, are transitive, and this is also the
case in Asamiya: rām-e bhai khāle is transitive, although it may be translated as
‘Ram got scared’. There are a few transitive light verbs in Hindi, but these never
occur with an intransitive main verb (Kellogg 1938: 260–261). For example, the
verb ḍālnā literally means ‘to throw’ or ‘to pour’, and in its function as a light
verb, it merely emphasizes or strengthens the meaning of a transitive main verb²⁰.
However, it is not impossible to find a combination of an intransitive main verb
with a transitive light verb, e.g., cal denā, lit. ‘go + give’ > ‘to go away’. Whenever
this light verb is used in a perfective construction, the A is in the ergative.

19 Exceptions to this agreement pattern exist. For instance, in Hindi, pasand ‘taste’ is femi-
nine, which would imply that in perfective constructions the verb karnā must always be f.sg.
However, this is not the case; e.g., in rājū=ne baccoṃ=ko pasand kiyā ‘Raju liked kids’, the
perfective verb kiyā is m.sg., not f.sg. (Klaiman 1990: 340).
20 E.g., the well-known song lyric in Hindi mār ḍālā, meaning, ‘it killed me’.
94 Indo-Aryan

This combination of an intransitive main verb and an ergative A (e.g., cal denā) is
an argument for Mohanan (1994: 74) to consider the ergative case as a semantic
case that only marks a voluntary agent. If the ergative case were a structural case,
so the argument goes, its occurrence would be strictly limited to the marking
of the A of a transitive main verb. However, rather than referring to a semantic
property, the occurrence of the ergative marker may be said to be syntactically
determined by the transitivity of the light verb. The following example from Hindi
with the intransitive light verb cuknā, lit. ‘to be finished’, in combination with the
lexical compound verb qatl karnā ‘to kill’, illustrates a construction in which ne is
absent, although the agent arguably carries out a deliberate action:

(54) maiṃ chaḥ ādmiyoṃ=ko qatl

I six man.pl.obl=obj killing
kar+cuk-ā+h-ūṃ
do+be finished-pst.m.sg+aux-prs.1sg
‘I have (already) killed six men.’ (Manto 1984)

One possible exception to this pattern is the so-called habitual construction,

which is formed by a verbal noun on -ā in combination with the verb karnā ‘to
do’. Because such combinations with karnā express a habitual or repeated action,
they do not occur often in a perfective. On the rare occasions when they do, there
is no ergative marking (Everaert 2004: 77), although karnā is a transitive verb.

2.4.2.3 Perfect, perfective, and past

The perfect aspect and the past tense have intermingled in most Modern IA lan-
guages as a consequence of disappearance of many tenses in the Middle Indo-
Aryan stage. The past tenses are based on a participle expressing perfect aspect.
In Hindi, constructions in which the participle is combined with a present form or
any other form of the verb ‘to be’ are dominant. The tense is either past or present,
but the aspect is always perfect. Compare, for instance, the following examples of
a present perfect and a past perfect, respectively:

(55) rām=ne kitāb paṛh-ī+h-ai

Ram=erg book[f] read-pst.f.sg+aux-prs.3sg
‘Ram has read a book.’

(56) rām=ne kitāb paṛh-ī+th-ī

Ram=erg book[f] read-pst.f.sg+aux.pst-f.sg
‘Ram had read a book.’
 Some key concepts of the Indo-Aryan languages 95

However, when the perfect participle is used as a finite verb (without an auxil-
iary), it is generally interpreted to indicate a simple past tense with perfective
aspect (cf. Davison 2002) as in the following example:

(57) rām=ne kitāb paṛh-ī

Ram=erg book[f] read-pst.f.sg
‘Ram read a book.’ (and finished it)

Montaut (2004), on the other hand, considers the verb form of the previous
example as an aorist, and argues that its primary function is to express the pret-
erit (past). At the same time, it can also occur when “the temporal meaning is
constructed by the syntactic context” (Montaut 2004: 106), as in her following
example.²¹ Observe that the participle gaī takes over the tense of the main verb
bacegā, i.e., future. Aspectually, though, the participle gaī is perfective, because
it indicates an action that is a condition for, and needs to be completed before, a
second action:

(58) billī rasoī=meṃ ghus+ga-ī to kuch

cat[f] kitchen=loc enter+go-pst.f.sg then something
nahīṃ bac-e-g-ā
not be saved-3sg-fut-m.sg
‘If the cat happens to enter the kitchen, nothing will be saved.’

Examples such as (57) show that tense and aspect overlap in IA to the extent that
the perfect participle is a necessary condition for the past tense as well as for the
present and past perfect. The tenses constructed with the present/imperfect par-
ticiple are imperfective, as in the following example:

(59) rām kitāb paṛh-tā+th-ā

Ram book[f] read-prs.m.sg+aux.pst-m.sg
‘Ram was reading a book.’

Imperfectives often alternate with a progressive tense that consists of a light verb
construction with rahnā.
Of course, the system of tenses and aspects is not the same in all IA languages.
There is considerable variation with respect to the combinations of participles,
main verbs, and person markers. Moreover, the elaborate system of light verbs

21 Montaut (2006b) also argues that these “aorists” can express certain evidential meanings.
96 Indo-Aryan

found in Hindi and in many Indo-Aryan languages (and certainly not restricted to
rahnā), the range of which is beyond this study, is often considered to be related
to aspect and Aktionsart (cf. Hook 1978; Nespital 1981).
Nevertheless, participial verb constructions sharing the same historical
origin are found intervening in the tense/aspect system in all IA languages. In the
remainder of this study, the forms that generally occur with an ergative construc-
tion will be labeled “perfective” and those that occur without ergative construc-
tion “imperfective”. Although these terms may not do full justice to the complex
tense/aspect system in Indo-Aryan, the term “perfective” is preferred above terms
such as “past (preterit) tense” or “perfect aspect”, because forms like the “present
perfect” or “simple past” are either perfect or past, but both can be perfective.
Moreover, the occurrence of an ergative pattern seems to be rather related to the
formal presence of the perfect participle, than to semantic distinctions related to
perfectivity (with the exception, to a certain extent, of Nepali, cf. 4.1.3.).

2.4.2.4 Passives
Hindi possesses an analytic passive that is made up of a perfect participle in com-
bination with the conjugated verb jānā ‘to go’ and that is attested in the present
as well as the past tense.²² The expression of the agent in the passive is optional,
but if it is expressed, it is marked by the ablative postpositions se or ke dwārā, as,
e.g., in ex. (60) (Montaut 2004: 132):

(60) cor mujh=se pakṛā+ga-yā

thief I.obl=abl catch+pass-pst.m.sg
‘The thief was caught by me.’

In certain passive constructions, the patient is marked with the objective postpo-
sition ko, e.g., (61), (Montaut 2004: 132).²³

(61) in donoṃ=ko pakṛā+ga-yā+th-ā

these two=obj catch+pass-pst.m.sg+aux.pst-m.sg
‘These two had been caught.’

22 The passive in Indo-Aryan follows the Indo-European tradition according to which an (in-
transitive) auxiliary is used to form the passive (Haspelmath 1990: 29).
23 According to Mohanan (1994: 94), whether the patient in a passive construction is in the
nominative or is marked with ko is a dialectal preference.
 Some key concepts of the Indo-Aryan languages 97

According to Montaut (2004: 132), the passive in this example “does not necessar-
ily promote the patient”, but primarily aims at “backgrounding the agent”. This
feature of “backgrounding the agent” is in fact the only general cross-linguistic
feature that allows us to identify a passive (cf. Andersen 1990: 198; Shibatani
1985; Siewierska 1984: 256; but see Haspelmath 1990, who argues that the inactiv-
ization of the verbal situation is the basic function of the passive). Both transitive
and intransitive verbs can take a passive form, although passives of intransitive
verbs are often impersonal.
All IA languages have some kind of passive. The attested passives can be split
into two main types. “Periphrastic passives” are based on a light verb construc-
tion in combination with a verb meaning ‘to go’, such as Hindi jānā. “Synthetic
passives” are constructed by inserting a sound based on -ij- (which can also be -i-,
-j- or -s(s)-, as in some Rajasthani languages) into the verb form.
A passive form can also be used in a meaning of capability, often in a negative
sense. The following examples illustrate this use of the passive in an impersonal
and a personal construction, respectively (Montaut 2004: 132, 208).

(62) aise ro-yā nahīṃ jā-tā

so cry-pst.m.sg not pass-prs.m.sg
‘One should not cry.’

(63) us=se mer-ā gham nahīṃ

he.obl=abl I.gen-m.sg sorrow[m] not
dekh-ā+ga-yā
see-pst.m.sg+pass-pst.m.sg
‘He could not bear to see my sorrow.’

2.4.2.5 Causatives
In all IA languages, causative verbs are formally indicated. Typically, a causative
verb has increased valency compared to the verb from which it is derived. So-
called “first causatives” are formed by adding -ā- to the root of the intransitive
verb. The added suffix may cause a sound change in the stem. For instance, the
causative of ṭuṭna ‘to be broken’ is toṛna ‘to break’. Bannā ‘to be made’ becomes
banānā ‘to make’. Verbs in IA also have a “second causative”, which is a causa-
tive of a verb form that is itself a causative (for instance banwānā ‘to cause to
make’, which is arrived at by causativizing banānā ‘to make’). khānā ‘to eat’ in
Hindi is causativized to khilānā ‘to feed’ and may become khilwānā ‘to cause to
feed’. However, because of their high token frequencies, verbs such as toṛnā and
banānā are no longer considered as derived verb forms. Conversely, many non-
98 Indo-Aryan

causative intransitives are relatively infrequent (e.g., sūjhnā ‘be thought’, biknā
‘be sold’). The process of causativization, through which basic verbs that are
intransitive are made transitive, is a frequent phenomenon in IA. Montaut (2007:
230) describes the direction of the transition in Hindi, i.e., from intransitive to
transitive, as follows:

L’argument principal en faveur de la dérivation des transitifs à partir des intransitifs, les
noms à dérivation zéro n’étant guère concluants, est d’une part la forme phonologique
(toujours plus lourde, ou demandant un effort articulatoire plus important, même sans
évoquer l’étymologie en sanscrit où la voyelle tendue est considérée comme le degré long
de la voyelle relâchée), d’autre part l’homologie fonctionnelle évidente entre les deux séries
a et b [two possible ways to form a causative, SV], et surtout le fait que la seconde dérivation
causative (ou factitive selon les verbes de base), par suffixation du morphème -vâ, procède
presque toujours de la forme intransitive.

Cross-linguistically, Hindi can be considered as a “transitivizing” language in the

terms of Haspelmath (1993) and Nichols et al. (2004) because intransitive verbs
are formally and semantically more unmarked than transitive ones (Montaut
2004: 85). Montaut (2007: 230–231) mentions only a few verbs that are “ambitran-
sitive” or “labile” in the sense that they can be used in intransitive and transitive
constructions. These verbs are khonā ‘to loose, to get lost’, badalnā ‘to change’,
bharnā ‘to fill’, and baṛhnā ‘to increase’.

2.4.3 Experiencer constructions

Non-canonical marking of subjects is the phenomenon that a small set of verbs

pertaining to a lexical group does not take a subject (A or S) marked in one of the
two structurally expected cases, viz. nominative (for intransitive and accusative
constructions) or ergative (for ergative constructions) (cf. Verma and Mohanan
1990; Aikhenvald et al. 2001; Bhaskararao 2004). Non-canonical subject marking
is found in many Indo-European languages (e.g., in present-day German mich
friert [with accusative], mir ist kalt [with dative], both meaning ‘I am cold’, or in
Old English me liketh ‘I like it’). In IA, the non-canonical marking comprises either
an objective or genitive marking of the subject, with the objective being the most
frequent case.²⁴ In general, the verbs constructed with a non-canonical subject

24 Because the objective case has often been called ‘dative’, the designation “dative subject”
has become common currency in IA linguistics to refer to this type of non-canonical case mark-
ing of the subject.
 Some key concepts of the Indo-Aryan languages 99

can be semantically grouped as “experiencer” verbs. Experiencer verbs may be

divided into two groups, viz. experience-based predicates (verbs of emotion, atti-
tudes, cognition, perception, bodily states, changes in bodily states) and hap-
penstance predicates (verbs of decline, failing/mistaking, success/performance,
ontological states, social interaction, gain, personal properties, verbs of land-
scape and nature). By extension, they are verbs of physical sensation, psychologi-
cal states, wanting, needing or obligation, and receiving, and they also include
verbs that express kinship relations (Barđdal 2011; Cole et al. 1980). Instead of
having the common semantic role of agent and patient, the arguments taken by
these verbs can be analyzed as instantiating the thematic roles of experiencer
and stimulus (cf. Dixon 2010, Chapter 1). The following examples are from Hindi:

(64) yah kitāb mujh-e pasand ā+rahī+h-ai

this book[f] I.obl-obj nice come+prog.f.sg+aux-prs.3sg
‘I like this book.’

(65) mujh-e būkh lag-ī

I.obl-obj hunger[f] attach-pst.f.sg
‘I am hungry.’

(66) rām mujh-e dikh-tā+h-ai

Ram[m] I.obl-obj appear-prs.m.sg+aux-prs.3sg
‘Ram appears to me.’ > ‘I see Ram.’

The pattern illustrated in these examples is quite straightforward. The expe-

riencer role, mujhe, is in the objective case, whereas the stimulus roles, kitāb,
būkh and rām control the agreement. Experiencer constructions have been dis-
cussed intensively, in particular with respect to the subject status of the expe-
riencer argument (cf. Cole et al. 1980; Andrews 1976; Bayer 2004; Barđdal 2011,
among others). Parallel to the discussion on the “subjectness” of oblique expe-
riencer arguments in Icelandic and other German languages (cf. Bayer 2004),
there used to be a debate on whether the experiencer controls certain syntactic
subject properties in IA languages and whether oblique experiencer arguments
can therefore be considered “subjects” (cf. Klaiman 1981; Mohanan 1994: 96;
Verma and Mohanan 1990; Montaut 2001, 2004b). The morphological proper-
ties of a subject, viz. controlling verb agreement and being the argument in an
unmarked case, are not typical of an experiencer in IA. On the other hand, it is
well-known that experiencers in Icelandic can control agreement under certain
circumstances; this has been considered as an acquired subject property (cf.
Bayer 2004). In IA too, an exceptional agreement pattern with experiencer verbs
100 Indo-Aryan

is found in two variants of the language Shina (spoken on the border between
India and Pakistan), i.e., the Shina of Dras, Skurda and Gultari (cf. Hook 1990,
1996). In these languages, the experiencer verb agrees with an object marked in
the objective case. According to Hook (1990: 82), this agreement pattern can be
considered to be the result of a transition of subject features to the experiencer
role. At least in the Shina of Gultari, the objective marking of the experiencer
alternates with an ergative marking, with “little or no change in meaning”, and
the stimulus, in Hindi always in the nominative case, can take objective marking
(Hook 1996). However, this pattern does not seem to occur in any other IA lan-
guage except for Shina. It is possible that other factors, such as the use of pro-
nominal suffixes, may also have contributed to the rise of experiencer agreement
in Shina (cf. Section 5.2).
Bickel and Yadava (2000: 364) compare the behavior of arguments, includ-
ing the experiencer, in a series of syntactic constructions such as “converbial
clause chaining, coordinate conjunction reduction, and reflexivization” in
Hindi, Maithili, and Nepali. Converbial clause chaining, coordinate conjunction
reduction and reflexivization are traditionally used as “syntactic diagnostics”
to establish whether there is a grammatical relationship of subject in a certain
syntactic construction. However, according to Bickel and Yadava, these tests
often fail for such a purpose in IA because, in these languages, the patterns
of coreference and reflexivization are determined by semantic constraints and
not by syntactic constraints of case marking or grammatical role. The following
two examples of elliptical constructions (67) and (68) are taken from Montaut
(2004: 259). One argument of the main verb of the second clause in the coordi-
nated sentence is not overtly expressed, and the context provides the necessary
information to infer the correct coreference. The ellipsis is not case-sensitive,
so the case marking of the omitted argument does not matter. For instance, in
ex. (67) the verb paṛhī is transitive and takes the ergative-marked A us=ne ‘he’.
The coordinated verb lagī is intransitive, which implies that its overt argument
would have to be in the nominative case. However, although the S verb lagī is not
overtly expressed in ex. (68), it is clear from the semantics that the intended S in
the second clause coincides with us=ne with the overtly expressed A in the first
clause. Ex. (68) is almost the same construction as ex. (67), only here the experi-
encer predicate yād āī normally takes an experiencer argument marked with ko,
which is not overtly expressed.

(67) us=ne ciṭṭhī paṛh-ī aur ro-ne+lag-ī

she.obl=erg letter[f] read-pst.f.sg and cry-inf+be attached-pst.f.sg
‘She read the letter and [she] started to cry.’
 Some key concepts of the Indo-Aryan languages 101

(68) us=ne ciṭṭhī paṛh-ī aur us din=kī

she.obl=erg letter[f] read-pst.f.sg and that.obl day=gen
yād ā-ī
memory[f] come-pst.f.sg
‘She read the letter and [she] remembered that day.’

Bickel and Yadava (2000: 364) offer the following examples (69)–(71) to illustrate
the “free” coreference and reflexivization pattern in Hindi. In the first two exam-
ples, Bickel and Yadava assume that there is coreference between S/A in the first
sentence and O in the second one. However, all of these examples have been con-
sidered unacceptable by native speakers of Hindi, in particular ex. (69) and (71),
and seem to have been derived from a highly marked register. In fact, ex. (69) is
clearly grammatically wrong, because the verb lagnā never takes an ergatively
marked subject, as is the case in this example.

(69) *rām ā-yā aur sab-oṃ=ne

ram come-pst.m.sg and all-obl=erg
piṭ-ne+lag-ā
beat-inf+be attached-pst.m.sg
‘Ram came and everyone started to beat [him].’

(70) ?maiṃ=ne glās pheṃk-ā aur

I=erg glass[m] throw-pst.m.sg and
ṭuṭ+ga-yā
break+go-pst.m.sg
‘I threw the glass and [it] broke.’

(71) ?rām=ne patnī=ko apnī sarī dī

Ram=erg wife[f]=obj one’s own sari[f] give.pst.f.sg
‘Ram gave his wife her own sari.’

According to Montaut (2004: 259), examples such as (70) can occur only when O
is high in topicality, but they are extremely rare. There may be dialectal variation
involved. Ex. (71) is not an elliptical construction, but illustrates the use of the
reflexive (if accepted as grammatical). The reflexive apnī refers to the O patni=ko,
but apnī could also refer to the A rām=ne. However, because of the context (in par-
ticular the fact that a sari normally belongs to a woman) the first interpretation is
the most obvious one for Bickel and Yadava. They argue that reflexivization of an
argument is also related to “linear precedence”, the reflexive normally referring
102 Indo-Aryan

to the “closest” argument in the sentence with which a coreferential relation can
be established (Bickel and Yadava 2000: 368, Montaut 2004: 271).
The above examples (67)–(68) lend support to the view that, for Hindi, it is
generally impossible to determine what the “subject” of a clause is purely on the
basis of such syntactic constructions. This conclusion holds for the IA languages
in general. Moreover, in non-coordinated sentences, it is also possible that S, A,
and O are not overtly expressed. In Hindi and other IA languages, this pattern of
free deletion of core argument abounds in question-answer pairs, yet in spoken
language, it also occurs frequently on other occasions as well. See, for instance,
the following example from Asamiya:

(72) kāli āpuni cakā meja-bor ān-il-e ne

yesterday you.h chair table-pl bring-pst-2h Q
hay, ān-il-oṃ
yes bring-pst-1
‘Did you bring chairs and tables yesterday? Yes, [I] brought [them].’
(Baruah 1980: 318)

In the next example from Kashmiri (Koul 2006:61), only the O is not overtly
expressed in the answer to the question:

(73) tohi por-v-ā akhbār?

you.erg read.pst.m.sg-2pl-q newspaper[m]
ā, me por
yes I.erg read.pst.m.sg
‘Did you read the newspaper? Yes, I read [it].

All of these patterns of ellipsis, in independent clauses and in coordination pat-

terns alike, appear to be determined by the discourse topicality of the arguments.
This means that these mechanisms are not completely adequate to indicate the
subject in this language. Nevertheless, constructions such as (69) and (70) are
considered ungrammatical, and S and A are often coreferential. Moreover, in
Hindi converbial constructions, the unexpressed agent of the converb is corefer-
ential with A/S of the main verb (cf. Montaut 2004: 182–183, Section 2.4.2.1). These
two factors indicate that, at least in Hindi, the identification of S with A is more
natural, and may suppose a subject category – which is in line with findings in
other Indo-European languages (cf. Eythórsson and Barđdal 2012).
It bears pointing out that the experiencer construction is an exceptional
construction in which the arguments behave idiosyncratically. In contrast to
“normal” transitive constructions, the marking and behavior of the arguments
 Some key concepts of the Indo-Aryan languages 103

of an experiencer construction are primarily determined by their semantic roles.

The question of whether an experiencer is a subject is ultimately irrelevant
because the morphosyntactic and semantic properties of the semantic roles of
experiencer/stimulus are different from those of A/O (Hook 1990: 320; Montaut
2004b: 43; Næss 2007: 190).
Although experiencer constructions are different from ergative constructions
in that the former are lexically determined (i.e., the meaning of the verb deter-
mines the argument) and occur in perfective as well as imperfective construc-
tions, the two have often been considered to be related. For instance, according
to Montaut (1991, 1996, 2001, 2004b, 2009), the ergative construction in Hindi is a
“localizing predication”, the ergative case is said to express a change of state. The
postposition ne indicates then an “oblique subject” in the same sense as an objec-
tive in experiencer constructions and as an instrumental in passive constructions
(Montaut 1996). According to Montaut, even a locative can be construed as a
subject, in constructions describing a state. Consider, e.g., the following example
from Mohanan (1994: 64):

(74) ravī=meṃ bilkul dayā nahīṃ th-ī

Ravi=loc totally mercy[f] not be.pst-f.sg
‘There was no mercy in Ravi’ > ‘Ravi did not have mercy’

Montaut (2009b) argues, from a cognitive perspective, that the ergative in Hindi
is not the mirror, or the “reversed trajectory”, of the accusative construction, but
that it patterns together with the experiencer construction because of the cogni-
tive similarity between the two constructions. Still, the ergative in Hindi has a
conceptual identity of its own, although this identity has a spatial basis in the
sense commonly assumed in the cognitive-linguistic framework.
Montaut refers to DeLancey (1981) who associated ergative and experiencer
verb constructions with a viewpoint that focuses on the result of an action.²⁵
According to Montaut and DeLancey, the source of the event is the natural start-
ing point of the action, but it is not the viewpoint of the speaker and thus remains
outside the predication. Only the last part of the action, the result, is profiled
“on stage” (cf. Section 1.2.1). Importantly, both the experiencer and the erga-
tive constructions are said to be closer to an intransitive construction than to a

25 This is in contrast to Langacker, who considers the experiencer as the source of a mental or
perceptual path that is not profiled on the onstage domain and is not the result of the action.
However, in Hindi, the main feature of the experiencer is that it is not an active/volitional agent,
given that the experiencer verb is rather intransitive (Montaut 1996: 355).
104 Indo-Aryan

transitive construction in which the accusative is used. As shown in Chapter 1,

according to cognitive linguists, the central relationship in an ergative pattern is
the one between the verb and O, viz. “the thematic relationship”. In an intran-
sitive construction, the relationship between the verb and S is also considered
to be thematic. This thematic relationship is called the “absolute construal” by
Langacker (1999: 84). According to Montaut (2009b), the absolute construal can
be observed both in the patient-verb relationship in ergative constructions and in
the theme-verb relationship in experiencer constructions (with the theme being
the stimulus in the latter construction). The semantic similarity between the erga-
tive and intransitive constructions is said to be dependent on the cognitive con-
ceptualization of the theme relationships in the intransitive construction (S-verb)
and in the transitive construction (O-verb). However, the focus on the theme in
the ergative construction seems to be obscured by the (often obligatory) presence
of the ergative A in the clause. An ergative construction represents an action with
two arguments and is transitive, whereas intransitive constructions express only
one of the two arguments. Furthermore, if a transitive (ergative) construction
alternates with an intransitive one (cf. the break examples in English discussed
in Section 1.2.1), the same verb cannot be used in transitive and intransitive con-
structions; thus, a causative verb must be involved:

(75) The vase broke: phūldān ṭuṭ-ā

vase[m] be broken-pst.m.sg
John broke the vase: John=ne phūldān toṛ-ā
John=erg vase [m] break-pst.m.sg

According to Montaut, the “absolute construal” (i.e., the nominative marking and
the agreement of S and O in perfective constructions) is a first similarity between
an ergative and an intransitive construction. Montaut believes that there is a
second reason to consider the ergative construction as bordering on the “intran-
sitive” in Hindi. She indicates that stative intransitive constructions are similar
to ergative constructions. For example, ex. (76) from Montaut (2009b) is a Hindi
stative construction with the stative verb form banāe hue consisting of a perfect
participle of the main verb and a perfect participle of the auxiliary. The stative
verb form banāe hue does not agree with any argument in the clause, but it is in
the so-called “default” form, i.e., oblique masculine singular.
 Some key concepts of the Indo-Aryan languages 105

(76) maiṃ un=se mitratā

I they.obl=abl friendship[f]
banā-e+hu-e h-ūṃ
make-pst.m.sg+aux.ptcp.prf-m.sg be-prs.1sg
‘I am friends with them’ < ‘I am in the state of having made friends with
them’

Montaut (2009b) considers the construction in ex. (76) to be almost identical to

the following example (77), which is a normal ergative perfective constructive
with a verb agreeing with the O, mitratā:

(77) maiṃ=ne un=se mitratā banā-ī+h-ai

I=erg they.obl=abl friendship[f] make-pst.f.sg+aux-prs.3sg
‘I have made friendship with them.’

However, Montaut’s analysis does not account for the fact that the two construc-
tions illustrated in the examples (76) and (77) are structurally very different,
despite the fact that the meaning of both constructions is similar. In ex. (76),
banāe hue is an independent stative participle in a construction where hūṃ is the
main verb. In ex. (77), banāī is not an independent stative participle but part of
the verbal construction banāī hai. Therefore, there is only one core argument in
ex. (76) that is determined by the substantive verb hūṃ, i.e., the unmarked maiṃ,
whereas in ex. (77), the core arguments maiṃ=ne and mitratā are determined by
the perfective transitive verb banāī hai. Consequently, in this transitive construc-
tion, maiṃ=ne is an ergative A. On the other hand, the construction banāe hue
in ex. (76) is an independent unit, but this does not need to be the case. A stative
construction in which a finite verb is involved is also possible, as, for instance, in
the following example (Montaut 2009b):

(78) sītā=ne sārī pahn-ī+hu-ī th-ī

Sita=erg sari[f] wear-pst.f.sg+aux.ptcp.prf-f.sg be-pst.f.sg
‘Sita was wearing a sari.’

In this example, the ergative postposition ne is not obligatory. Nevertheless, in

the cases where it occurs marking A, there is always a stative finite verb form,
e.g., pahnī huī thī ‘was [in the state of] wearing’ in ex. (78). Conversely, when ne
is absent, as in ex. (76), the stative participle is syntactically an independent part
of the sentence. The argument structure patterns intransitively, because of the
main verb meaning ‘to be’, and the stative infinite verb pahnī huī functions as an
adjective meaning ‘wearing’. However, in ex. (78), the stative participle pahnī huī
106 Indo-Aryan

is part of the verb phrase pahnī huī thī and therefore the construction is transi-
tive. The resemblance between a stative and an ergative construction thus exists
merely on the level of semantics, not on a syntactic level. Both stative and perfec-
tive constructions focus on the result of an action, but structurally the two con-
structions pattern differently with respect to case marking and verb agreement.
On the other hand, there are other structural and semantic reasons to con-
sider experiencer constructions as structurally and semantically similar to intran-
sitive constructions. In contrast to the A in the ergative, the experiencer argument
is often not overtly expressed, e.g., kitāb acchī lag rahī hai ‘the book pleases [me]’.
More generally, experiencer constructions are said to lack certain semantic fea-
tures of transitive constructions (Hopper and Thompson 1980, Section 1.2.3.2).
Hook (1990: 325) argues that although experiencer constructions are lexically
transitive, in the sense that there is a patient role involved, they are grammati-
cally intransitive.²⁶
In some constructions, the ergative and objective postposition are apparently
exchangeable. This similarity in use has led linguists to assume that there exists
a semantic connection between the ergative and objective cases. The Hindi post-
position ne is primarily the ergative marker, but sometimes ne also occurs with
verbs of obligation, perhaps indicating an influence of Punjabi (cf. Section 6.2.4).
Normally, obligations are expressed by means of an experiencer construction in
Hindi, as, e.g., in the following example:

(79) āp=ko hindī bol-nī h-ai

you=obj Hindi[f] speak-inf.f.sg aux-prs.3sg
‘You have to speak Hindi.’

However, Butt (2007; Masica 1990; Bashir 1999) reported that in some dialects of
Hindi (especially the Hindi spoken around Delhi), ko may be replaced by ne in an
obligational construction, e.g.,:

(80) nādyā=ne skūl jā-nā h-ai

Nadya=erg school go-inf aux-prs.3sg
‘Nadya must go to school.’

Butt (2007) takes the variation in the marking of the experiencer in the previous
sentences as evidence that the ergative postposition ne is the marker of a seman-

26 For the same reasons, Pandharipande (1990: 161) considers the marker of the experiencer
argument in Marathi to be a “locative case”, and the experiencer a “locational NP”.
 Some key concepts of the Indo-Aryan languages 107

tic case that expresses volitionality and agentivity. She also maintains that ne in
such constructions indicates that the action involves a conscious choice, whereas
the corresponding construction with ko is more neutral with respect to volitional-
ity of the experiencer/agent. Montaut (2009b: 310–311) rejects the view that ne in
Hindi indicates volition or control. According to her, quite rightly, it is the tran-
sitivity of the verb that transfers the meaning of volitionality/conscious choice/
control, not the particular marking of the arguments. Montaut agrees with Bashir
that the use of ne in constructions such as ex. (80) is not structural but is instead
related to other semantic categories. For instance, she argues that an experiencer
construction with a ne-marked experiencer conveys an injunctive or prospective
meaning. Nevertheless, like Butt, Montaut maintains that the ergative and the
objective case markings are semantically related because they can occur in differ-
ent constructions with the same meaning. For instance, the verb milnā meaning
‘to get’, is always constructed with an experiencer construction, whereas pānā,
which also means ‘to get’, is a normal transitive verb and therefore is always con-
structed with an ergative A in the perfective.
The apparent similarity between the ergative and the objective cases has
led linguists to assume a common origin of the two case markers. For instance,
according to Butt (2007), the Hindi ergative marker ne must have originated from
an objective marker neṃ. In her view, the ergative and objective cases are seman-
tically related, with both being cases expressing a “result”. The ergative expresses
a result on a “control dimension” – the action is the result of the ergative argu-
ment that started the action. By contrast, the objective expresses a result on a
“spatial/path dimension”. This is particularly clear when the objective marks an
IO and when the IO is at the end of a displacement along a path (e.g., ‘I gave it to
him’). Butt’s approach to the concept of “result” on a control vs. spatial dimen-
sion is arguably highly metaphorical. Moreover, it ignores the influence of verb
meaning on the constructions. For instance, Davison (1990) gives the following
two examples:

(81) pitā=ne apne beṭ-e=par taras khā-ī

father=erg self’s son-obl=loc pity[f] eat-pst.f.sg
‘The father felt/showed pity for his son.’

(82) pitā=ko apne beṭ-e=par taras ā-ī

father=obj self’s son-obl=loc pity[f] come-pst.f.sg
‘The father felt pity for his son.’

The first example, (81) (Davison 1990: 356) is an ergative construction, with a
transitive verb. The second one, (82) (Davison 1990: 357), is an experiencer con-
108 Indo-Aryan

struction. According to Davison, in ex. (81), pitā ‘father’ may have some control of
the action, whereas in ex. (82), pitā has no control whatsoever over the emotion.
However, note that the verbs are fundamentally different in both example sen-
tences. In ex. (81), the verb khānā is transitive, but in ex. (82), the verb is a typical
lexical compound verb with the intransitive verb ānā.
In conclusion, the main similarity between the ergative and the experiencer
constructions is formal: both have a non-nominative A, and the verb agrees with
O in both constructions. Semantically though, experiencer constructions are con-
structions with a reduced transitivity, with the A expressing a semantic role with
diminished control of the action, whereas ergative constructions are fully transi-
tive, typically, but not necessarily, occurring with an agentive, controlling A.
Experiencer constructions are particularly frequent in Eastern IA and will be
discussed in more detail in Section 3.1.3.3. I will return to the issue of the similar-
ity between the ergative and objective case in Chapter 6.

2.4.4 Unergatives

All IA languages have unergative verbs. Like experiencer verb constructions,

they constitute a semantically determined set of verbs that show deviating align-
ment patterns. The terminology “unergative” stems from Perlmutter and Postal
(1984).²⁷ Unergatives are intransitive verbs with an agentive S, i.e., an S that con-
trols the situation. However, with respect to IA, unergatives can be described in
a narrower sense as verbs that can hypothetically take an O, as opposed to so-
called “unaccusative” verbs, which are intransitive verbs that cannot take an O.
Unergative verbs go unnoticed in a fully accusative language, given that in
these languages there is no different marking of the subject. However, in Hindi,
transitive verbs take an ergative A in perfective constructions. Examples of IA
unergatives include verbs with meanings such as ‘to sleep’, ‘to dance’, ‘to bathe’,
all kinds of verbs referring to bodily excretions (such as the Hindi verbs chīṃknā
‘to sneeze’, khāṃsnā ‘to cough’, mūtnā ‘to urinate’, mitlānā ‘to vomit’), and verbs
such as bhauṃknā ‘to bark’, gānā ‘to sing’ etc. (Montaut 2004: 181). These lexical
items have traditionally been analyzed as verbs that may take a “cognate” object.

27 Although the terminology “unergative” is taken from Perlmutter and Postal, I do not share
their view on ergativity (cf. Chapter 1). Another term that is often used to refer to an “unerga-
tive verb” is “anticausative verb” (e.g., Bynon 2005; Masica 1976 used this term in a different
sense). The term “anticausative” indicates that this type of verb forms the opposite of the
“causative” verb type, i.e.,it is de-transitivized instead of “transitivized” or causativized.
 Some key concepts of the Indo-Aryan languages 109

Hook and Koul (2004) argue that they all are in fact transitive verbs, of which the
O is most often left unexpressed. Most of these verbs may be combined with an
unexpected object as well (for instance in a construction meaning ‘to spit blood’)
and then it is entirely clear that they are, in effect, transitive verbs. The presence
of an O does not determine the occurrence of an ergative subject, given that the
ergative subject occurs anyway. It is more likely that it is the very possibility of an
O, rather than its actual occurrence, that renders the unergative verbs transitive.
These verbs may therefore well be called exceptional on the basis of their lexical
content, but they are definitely not exceptional as far as IA is concerned (cf. Hook
and Koul 2004). Moreover, note, that some verbs can take either an ergative or
a nominative A (e.g., in Hindi cillānā ‘to shout’, cīkhnā ‘to scream’, nācnā ‘to
dance’, gānā ‘to sing’, khelnā ‘to play’, laṛnā ‘to fight’²⁸, ronā ‘to cry’, muskarānā
‘to smile’, and samajhnā ‘to understand’).
Because, in Hindi, several verbs can take both an ergative marked A and
a non-marked A, Mohanan (1994) assumes that the ergative case is semanti-
cally determined; more specifically, it indicates “conscious awareness”.²⁹ In
Mohanan’s opinion, perfective verb forms do not take an ergative marked A
when they are transitive but when A designates a deliberate agent. If this view
were correct, Hindi could be considered as an “active” language, because then
its case marking would be semantically determined (cf. Dixon 1994: 28–29, and
Chapter 1). However, Mohanan also points out that there are many exceptions to
the alleged pattern based on “conscious awareness”. For instance, she observes
that the verb cillānā ‘to scream’ can pattern ergatively or accusatively, accord-
ing to the volitionality of A. However, “all instances of verbs like cillaa ‘scream’
cooccurring with ERG subjects are not necessarily associated with deliberate-
ness; there are a few instances that do not always conform to the pattern in terms
of conscious choice” (Mohanan 1994: 73). Indeed, there are more verbs whose
A-argument is not determined by semantic categories than there are verbs with
an A having a semantic connotation of deliberateness. Furthermore, the verbs
bolnā ‘to speak’ and lānā ‘to bring’ never take an ergative A, although the actions
of speaking and bringing can arguably be deliberate (Kachru 1980). An even more
important counterargument to Mohanan’s idea of “conscious awareness” is that
Mohanan bases her argument that the ergative case is semantic on the align-

28 According to Kellogg (1938: 240), this verb is always intransitive, and if a direct object is
mentioned, it takes a passive construction.
29 The discussion on whether the ergative is a lexical or structural case returns regularly in
generative accounts, often in connection with the difference between case and agreement pat-
terns, e.g., Woolford (1997, 2006). Cf. also Chapter 6.
110 Indo-Aryan

ment of exceptional verbs. Most verbs however follow a regular structural pattern
of marking; that is, many transitive verbs do not refer to deliberate actions but
are patterned with an ergative, and many intransitive verbs express deliberate
actions but pattern with a nominative S, without an ergative.
Davison (1999: 197), in contrast to Mohanan (1994), considers the ergative to
be a structural case with an occurrence pattern that is not determined by seman-
tic factors. She arrives at this conclusion after investigating the same unerga-
tive and anomalous intransitive verbs in Hindi as Mohanan (1994). According to
Davison, the verbs jītna ‘to win’ and hārnā ‘to lose’, bolnā ‘to talk’ and baknā ‘to
ramble on’, and paṛhnā (not in the sense of ‘to read’, but in the sense of ‘to take
a course’), among others, take the intransitive pattern. This is not because their
subjects are not volitional agents but because an O is lacking, and they are there-
fore semantically intransitive. The following example from Hindi illustrates the
intransitive use of the verb bolnā ‘to talk’ :

(83) mujh=se ek ādmī bol-ā

I.obl=abl one man say-pst.m.sg
‘A man said to me…’

Another pair of transitive verbs in Hindi without an ergative pattern is lānā ‘to
bring’ and bhūlnā ‘to forget’, although their agents can be volitional. Mohanan
argues that these verbs are lexically idiosyncratic, i.e., they are exceptional in
patterning only in Hindi. However, another explanation is at least equally plausi-
ble. The verb bhūlnā frequently occurs in a light verb construction with an intran-
sitive verb, and the verb lānā seems to be derived from such a light construction.
The verb bhūlnā, for instance, is most often used in the compound bhūl jānā.
The intransitive verb jānā means ‘to go’, and when used as a light verb, it adds
the connotation of a transition to the main verb; thus, bhūl jānā also means ‘to
forget’. The verb lānā appears to be derived from the compound verb construction
le ānā, which means ‘to come after having taken something’. The verb ānā ‘to
come’ is the intransitive light verb (cf. Montaut 2004: 180–181; Butt 2006). Hence,
in analogy to their more frequently used compound variants, lānā and bhūlnā
may also be analyzed as intransitives, with the result that they cannot take an
ergative A (cf. Creissels 2008: 13 on light verb patterns; Chapter 1).
3 Eastern Indo-Aryan: Asamiya
Eastern Indo-Aryan languages are spoken in the easternmost states of India,
Bangladesh, Bhutan and Nepal. Traditionally, three languages form the core of
Eastern IA, i.e. Asamiya, Bangla and Oriya. They are predominantly spoken in the
states of West-Bengal, Assam and Orissa in Eastern India and Bangladesh. Maith-
ili, Magahi and Bhojpuri are three Eastern IA languages that are spoken around
Central India. All of the Eastern IA languages are closely related to one another
and are said to derive from one particular branch of the Prakrit languages, i.e.,
the Magadhan branch (cf. Grierson 1903 V, I: 5; Chatterji 1926: 6). Eastern IA
languages have rarely been discussed in the linguistic literature. The exception
is Bangla, which has been investigated with regard to its complex non-canon-
ical subject constructions (cf. Klaiman 1976, 1980; Dasgupta 2004), and, more
recently, Maithili, which has attracted attention due to its exceptional agreement
pattern (cf. Stump and Yadav 1988; Bickel et al. 1999; Yadava 1999; Bickel and
Yadava 2000; Bickel 2003, 2004). Asamiya is less known in linguistics, yet it is
the most eastern of the IA languages, and thus represents the border of the Indo-
Aryan language area. In India only, around 13,168,500 people speak Asamiya
(Census 2001). The language is primarily spoken in the state of Assam in Eastern
India, and secondarily, it is spoken in the states Meghalaya, West Bengal and
Arunachal Pradesh, and in the neighboring countries of Bhutan and Bangladesh.
In this area, it is surrounded by non-IA languages, including Tibeto-Burman, Tai,
and Austro-Asiatic. Bangla and Oriya are also in linguistic contact with non-IA
languages but to a far lesser extent. In Asamiya, the typical Indo-Aryan gram-
matical properties can be observed, although many of them have been modified
through historical evolution and language contact.
The earliest grammars of Asamiya were composed at the time of British colo-
nialism e.g., Robinson (1839), Brown (1848), Barua (1859), Nicholl (1885), Medhi
(1936). Grierson’s Linguistic Survey (1903) attributed two volumes to Eastern
IA, and Chatterji (1926) compared Bangla with Asamiya and other Eastern IA
languages in his impressive description of the Bangla language. More recently,
important works have been produced by Kakati (1962), U. Goswami (1978) and
G. Goswami (1982). Furthermore, teaching material is found in Das Gupta (1981),
Sharma (1963) and Baruah (1980). For the following chapter, original source
material has been found in Baruah (1980), children’s stories (Bhattacharyya 1997)
112 Eastern Indo-Aryan: Asamiya

and a short story by Kakoti (2011), translated by Meenaxi Barkataki-Ruscheweyh,

who also functioned as the main informant for the examples.¹
On the western border of the Eastern IA language area, the Bihari languages are
spoken. Grierson (1883) distinguished three Bihari languages: Magahi, Maithili
and Bhojpuri.² The Bihari languages are geographically situated between the
Eastern and Central region, which has resulted in ongoing debate on the issue of
whether they are more closely related to Bangla, Asamiya and Oriya, or to Hindi
(Verma 2003: 499; Cardona 1974; Jeffers 1976). The main motivation to include
them in this Chapter on Eastern IA is the fact that the alignment patterns of
the Bihari languages are more interesting to compare with the patterns of the
core Eastern IA languages than with the patterns of most Central IA languages,
because they seem to show more similarities with the Eastern IA languages.
For this reason only, Awadhi, a language not included in the Bihari languages
but typically considered to be an Eastern variety of Hindi, is discussed in this
Section and not in the Chapter on Central Indo-Aryan. For other topics, such as
phonetics or the lexicon, it is very well possible that the Bihari languages, and
certainly Awadhi, tend to be closer to Central IA.
The following Section 3.1 begins with a general description of Asamiya,
focusing on the case marking of the core arguments (3.1.1–3.1.3) and on verb
agreement (3.1.4). In Section 3.2, the indication of transitivity in the verb conjuga-
tion in Asamiya and other Eastern IA languages is discussed. A third point to be
raised is the influence of non-IA on the case marking in Eastern IA (Section 3.3).
The speakers of Eastern IA languages are among those remotest from the Central
Indo-Aryan region. The opportunities to exchange thoughts and language with
the indigenous population of East India throughout the course of time were mani-
fold. It is therefore not surprising that in the phonology, morphology and, to a
lesser extent, the syntax of Eastern IA, the influences of non-IA languages are
apparent.

1 An electronic corpus of Asamiya has been developed by the University of Lancaster and the
University of Guwahati, as part of the EMILLE corpus (http://www.lancs.ac.uk/fass/projects/
corpus/emille/). However, there are still several problems with the typography and grammar
of the texts, and the use is further complicated by the lack of annotation. Therefore, the more
traditional analysis of written texts has been preferred in this chapter.
2 Althernatively, the Bihari languages were called Eastern Hindi or the Gaudian languages.
Grierson determined the name Bihari on the basis of the name of the area where the languages
were predominantly spoken, i.e., the state of Bihar in India (Grierson 1883: 2).
 Asamiya: description 113

3.1 Asamiya: description³

3.1.1 Case marking in Asamiya

Asamiya makes no syntactic differentiation between the singular and plural

forms. It is possible, but not obligatory, to mark the plural by means of suffixes,
e.g. -lok, -haṃt or -bor among others (Goswami 1978: 34). Some are typically used
with inanimate nouns, others are only used with animate nouns. These plural
suffixes are not obligatory, and a noun occurring without a suffix may even be
interpreted as a plural if such an interpretation fits the context. Asamiya has also
lost the grammatical gender distinction, and there is no indication of gender
concord between nouns and adjectives.
Case marking in Asamiya only applies to nouns and pronouns (hence not to
adjectives). The language does not show a distinct opposition between a nomina-
tive and an oblique inflectional case, and it does not strictly follow Masica’s layer
division (cf. Masica 1991, Section 2.4.1). There are five cases with different case
markings, i.e., nominative, ergative, objective, instrumental, genitive and loca-
tive, as illustrated in Table 3.1.

erg -e
obj -k/-lai/-k-lai
ins -ere/-edi-/-ra dwārā
gen -r
loc -t

Table 3.1 Case in Asamiya

The nominative is the unmarked case. The other case forms may be considered
inflectional, although the difference with postpositional marking is not entirely
clear. According to Grierson (1903, V, I: 13), the terminations -k, -r, and -t were
originally postpositions but now have been “worn away by centuries of attrition”.
The objective ending lai is postpositional, while the instrumental ending -ere

3 For the transcription of Asamiya I follow the indological convention of the transcription for
Devanagari. This implies that it is not a phonetic but a literal transcription, and some phono-
logical rules are not followed. <a> in Asamiya is, for instance, often pronounced as [o]. Asamiya
also lacks cerebral and palatal sounds, although they still occur in the script. <c> and <ch> are
pronounced as [s], while <j> and <jh> are pronounced as [z]. The cerebral plosives are all pro-
nounced as dental plosives. The fricatives from IA are in Asamiya replaced by a voiceless velar
fricative [x].
114 Eastern Indo-Aryan: Asamiya

can be split into the inflectional ending -e, combined with a postposition re. The
ending -e could be considered as a marker of the oblique case, because the post-
positions re, di or ra can be added. Occasionally, another postposition is added
to a noun taking the ending -e, resulting in an adverbial construction, such as
rām-e=saite ‘with Ram’ (cf. Kakati 1962: 286). However, -e is not added to adjec-
tives, and the case terminations of the locative and genitive are directly added to
the nominative form, not to the form with -e (as would be expected if -e were the
marker of the oblique case). Historically, -e was a former instrumental marker,
but for contemporary instrumentals, the ending has been reinforced by adding
another suffix (cf. Kakati 1962: 286). Occasionally, -e still occurs as the ending of a
true instrumental. These instances seem to be highly idiomatic, though, forming
adverbial constructions, e.g., anāyās-e ‘with ease’, lāh-e lāh-e ‘(very) slowly’ (cf.
Kakati 1962: 286). Of the other instrumental endings, -ere and -edi are the instru-
mental suffixes of non-human entities (cf. Baruah 1980: 523). Whereas -ere is
more often used in the Eastern part of the Asamiya area, -edi is a recently devel-
oped instrumental marker used in Western Asamiya (cf. Post 2008: 68).⁴ -ere and
-edi are shortened to -re and di when the preceding sound is a vowel.
The instrumental -ra dwārā marks a human agent of a passive construction,
except when there is a perfect participle in the construction. In that instance, -e
is the marker of the agent of this kind of attributive construction with a perfect
participle, e.g., Gopāl-e likhā puthi, ‘the book written by Gopal’ and rām-e karā
kām, ‘a work done by Ram (cf. Kakati 1962: 285). Again, this use goes back to the
old instrumental function of the case ending -e, which seems to have survived
in this construction. The perfect participle used non-attributively does not occur
frequently in Asamiya; instead, a construction with a synthetic verb expressing
the perfect aspect is preferred.⁵
The case markings of our interest are those which indicate the core argu-
ments of a sentence. They will be discussed in the following Sections 3.1.2–3.1.3.

4 -edi may have originated from the combination of the original instrumental marker -e and the
serial verb di ‘to give’, which has grammaticalized into a case marker of means (cf. Post 2008:
68).
5 When asked to translate the English sentence ‘The book was written by Ram’, an informant
translated the sentence in an active way as kitāb-khan rām-e likh-ich-e (book-def Ram-erg
write-prf-3.tr) ‘Ram has written the book.’. When asked to translate the sentence ‘Ram wrote
the book’, to find a difference between the two constructions, the same translation was given,
only this time a past perfect form of the verb was used: kitāb-khan rām-e likh-ich-il (book-def
Ram-erg write-pst-prf.3) ‘Ram had written the book.’.
 Asamiya: description 115

3.1.2 Case marking of the subject

In various Asamiya grammars, the language has been analyzed as having a typo-
logically rare subject marking, with an unmarked as well as a “marked” nomi-
native case (cf. Goswami 1978: 40). Many authors consider the “ergative” case
ending -e not as the ergative case, but as a variant of the unmarked nominative,
which they leave undefined as the default form of the noun (cf. Grierson 1903;
Kakati 1962; Sharma 1963; Goswami 1978; Das Gupta 1981; Goswami and Tamuli
2003). The marked form on -e, however, is only used for A. S is always indicated
in the unmarked stem form; cf. the following example:

(84) bubul-e paṛh-i+āch-e, bubul sowā nāi

bubul-erg read-prog+aux-prs.3 bubul sleeping not be.prs
‘Bubul is reading, Bubul is not sleeping. (Baruah 1980: 230)

The term “nominative” for this -e-marking is confusing, because there is an

unmarked nominative case form in Asamiya. Typologically, a case form which is
only used to mark A traditionally represents an ergative case. To avoid any confu-
sion, I prefer to use the term ergative (cf. also Amritavalli and Sarma 2002). All
nouns display the addition of -e when they are used as A of a finite verb. Some
phonological restrictions apply: for nouns ending in -e or -ā, the -e alternates
with -i (cf. Baruah 1980: 251), e.g., Mainā-i, ‘Moina’ (boy’s name), rajā-i, ‘king’.
Personal pronouns behave slightly differently for subject marking, as is illus-
trated in the following Table.

1sg 2sg 3sg.m 1pl 2pl 3pl 2hon

nom maï taï/tumi si/teoṃ āmi tahaṃt sihaṃt āpuni

erg maï taï/tumi si/teoṃ āmi tahaṃt-e sihaṃt-e āpuni
obj mo-k/ to(mā)-k/ tā-k/ āmā-k/ tahaṃta-k/ sihaṃt-ak/ āponā-k/
mo-lai to(mā)-lai tā-lai āmā-lai tahaṃta-lai sihaṃta-lai āponā-lai

Table 3.2 Pronouns in Asamiya

Note that although the number distinction is not grammatically indicated in

nouns in Asamiya, the pronouns have a different form for the singular and the
plural. However, only the first person plural is an independent form; the second
and third person plurals are the singular forms with a plural affix like -haṃt or
-lok (e.g., tumi-lok) added to them. The conjugation of these two forms follows the
116 Eastern Indo-Aryan: Asamiya

paradigm of the nouns, as exemplified in ex. (85), where the second person plural
pronoun is marked with a plural lexical marker lok ‘people’ and the -e suffix.

(85) toma-lok-e tā-k eko no-kow-ā

you-pl-erg he-obj a not-say-prs.2
‘You don’t tell him anything.’ (Baruah 1980: 139)

The three forms for the third person pronoun, si, teoṃ and tekhet, correspond to
the three second person forms taï, tumi and āpuni in grades of honorific quality,
with the former implying inferiority/equality, and the last one implying a higher
grade of honorificity (cf. Sharma 1963: 93). For the singular pronouns and the first
person plural, the form of S is the same as that of A.
Other pronominal forms tend to end more often on -e, even when used as S,
for instance āṭāi/āṭāye and sakale/sakalowe, both meaning ‘all’ (cf. Baruah 1980:
226).

(86) ne sakalow-e ekailage eke-khan cāṇa-t bahib+nowārā

or all-erg together one-def meeting place-loc sit+not being able
ha-ich-e?
be-prf-3
‘Or couldn’t all sit together at one place?’ (Kakoti 2011: 3)

(87) āmi sakalow-e jagannāth mandira-t somā-m

we all-erg Jagannath temple-loc go-fut.1
‘All of us will go into the Jagannath temple.’ (Baruah 1980: 574)

The marking with -e is not only limited to subjects of finite verbs. Non-finite verbs
such as converbs, which occur regularly in Asamiya (cf. Post 2008), also take A’s
marked with -e, cf. the following example.

(88) mīnuw-e chabi āṃkiba-lai kal-at si chabi

Minu-erg picture draw.inf-obj ask-cvb he picture
āṃkiba-lai dhar-il-e
draw.inf-obj begin-pst-tr.3sg
‘After Minu had asked [him] to draw a picture, he started to draw a picture.’
(Baruah 1980: 865)

If the subject of both the converb and the main verb is the same, Asamiya seems
to prefer that the subject is marked according to the verb it precedes, irrespective
 Asamiya: description 117

of whether that verb is transitive or intransitive.⁶ Hence, it can sometimes appear

as if an intransitive main verb takes an ergative-marked A, whereas, instead, the
ergative marking is determined by the transitivity of the converb, as illustrated by
the following examples.

(89) amṛt-e eko hā-nā na-kar-i āh-i+āch-e

Amrit-erg one yes-no not-do-cvb come-prog+aux-prs.3
‘Without responding, Amrit keeps walking.’ (Kakoti 2011: 16)

(90) la’rāṭow-e āponā-lai bāṭ+cā-i+cā-i alap āgate

son-erg you.h-obj way+look-cvb+look-cvb little ago
su-i+pari-ch-e
sleep-cvb+fall-prf-3
‘After the boy waited and waited for you, he fell asleep a little while ago.’
(Baruah 1980: 840)

The general rule for the marking of the subject in Asamiya is that A is marked and
S is unmarked. However, there are counterexamples to this rule. Unergative verbs
(which are particularly common in IA, cf. Chapter 3) generally take a marked
subject. Some of the most common unergative verbs in Asamiya are nāc ‘dance’,
hāṃh ‘laugh’, kānd ‘cry’, and daur ‘run’ (cf. Baruah 1980: 482; Amritavalli and
Sarma 2002). The examples below illustrate some unergative patterns:

(91) ākāsa-t megh-e garaj-ich-il

sky-loc cloud-erg roar-prf-pst.3
‘Clouds roared in the sky.’ (Baruah 1980: 459)

(92) ingrāj-e bor beyākoi nāc-e

Englishman-erg very bad dance-prs.3
‘Englishmen dance very bad.’

The range of unergative verbs in Asamiya is quite extensive and includes verbs
that are not commonly considered as typically unergative. Compare, e.g., the fol-
lowing example of Goswami and Tamuli (2003: 432) with the verb ‘to swim’⁷:

6 This is the same rule as in Gultari Shina (Hook 1996), but exceptions can occur.
7 Note that in Hindi, the verb tairnā ‘to swim’ can be used intransitively as well as transitively,
meaning ‘to swim across’.
118 Eastern Indo-Aryan: Asamiya

(93) ram-e saṃtur-il-e

Ram-erg strike-pst-3tr
‘Ram swam.’

The following expression is also quite common in which the concept ‘to swim’ is
expressed by a transitive, lexical compound construction:

(94) ram-e saṃtor mar-il-e

Ram-erg swim beat-pst-3tr
‘Ram took a swim.’

The verb meaning ‘to sleep’ also regularly occurs with a subject marked with -e
in Asamiya, e.g.,

(95) rām-e āji derikoi su-l-e

Ram-erg today late sleep-pst-3tr
‘Ram slept/went to bed late today.’

In comparison with the other IA languages, the verb meaning ‘to sleep’ behaves
more often as an unergative verb in Asamiya.
Intransitive verbs that take an -e marked subject also show the transitive con-
jugation. Asamiya has a different conjugation for transitive and intransitive verbs
in the third person past tense (Section 3.2.1). Kakati (1962: 334) points out that
some verbs in Asamiya always take a transitive ending in the third person, and
he argues that in all of these verbs, a cognate object is implied. His examples are
kāhile ‘he coughed’, larile ‘he ran’, and sule ‘he slept’. These verbs also take an
ergative-marked subject (cf. ex. (95)).
An even more intriguing example is the following one, which is also found in
Goswami and Tamuli (2003: 425):

(96) ram-e sadae saṃtur-iba-lai za-e

Ram-erg daily swim-inf-obj go-prs.3
‘Ram goes to swim every day.’

The main verb za-e is intransitive, just like the infinitive saṃtur-iba, but, ‘to swim’
is unergative. The subject is marked because it is either the subject of the infini-
tive, assuming that an infinitive form can control the marking of its subject (as
a converb) or because it is simply an exceptional marked subject of the clearly
intransitive main verb za-e meaning ‘to go’. Goswami and Tamuli do not elabo-
rate on this example, but in general, the apparent ungrammatical marking of an
 Asamiya: description 119

intransitive subject is quite common in colloquial Asamiya. Kakati (1962: 285), for
instance, points out that S sometimes takes an -e ending when the argument is
emphasized, thus implying a certain discursive function of -e. Additionally, -e fre-
quently occurs as a marker of the subject of a non-finite verb construction, which
again is evidence of the extensive distribution of the marker (cf. 3.1.2). Moreover,
the use of a marked A across all tenses is unusual in non-Eastern IA languages,
where the marking is generally restricted to the perfective. The frequent distribu-
tion of -e has contributed to the general view that -e is a variant of the unmarked
nominative, though this view ignores the predominant distribution pattern of -e
as an ergative marker.

3.1.3 Differential object marking and the dual marking of IO

3.1.3.1 O-marking
In Asamiya, the objective marker of O and IO is -k. O is only marked when it is
animate and/or definite. If O is inanimate, then it only receives a marking when
its definiteness is emphasized. On the other hand, an animate third person O may
not take the marking when it is explicitly indefinite. The O in ex. (97), inginiyāriḥ,
for instance, is indefinite and hence also unmarked. The O in ex. (98) is animate;
however, the ‘new wife’ has not been referred to earlier in the discourse. It is
therefore not possible for the hearer to readily identify her, and this is precisely
the reason why mānuh occurs in the nominative, not in the objective.⁸

(97) taï etiyā inginiyāriḥ na-paṛ-a ne?

you now engineering not-do-prs.2 q
‘Don’t you do engineering now?’ (Baruah 1980: 156)

8 Note that it is easy to consider the O in these examples as incorporated, in particular in ex.
(98), which is a kind of double object construction. biyā ‘wedding’ is one of the two O’s, but
it forms a lexical compound with the verb karā, yielding the meaning ‘to marry’. The person
with whom the marriage is contracted, is the second O, in this example mānuh ‘wife’. For a
discussion of incorporation in Hindi and the reasons why a similar unmarked O in that language
cannot really be considered to be an incorporated noun, see Section 2.4.2.2 (cf. also Klaiman
1990). The syntactic incorporation of an unmarked O in Asamiya is more difficult to assess than
it is in Hindi, because Asamiya lacks, for instance, O-agreement, which is a mechanism that can
be taken into account as syntactic evidence for or against the incorporation of O (depending on
whether there is agreement with O or not).
120 Eastern Indo-Aryan: Asamiya

(98) rajā-i ejanī natun mānuh⁹ biyā karā-l-e

king-erg a.f new wife[f] wedding do-pst-3tr
‘The king married a new wife.’ (Bhattacharyya 1997: 54)

Compare ex. (98) with the following example (99). The verb has the same meaning,
but O is now a definite and given referent, which has already been mentioned
earlier in the discourse:

(99) rajā narakāsur-e dewī-k biyā karā-ba-lai

king Narakasur-erg goddess-obj wedding do-inf-obj
khuj-ich-il
want-prf-pst.3
‘King Narakasur wanted to marry the goddess.’ (Baruah 1980: 652)

In the next example, the inanimate O kathāke is definite, determined by the

adjective ‘all’, and marked with the postposition -k:

(100) si ghara-r sakalo kathā-k-e sambhāli la-l-e

he house-gen all thing-obj-emph charge take-pst-3tr
‘He took charge of every single thing in the house.’ (Baruah 1980: 1021)

First and second pronouns used as O are always marked, because they are mostly
animate, and they are never indefinite. The form of the pronouns marked with -k
shows a modification of the stem vowel, e.g., maï becomes mo- and taï becomes
to- (cf. Table 3.2).

(101) āpuni mo-k kenekai cin-e

you.h me-obj how know-prs.2h
‘How do you know me?’ (Baruah 1980: 156)

Third person pronouns, e.g. tā-k, are also almost always marked, because their
referents tend to be identified previously in the discourse.

9 The use of mānuh is slightly unusual in this example, because normally mānuh refers to a
man, although it can be used in the meaning of ‘person’.
 Asamiya: description 121

3.1.3.2 Dual marking of IO

Whereas the marking of O is dependent on its semantics, the IO is obligatorily
marked in Asamiya. The marking of an IO is the same as that of an animate/definite
O, viz. -k. Nevertheless, unlike other IA languages, Asamiya differentiates between
O and IO by means of a second marker which only functions as marker of IO, i.e.,
-lai. -k is the most common marker of an IO functioning as a recipient in general,
whereas -lai is the marker of the beneficiary role (Kittilä 2005b). According to Baruah
(1980: 211), -lai is only used for some categories of IO, namely, for the so-called
“destinational dative” (what Kittilä 2005b calls the “recipient-beneficiary”)¹⁰. It
denotes someone to whom something is given for his/her benefit (Das Gupta 1981:
92); or, it may as well denote an action performed for the benefit of the lai-marked
argument. The following examples display appropriate instances to use -lai.¹¹

(102) maï tāi-lai eikhan cāikel ān-ich-oṃ

I her-obj this cycle bring-prf-1
‘I have brought this cycle for her.’ (Baruah 1980: 193)

(103) āji puwā bhāiṭi-lai-o eṭā pāijāmā cilā-l-e

today morning Bhaiti-obj-emph a trousers stitch-pst-3tr
‘This morning [she] stitched trousers for Bhaiti also.’ (Baruah 1980: 406)

In a sentence in which the IO is exclusively the recipient and not the beneficiary,
the neutral marking -k is preferred, such as with speaking verbs and neutral
transfer verbs. In ex. (104) and (105) both -k or -lai are possible, depending on
whether the emphasis is on the mere transfer of the basket or on the benefit of the
transfer for the girl.

(104) mālik-e āpunā-k pāthar-bor ka’t āch-e

master-erg you.h-obj stone-pl where be-prs.3
āgatei ka-l-e
earlier tell-pst-3tr
‘The master already told you where the stones are.’

10 According to Kittilä (2005b: 270–271), the difference between the semantic roles of ben-
eficiary and recipient-beneficiary is that the role of beneficiary simply benefits from an action,
whereas in a construction with the role of recipient-beneficiary, there is a transfer that brings
benefit to the recipient-beneficiary.
11 In Hindi, the lai marked argument in these examples may be indicated by ke lie, though the
objective postposition ko is not impossible.
122 Eastern Indo-Aryan: Asamiya

(105) āpuni chowālī-janī-k hāladhīyā phula-r tukuri eṭā di-ch-il

you.h girl-def-obj yellow flower-gen basket a give-prf-pst.3
‘You had given the girl a basket with yellow flowers.’

According to Das Gupta (1981: 92), -lai marking the IO of a variant of the verb
meaning ‘to give’ indicates a beneficiary gift. Conversely, when -k marks the IO,
it denotes an unreserved gift, which cannot be taken back, and is not necessarily
for the good of the recipient. Therefore, the receivers of untouchable or untenable
entities, such as messages, advice, news etc., are marked with -k, as illustrated in
the following examples.

(106) āpuni bipula-r bandhu-jana-k khabaraṭo di-ch-e ne?

you.h Bipul-gen friend-def-obj message give-prf-2h q
‘Have you given the message to Bipul’s friend?’ (Baruah 1980: 229)

(107) rabibāre maï sihaṃta-k kathāchabi dekhu-wā-oṃ

Sunday I they-obj picture see-caus-prs.1
‘I show them pictures on Sunday.’ (Baruah 1980: 709)

If it is impossible to replace -lai with -k, -lai can be replaced with a longer postpo-
sition, as for instance in sentences such as the following, where the genitive case
form combined with kāraṇe or bābe(i) functions in the same way as -lai in ex.
(108)–(109). Again the meaning of a recipient-beneficiary is implied, e.g.:

(108) maï tahaṃta-r=kārane natun kitāp ān-ich-oṃ

I you-gen=for new book bring-prf-1
‘I have brought new books for you.’ (Baruah 1980: 252)

(109) teoṃ-r=kāraṇe mo-k eṭā ausadh lāg-e

he-gen=for I-obj a medicine need-prs.3
‘I need a medicine for him.’ (Baruah 1980: 157)

Baruah (1980: 211) further mentions that gen=kāraṇe/bābei is the only option for
marking a beneficiary when there is no sense of ‘motion’ implied. In the following
example, gen=bābei is not likely to be replaced by -lai:

(110) āru ājhikopati la’rā-keiṭā-r=bābei sakalo nijhei kar-i+ga’-l

and until today boy-pl-gen=for everything self do-cvb+go-pst.3
‘And till today he did everything for the boys himself.’ (Kakoti 2011: 4)
 Asamiya: description 123

The postposition -lai indicated originally a locative and is still used in this sense,
with various degrees of abstractness, compare ex. (111)–(115).

(111) edin camunḍī mandira-lai ga-ich-e

one day chamundi temple-to go-prf-3
‘One day they have gone to Chamundi temple.’ (Baruah 1980: 212)

(112) tekhet awase’ sesa-lai na-r-’l

she of course end-to not-wait-pst.3
‘She, of course, did not wait for the end [to come].’ (Baruah 1980: 342)

(113) maïnā-lai tā-r maram lāg-i+ga’-l

Moina-to he-gen love attach-cvb+go-pst.3
‘He felt a wave of love for Moina.’ lit.:
‘Love directed at Moina came over him.’ (Kakoti 2011: 6)

(114) māhīmāk-e tejīmalā-k biyā-lai pathā-l-e

stepmother-erg Tejimala-obj marriage-to send-pst-3tr
‘The stepmother sent Tejimala to the marriage.’ (Baruah 1980: 770)

(115) si mainā-r mukha-lai cā-l-e

he Moina-gen face-to look-pst-3tr
‘He looked at Moina’s face.’ (Kakoti 2011: 6)

The postposition lai also marks the goal-infinitive, as in ex. (116):

(116) teoṃ āmā-k khāb-lai di-ch-il

he us-obj eat.inf-to give-prf-pst.3
‘He gave us to eat.’ (Baruah 1980: 666)

The different ditransitive verbs may thus be split into two groups, 1/ the verbs
indicating gifts or displacements, e.g., ‘to give’, ‘to bring’ etc., and 2/ verbs indi-
cating a displacement of something untouchable and irretrievable, such as ‘to
say’, ‘to tell’, ‘to betray’ etc. For the first group, there is an option to use -k or -lai
as markers of the IO; in the second group, the IO is invariantly marked with -k.
Note that the verb meaning ‘to give’ occurs equally with an IO marked with -k as
with -lai.
An earlier variant of -lai, the double marking -k-lai, is still occasionally
attested, however, it is less frequent (cf. Kakati 1962: 286). -k-lai is not a combina-
tion of the objective suffixes -k and -lai. According to Kakati (1962: 289), it is his-
124 Eastern Indo-Aryan: Asamiya

torically the former genitive suffix -k to which -lai is added, not the contemporary
objective suffix -k. Clearly, -k-lai is reminiscent of the Hindi complex postposition
ke lie. The meaning of -k-lai is, as expected, close to the meaning of -lai.

(117) amrita-k-lai gāwaṃ-r mānuh-e bahut āsā kar-ich-il

Amrit-for village-gen people-erg a lot hope do-prf-pst.3
‘The villagers had hoped a lot for Amrit.’ (Kakoti 2011: 4)

3.1.3.3 Experiencer constructions

The experiencer construction, or non-canonical subject construction with expe-
riencer verbs, frequently occurs in IA languages (cf. Verma and Mohanan 1991;
Bhaskararao and Subbarao 2004, Section 2.4.3). In all IA languages, experiencer
verbs show an alignment that is different from the other verbs. Commonly occur-
ring examples of experiencer verbs in IA are verbs meaning ‘to know’, ‘to think’,
‘to consider’, ‘to become visible’, ‘to meet’¹², ‘to like’, etc. Many of these verbs are
compound verbs, consisting of a noun or adverb combined with a verb.
In Asamiya, the objective marker -k is also the suffix marking the experiencer.
The verb lāg, literally meaning ‘to be attached to’, frequently occurs in Asamiya
in an experiencer construction. The verb appears in the third person, while the
experiencer takes the postposition -k, as illustrated in the following example.

(118) teoṃ-r=kārane mo-k etā ausadh lāg-e

he-gen=for I-obj now medicine be attached-prs.3
‘I need a medicine for him.’ (Baruah 1980: 157)

In contrast to the other IA languages, the genitive case marker also regularly func-
tions as the indicator of the experiencer in Eastern IA. In Central and Western IA,
this is not a function of the genitive.

(119) cowāc-on, tomā-r pacand ha-ich-e ne?

look-imp you-gen like be-prf-3 emph
‘Look, do you like it?’ (Baruah 1980: 193)

12 The verb meaning ‘to meet’ in IA could be better paraphrased as ‘to stumble upon’, ‘to be
available to someone’. In Tsez, the subject of this verb is expressed with an allative case (cf.
Cysouw and Forker 2009).
 Asamiya: description 125

(120) rātul rinṭi-r bar ānando lāg-ich-e

Ratul Rinti-gen very joy be attached-prf-3
‘Ratul and Rinti are feeling very happy.’ (Baruah 1980: 212)

(121) tā-r haṭhāt mana-t par-i+ga’-l ratnezwa-r

he-gen suddenly mind-loc come-cvb+go-pst.3 Ratneshwa-gen
dedāi-r kathā
brother-gen word
‘He suddenly remembered the words of Ratneshwa-Dedai.’ (Kakoti 2011: 6)

The use of the genitive is the general way to mark the experiencer in Bangla, e.g.
ex. (122) (Radice 2007: 145).

(122) ama-r am bhalo lag-e

I-gen mango good be attached-prs-3
‘I like mangoes.’

In Oriya, the same non-canonical construction also occurs, with either an objec-
tive marker or a genitive, cf. ex. (123) (Ray 2003: 467).

(123) rama=ku bhoka lag-u-ch-i

rama=obj hunger be attached-prog-prs-3sg
‘Rama is hungry.’

In most IA languages, including Asamiya, obligational constructions take the

same form as experiencer constructions (cf. Chapter 2). Masica (1990: 336),
however, points out that in Asamiya, an obligation can also be expressed with an
impersonal construction, as in the following example. The verb in this example
does not agree with maï, although maï is a nominative form:

(124) maï zā-bo+lāg-e

I go-inf+be attached-prs.3
‘I have to go.’

The use of non-canonical subjects is determined by the semantics of the verb

(cf. Section 2.4.3). While in an ergative pattern, the marking is determined by the
transitivity of the verb, non-canonical subjects exclusively occur with specific
verbs in the lexicon, viz. verbs with an experiencer meaning. As a rule, the verb
does not agree with the experiencer argument but with the stimulus role, which
126 Eastern Indo-Aryan: Asamiya

is in the unmarked case form. Whereas the experiencer role is semantically deter-
mined by the meaning of the verb, ergative subjects are syntactically determined
by the form and tense of the verb. This position concurs with Genetti’s (2007:
294) view that experiencer constructions have “a different argument structure
altogether”. The occurrence or non-occurrence of experiencer constructions in
a language is not influenced by the main alignment pattern of the language, nor
does it influence that pattern.

3.1.4 Verb agreement in Asamiya

In Eastern IA, the copula has merged with the participial form, thus forming
a single verb, consisting of a stem, a tense/aspect infix (the former copula),
and a personal ending. Asamiya has an extensive system to indicate the honor-
ific grade of the second person. Three possible ranks of honorificity are distin-
guished in the verb conjugation, viz. inferior, familiar, and honorific. Note that
number and gender are not grammatically indicated in Asamiya, neither the
nouns nor the verbs indicate these categories. The following simple conjuga-
tions are there:
1. Present conjugation
2. Future conjugation
3. Perfect conjugation (can also be used to express the progressive aspect, cf.
Goswami 1981: 112): the infix -ich- to which the present tense endings are
added.
4. Past conjugation ending on the past suffix -il-.
5. Past perfect conjugation: the infix -ich- to which the past tense endings on
-il- are added.

The latter three are of particular interest for determining alignment in Asamiya,
therefore their endings are represented in more detail in the following tables.

1 -ich-o
2 inferior -ich-[a]
2 familiar -ich-a
2 honorific -ich-e
3 -ich-e

Table 3.3 Endings of the present perfect in Asamiya

 The transitive verb 127

1 -il-o
2 inferior -il-i
2 familiar -il-ā
2 honorific -il-e
3 -il-e/-il

Table 3.4 Endings of the past in Asamiya

1 ich-il-o
2 inferior ich-il-i
2 familiar ich-il-ā
2 honorific ich-il-e
3 ich-il

Table 3.5 Endings of the past perfect in Asamiya

Besides these simple conjugations, there are a number of complex conjugations

taking an auxiliary, which express various kinds of aspect. For instance, the con-
jugated future forms of the verbs acch or thāk are added to form the future con-
tinuous, and similarly, the past continuous is formed with the past tenses of these
verbs (cf. Das Gupta 1981: 120). Subjunctive forms are formed with the invariable
suffix heten which is added to the conjugated verb.
The agreement of the verb in Asamiya is with the subject, under all cir-
cumstances. However, the past tense of the third person ending on the suffix
-il displays a remarkable pattern depending on the transitivity of the verb. An
intransitive verb displays no ending, and a transitive verb adds the ending -e. The
following Section 3.2 focuses on this unusual feature of the verb in Asamiya, and
on similar patterns in other Eastern IA languages.

3.2 The transitive verb

In all Eastern IA languages, there is a remarkable emphasis on the distinction

between transitive and intransitive verbs in the past tense of the third person.
The different marking of transitive and intransitive verbs is typologically rare, but
it is attested in other languages. We find it, for instance, also in Tibeto-Burman
languages such as Dolakha Newar (cf. Genetti 2007), and in Austronesian Pileni
(Næss 2002), amongst others. In Asamiya, there is an ending -e that is added only
to the third person of transitive verbs, which indicates the difference between
transitive and intransitive third person conjugations. In other Eastern IA lan-
128 Eastern Indo-Aryan: Asamiya

guages, different mechanisms are used to indicate whether a verb is transitive or

not. Note that Asamiya is a language that displays person agreement on the verb.
In fact, having lost grammatical gender and number, the person of the subject
argument is, besides the indication of the honorific grade, the only category that
is cross-referenced on the verb.

3.2.1 The transitive verb in Eastern Indo-Aryan

In Asamiya, the third person transitive of the verb in the past tense takes -e as
ending, whereas the third person intransitive has no marking. Compare examples
(125) and (126). The first one is a transitive construction, and the verb takes the
ending -e, whereas the second one is an intransitive construction in which the -e
is absent from the verb. Note that this alternation only occurs with the past tense.
In the present perfect constructed with -ich-, the third person ending is always
-e, regardless of whether the verb is transitive or intransitive, cf. ex. (127). In the
future and the past perfect, on the other hand, the third person form does not
take a marking.¹³

(125) dām kimān la-l-e?

price how much take-pst-3tr
‘How much (price) did he take?’ (Baruah 1980: 318)

(126) seikhan eghār satuāh cal-il

that eleven weeks run-pst.3
‘That [movie] ran for eleven weeks.’ (Baruah 1980: 341)

(127) bipul kalai ga-ich-e?

Bipul where go-prf-3
‘Where has Bipul gone?’ (Baruah 1980: 229)

13 Grierson (1903 V, I: 403) notes that past perfect verbs may take an -e ending when they are
transitive, e.g., bulichile ‘he had spoken’. He also attested an intransitive verb marker with
-e, i.e., āhile ‘he came’. In later grammars, these exceptions are no longer mentioned. Kakati
(1962: 344), however, points out that the form āhile is used in Asamiya but not as a finite past
tense verb. The suffix -ile denotes a converb deriving from a past participle in -il to which the
locative suffix -e is added. The translation is not ‘he came’, but, e.g., maï āhile ‘upon my com-
ing’.
 The transitive verb 129

Perfective verb forms in Asamiya are historically derived from the OIA past par-
ticiple (Chatterji 1926: 940–941). The perfect verb form in Asamiya evolved from
a form of the main verb combined with the auxiliary verb (a)ch- ‘to be’ in the
present tense (Kakati 1962: 347). The auxiliary form takes the present endings,
and -e is the standard ending of a third person in the present tense. In contrast,
the past tense form on -il- was originally a past participle. According to Kakati
(1962: 330), forms on -il used to occur as adjective participles in Early Asamiya;
only in modern Asamiya has the form on -il- “been restricted to the predicate”.
He also adds that person agreement only appeared after this restriction in usage
was established. Synchronically, the different marking is a kind of indication of
transitivity of the verb, which is not always as clear. Kakati (1962) for instance
gives the example of maril-, a past tense of the verb mar, which could mean the
intransitive ‘died’ or transitive ‘beat’. Adding -e to maril makes it clear that the
verb is transitive.

(128) māhīmāk-e tejimalā-k barakai mār-il-e

stepmother-erg Tejimala-obj much beat-pst-3tr
‘The stepmother had beaten Tejimala a lot.’ (Baruah 1980: 771)

(129) becerī tejimalā mar-il

poor Tejimala die-pst-3
‘Poor Tejimala died’ (Baruah 1980: 771)

The past tense in Asamiya expresses a strict past meaning and lacks an explicitly
overt copula that expresses aspect (cf. Davison 2002). This past tense category
exemplifies the true counterpart of the past tenses based on the participle known
from the other IA languages. As in other IA languages, the forms on -il- display a
deviation of the alignment pattern that is normally found with the other tenses.
In Asamiya, it is only in the past tense that there are two different endings for a
transitive and an intransitive verb. According to Kakati (1962: 333, following Chat-
terji 1926: 983), the -e ending of a transitive past third person verb is an exten-
sion from the -e ending that is obligatory for the third person of the present and
perfect tenses. The past perfect, ending in -ichil, however, does not take any addi-
tional ending in the third person. This past perfect form is derived from a form of
the main verb combined with a past tense form of the copula, viz., (a)ch-il. This
copula is intransitive. Hence, the past perfect takes the endings of an intransitive
past tense on -il, which means that in the third person, there is no transitive -e
suffix added after -il.
In the other Eastern languages, Bangla and Oriya, a systematically different
ending for a transitive or intransitive past tense third person verb is absent. In
130 Eastern Indo-Aryan: Asamiya

1 bol-l-am
2nh bol-l-e
2h bol-l-en
3nh bol-l-o
3h bol-l-en

Table 3.6 Past conjugation of the verb bol ‘to talk’ in Bangla (Radice 2007: 167)

Bangla, the past tense also consists of an old participial form on -l- combined
with personal endings (cf. Chatterji 1926: 941).
However, Grierson (1903 V, I: 13) observes that even in standard Bangla, vari-
ation exists for the third person form of a verb in the past tense. In particular,
the ending of a transitive verb may be -e instead of -o/-a¹⁴. He gives the examples
khāle ‘he ate’, and puchhle ‘he asked’. For intransitive verbs, the ending remains
as -o/-a or zero, e.g., gela (gelo), or gel, ‘he went’. In addition, Grierson maintains
that in written texts, a transitive verb may end on -e, such as dile instead of the
regular dila ‘he gave’ (1903 V, I: 46, 65). Furthermore, in his discussion of the
dialects of Bangla, the marking of a third person transitive verb is invariantly
mentioned. The following table gives an overview of the different markings of
the third person verb form in the past tense in the various dialects described in
Grierson’s Linguistic Survey.

language tr itr

Standard (Katwa/Burdan) -e -a
Western -e/-ek -o
Western Saraki (Ranchi) -ek -ek
Western Khariathar (Manbhum) -ek/-ak -ek/-ak
Western Mal-Paharia (Sonthal Parganas) -aṃ -aṃ
Southwestern (Midnapore) -a -a
Northern (Dinapore) -e Ø
Northern (Malda) -e -o
Northern Koch (Malda) Ø Ø
Eastern (Dacca) -ā/-o -ā/-o
Eastern (Mymensingh) -ā/-o-/Ø -ā/-o/Ø
Eastern (Jessore) -e -o
Southeastern (Chittagong) -a Ø

Table 3.7 Third person verb endings in the past tense in dialects of Bangla (Grierson V, I)

14 The variation between -o and -a is probably due to pronunciation and is found in spelling
(Radice 2007: 167). Similarly to Asamiya, <a> is often pronounced as [o] in Bangla.
 The transitive verb 131

The variability can be categorized according to the geographical situation of the

dialects. Eastern Bangla dialects, spoken in contemporary Bangladesh display
markedly less variation: transitive and intransitive verbs are marked in the same
way, except for the Jessore dialect, which appears more westwards. In Western
dialects, there is a difference, and often, the -k sound in the marking is attested.
This marking is also found in the Bihari languages, suggesting that it is a typical
feature in this area. Northern Bangla dialects are spoken around the area of
Asamiya, and tend to show a preference for an -e marking on transitive verbs, as
in Asamiya. Although it is clear that in Bangla the verb can be differently marked
for transitivity, it is also obvious that this marking is far less systematic than it
is in Asamiya. Because in contemporary standard Bangla the different marking
is apparently absent (cf. Radice 2007; Nasrin and van der Wurff 2009), it can be
expected that this difference will also fade away in the non-standardized dialects,
if it has not already done so.

3.2.1.1 The transitive verb in the Bihari languages

The variety in agreement patterns is at least as great in the Bihari languages. In
the following Table 3.8, the differential markings on the verbs determined by
transitivity are outlined. These differences do not necessarily occur on the level
of the conjugational ending; they may also occur in the tense marking infix.

language tr itr

Asamiya 3rd p. past -e 3rd p. past Ø

Magahi 3rd p. past -lk- 3rd p. past -l-

Bhojpuri 3rd p. past -al-as 3rd p. past -al

Maithili 3rd p. past -ainh/-ak 3rd p. past -ah/Ø

Table 3.8 Differences between transitive and intransitive conjugations in Eastern IA

Magahi, Bhojpuri and Maithili have different markings for the third person past
tense form according to the transitivity of the verb. In Bhojpuri, -as is added to
the past form of transitive verbs (on -al), whereas the Bhojpuri intransitive third
person verb does not take an additional ending. In Magahi, the difference is
found in the tense-indicating suffix -l- to which -k is added only when the verb is
transitive (cf. Verma 2003: 511).
Besides this different marking for the third person, Magahi and Bhojpuri have
a separate marking for an intransitive or a transitive past participle, indicated by
132 Eastern Indo-Aryan: Asamiya

-l-. In Magahi for instance, a past participle may be used in combination with
person agreement and a copula, forming a present perfect construction, such as
in the following ex. (130) from Verma (2003: 511) (note that gender and number
agreement have disappeared from Magahi):

(130) ham sut-l-i+h-e

I sleep-ptcp-1+aux-prs
‘I have slept.’

(131) tu sut-l-a+h-e
you sleep-ptcp-2+aux-prs
‘You have slept.’

In these examples, the copula is invariable, or ‘frozen’, and the person agreement
occurs with the participle. The construction ham sutli can also occur indepen-
dently as a past tense, meaning ‘I slept’ (cf. Verma 2003: 511). However, there is a
second construction in Magahi in which the copula indicates the person and the
participle is invariable. Two examples are shown below (Verma 2003: 512):

(132) ham sut-al+h-i

I sleep-ptcp+aux-prs.1
‘I am asleep.’

(133) tu sut-al+h-a
you sleep-ptcp+aux-prs.2
‘You are asleep.’

(134) ham khai-le+h-i

I eat-ptcp.tr+aux-prs.1
‘I have eaten.’

The construction with an intransitive verb, ex. (132), indicates the stative aspect.
However, with a transitive verb (ex. (134)) it indicates, in the words of Verma
(2003: 512), “I am in the state of having had the experience of eating it”.¹⁵ The

15 Note that this Asamiya construction resembles the stative constructions in Hindi (Chapter 2).
If a transitive stative participle in Hindi is combined with a copula, and A is not marked with the
ergative postposition (e.g., sītā sārī pahnī huī hai ‘Sita is [in the state of] wearing a sari’), then
the meaning is stative, as in the Magahi sentence ham khaile hi ‘I am in the state of having eaten
 The transitive verb 133

meaning of a transitive verb is more difficult to reconcile with a stative interpre-

tation, so one harks back to what is basically a perfect aspect meaning. Verma
further notes that an intransitive participle may also take the stative construction
of a transitive verb in which case an “extra stative meaning resulting in a contras-
tive finite string” is intended. She gives the following example:

(135) ham sut-le+h-i

I sleep-ptcp+aux-prs.1
‘I have had the experience of sleeping.’

In Bhojpuri, the third person past tense displays a difference in endings between
transitive and intransitive verbs, as in the other Eastern IA languages. Whereas
the third person ending of a transitive verb is -as, e.g. dekh-al-as ‘he saw’, the
intransitive third person consist only of the marker of the past -al, e.g. cal-al ‘he
went’ (cf. Tiwari 1960: xli). Secondly, both transitive and intransitive participles
are formed with -l. However, from the moment the transitive participle is com-
bined with an auxiliary, it takes the ending -e, e.g., u dekh-(a)l-e bā ‘he has seen
it’, whereas the intransitive participle remains invariable. As in Magahi, an intran-
sitive participle combined with a copula expresses a stative (cf. Verma 2003: 528).
According to Tiwari (1960: 183), the e-ending of the transitive participle is an old
locative suffix. In his view, the intransitive participle is adjectival, and attribu-
tively used to determine the subject, whereas the -e ending of the transitive verb
indicates a more ‘nominal’ meaning.¹⁶
The past participles used in combination with a copula also take different
markings depending on the transitivity of the verb in Maithili, viz. -ne (trans.) and
-l (intrans.). The following examples are from Yadav (2003: 487):

(136) ram khai-ne+ch-aith

Ram eat-pst.tr+aux.prs-3h
‘Ram has eaten.’

(137) ram ge-l+ch-aith

Ram go-pst.itr+aux.prs-3h
‘Ram has gone.’

it’. If the Hindi stative participle is, however, combined with a copula in an ergative construction
(e.g., sītā=ne sārī pahne hue hai ‘Sita has been wearing a sari’), the construction is closer to the
Magahi construction with the invariant copula, e.g., ham sutli he ‘I have slept’.
16 This is reminiscent of the use of participles, such as āhile in Asamiya.
134 Eastern Indo-Aryan: Asamiya

3.2.1.2 Honorific agreement in Maithili

As one of the few Eastern IA languages’ verbs, Maithili verbs display agree-
ment with O. The indication of honorific agreement plays an important role in
the complex agreement system of Maithili (cf. Grierson 1903 V, I: 24; Stump and
Yadav 1988; Bickel et al. 1999; Yadava 1999; Bickel 2003; Bickel and Yadava 2000).
Only the agreement system in Magahi resembles the agreement in Maithili, but in
Magahi, it is far less complex (cf. Verma 1991: 129).
Maithili has a conjugational system for person and honorific agreement. Number
and gender are not grammatically indicated in the language.¹⁷ Verbs in Maithili
agree with the nominative subject in person and honorific grade. Tense is indi-
cated on the stem. The conjugated past is formed on the basis of a participle with
-l-, e.g., ge-l-i ‘I went’.

imperfective perfective

1 -i -ahuṃ
2h -i -ahuṃ
2mh -ah -ah
2nh -e(ṃ) -e(ṃ)
3h -aith, -athinh -ainh/-ah
3nh -aik -ak

Table 3.9 Maithili verb endings displaying subject agreement

Verbs in Maithili may additionally agree with a second argument. Oddly, the verb
does not agree with an unmarked O, but with any marked object from the sen-
tence. This may be a marked O, an IO, a possessive form added to the subject or to
O, or an experiencer; in other words, a non-nominative argument (cf. Bickel et al.
1999:491; Mishra 1990: 110). According to Jha (1958: 472, qtd. in Stump and Yadav
1988), the verb displays secondary agreement with the “most prominent NP”.
With prominence, Jha means three things: the NP is either emphasized, highly
honorific, or highly animate. Stump and Yadav (1988: 308) suggest a hierarchy of
agreement: on the left side, agreement with the argument is possible; the argu-
ments on the right side are not likely to cause secondary agreement.

3H < 2MH < 2NH, 3NH < 1, 2H

17 Gender seems to have disappeared only recently in spoken Maithili, and the remains of
gender agreement are still found in formal and literary writing. They appear to be particularly
persistent in combination with the third person, where there are different endings for mascu-
line and feminine gender (cf. Stump and Yadav 1988).
 The transitive verb 135

The ranking of features of arguments in a hierarchy is of course reminiscent of

Silverstein’s hierarchy (cf. Chapter 2). Following Silverstein, speech act partici-
pants (henceforth SAP) are associated with accusative alignment, whereas other
arguments seem to prefer ergative alignment. With respect to verb agreement,
this implies that, following Silverstein, secondary agreement with an objective
first or second person form would be unexpected, and agreement with a third
person form would be expected. Stump and Yadav’s hierarchy indeed indicates
that third person forms tend to control secondary, ‘non-nominative’ agreement
more often than first and second person forms do. However, it is also clear that
the honorific grade plays an equally important role in determining agreement in
Maithili. As Stump and Yadav (1988) point out, the higher the honorific grade of
the argument is, the more likely it is that secondary agreement follows as well.
However, the second honorific person is an exception, because this argument will
almost never control secondary agreement. The 2h pronoun follows the pattern
of the first person entirely.
Secondary agreement is illustrated by the forms listed in Table 3.10. Some are
transparent, and one can notice the separate markers for primary and secondary
agreement (e.g. -i-ainh, 1-3h), others are entirely merged into one form. Only in 3h
the present and past tense take different endings. Grierson (1903 V, I: 24) discerns
some regularities:
– combinations of a subject with an honorific second argument requires the
addition of -nh;
– an honorific second person subject triggers the same verb conjugation as a
first person subject;
– the subject marker -ai/-ainh in combination with a second person object
becomes -au/-aunh, -ah for a first person;
– -k can be added to all forms on -ai or -au, without changing the meaning.

object 1 object 2h object 2mh object 2nh object 3h object 3nh

A/S 1 -iah -iauk -iainh -iaik

A/S 2h -iainh -iaik
A/S 2mh -ahunh -ahak
A/S 2nh -ahunh -ahik
A/S 3h past: -ainh past: -ainh -athinh, -athinh, -athunh, -athunh,
-akhinh -akhinh -akhunh, -akhunh,
-ainh -ainh
A/S 3nh -(ak)ah -(ak)auk -(ak)ainh -(ak)aik

Table 3.10 Secondary agreement in Maithili

136 Eastern Indo-Aryan: Asamiya

The occurrence of -ak- when the subject is 3nh only happens in the past tenses
and with transitive verbs (cf. Yadava 1999: 151). Examples of agreement are found
below (cf. Bickel et al. 1999). In the first example, the intransitive verb refers to an
unexpressed possessor.

(138) barad bhāg-al jā-it+ch-auk

ox run-ptcp go-prs+aux.prs-3nh.2nh
‘Your ox is running away.’

(139) hun-kā=saṃ kitāb nahi paḍha-l ge-l-ainh

he-obl.h=abl book not read-ptcp pass-pst-3nh.3h
‘The book could not be read by him.’

In the last example, a passive verb, paḍhal gelainh, agrees with the subject kitāb,
3nh, and with the ablative of the agent-object, which is 3h, hunkā=saṃ. The next
example is one of the rare occasions when there is three-fold agreement, which
only occurs when there is a combination of the following arguments: 1/2nh/3h:

(140) ham to=rā kaniya=keṃ dekh-au-l-iaunh

I you=obj bride=obj show-caus-pst-1.2nh.3h
‘I showed you the bride.’

Secondary agreement has been discussed in detail by Bickel et al. (1999: 497)
by making use of the social concepts of FACE and EMPATHY. FACE is defined as
“a common interest in maintaining each other’s FACE” shared by the people (cf.
Brown and Levinson 1987). For Maithili, this means that a first person speaker
should treat a second person honorific speech participant the same way as when
(s)he would refer to him-/herself so that no (social) difference is expressed in the
words of the speaker in an conversation between the speaker and speech partner.
Secondly, in conversation, all higher honorific persons should be overtly marked
on the verb to indicate their social status. This is a possible explanation for Stump
and Yadav’s hierarchy (1988): the higher the honorific quality of the argument,
the more likely it is that it controls secondary agreement. The second honorific
person constitutes an exception because as the first person, it rarely controls sec-
ondary agreement (and is therefore situated at the right side of the hierarchy).
The second concept, EMPATHY, refers to the third persons that can control
agreement. The definition of empathy is taken from Kuno (1987: 206): “EMPATHY
is the speaker’s identification, which may vary in degrees, with a person or thing
that participates in the event or state that he describes in sentence” (Bickel et al.
1999: 503). Third persons are not SAP but are incorporated into the conversation
 The transitive verb 137

through the “empathy” of the speakers. The higher the third person is on the social
hierarchy, the stronger the need is that (s)he is involved in the speech act and that,
(s)he is referred to by the verb. The hierarchy constituted by the concepts of “face”
and “empathy” is actually a variation on Silverstein’s animacy/SAP hierarchy, but
on the basis of the feature of honorificity. Honorificity is included as a feature that
can determine referential hierarchy in Bickel’s (2011) overview of possible features
of referential hierarchies. Nevertheless, as Bickel (2008) points out, the mutual
influence between hierarchies of features and types of alignment, particularly
agreement patterns, should not be overestimated. Maithili is a case in point, given
that the conventional association between SAP and accusative alignment is not
completely followed. Firstly, honorificity and other social concepts are at least as
important as the difference between pronouns and nouns. Secondly, the second-
ary agreement found in Maithili is not ergative agreement in the sense that the
verb always agrees with S/O; on the contrary, secondary agreement in Maithili
seems to be with the most “prominent”, objective-marked argument.

3.2.1.3 Honorific agreement in Magahi

The agreement pattern in Magahi is similar, but less complex, than the one found
in Maithili. Verbs in Magahi agree with the subject argument, but secondary
agreement, in which the verb can agree with any marked object is equally pos-
sible. In Table 3.11 and Table 3.12, the past tense endings in Magahi are given.

1 -i/-io/-iau
2h -a
2nh -e
3h -thin/-thun
3nh -ai/-o/-au

Table 3.11 Verb endings in Magahi, agreement with subject

object 2h object 2nh object 3h object 3nh

A/S 1 -io -iau -iain -iai

A/S 3h -ko -kau -kain -kai

A/S 3nh -ko -kau -kain -kai

Table 3.12 Verb endings in Magahi, secondary agreement

138 Eastern Indo-Aryan: Asamiya

Table 3.12 is an overview of the secondary agreement suffixes in Magahi. When

the subject is a second person, there are no secondary agreement suffixes. Only
first and third person subjects allow object indication on the verb. The suffix is
then -o or -au. Verma (1991) observes that this suffix is sometimes used without
the overt presence of a second person and calls this the “addressee component”
(comparable to the so-called dativus ethicus, cf. Section 5.2.2). Ex. (141) and (142)
illustrates the consequence of this agreement: (142) is directed at someone who is
not overtly referred to in the conversation:

(141) ham okrā dekh-al-iai

I he.nh.obj see-pst-1.3nh
‘I saw him.’

(142) ham okrā dekh-al-io

I he.nh.obj see-pst-1.2h
‘[Let me tell you,] I saw him.’

The agreement pattern of Maithili and Magahi is exceptional among the Indo-
Aryan languages. Firstly, it is unusual that the verb only agrees with a marked
object. Secondly, agreement with an unexpressed “addressee component” seems
to be unattested in any other IA language (but see Kashmiri, Chapter 5). The
parameter of honorificity, though, is also well-observable in other IA languages.
Nepali is the best known example, with its separate verb conjugation for kings
and noblemen. However, in Nepali, the honorificity of a certain argument does
not determine the verb agreement.
It is no coincidence that the regions with Maithili speakers and Nepali speak-
ers border each other. I concur with Bickel et al. (1999) who claim that the influ-
ence of the honorific grade of the arguments on the agreement in Maithili is a result
of the social stratification in this area. As Masica (1991: 344) observes, following
Jha (1958), the secondary agreement in Maithili is of recent origin. It is reasonable
to assume that former verb endings, which used to indicate gender and number,
shifted to indicators of honorificity when gender and number disappeared as
grammatical categories. Whereas social stratification may be considered as the
trigger of the honorific agreement pattern, evolutions in verb agreement found in
the other IA languages should also be taken into account. The possibility of object
agreement is present in Central and Western IA languages. Even though this pos-
sibility is far less outspoken in the other Eastern IA languages, it constitutes the
background of secondary agreement in Maithili and Magahi.
 The transitive verb 139

3.2.2 The transitive verb in an Eastern Hindi variety

In the Eastern IA languages, the participial origin of the perfective verb forms is
less noticeable than it is in other IA languages, because the person agreement
is expressed by the finite verb. An interesting language for a comparative case
study in this respect is Awadhi. Awadhi is spoken in the central states of Uttar
and Madhya Pradesh and has been considered as an Eastern dialect of Hindi.
However, it also shares properties with Bhojpuri and other Eastern IA languages.
In Awadhi, one finds a tense system based on participles (as in Central IA), but
the verb forms all display person agreement with the subject (as in Eastern IA).
Furthermore, Awadhi has another feature in common with Eastern IA, i.e., the
language displays a different pattern for transitive and intransitive verbs in the
past tense (referred to as conjugation t & i below). Besides this different conjuga-
tion based on the transitivity of the verb, Awadhi also possesses two morphologi-
cally different conjugations for the intransitive (conjugation i1 and i2) and transi-
tive (conjugation t1 and t2) past tense verbs. The first conjugation displays person
marking on the copula, as in Hindi and other Central IA languages, the second
conjugation indicates the person in the participial form, aligning Awadhi with
the Bihari languages.¹⁸ Awadhi is therefore illustrative to compare the difference
between a tense system based on participial forms and one based on finite verbal
forms that indicate person agreement.
Table 3.13 (based on Saksena [1937] 1971: 248, 260) illustrates the first kind of
conjugation. The copula agrees in person with the subject in the present tense.
When the verb is intransitive (i1), the ending of the participial form, e.g., marā,
‘died’, agrees in gender and number with S, m.sg. -ā, m.pl. -e, f.sg. and pl. -ī (īṃ
for third person plurals). For transitive verbs (t1), the oblique -e ending is added
to the participle, e.g., dekhe ‘seen’, no matter what the number or gender of A is.

18 Masica (1991: 477) refers to Smith (1974), who describes patterns of agreement with the
subject as well as with O in Awadhi, particularly in the language spoken on the border areas
between Awadhi and Kanauji (the latter language normally only displays agreement with S/O).
140 Eastern Indo-Aryan: Asamiya

tr t1 itr i1
sg pl sg pl

1m dekhe hauṃ dekhe han marā hauṃ mare han

1f dekhe hauṃ dekhe han marī hauṃ marī han
2m dekhe hai dekhe hau marā hai mare hau
2f dekhe hai dekhe hau marī hai marī hau
3m dekhe hai marā hai mare haiṃ
3f dekhe hai marī hai marīṃ haiṃ

Table 3.13 t1 and i1

Ex. (143) illustrates the agreement pattern of the past tense verb (Saksena [1937]
1971: 432–433).

(143) kā ī=ma ke gulgul-ā tum lih-e+h-au?

q this=from some sweet-pl you.f take-pst.tr+aux-prs.2pl
‘Have you taken some sweets from this?’

It should be noted that a transitive verb can also take the intransitive conjuga-
tion in some circumstances. In Saksena ([1937] 1971: 248) the examples mai dekhā
(m.sg.) hauṃ and mai dekhī (f.sg.) hauṃ are mentioned, where the transitive par-
ticiple dekhā resp. dekhī agrees in gender and number with its subject, as if it were
an intransitive participle. The meaning of the participle is, in this case, different.
Saksena translates these sentences as: “I am the one who has been seen” instead
of “I have seen”. The participle, conjugated in this manner, lacks the predicative,
finite verb meaning that it normally has as a verb and rather is considered as an
adjectival participle determining the first person pronoun. This is reminiscent of
the intransitive past participles in Magahi, e.g., sutle hi (Section 3.2.1.1), that take
the transitive suffix, which subsequently leads to a change in the meaning.
The second conjugation of the participle in Awadhi (i2 for intransitive and
t2 for transitive verbs) is closer to the person conjugation of the participles in
the Eastern IA languages. In contrast to the first conjugation, the participial form
agrees in person; when the optional copula is added, it remains in the invariant
third person singular form hai in all circumstances. This is the ‘frozen’ form of
the copula, which we also find in other conjugations in Magahi, Bhojpuri and
Maithili (cf. Verma 2003: 511). The difference between the intransitive and transi-
tive conjugations is especially clear when it comes to gender agreement. Transi-
tive forms only agree in number and person with the subject, whereas intransitive
verbs do have a different ending when the subject is masculine or feminine. In
example (144) a transitive verb is followed by an intransitive verb, both having
 The transitive verb 141

a feminine subject. The form lihisi (t2) is the same for the masculine and singu-
lar paradigm, while lauṭaīṃ (i1) is a typical intransitive feminine form (Saksena
[1937] 1971: 432–433).

(144) bahū jale ui lauṭa-īṃ tale

daughter-in-law when she.pl return-pst.f.pl then
tīni gulgul-ā nikāri lih-isi
three sweet-pl take out.inf take-pst.3sg
‘Before she [the old lady] returned, the daughter-in-law took out three
sweets.’

The conjugations t2 and i2 are exemplified in Table 3.14.¹⁹

tr t2 itr i2

sg pl sg pl

1m dekheuṃ hai dekhen hai mareuṃ hai maren hai

1f dekheuṃ hai dekhen hai mariuṃ hai maren hai
2m dekhisi hai dekheu hai mare hai mareu hai
2f dekhisi hai dekheu hai mare hai mariu hai
3m dekhisi hai dekhini hai
3f dekhisi hai dekhini hai

Table 3.14 t2 and i2

In conclusion, in Awadhi, there are two past conjugations, one of which is closer
to Hindi and its varieties (t1 and i1) and the other one of which is closer to the
Eastern languages (t2 and i2). Awadhi does not take the infix -l- to indicate a
past tense like the Eastern IA languages do. The conjugations differ according to
whether the verb is transitive or intransitive. Hence, in Awadhi, although the case
marking in itself is not exceptional, the indication of the transitivity of the verb
for the past tenses is an important factor influencing the formal expression of the
verb in the clause.

19 In the third person, the following forms are restricted to one conjugation: the third person
plural of a transitive verb always ends with -ini (conjugation t2), while the third person forms of
the intransitive verbs always end, according to the gender and number of S, with ā, ī, e or īṃ,
with nasalization of the copula haiṃ, as in i1.
142 Eastern Indo-Aryan: Asamiya

3.3 Ergative marking in Eastern Indo-Aryan

Asamiya is the only Eastern IA language where the marking of A occurs system-
atically (cf. Section 3.1.2). In modern grammars of Bangla (e.g. Radice 2007) and
Oriya (Neukom and Patnaik 2003), there is generally no mention of an -e marking
for the subject. Grierson (1903 V, I: 14, 34) points out that the -e ending is some-
times used as the ending of human subjects of transitive verbs in Bangla (though
not obligatorily), and Chatterji (1926: 741) regards -e as a nominative form that
was already obsolete in Western Bangla during his time, but occurred commonly
in Eastern Bangla (and in Asamiya). The -e ending in Bangla is no longer related
to a specific argument role but tends to occur, albeit exceptionally, with human
nouns, particularly when the nouns are plural and indefinite (cf. Nasrin and van
der Wurff 2009: 140), see the following example (Chatterjee 2008: 105):

(145) lok-e brisṭi-r por kokhono+kokhono akas-e

people-erg rain-gen after sometimes sky-loc
ramdhonu dakh-e
rainbow see-prs.3
‘People sometimes see a rainbow in the sky after the rain.’

The form lok-e is reminiscent of the pronominal forms ending in -e that are also
found in Asamiya and always take -e, irrespective of whether they are A or S
(cf. 3.1.2).
In Oriya, however, the subject of any verb is unmarked, though again, certain
pronouns retain the ending -e, e.g., tume ‘you’, or the marker of indefiniteness
added to a singular noun jaṇe ‘one person’ (cf. Ray 2003: 454–457). However,–e is
not associated with a particular argument role. This is also the thrust of Neukom
and Paitnik (2003: 47). They discern a nominative case with a zero ending.
However, later in their study, they discuss the nominative case separately: “Many
nominals with human referents show a suffix -e in the nominative case” with
examples of the plural suffix -mane, and some other pronominal forms such as
kie ‘who?’, anek-e ‘many’ etc. Note that these forms ending with -e do not have an
unmarked nominative form. Regardless of whether they are used as A or S, they
always occur with -e.²⁰

20 Neukom and Paitnik (2003: 35) also analyze the suffix -e that is added to nouns as a marker
of indefiniteness and is often in the function of O, e.g., ghaṇṭa-e ‘an hour’, pahar-e deba ‘to
give a blow’. They distinguish this use of -e from the -e of the nominative, for the justifiable rea-
son that the indefinite -e may be added before a case marker, e.g., akha-ṭi-e-re ‘in a sackcloth’,
 Ergative marking in Eastern Indo-Aryan 143

Whereas the suffix -e has merged with the nominative case and is no longer rec-
ognizable as an ergative marker in Bangla and Oriya, in Asamiya -e clearly func-
tions as a marker of A. Since Asamiya is the most eastward spoken IA language,
influence of Central IA on the ergative marking seems to be improbable and can
hardly explain the different evolution of Asamiya in comparison with its closest
related languages. A hypothesis worth taking seriously is that the ergative case
marking in Asamiya was reinforced through language contact. Asamiya is spoken
in an area in which there is contact with a wide array of languages of at least
three non-Indo-European families. Some of the most wide-spread languages are
Bodo, Garo and Koch (Tibeto-Burman), Mundari, Santali and Malayan (Austro-
Asiatic) and Khasi (Mon-Khmer), amongst others (Post 2008: 63; Kakati 1962: 32).
Many inhabitants of Assam are bilingual, using Asamiya outside the home and
a non-IA language as the language inside the home. Besides the possible influ-
ence on the ergative case marking, the linguistic contact situation has led to some
other extraordinary deviations from most IA languages. For instance, the use of
classifiers to indicate plurality, gender, diminutives, and definiteness, and the
indication of possession on the relational and kinship nouns is obviously due to
the Tibetan and Austro-Asiatic influence. Furthermore, there are certain parallels
between the ergative marking in Asamiya and Tibetan. Tibetan is a language with
an ergative marking that is optional in the present tenses but obligatory in the
perfect constructions. Particularly revealing is the fact that the ergative marking
often assumes a discursive function (cf. Tournadre 1996: 372 la visée communica-
tive) by occurring in an unexpected position, such as a marker of S, cf. ex. (146)
from Tibetan (Tournadre 1996: 300):²¹

where ṭi is the postposed article, -e- emphasizes the indefiniteness, and -re is the locative case
marker. However, in other instances the use of -e as an indefinite marker is not all that different
from its use as a nominative marker. The classifiers jaṇ, ‘a person’, and goṭa or goṭi ‘a, some-
thing’ may also be combined with the indefinite suffix -e, and they express specificity (Neukom
and Paitnik 2003: 120). Inanimate nouns cannot take the suffix -e (examples given in Neukom
and Paitnik 2003: 50). Interestingly, in Oriya the subject can also be expressed in the genitive
case form when it is animate and coordinated. Neukom and Paitnik (2003: 59) give the follow-
ing example: mor tankar e kamṭa kalu ‘I and he did this work’, where mor as well as tankar are
in the genitive form. However, this seems more of an extension of the experiencer construction
with a genitive experiencer, than of a kind of ergative marker.
21 The ergative marker in Tibetan is also associated with volition, occurring as a marker of a
volitional S in an intransitive construction. Although I have not found any evidence of -e as the
marker of a volitional S in Asamiya, it is not unconceivable, and it has been reported to me by J.
Tamuli (p.c.). However, verbs analyzed as volitional are in fact often unergative (e.g., ‘to sleep’).
144 Eastern Indo-Aryan: Asamiya

(146) kho-s lo nyishu btsonkhang nang-la bsdad-pa-red

he-erg year twenty prison in-obl stay-pst-aor
‘He, [he] has stayed twenty years in prison.’

In sum, unexpected occurrences of -e as in the above example (96) or when it

marks an “emphasized” S could be attributed to the influence of Tibetan lan-
guages on Asamiya. In general, the extended use of -e for all A’s, and not only
for A’s of perfective verbs, may be due to the influence of surrounding languages,
where ergative marking is standard for A.

3.4 Conclusion

With regard to alignment, there are several features that unite the Eastern IA lan-
guages and contrast them with the other IA languages. First of all, in all Eastern
IA languages, verbs agree with A/S. Importantly, the agreement is in person.
Gender and number agreement may also occur but only in the Eastern IA lan-
guages where these categories have not disappeared. Secondly, in many Eastern
languages, there is no ergative marking attested. Although Asamiya can be con-
sidered an exception to this generalization, even in this language, the ergative
marking -e does not occur in the same distributional pattern as the postposition
ne does in Hindi (cf. Section 3.1.2). The differences are that the use of -e is not
bound to a particular tense/aspect of the verb. Rather, it occurs much more reg-
ularly in unergative constructions, although many of the unergative verbs have
the same meaning as the unergative verbs in Hindi. The ergative marker also
occurs as marker of the A of a non-finite verb (cf. Section 3.1.2). In comparison to
Hindi, the ergative marker in Asamiya occurs much more often, and it is possible,
and quite probable, that this spread happened under the influence of language
contact with a non-IA language (cf. Section 3.3).
Asamiya is spoken at the eastern border of India, and hence, it is the remotest
from the “ergative” Central and Western IA languages. The ergative marking in
Asamiya is clear evidence that the “ergative case” is not a feature that exclusively
belongs to Central and Western IA.
Another feature making the transitive constructions of Eastern IA differ from
the classical accusative pattern, is that certain Eastern IA verb forms do agree
with O. The agreement is, however, less straightforward than it is in Hindi. The
-e suffix added to the third person past tense in Asamiya, can be considered to
be an agreement marker that cross-references to O, because it only occurs when
the verb is transitive. Because gender and number are no longer grammatically
encoded in Asamiya, O-agreement necessarily takes a simplified form. In particu-
 Conclusion 145

lar, in Asamiya, there is no possibility that the verb cross-references to the gender
and number of O.
In the literature, two explanations have been proposed for the historical
development of the -e marking of transitive past tense verbs in Asamiya. Either
the transitive verb received an additional marking, or the intransitive verb lost its
marking. Today, this first explanation is the preferred one among most scholars
of IA, and it is indeed the most plausible. The -e ending in Asamiya has been
explained as an extension of the present tense ending. For other languages, it
has been suggested that the transitive ending is derived from a locative suffix
added to a participle. According to Chatterji (1926), the zero-marked intransitive
third person verb form is associated with an adjectival function. He claims that
this verb form is a participial form that lacks person agreement. The e of a tran-
sitive third person past verb form is said to indicate the “verbality” of the form,
amounting to the possibility of expressing person agreement on the verb. This
explains, according to Chatterji, the discrepancy that exists between transitive
and intransitive third persons.
Although it seems rather farfetched to consider the third person of an intran-
sitive verb in the past tense as the only “participial” form in the conjugational
paradigm, it stands to reason that there is a link between the person agreement of
the verb, i.e., the “verbal” rather than the “adjectival” nature of the verb, and A/S
agreement. Dixon (1994), for instance, suggests that person agreement, which he
calls “bound pronouns”, displays a cross-linguistic tendency to pattern accusa-
tively, whereas gender and number agreement do not show such a preference. On
this view, the -e suffix on the transitive past verbs in Asamiya must be interpreted
as a historical remnant of the participial origin of the past tense, and as indicat-
ing agreement (but not necessarily person agreement) with O. Importantly, the
distinction in conjugation between a transitive and an intransitive past tense verb
is not exclusively a property of Asamiya, but occurs in the Bihari languages as
well (cf. Section 3.2.1). In Maithili and Magahi, one even finds object agreement,
although the pattern is of a different nature from what is commonly known as
O-agreement in ergative constructions. Verbs in Maithili and Magahi may display
secondary agreement with a marked argument that is high in animacy, definite-
ness, and honorific grade. Although the secondary agreement pattern seems to
be determined either by social factors or by discourse elements, the possibility
of secondary agreement with O is a typical feature of IA languages in general.
However, this kind of agreement is not the type of agreement with {S, O} that is
part of the definition of ergativity.
4 Northern Indo-Aryan: Nepali
Nepali is the most important of the Northern IA languages spoken in the Himala-
yas. Nepali is also known as Khas Kura or Khas Bhasa, the ‘language of the Khas’,
a Nepali-speaking ethnic group. The name Gorkhali is also used, and refers to the
language of the city and the people of Gorkha (Riccardi 2003: 540). The Nepali
themselves often call their language Pahari or Parbartiya, meaning ‘the language
of the mountains’ (Acharya 1991: 2). The name Pahari also refers to a group of
languages spoken in the mountains, of which Nepali is only one representative
(others are for instance Kumaoni, Garhwali). Today, Nepali is spoken by appr.
17,209,000 people all over the world, appr. 11,053,000 of whom reside in Nepal
(Ethnologue 2010, Census 2001). Nepali’s most important competitors in Nepal
are Tibeto-Burman languages, particularly the group of the Newari languages.
Newari was originally named Nepal Bhasa (‘the Nepal language’). Although
a Newari language was once the language of the capital Kathmandu, Nepali is
currently the dominant language in Nepal. Nevertheless, many speakers are of
course bilingual. Some Newari dialects have been heavily influenced by Nepali;
influence in the other direction is less clear (cf. Riccardi 2003: 546).
Nepali shares many resemblances with Eastern IA languages with regard to
its alignment. It has also been argued that Nepali originated in the language of
the Rajput people of the Indian state of Rajasthan in the West, who would have
entered Nepal between the 11th and 15th century (Wallace 1982: 152–153). The
oldest inscriptions found in Nepal are dated 1255 AD (Riccardi 2003: 543). It is
believed that they were made by an emperor of the Khas, and indeed, it seems
that this group of people was not native to the mountains, but entered Nepal
from the West, bringing with them an Indo-Aryan language. The language of the
inscriptions is related to Nepali, but it is only in the 17th century that inscriptions
were made in a language that can be considered the ancestor of modern Nepali.
From that date onward, a considerable amount of literature has been composed
in Nepali (Riccardi 2003: 544). Whether or not historical sources will confirm the
relation with Western IA, on a synchronic level, some of the Nepali morphology
seems to be closely related to the morphology of the Rajasthani languages. As
such, Nepali represents a link between Western and Eastern IA.
The examples in the following Chapter derive from the short stories assem-
bled in Hutt (1997) and Acharya (1991), children’s stories (Thapa 2001) and a
questionnaire conducted in various places in Nepal (2009).
 Nepali: description 147

4.1 Nepali: description

4.1.1 Case marking in Nepali

Like Asamiya, Nepali has no gender or number marked on the noun and the adjec-
tive. The plural is optionally indicated by means of a lexical suffix -harū which
in turn may be marked for case. In contemporary Nepali, nouns are not generally
inflected and only have an unmarked (nominative) form, although there is an
older, oblique form. The nominative form on -o is used before a postposition, e.g.,
lāuro=le ‘the stick’. If an adjective or demonstrative determines such a noun in
combination with a postposition, then the adjective/demonstrative is normally in
the oblique case, e.g., yo śahar ‘this town’, yas śahar=mā ‘in this town (Hutt and
Subedi 1999: 68). However, in the colloquial language, the demonstrative often
retains its nominative form. Certain nouns ending in a vowel still display the
oblique before a postposition, e.g., nom. choro ‘boy’, obl. chorā (cf. Srivastava
1962: 91). The postpositions given in Table 4.1 function as case markers and were
also defined as such by previous scholars (e.g., Turnbull [1923] 1982; Morland-
Hughes 1947).

nom Ø
erg le
obj lāī
gen ko
loc mā
ins le

Table 4.1 Case marking in Nepali

The postposition le marks A and it also functions as the marker of the instrumen-
tal case. The postposition lāī marks O when it is animate and/or definite, and it is
also the marker of IO. According to Turnbull ([1923] 1982), only one argument can
be marked with lāī in a sentence. When both O and IO occur in a sentence, the
preference is to mark the IO.
The paradigm for pronouns is only slightly different from that of nouns. In
contrast to nouns, most pronouns have a distinct oblique form in combination
with the postposition le. The first person singular is ma but when the le-postpo-
sition is added it becomes the form maiṃ=le. However, when other postpositions
are added, it remains ma, e.g., ma=lāī. The genitive of personal pronouns is a sep-
arate possessive form, e.g., mero ‘my’. Aside from these minor differences, pro-
nouns pattern identically to nouns. Like Asamiya, Nepali differentiates between
148 Northern Indo-Aryan: Nepali

honorific and non-honorific forms. The second and third persons have several
forms according to the honorific grade of the argument.¹

4.1.2 Case marking of the core arguments

A and S in the present tense take the nominative marking. In a construction with
a perfective verb form, A takes the ergative marker le, and S is unmarked. Below
are examples of the classic case marking pattern in Nepali. The next example is a
sentence with an animate A marked by le, and an unmarked O; while in ex. (148),
there is an inanimate A and an animate, marked O. Note that animacy is only a
parameter in the marking of O. Also consider that O need not be present for a
sentence to have a transitive verb and a marked A. O may also be a clausal object,
as in ex. (149).

(147) rājā=le mukh bharī pāṇī hāl-e

king=erg mouth full water put-pst.3sg.h
‘The king put his mouth full [with] water.’ (Thapa 2001: 12)

(148) baigun=ko kāṃḍā=le tyas=lāī ghoc-yo

aubergine=gen thorn[f]=erg she=obj prick-pst.3sg
‘A thorn of the aubergine-plant pricked her.’ (Thapa 2001: 6)

(149) lāuro=le bhan-yo, ko h-o?

stick=erg say-pst.3sg who be-prs.3sg
‘The stick said, who is it?’ (Thapa 2001: 7)

Just like in Asamiya, some unergative verbs can take a cognate object. These
are, in particular, the so-called verbs of “bodily emission”, such as ‘to vomit’,
‘to urinate’, ‘to cough’, etc. The use of a transitive subject in combination with
such verbs is a wide spread phenomenon in IA and is not limited to IA alone (cf.
Section 2.4.4).
However, in Nepali, le also occurs in other constructions. First, le is the marker
of the instrumental case, used to express a means or an inanimate agent. The
head to which it is added can be a noun, participle or infinitive, e.g., hunā=le, lit.:
‘because of being’ > ‘because of this situation’. Whereas in Asamiya, the ergative

1 Nepali grammaticalizes different layers of honorificity; for instance, certain forms are only
used to address or refer to members of the royal family.
 Nepali: description 149

marking -e marks the agent in a passive sentence (cf. Chapter 3), in Nepali, the
postposition bāta is used. However, the passive rarely occurs with an overt agent.
Impersonal passives are much more frequent in Nepali, and often, the difference
between an experiencer construction and a passive construction is semantically
not clear (cf. Clark 1963: 107–108, ex. (150)).

(150) tyo timī=lāī cāh-in-ch-a?

that you=obj want-pass-prs-3sg
‘Do you want that?’ (< ‘Is this wanted for you?’)

Verma (1976: 277) analyzed kocā as an agent in the following passive sentence; it
is more probable, however, that kocā is a pure instrumental rather than an agent.
Considering that the verb is causative, an inanimate A would be quite unusual.

(151) kocā=le yo kothā saphā gar-ā-i-yo

broom=erg this room clean make-caus-pass-pst.3sg
‘This room was (caused to be) made clean with a broom.’

The following example from Abadie (1974) illustrates that le can occur twice in a
sentence, simultaneously marking both A and the instrumental. In comparison,
the objective marker lāī can occur only once in a sentence.

(152) lathi=le maiṃ=le sisi phuṭa-eṃ

stick=ins I=erg bottle break-pst.1sg
‘I broke the bottle with the stick.’

Although there are differences in the marking of the passive agent, the ergative
postpositions in both Nepali and Asamiya obviously share the feature that they
are used to mark the instrumental as well.
Second, in Nepali, independently used transitive participles, transitive con-
verbs and infinitives may take a le-marked A (cf. Bickel and Yadava 2000: 354).²
In ex. (153), the perfect participle gar-e-ko (‘made’) is used as a determiner of
āwāj (‘sound’), which is the O of another determining participle, sundā. The A of
gareko, unīharū=le (‘they’), differs from that of the intransitive main verb ga-yo,
and it is indicated with le.

2 This stands in contrast to Hindi, in which the overt S/A of a non-finite verb form must be
either in the objective or genitive case (cf. Bickel and Yadava 2000, Section 2.4.2.1).
150 Northern Indo-Aryan: Nepali

(153) unī-harū=le gar-eko karākakuruk-kacyākakucuk=ko āwāj

he-pl=erg make-prf.ptcp noise=gen sound
sun-dā poi=ko man=mā pani kehī ciso
hear-prs.ptcp.m.sg husband=gen mind=loc too little chill
pas-na ga-yo
enter-inf go-pst.3sg
‘Hearing the sound of the noises they made, a little chill entered the hus-
band’s mind too.’ (Hutt 1997: 53)

Wallace (1982) draws attention to the behavior of Nepali converbs (or “Conjunc-
tive participles” in his terminology). In his opinion, converbs on -i or on -era are,
as per definition, perfective; thus, if they are transitive, their A should be marked
with le. Peterson (2002) supports this idea by suggesting that the infix -e- found in
the ending -era is the perfective marker. In many cases, the subject of the converb
is different from that of the main verb, which explains why there is no con-
junction reduction and why both subjects are marked according to their verbs.
However, if the main verb and the converb share the same subject, the subject is
only expressed once in the sentence, and its case marking is either determined
by the main verb or by the converb. Below are two of Wallace’s examples (1982:
168) in which the converb is transitive and the main verb intransitive. As in the
other IA languages, A and S are treated in the same way in coordination patterns.
However, because A and S do not share the same morphological marking, there is
a possible conflict if one verb is transitive and the other intransitive. The subject
may be marked according to the converb or according to the main verb. In the fol-
lowing examples, the le-marking is controlled by the transitive converbs bheṭera
and khāera.

(154) mai=le tyas=lāi bheṭ-era ghar=mā ga-eṃ

I=erg he=obj meet-cvb house=loc go-pst.1sg
‘I met him and went home.’

(155) us=le bikh khā-era mar-yo

he=erg poison eat-cvb die-pst.3sg
‘He died while eating poison.’

There are also examples of the opposite situation; that is, where the subject is
in the nominative because of the intransitive main verb, although the converb is
transitive. This pattern is the most frequent (cf. Wallace 1982: 172). Note that in
 Nepali: description 151

the latter example, it is not clear whether unī is controlled by the second converb,
the intransitive gaera, or by the main verb ḍaṭin.³

(156) chorā mahīnā din=ko bidā pā-era kalkaṭṭa=bāṭa

son month day=gen leave get-cvb Kolkata=abl
ghar ā-e
home come-pst.3sg
‘The son came home from Kolkata after getting a month’s leave.’ (Wallace
1982: 168)

(157) unī korībāṭī gar-era jhyāl=mā ga-era ḍaṭ-in

she make-up do-cvb window=loc go-cvb stand-pst.f.3sg
‘After she put on make-up, she went to the window and stood there.’ (Hutt
1997: 49)

Third, le can function as the marker of the experiencer argument in modal con-
structions of obligation or permission, for instance with the verb par ‘must’. It
occurs in variation with the lāī-postposition, which, as in Hindi, marks the expe-
riencer argument of an experiencer construction. Less frequently, the argument
of a construction of obligation can also be left in the nominative case (Clark 1963:
146). According to Abadie (1974), when le is used, the modal should be interpreted
as having an “internalized necessity” with a greater obligatory power than when
lāī is used. Below are two of her examples illustrating the difference.

3 The reason why the converb determines the marking of the subject in some sentences,
whereas in other sentences the main verb is determining, has not yet been systematically in-
vestigated in Nepali. An interesting hypothesis for Newari has been brought forward by Genetti
(1988). Newari is a Tibeto-Burman language, spoken in close contact with Nepali. Quite the
same situation of conjunction reduction is found in this language, i.e., sometimes the subject
of the main verb and sometimes that of the converb is deleted. Genetti explains this by arguing
that it is always the subject in a topic role that is maintained. It is left for further investigation
if this is also the reason for conjunction reduction in neighbouring Nepali. Note that in Indo-
Aryan in general, there is quite some variation on this rule. In Hindi, it is always the main verb
that determines the case marking of the subject. In Asamiya, the general rule seems to be that
the case marking of the subject is determined by the verb which it precedes, irrespective of
whether that verb is transitive or intransitive (cf. Chapter 4). In Gultari Shina, the subject case
is usually determined by the converb (if the converb is an experiencer verb, the experiencer/
subject takes the objective case), unless the subject follows the converb because it then takes
the case determined by the main verb (Hook 1996).
152 Northern Indo-Aryan: Nepali

(158) mai=le mantri hunu+par-ch-a

I=erg minister be+must-prs-3sg
‘I should be a minister.’

(159) hami=lāī ta jhan rittai hinnu+par-ch-a ke

we=obj more empty walk+must-prs-3sg what
‘At least we can walk empty [without a load].’

On the basis of data from Kathmandu, Clark (1963: 146) maintains that le and,
albeit to a lesser extent, lāī are used with a transitive infinitive verb. In combina-
tion with intransitive infinitives, lāī is the least common marking of the experi-
encer and le the most common marking; the nominative occupies an intermediate
position between these two in terms of frequency of occurrence. For Hutt and
Subedi (1999), intransitive verbs take a lāī-marked argument and transitive infini-
tives a le-marked argument, which is almost the opposite of Clark’s observation.
By all accounts, it is clear that a great deal of regional variation is involved (Karki
and Shrestha 1974: 73; Bickel and Yadava 2000: 359) and that there is no clear
preference for either le, lāī or nominative marking. Nevertheless, the le-marked
experiencers of modal constructions of obligation/permission do not act in the
same manner as the le-marked A’s of perfectives, in the sense that the former
do not control agreement. Nor is there agreement with the lāī-marked argument;
however this is as expected because the verb never agrees with an objective-
marked argument in Nepali. In these kinds of expressions, either le is used in its
modality of instrumental marker, or (non-native?) speakers of Nepali have found
it acceptable to replace the object-marking postposition lāī with the subject-
marking postposition le. In Hindi, one finds a similar extension of the ergative
marker (cf. Section 6.2.4).
Interestingly, the ergative marking le is also often added to A of an imperfec-
tive construction. Because this is an exceptional pattern in IA, it will be discussed
in Section 4.2.

4.1.3 Verb agreement in Nepali

Nepali has an extended verb system to express various nuances of tense and
aspect use, and the forms in this system have been labelled in different ways in
the literature (cf. Morland-Hughes 1947; Srivastava 1962; Clark 1963; Verma and
Sharma 1979; Wallace 1982; Matthews 1984; Acharya 1991 and Riccardi 2003).
 Nepali: description 153

In particular the perfective verb forms have been given many different names.⁴
The Nepali simple past is a form consisting of the root combined with inflective
person endings. Other forms are created by combining the root or the participle
of the main verb with the copula meaning ‘to be’. There are two possible ways to
form the perfect and past perfect; one of them periphrastic, the other synthetic.
The two forms do not occur in the same circumstances; for instance, Acharya
(1991) considers the periphrastic perfect to be a present perfective and the second
perfect a past unknown.⁵ The constructions with these forms do not, however,
display a different case marking and agreement pattern.

simple present root-chu gar-chu

imperfect root-theṃ gar-theṃ
present continuous pres.part.-dai+chu gar-dai+chu
present subjunctive root-um gar-uṃ
simple past root-eṃ gar-eṃ
present perfect past.part.-eko+chu gar-eko+chu
second present perfect root-e-chu gar-e-chu
past perfect past.part.-eko+thieṃ gar-eko+thieṃ
second past perfect root-e-theṃ gar-e-theṃ
past continuous pres.part.-dai+thieṃ gar-dai+thieṃ
definite future inf.-ne+chu gar-ne+chu
indefinite future root-uṃlā/uṃlī gar-uṃlā
conditional future inf.-ne+thieṃ gar-ne+thieṃ

Table 4.2 Tenses in Nepali (1st person singular)

Verb agreement in Nepali is straightforward: the verb always agrees with A/S,
even when A is ergative. This is illustrated in ex. (160), where banāī is f.sg., in
agreement with A.

(160) ṭunṭunī=le āphno guṃḍ banā-ī

bird[f]=erg one’s own nest[m] make-pst.f.3sg.
‘The bird made its nest.’ (Thapa 2001: 5)

Agreement is equal for all tenses, and O-agreement does not occur in Nepali. In
general, Nepali shows agreement in person and number. The copula chV ‘to be’

4 For instance, the form ‘root-e-chu’ has been called ‘present unknown’, ‘second perfect’, ‘nar-
rative’, ‘short completed present’, ‘unknown past’ and ‘present perfect’.
5 A more extensive semantic discussion of this “second perfect” is found in Michailovsky
(1996).
154 Northern Indo-Aryan: Nepali

sg.m sg.f pl.m/f

1 chu chu chauṃ

2 chas ches chau

3 cha che chau

Table 4.3 Conjugation of chV

sg pl

1 -eṃ -yauṃ
2 -is -yau
3 -yo (m), -ī (f) -e

Table 4.4 Endings of the simple past

also displays gender agreement (cf. Table 4.3), just like certain forms of the future
and the simple past.
The present participle on -da(i) is invariant in the present continuous tense.
Gender agreement of the participle on -eko, but also of the copula chV, is often
absent in spoken languages, and sometimes in written languages as well. Accord-
ing to Genetti (1999: 545), agreement in number can also disappear in the spoken
language, thus creating an agreement pattern identical to that of the Eastern IA
languages, in which there is only agreement in person.
Scholars have sought to explain this subject agreement pattern, because, con-
sidering the different case marking of S and A, it is not the pattern one expects to
find. The Nepali agreement pattern differs considerably from the Hindi agreement
pattern, in which the perfective verb agrees with an unmarked O. In Hindi, if O is
marked, there is no agreement in gender, number and (optionally) person. Wallace
(1982: 188) offers a historical explanation for the Nepali pattern on the basis of a com-
parison with the pattern found in Hindi. He surmises that, in an undefined previous
stage of the Nepali language, O-agreement occurred in perfective constructions. He
assumes that this “Old Nepali” agreement was identical to the one in Hindi, i.e.,
there used to be agreement with an unmarked O. In his opinion, the ultimate cause
of the contemporary subject agreement and the absence of O-agreement is the
disappearance of grammatical gender. In modern Nepali, only nouns with a clear
female human reference are grammatically feminine. Consequently, all feminine
O’s are marked with lāī because they necessarily refer to an animate being. Wallace
argues that, analogously to Hindi, there would never have been gender and number
agreement with a feminine O in a hypothetical Nepali O-agreement system because
 Nepali: description 155

all feminine O-arguments are marked, and marked O’s cannot control agreement.
All perfective verb forms would take the masculine singular form, which, accord-
ing to Wallace, is the default form of the verb when there is no agreement with
an argument. He further contends that the gender/number agreement of the finite
verb is additionally combined with so-called “pronominal affixes”. These affixes
display person agreement with the subject. According to Wallace, because most of
the conjugated verbs displayed masculine singular agreement in combination with
a pronominal affix referring to the A, gender/number agreement in general was
reanalyzed as being controlled by A, not by O. This reanalysis subsequently spread
to every transitive construction, and O-agreement was lost.
There are several problems with Wallace’s account. First, it is not certain that
Old Nepali patterned in the same way as Hindi or that the Nepali objective postpo-
sition lāī prevented agreement in the manner of the Hindi objective postposition
ko. On the contrary, in Rajasthani, the finite verb consistently agrees with O, irre-
spective of the marking of that argument (cf. Chapter 6). On the basis of modern
Nepali, it is impossible to determine which type of verb agreement dominated in
the older stages of the language, but, in general, the Nepali language seems to be
more closely related to Rajasthani than to Hindi (e.g., the Nepali ch-form of the
copula is identical to Harauti, in Hindi the copula has the onset h-; the postposi-
tion lāī is more closely related to Rajasthani naiṃ than to Hindi ko; etc.). Second,
Wallace’s “pronominal affixes”, which are controlled by subject argument, are
not well-defined, and he himself admits that “there is really no evidence to indi-
cate definitely one way or another what the origin of the Nepali personal suffixes
is” (Wallace 1982: 186). The pronominal affixes he refers to are the person endings
of the simple past, similar to the endings of the copula chV (Table 4.3). Both finite
verbs in the simple past and periphrastic verb constructions with chV agree in
person with the subject in contemporary Nepali. Because Wallace cannot define
the origin of the pronominal affixes, it is possible that the finite verb has always
agreed with {A, S}, in gender, number and person, by means of these pronominal
affixes; therefore, there is no need to assume a previous stage in which there was
O-agreement. Another counterargument against the view that the disappearance
of grammatical gender led to subject agreement is given by Beames (1966: 179).
Although Beames deals in particular with Bangla, the argument can also apply
to Nepali. He maintains that the disappearance of grammatical gender differen-
tiation in Bangla was an effect of the language’s use of tense conjugations with
personal agreement as opposed to participles. Because the conjugation in Bangla
did not display gender agreement, the only remaining formal indication of gram-
matical gender on nouns and pronouns came to be considered obsolete. Beames’
view of the language change at issue is the opposite of Wallace’s argument. Obvi-
ously, it is only possible to resolve the matter by investigating Old Nepali sources.
156 Northern Indo-Aryan: Nepali

Cross-linguistically, a mismatch in alignment between case marking and agree-

ment patterns such as the one in Nepali is not exceptional (cf. Section 1.1.4.2).
The pairing of an ergative case marking and an accusative agreement pattern has
been related to the assumption that person agreement derives from bound pro-
nouns (Dixon 1994). Free, lexical pronouns display the cross-linguistic tendency
to pattern accusatively.⁶ Therefore, bound pronouns on the verb (i.e., person
agreement) are also expected to display an accusative pattern, irrespective of the
marking of the overt argument with which they agree. In contrast to the simple
past forms in Hindi and Rajasthani, which only agree in gender and number,
Nepali does display person agreement with the controlling argument. Hence, the
mismatch between the case marking and the agreement pattern in Nepali does
not need to be “caused” by gender simplification but may very well be related to
the occurrence of person agreement (Wallace’s “pronominal affixes”). In the IA
languages, person agreement seems to have the effect that gender/number agree-
ment on the verb is often absent. In modern colloquial Nepali, person agreement
is consistently followed, but gender and number agreement is less strictly applied
(Genetti 1999). In Asamiya, one finds a similar pattern to the one in Nepali: there
is consistent accusative verb agreement combined with the consistent ergative
marking of A. In Asamiya, grammatical gender and gender/number agreement
have completely disappeared.

4.2 Reanalysis and extension of the postposition le

The postposition le marking A can appear in imperfective constructions in Nepali,

that is, in constructions without a perfective marker -e-. This phenomenon has
been described by several authors, including Turnbull ([1923] 1982), Clark (1963),
Verma (1976), Kachru and Pandharipande (1979), Wallace (1982, 1985) and, most
recently, Li (2007). Below are a few attested examples of this rather unusual
occurrence of le in several imperfective constructions.

Present
(161) tapāīṃ=ko hajām=le mero ghāu kina kāṭ-i-diṃ-dai-na
you.h=gen barber=erg my wound why stitch-lnk-give-prs-neg.3sg
‘Why doesn’t your barber stitch my wound?’ (Thapa 2001: 6)

6 However, as mentioned in Chapter 1, the pronouns, therefore, do not need to have a nomi-
native form for A and S; a pronominal A and S can show the same oblique/ergative form.
 Reanalysis and extension of the postposition le 157

Imperfective
(162) dinhauṃ ṭunṭunī=le tyas=lāī praṇām gar-th-ī
day.pl bird[f]=erg she=obj greeting do-iprf-f.sg
‘[there were] days when the bird greeted her.’ (Thapa 2001: 5)

Present Continuous
(163) yas=le bhan-dai+ch-e
she=erg say-cont+prs-3f.sg
‘She is saying.’ (Thapa 2001: 10)

Present Subjunctive
(164) kina hāmī=lāī kasai=le samjh-os?
why we=obj someone=erg remember-sbjv.3sg
‘Why would anyone remember us?’ (Hutt 1997: 62)

Past Continuous
(165) tyas=le keval ṭunṭunī=ko baccā khān-ch-u bhanera
she=erg only bird=gen child eat-prs-1sg quot
soc-da+th-ī
think-cont+pst-f.sg
‘She was only [repeatedly] thinking “I eat the children of the bird”.’ (Thapa
2001: 5)

Definite Future
(166) ṭunṭunī=lāī nai khā-e-ṃ bhanera rājā=le soc-ne+ch-an
bird=obj emph eat-pst-1sg quot king=erg think-fut+prs-3sg.h
‘The king will think “I ate the bird”.’ (Thapa 2001: 11)

Indefinite Future
(167) tarsā-eko ra na-tarsā-eko hajūr=le thāhā
terrify-prf.ptcp or neg-terrify-prf.ptcp you.h=erg knowledge
pā-i-bakse-lā!
get-lnk-royal.h-fut.3sg
‘Terrified or not terrified, you will find out.’ (Hutt 1997: 52)

The postposition le can be used with every Nepali tense but the frequency of
occurrence varies. Moreover, le is not mandatory in any of the previous examples.
Use of le is entirely optional for speakers; the examples without le are considered
as grammatical as those with the postposition.
158 Northern Indo-Aryan: Nepali

In addition to these occurrences of le with a simple tense, the postposition also

occurs in combination with tenses using a compound verb construction; e.g., the
intransitive verb rah ‘to stay’ is used to indicate the progressive. The copula rah is
intransitive, but if the main verb is a transitive verb, then le can again be used as
in the examples below:

(168) her-ūṃ na timī=le mero ghāu kasarī na

see-prs.1sg not you=erg my wound how not
kāṭ-da-rahe-ch-au?
stitch-cont-prog-prs-2sg
‘I don’t understand how you cannot stitch my wound?’ (Thapa 2001: 6)

(169) svara-harū=le mānauṃ us=lāī bolā-i-rah-ekā+ch-an

voice-pl=erg as if he=obj call-lnk-prog-prf.ptcp.pl+prs-3pl
‘[it was] as if voices were calling him.’ (Hutt 1997: 75)

If these sentences are translated into Hindi, they are constructed with a nomina-
tive A. However, in Nepali, rah forms an integral morphological part of the verb
(e.g., gar-e-rah-e-chu ‘I was doing’), while in Hindi, rah is morphologically more
independent in the verb construction (e.g. kar rahā/rahī hūṃ ‘I (m/f) am doing’,
ham kar rahe haiṃ, ‘we are doing’, cf. Chapter 2). It could be argued that in the
Nepali rah-construction, the auxiliary verb is formally more grammaticalized
than in Hindi. Consequently, it is possible in Nepali to overrule the argument from
Hindi that rah is intransitive and therefore does not take an ergatively marked A
because Nepali rah has completely lost its lexical meaning in this construction.
The transitivity of the construction is determined by the main verb rather than the
lexically empty auxiliary. Nevertheless, the fact remains that constructions with
rah are imperfective, and le only optionally marks A in this kind of construction.
In this respect, Nepali deviates entirely from most Indo-Aryan languages. The
postposition ne in Hindi is never used in combination with a present, progressive
or continuous tense. In Asamiya, the ergative marker -e does occur in imperfec-
tive constructions, but obligatorily (cf. Section 3.1.2), whereas in Nepali speakers
seem to have the choice to use it or not. This apparently unmotivated extension
of the postposition in Nepali has led to many accounts that attempt to explain the
remarkable distribution of le, but a unified explanation has not yet been attained.
 Reanalysis and extension of the postposition le 159

4.2.1 Previous accounts

4.2.1.1 Focalization
Traditional grammarians and most teachers’ books of Nepali explain the pecu-
liar occurrence of le marking A with imperfectives by arguing that le is used to
stress A. This account would provide a fairly straightforward explanation for
the optional nature of the postposition, under the condition that it is only used
when emphasis is involved. However, native speakers denied that le is a marker
of emphasis on A (Abadie 1974). Moreover, the emphasis explanation raises the
question of precisely why A is emphasized or must be emphasized. Hutt and
Subedi (1999: 116–117) observe that le is used outside its normal range in the fol-
lowing three situations:
(i) “if the sentence says that it is part of the natural order of things for the
subject to perform the verb, and therefore states that this is a role that is
specific to the subject”,
(ii) “if the sentence is a question asking who or what is the subject of a transi-
tive verb”, or
(iii) “if the sentence is a response to a question […], or focuses in any way upon
the subject of the verb”.

The latter two occurrences suggest that a ‘focalization’ of A may be the factor
that causes the addition of the postposition le. ‘Focus’ is defined in Dik (1989:
277) as a pragmatic function indicating “that information which is relatively the
most important or salient in the given communicative setting, and considered by
S [Speaker] to be most essential for A [Addressee] to integrate into his pragmatic
information”. The speaker may either “add pieces of information” or “replace”
them. It is implied that there is a setting/background in which the conversa-
tion takes place and that the hearer/addressee already has some information at
his/her disposal. In the first instance, the information in focus is new and adds
something to the information already available; in the second case, the speaker
contrasts the focused information with already known information and conse-
quently changes the available information. Question-answer pairs are ideal con-
structions for illustrating focus, according to Dik (1989: 279), and this observation
is illustrated by the occurrence of le in Nepali (consider situation (ii) and (iii) in
Hutt and Subedi 1999). The example given by Clark (1963: 279), repeated in Bickel
(2011), illustrates that the focus is on karmiharu ‘the workmen’ in the answer to
the question.
160 Northern Indo-Aryan: Nepali

(170) bāhira ke-ko khalbal? – karmi-haru=le chānā hāli-rahe-ch-an

outside what-gen noise worker-pl=erg roof lay-prog-prs-3pl
‘What’s the noise about outside? – It’s the workmen, laying the roof.’

In ex. (170), karmiharu is new information that adds up to the already known
information (there is noise), but le may as well be used when another A-argument
is introduced that differs from an A-argument in a prior clause. This A-argument
need not be new; it can be an argument that has thus far been unimportant to the
discourse (Tichekhoff 1978: 156–157). For instance, in ex. (171), saubhāgyavatī=le
replaces guṃḍo as the new central argument, although it is arguably not a totally
new argument in the discourse:

(171) yo sun-ī guṃḍo bhan-da+ch-a […] ani

this hear-cvb rascal[m] say-cont+prs-3sg.m then
Saubhāgyavatī=le bhan-da+ch-in […]
Saubhagyavati[f]=erg say-cont+prs-3sg.f
‘After he heard this, the rascal is saying […] and then Saubhagyavati is
saying […].’ (Hutt 1997: 51)

In ex. (172), the le-marking seems to be simply emphasizing the A-argument, but
instances such as these are rather rare.

(172) hajūr=ko sammān puryā-una ma jābī=le ke

you.h=obj sufficient honour-inf I paltry woman=erg how
sak-ta+th-eṃ
can-cont+pst-3sg
‘How can a paltry woman such as I honor you sufficiently?’ (Hutt 1997: 50)

Although acceptable for the previous example, the focus/emphasis approach is

not applicable to all examples. For instance, a number of Nepali native speakers
produced the following sentence when asked to translate the English sentence
The workmen are laying the roof (note that the informants used a slightly different
lexical expression for workmen, but the construction is basically the same as in
ex. (170)).

(173) jyāmi-haru=le chānā lagā-dai+ch-an

worker-pl=erg roof put-cont+prs-3pl
‘The workmen are laying the roof.’
 Reanalysis and extension of the postposition le 161

No additional context was supplied; no question is involved, nor is there any

extra-lingual emphasis on A. Furthermore, because the expression has no well-
described context, there is no background situation and the speaker does not
need to add or replace particular information. Neither focus nor emphasis is in
evidence.
Many examples can be explained as related to the marking of focus, either
because A is simply emphasized (cf. ex. (172)) or because a new A is introduced
or stands in opposition to a previous A (cf. ex. (170)). However, many other sen-
tences do not seem to have a focalized A whatsoever, even though the A-argument
is marked with le. In other words, emphasis/focus may be the motivation of the
occurrence of le when a new A that is different from the previous subject is intro-
duced; in other instances, emphasis/focus is clearly not the decisive factor.

4.2.1.2 Syntactic restrictions

Kachru and Pandharipande (1979; followed by Riccardi 2003: 557) offer an alter-
native explanation based on syntax for le marking the A of imperfective verb
constructions. Their hypothesis is easily refuted regarding the attested examples.
They claim that le is the marker of all imperfective verb forms, i.e., perfective,
future and subjunctive forms. In contrast to what they claim, however, simple
present tenses may also take A marked with le (cf. ex. (161)). Furthermore, most
examples in the data are present tense constructions. Although le occasionally
appears in combination with future or subjunctive forms, this combination is by
no means obligatory. Combining a future or subjunctive form with an A marked
with le is a matter of choice rather than obligatory. Similarly, Abadie (1974) argues
that le never occurs with progressives; however, it is not difficult to find exam-
ples where le does occur as the marker of A of a continuous verb (cf. ex. (163)).
Informed by the Tibeto-Burman ergative pattern, Abadie further suggests that the
presence of le may indicate a greater volitionality of A. However, convincing evi-
dence for this opinion is not found in the texts.

4.2.1.3 Disambiguation hypothesis

Abadie (1974) examines the disambiguating or distinguishing functions of case
markers. She suggests that A can be marked with le when it is impossible for
the hearer to identify the agent of a clause. The postposition then functions as a
disambiguating suffix. This idea is similar to the already discussed focalization
function of le, particularly the instances when le is used to introduce a different
A. The disambiguation hypothesis has been further explored and extended by
Verma (1976) and later by Li (2007). Verma (1976) restricts le-marking to focaliza-
162 Northern Indo-Aryan: Nepali

tion contexts but at the same time argues that every inanimate A is obligatorily
marked with le. In a markedness-based framework, an inanimate A is atypical,
and according to Verma, le therefore functions as a distinguishing device only to
identify an otherwise unrecognizable (due to its inanimate nature) A. This view
on DSM in Nepali is related to a common view on DOM, in which an atypical
(hence functionally marked) O also needs to be formally marked (Aissen 2003,
Section 1.1.3.2). Bickel (2011: 406), following Clark (1963) and Pokharel (1998),
endorses this analysis and argues that “the odds for this {S, A}-status decrease
if the A is abstract or inanimate or non-topical; in all these cases, A arguments
are more likely to be projected into an ergative {A}-relation even in the nonpast”.
The assertion that non-topical arguments tend to be marked with le is some-
what contradictory to the view that le-marked arguments express focalization,
given that there is a continuum between focal and topical arguments (Dik 1989:
266, who defines topic broadly as “the entity of a discourse”). However, Bickel
seems to refer to a narrower definition of topicality that excludes focus. Abstract
A’s indeed occur regularly with le, yet there may be another explanation for this
observation that is in line with Hutt and Subedi (1999)’s situation (i): “if the sen-
tence says that it is part of the natural order of things for the subject to perform
the [action expressed in the, SV] verb, and therefore states that this is a role that is
specific to the subject”. For instance, ex. (174) from Bickel (2011) is non-perfective,
but constructions such as these generally occur with le added to the inanimate A.

(174) dhumrapān=le aru=lāī kharab gar-ch-a

smoking=erg other=obj harm do-prs-3sg
‘Smoking harms others.’

A similar instance is found in Verma (1976: 276); see ex. (175):

(175) curut=le dhit mār-ch-a

cigarette=erg appetite kill-prs-3sg
‘Cigarettes kill appetite.’

However, the nature of the expression in ex. (174) and (175) is such that it might
also be analyzed as expressing a general truth, and this condition has been used
to explain the presence of le (cf. Hutt and Subedi 1999). Thus, A may be marked
because it performs “the natural order of things”, not because it is abstract or
inanimate.
Non-topical and, to a certain extent, abstract concepts are ‘interpretable’ in
the sense that much depends on the context and the speaker to decide whether
s/he wishes to interpret or convey an argument as being focused, emphasized
 Reanalysis and extension of the postposition le 163

or fully abstract. Inanimates, however, form an undisputable referential cat-

egory and thus are likely to be marked in a consistent manner. However, Abadie
(1974) provides examples of inanimate A’s that do not take le, such as yo kathā
bhan-da+ch-u ‘this story says’; similar constructions can also be found in other
sources, e.g.:

(176) sireṭo muṭu cheḍ-lā bhane jasto gar-th-yo

wind heart pierce-fut.3sg quot if do-iprf-3sg
‘The wind blew (acted) as if it would pierce the heart.’ (Acharya 1991: 191)

(177) santān vinā svarga=ko bāṭo chek-in-ch-a bhanne

children without heaven=gen way bar-pass-prs-3sg quot
hindū dharma jān-os
hindu belief/rule know-sbjv.3sg
‘May the Hindu belief/rule, saying that the way to heaven is barred to one
if he does not have children, know it!’ (Acharya 1991: 192)

On the other hand, counterexamples in which A is animate but marked with le are
abundant; for instance, consider ex. (161)–(168) and the following one.

(178) mero sāthi=le momo khā-i-rahe-ch-a

my friend=erg momo eat-lnk-prog-prs-3sg
‘My friend is eating momo’s.’

Some informants constructed this with le and some without le, even though ‘my
friend’ is obviously not inanimate.

4.2.1.4 Stage- vs. individual-level predicates

A new hypothesis has recently been advanced by Butt and Poudel (2007). The
authors attribute the appearance of le to a conceptual distinction between stage-
level and individual-level expressions. It had been established earlier that selec-
tion of the copula is dependent on this difference. Nepali has two forms of the verb
‘to be’: ho is used for “a lasting/inherent property of a referent” (Butt and Poudel
2007: 5) and constitutes the individual level; conversely, chV expresses unique
states limited in time and place. In Butt and Poudel’s words, this expression is
“a property of a referent that holds for a slice of that referent’s spatio-temporal
existence”, i.e., looked upon as if they were performed on a stage (hence stage-
level). This verbal distinction is extrapolated by Butt and Poudel to the distribu-
tion pattern of le. Imperfectives with le are considered individual-level expres-
164 Northern Indo-Aryan: Nepali

sions, whereas the A-argument of a stage-level expression is said to be normally

indicated without le; see the following examples from Butt and Poudel (2007: 7):

(179) cālak=le gāṛi calāu-ch-a

driver=erg car drive-prs-3sg
‘The driver drives the vehicles.’ or: ‘A driver drives a vehicle.’

(180) guru gāṛi calāu-ch-a

teacher car drive-prs.3sg
‘The teacher drives the car [today].’

Butt and Poudel’s account focuses on the instantiations of le in a sentence where

“it is part of the natural order of things for the subject to perform the verb, and
therefore states that this is a role that is specific to the subject” (cf. Hutt and
Subedi 1999: 116). Indeed, it is natural for a driver to drive cars. “Kind/generic
NP’s” are considered typical instantiations of the individual-level and should
always occur with le marking. Butt and Poudel further argue that “[i]n the past
tense, the distinction collapses, as the past tense renders everything into a stage-
level predication” (Butt and Poudel 2007: 9). Thus, according to Butt and Poudel,
every past tense construction is necessarily an expression on a stage level, and
general properties of an A-argument (the individual level) are always expressed
with a non-past tense. However, an A in a perfective construction always happens
to take the postposition, which contradicts Butt and Poudel’s claim that stage-
level predicates never display a marked A. There does not seem to be any solid
reason why le would indicate the individual level in the imperfective and the
stage level in the perfective. On the other hand, Butt and Poudel (2007: 9) cor-
rectly observe that habitual expressions should “not be confused with individual-
level predication”. Accordingly, they claim that habitual expressions normally do
not occur with a le-marked A.
The stage- vs. individual-level parameter may be relatively straightforward
with respect to constructed examples, such as those cited by Butt and Poudel
(2007), but with respect to longer narratives, it is more difficult to explain the odd
occurrence of le on the basis of this parameter. In some examples, the content
of the predicate clearly refers to unique actions that are certainly not the usual
behavior or condition of the subjects (cf. ex. (163) and the examples below). Spe-
cifically, a le-marked A often occurs in combination with verbs meaning ‘to say’
or ‘to do’, which are difficult to interpret as individual-level predicates or typical
actions. Of the following examples, ex. (183) is particularly interesting because
dinthyo is an imperfective verb form expressing a habitual action in the past. Butt
and Poudel (2007) claim that this verb does not normally occur with a le-marked
 Reanalysis and extension of the postposition le 165

A. In ex. (184), the men referred to are workers, not professional singers, despite
the implication of the latter by the use of the postposition.

(181) rājā=le sodh-e han yas=le pheri ke bhan-dai+ch-e

king=erg ask-pst.3sg q she=erg again what say-cont+prs-3sg.f
‘The king asked: What is she saying?’ (Thapa 2001: 10)

(182) āphu=le na-bok-era kas=le bok-i-din-ch-a?

oneself=erg neg-carry-cvb who=erg carry-lnk-give-prs-3sg
‘If I don’t carry it myself, who will carry it?’ (Hutt 1997: 67)

(183) bhaiṃsī=le tyām-tyām=mā dūdh din-th-yo

buffalo=erg regularly=loc milk give-pst-3sg
‘The buffalo used to give milk regularly.’ (Hutt 1997: 70)

(184) kasai=le dāiṃ+gar-ne belā=mā git gāuṃ-ch-an

some=erg threshing+do-inf time=loc song sing-prs-3pl
‘Some sing at the time of treshing.’ (Hale 1973: 39)

Although Butt and Poudel do not include habituals in their definition of an indi-
vidual-level predicate and not all habituals are indicated with le, it is clear from
the examples that le does occur in sentences with a progressive tense.⁷

4.2.2 Perfective aspect expressed by le

While the focus/emphasis hypothesis seems particularly useful for interpreting

many occurrences of le, the hypothesis cannot explain each and every occur-
rence of le in our data. Moreover, if one would rely on the rather vague notion
of focus/emphasis as the only remaining explanation for the postposition, this

7 As an illustration of the frequency of the le-marker in combination with an imperfective

verb form: from 8 different short stories, 270 finite imperfective verb constructions, transitive
and intransitive, were analyzed. In 52 of them, all transitive, le marked A. It is clear that the
le marking combined with imperfectives is certainly not a general pattern, and imperfective
constructions without a le-marked argument are far more common. The main part of verbs with
a le-marked A are in the simple present (26), which is as expected, because the simple present
is the most frequent of the imperfectives in narrative literature. 11 verbs with a marked le are in
a continuous tense, 4 in the imperfect, 4 in the present subjunctive and 7 in the future.
166 Northern Indo-Aryan: Nepali

would entail that the explanatory power of the focus/emphasis hypothesis would
be reduced considerably.
On the other hand, Butt and Poudel (2007)’s account is remarkable in that
they essentially transfer a property of a verb (i.e., the possibility of indicating
whether an expression is on a stage level or an individual level) to an argument.
Although it has been shown above that this property is not always discernable in
corpus examples, the idea that le expresses a semantic feature normally attrib-
uted to the verb rather than the construction is worthy of further exploration. The
normal use of le being to mark A of a perfective, I suggest that the optional use of
le adds a sense of perfectivity, or completion, to the action, even though the verb
is not in a perfective form (cf. Tchekhoff 1978: 160). This sense of perfectivity is
related to the telicity of the action expressed by the verb, telicity being one of the
properties of semantic transitivity (Hopper and Thompson 1980).
In traditional linguistics, the perfective aspect of the verb is defined as indi-
cating a completed action; it represents “a situation as a single whole”, denot-
ing “a beginning, middle, and end” (Comrie 1976: 17–18). As an extension of
this interpretation, the perfective designates a state of affairs that is ‘certain’ or
‘factual’ because the action is completed and irreversible. On this view, le is used
to indicate a telic aspect whenever it is impossible for the verb to be in a perfec-
tive tense form. Note that tense and aspect in Indo-Aryan in general are closely
intertwined and connected with the use of ergative marking (cf. Davison 2002,
Section 2.4.2.3). Diachronically, the meaning of le undergoes reanalysis (cf. Harris
and Campbell 1995) in the sense that the postposition was originally used as the
indicator of A in a perfective construction, i.e., as an postposition of the ergative
case, but when used in combination with an imperfective verb form, the meaning
of the postposition le is reanalyzed and extended to indicating a telic aspect,
without losing its ergative meaning of exclusively marking A and not S.
In the following example, the telic aspect is clearly conveyed by the le-mark-
ing:

(185) timī=le tyo musā=lāī mār-i-din-ch-au bhane ma bhāt

you=erg that mouse=obj kill-lnk-give-prs-2sg if I food
khā-ne+ch-u
eat-fut+prs-1sg
‘If you kill the mouse, then I will eat.’ (Thapa 2001: 7)

While both finite verbs in ex. (185) are imperfective, it is quite clear that
māridinchau ‘you kill’ is supposed to be completed before, and is actually a con-
dition for, khānechu ‘I will eat’. In accordance with the aforementioned hypoth-
esis, I surmise that the examples should be interpreted as actions perceived as
 Reanalysis and extension of the postposition le 167

complete or as established “facts” in the mind of the speaker, notwithstand-

ing the fact that all examples invariably refer to the present or the future. For
instance, the speaker conceives of the stitching in ex. (186) as already complete
and expresses her amazement that the action has not yet been performed.

(186) tapāīṃ=ko hajām=le mero ghāu kina

you.h=gen barber=erg my wound why
kāṭ-i-diṃ-dai-na
stitch-lnk-give-prs-neg.3sg
‘Why doesn’t your barber stitch my wound?’ (Thapa 2001: 6)

In most cases, imperfectives in which A is marked with le refer to repeated

actions, but they are always finished in the present. Compare ex. (187) and (188),
both containing marked A and referring to actions that are completed, with ex.
(189), in which A is unmarked as if the action can be endlessly repeated, up to
and including the present. There is no endpoint implied.

(187) dinhauṃ ṭunṭunī=le tyas=lāī praṇām gar-th-ī. […] ab

day.pl bird[f]=erg she=obj greeting do-iprf-f.3sg now
tinī-harū āṃkhā band gar-era bas-dai-nan.
he-pl eye close do-cvb sit-prs-neg.3pl
‘[there were] days when the bird was greeting her. […] Now they don’t sit
with their eyes closed [anymore].’ (Thapa 2001: 5)

(188) ghar=mā pani tyahī dhan ch-a bhanera yas=le

house=loc now that wealth be.prs-3sg quot she=erg
bhan-dai+ch-e
say-cont+prs-3f.sg
‘“That wealth is now in the house”, she is saying.’ (Thapa 2001: 10)

(189) jyotiṣi-hāru devīramaṇa=lāī sallāha din-th-e

astrologer-pl D.=obj advice give-iprf-3pl
‘Astrologers were offering advice to Deviramana.’ (Acharya 1991: 191)

An obvious counterargument to the view proposed here is that not all repeated
actions with an end point in the past have an A marked with le. However, as was
argued earlier, le is by no means obligatory. Using le is optional; that is, it is used
if a speaker wishes to emphasize the completion of an action. In the aforemen-
tioned examples, the speakers apparently considered le to be necessary, thus
168 Northern Indo-Aryan: Nepali

emphasizing the certainty and wholeness of the actions. Note that le also remains
to be associated with transitive verbs in its function of ergative marker.
In addition, when used with future and subjunctive verb forms, le adds the
meaning of irreversibility and lack of choice to the expression. A le-marked A often
occurs in rhetorical questions in which the speaker already knows the answer
before posing the question. Either the speaker has no doubt about the answer or s/
he wants to convey this impression, as for instance in ex. (190) and (191):

(190) kas=le timī=lāī birsa-na sak-ch-a?

who=erg you=obj forget-inf can-prs-3sg
‘Who can [possibly] forget you?’ (Hutt 1997: 62)

(191) kina hāmī=lāī kasai=le samjh-os?

why we=obj someone=erg remember-sbjv.3sg
‘Why would anyone remember us?’ (Hutt 1997: 62)

In ex. (191), the context is such that a woman is angry with a man because he
appears to have forgotten her. In other words, she is convinced that there is no
reason for anyone to remember her and her family, and she is denying the exist-
ence of such a reason before he can provide any arguments to convince her oth-
erwise. In ex. (190), the man wants the woman to firmly believe that he can never
forget her. Both speakers use a question, but by using le they leave no room for
another than rhetorical interpretation of their question; in other words, they are
already presupposing the answer. The future tense constructions in the following
examples express the same certainty.

(192) ṭunṭunī=lāī nai khā-e-ṃ bhanera rājā=le soc-ne+ch-an

bird=obj emph eat-pst-1sg quot king=erg think-fut+prs-3sg.
‘The king will (certainly) think “I ate the bird.”’ (Thapa 2001: 11)

(193) tarsā-eko ra na-tarsā-eko hajūra=le thāhā

terrify-prf.ptcp or neg-terrify-prf.ptcp you.h=erg knowledge
pā-i-bakse-lā!
get-lnk-royal.h-fut.3sg
‘Terrified or not terrified, you will (certainly) find out.’ (Hutt 1997: 52)

The extension of the telic aspect can be applied to a great number of examples.
Moreover, the hypothesis advanced here is in line with some of the other accounts
discussed above. First, it is compatible with the focus/emphasis account of the
descriptive grammarians of Nepali (Clark 1963; Matthews 1984; Hutt and Subedi
 Reanalysis and extension of the postposition le 169

1999), which was shown to hold for many instantiations. The occurrence of le is
demonstrably linked to extra emphasis in the expression but le also occurs when
the completion of the action is emphasized. The emphasis is then not on the subject
but on the action that is conceived of as complete in the mind of the speaker.
Secondly, Butt and Poudel’s division of unmarked stage-level predicates
vs. le-marked individual-level predicates can also be integrated in the present
account. The postposition le can be used to express an action on an individual
level because that kind of action is, in a sense, “tense-less”, i.e. the action has
already been completed at least once in the past. Conversely, actions on a stage
level expressed in an imperfective verb form are often not completed (e.g., the
teacher is driving the car right now), therefore A tends to remain unmarked. The
perfectivity hypothesis avoids the contradiction that arises from the observation
that le is used as a marker of A of a perfective stage-level predicate, although on
Butt and Poudel’s account, le-marking is normally avoided on stage level. But if le
is analyzed as indicating telicity, then le used with the A of a perfective verb form
is only natural. The action expressed by an imperfective verb form may be on the
individual level, but the crucial point is that the predicate refers to a completed
action, or an action conceived of as such. In the expression of general truths, time
does not matter.
The remaining question is why the reanalysis and extension of le happened
only in Nepali and not in other Indo-Aryan languages. Possibly, the extension of
le is related to the A/S agreement one finds in Nepali. Similar to what has been
observed for Asamiya, the strict relation between a perfective verb form and the
ergative case marker may have been weakened from the moment the perfective
verb displays person agreement instead of participial gender/number agreement.
Thus, the space has been provided for the ergative marker either to spread to
other tense constructions, or to disappear (as in Bangla). According to Wallace
(1982: 166), le spread from the perfective to other tenses from the 18th century
onward. Wallace offers examples of the earliest spread of le that are in accord-
ance with the above account of the semantics of the postposition: le indicates
the telicity and certainty of a specific expression. The following examples pose
no problems for this account: the speaker wishes to express an indisputable fact.

(194) kaliñga deśa=ko tyo rājā=kana sabai=le māṃ-dai+ch-an

Kalinga land=gen he king=obj⁸ all=erg consider-cont+prs-3pl
‘Everyone knows that he is king of Kalinga.’

8 kana is the old suffix for the O. Nowadays it is still used as a respectful variant of lāī (cf.
Turnbull 1982 [1923]: 12).
170 Northern Indo-Aryan: Nepali

(195) mai=le huṃkāra gar-ā-uṃlā

I=erg command do-caus-fut.m.sg
‘I will have you do my command.’

The spread of the postposition le in Nepali is of a different nature than the erga-
tive marker -e in Asamiya. The use of the ergative marker -e in Asamiya is syntac-
tically obligatory with any A, whereas the occurrence of the Nepali marker le in
imperfective constructions is determined by semantic factors, in particular the
perfectivity of the action. Syntactically, the postposition le in Nepali shows more
similarities with the Hindi postposition ne, which is restricted to perfective con-
structions, than to the Asamiya ergative marker -e.

4.3 Language contact

The extension of the use of le may have been influenced by language contact.
Nepali is spoken in an environment in which there has been extensive contact
with languages from other families, first and foremost Tibeto-Burman lan-
guages. Inhabitants of Nepal are often bilingual in Nepali and a Tibeto-Burman
language such as Newari or one of the many other Tibeto-Burman languages
spoken throughout the country. It is plausible that the ergative marking in Nepali
has spread to imperfective constructions through the bilingual language use of
Tibeto-Burman speakers – a hypothesis which seems to be confirmed by evidence
from the Shina language.
There are several regional variants of Newari. Because Nepali is now the offi-
cial language of administration and media, the use of Tibeto-Burman dialects is
not encouraged by the government or the media. Newari, like Tibetan, displays
nominal ergativity (an ergative case) without verbal ergativity (O-agreement) (cf.
Tournadre 1996: 18; Genetti 2007: 308), see the following example from Dolakha
Newar (Genetti 2007: 309):

(196) jin galti yet-ki

I.erg mistake do-pst.1sg
‘I made a mistake.’

In Dolakha Newar, a dialect of Newari spoken outside the Kathmandu valley (cf.
Genetti 2007: 1), every A is marked with an ergative suffix. However, in the Newari
variant spoken in the Kathmandu valley, the ergative marking is obligatory with
perfectives. With other tenses, in particular progressive ones, its use is optional
(Genetti 1988: 30).
 Language contact 171

Because there is no verbal ergativity, agreement is always with the subject in

Dolakha Newar (cf. Genetti 2007: 308). Note, however, that just like in Asamiya,
third person past forms of the verb may take a different ending depending on
whether the verb is transitive or intransitive. According to Genetti (2007: 107),
“[t]he suffix is -a when the verb is intransitive and -ju when the verb is transi-
tive. Other differences are found in negative, imperative, prohibitive and optative
forms”.⁹
There are other exceptions on the subject agreement pattern in Dolakha
Newar; for instance, the verb may be marked for the third person while A is a
first person. This deviation of the expected syntactic pattern is generally attrib-
uted to the notion of “control”; in particular, “the subject’s lack of control over
the events denoted in the clause” (cf. Genetti 2007: 310), which is reminiscent
of an impersonal experiencer construction (cf. Malchukov 2008, Section 2.4.3).
In those exceptional cases, the finite verb does not agree with any other argu-
ment in the sentence but simply lacks agreement altogether. In spoken Nepali,
number agreement is frequently avoided, and Genetti (2007: 175, 1999) attributes
the absence of agreement in Dolakha Newar to the influence of Nepali. This is
a plausible hypothesis; however, semantic notions such as control do have an
influence on the agreement pattern in other Tibeto-Burman languages as well (cf.
also for Tibetan: Tournadre 1996).
Because of the long period of contact, mutual influence between Nepali and
the Newari dialects is fairly obvious. However, the extent to which the alignment
pattern of both languages has been influenced by this contact situation remains
unknown. The influence may have gone in either direction. For instance, Kath-
mandu Newari may have lost the ergative marking with imperfective verbs as a
consequence of language contact with Nepali, but it is also possible, and equally
likely, that Nepali has extended the ergative le-marking to imperfective verb con-
structions under the influence of Newari. Ergative marking proves to be quite
consistent in Newari, in spite of some exceptional patterns, and bilingual speak-
ers must switch between Nepali imperfective constructions with an unmarked A
and Newari imperfective constructions with a marked A. Note that other Pahari
languages such as Kumaoni and Gahrwali, spoken more toward the southwest,

9 In Kathmandu Newar, a similar phenomenon is attested to an even greater extent. Har-

greaves (2005) distinguishes the “conjunct” and “disjunct” conjugations in Kathmandu Newar,
which are not only determined by the transitivity of the verb, but also by “agency” and “in-
tentionality”. This ‘split’ pattern in conjugation seems, however, to be more closely related to
evidentiality than to ergativity. It appears to be a Tibeto-Burman phenomenon as it also occurs
in other Tibeto-Burman languages, as, for instance in Lhasa Tibetan (DeLancey 1985, 1986).
172 Northern Indo-Aryan: Nepali

do not display a marking of A with an imperfective finite verb. An Indo-Aryan lan-

guage displaying this extension is Eastern IA Asamiya, which is also spoken in an
area where Tibeto-Burman languages are also common. The ergative -e marking
in Asamiya occurs only in this Eastern language, not in Bangla or Oriya, which
are in no contact with Tibeto-Burman languages.
The Northwestern language Shina is another example of an Indo-Aryan lan-
guage which has evolved in a bilingual setting with a Tibeto-Burman language.
It is remarkable that Shina shows similar features to Nepali, though it is spoken
much more westwards in the Himalayas. These features include ergative marking
in imperfective constructions, and a perfectly accusative pattern of verb agree-
ment. Shina is a name that denotes a group of dialects of which four variants
have been studied so far, Kohistani, Gilgiti, Guresi, and Drasi (cf. Bailey 1924;
Strand 1973; Bashir 2003: 878; Schmidt 2004; Schmidt and Kohistani 2008).
Structure-wise, Shina is related to Kashmiri. It has been established that Shina is
an IA language (cf. Masica 1991: 462), but it is spoken on the border of the IA lan-
guage family and deviates in many respects from the better known IA languages.
Some dialects of Shina display archaic IA features (Guresi and Gilgiti), while the
innovations found in the other dialects obscure the IA genetic relationship (cf.
Schmidt and Kohistani 2008: 3). With regard to alignment, Shina is unique, and
the different dialects in Shina display minor differences in alignment patterns.
Gilgiti Shina (Bashir 2003) is the best known variety of Shina, and more
recently, Kohistani Shina has also been the subject of renewed investigation
(cf. Schmidt and Kohistani 2008). Primarily, in all dialects of Shina, the noun,
pronoun and adjective have two inflectional cases: the unmarked or nomina-
tive case and the oblique. As was pointed out by Schmidt and Kohistani (2008:
51), Kohistani and Guresi Shina possess two ways to mark A. The ending -s(a) is
used for the “Ag.-I” (Agent-Imperfective), or the first ergative case form (ex. (197)),
while there is a second ending -e/-i, which forms the “Ag.-P” (Agent-Perfective), or
the second ergative case form (ex. (198)). The ending of Ag.-I is added to A, noun
or pronoun when the verb is imperfective; the ending Ag.-P is added to the transi-
tive subjects of perfective. In Gilgiti Shina, -s(a) is the only ergative marking.

(197) be-s jo th-ooṇ-as kapar-y-o-gi?

we-erg what do-1pl-aux.prs potsherd-pl-obj-with
‘What are we to do with potsherds?’ (Schmidt et al. 2008: 71)

(198) ich-i se-ṇo gooz-a uch-y-ā-o

bear-erg he-obj.pl home-obl take-caus-prf.m.sg-3
‘The bear took them home.’ (Schmidt et al. 2008: 82)
 Conclusion: alignment in Nepali 173

Since the first linguistic accounts, it has been suggested that Ag.-I. -s(a) has
been borrowed from Tibetan (cf. Bailey 1924: 211; Hook and Koul 2004) because
the marking of A in imperfective verb constructions is unusual in Western and
Central IA, yet similar to Northern languages such as Nepali. Second, an ending
on -s is not a frequent form of marking A in the neighboring IA languages, while
it acts as the ending of the ergative form in Tibetan languages (cf. Bailey 1924: 11,
qtd in Schmidt and Kohistani 2008: 51). The marking of A of an imperfective verb
in Kohistani Shina is therefore quite exceptional. The marking of A of a perfective
verb, on the other hand, is a feature often attested to in IA. The ending Ag.-P, -e/-i
in Kohistani Shina may be considered as a kind of vestige of an old oblique form.
Contrary to -s, -e/-i is a sound common to IA A-markings. In sum, the marking of
A with a past tense verb is a regular feature of Western IA, while the marking of A
with an imperfective seems to have been borrowed from the surrounding Tibetan
languages into Kohistani Shina. In a parallel fashion, it is quite plausible that the
ergative marking in Nepali and Asamiya has spread as well to imperfective tense
constructions under influence of the bilingual language use of Tibeto-Burman
speakers.
Both Newari and Nepali do not show verbal ergativity. Whereas this is
a common feature of the Tibeto-Burman languages, it is rather exceptional
for Indo-Aryan. Nepali shares the kind of person conjugation of the verb with
Eastern IA. However, the copula in Nepali is a more independent entity than the
copula in Asamiya, which is particularly clear in the tenses with a perfect aspect.
Compare, e.g., the Nepali present perfect gar-eko+chu and its shorter form, the
second perfect gar-e-chu ‘I have done’, which is (in formal language) conjugated
for gender, number and person, with the Asamiya perfect kar-ich-e ‘I have done’,
which only conjugates for person (cf. Section 3.1.4). In the Eastern IA languages in
general, agreement is also consistent with S/A. Therefore, it is realistic to assume
that Nepali has undergone the same evolution as Eastern IA, in which – due to
the system of person agreement – agreement developed in such a way so as to
occur consistently with A/S. Again, there is a similarity with Shina. All verbs in
Kohistani and Gilgiti Shina agree with the subject, irrespective of whether the
subject is in the nominative case or the ergative case.

4.4 Conclusion: alignment in Nepali

Nepali displays an alignment pattern that maintains the middle ground between
the Eastern and Central IA languages. First, as in Asamiya, agreement in Nepali
is with A/S. However, in contrast to the person agreement in Asamiya, Nepali
displays gender, number and person agreement. Gender/number agreement is
174 Northern Indo-Aryan: Nepali

expressed on the copula and on some (third-person) forms of the tenses without
a copula, but in colloquial language, it often appears to be neglected.¹⁰ The par-
ticipial form, on which most of the verb tenses are based, is much more distin-
guishable in Nepali than it is in Asamiya; in particular, in the tense/aspect forms
constructed with the perfect participle on -e(ko) and the present participle on
-dai. The overt combination of a participle and a copula in Nepali is similar to the
verb formation in Central IA.
As in Hindi, the A of a perfective verb is always marked with an ergative par-
ticle in Nepali, the postposition le. The marking with le is similar to the erga-
tive pattern in Hindi because it is primarily restricted to perfective verb construc-
tions. In Hindi, this restriction must be observed in all circumstances, whereas
in Nepali, the use of le is freer, and under certain circumstances, le in Nepali can
mark the A of an imperfective verb. However, in contrast to the ergative marker -e
in Asamiya, Nepali le is clearly related to the perfectivity of the construction. This
is demonstrated by the fact that le only occurs in constructions in which there is
a connotation that the action has ended, or is conceived of as such (i.e., entailing
an ending). Therefore, the use of le in combination with imperfective construc-
tions is not as structurally regular as in Asamiya; Nepali le occurs only occasion-
ally in such constructions. Another similarity with the Hindi ne is that the Nepali
postposition le can be used as the marker of the experiencer in a construction
expressing an obligation instead of the objective postposition lāī, although this
may be a dialectal preference. Whereas this particular use of ne is much less
common in Hindi, it does occur in certain areas around Delhi and Lahore and
thus also seems to be a regional variant (cf. Section 6.2.4).
The postposition le resembles Asamiya -e in that it can mark the A of non-
finite constructions, viz. converbs, participles, and infinitival verbs. In Hindi, the
ergative postposition cannot mark the A of such non-finite verbs. Nepali align-
ment is therefore somewhere in between these two strong currents of Eastern IA
and Central IA.
In the next chapter, a geographical leap takes us to Western Indo-Aryan lan-
guages, although we focus on the Northern part of the area. Spoken in the outer
Western border of the IA language area, it is interesting to see how Western IA is
related to Central IA and to examine whether the influences of Central IA have the
same effect as in Asamiya and Nepali.

10 I do not mention honorific agreement explicitly in this discussion because honorific agree-
ment in Nepali does not seem to influence the general alignment and is therefore included
under person agreement.
5 Western Indo-Aryan: Kashmiri
Kashmiri and the Northwestern IA languages in general have long been con-
sidered to belong to a language family other than IA, namely the Dardic family.
The Dardic family was supposed to include Kashmiri and many less frequently
spoken languages such as Shina, Pashai, Dameli and Kohistani among others (for
an overview, see Bashir 2003), all of which are spoken in the mountainous region
on the borders of the Indo-Aryan language area, close to Iranian languages. Since
Morgenstierne (1961), however, “Dardic” has been reduced to a purely geographi-
cal concept, and linguistically, the Dardic languages form several subdivisions of
Indo-Aryan (Koul 2003: 898). In this chapter, Kashmiri, the best studied variety
of the Northwestern IA languages, is the central language. On a superficial level,
Kashmiri and the Dardic languages seem to be very distinct from the Central and
Eastern IA languages in terms of phonology and morphosyntax. For instance,
the northwestern languages use pronominal suffixes (cf. Section 5.2) and Kash-
miri and Poguli in particular display inverse transition. Kashmiri displays a verb
second word order (V2), an exception to any Indo-Aryan language. However, in
other, less extraordinary ways, Northwestern IA resembles Eastern and Central
IA, for instance, in the case pattern and relation between the verb and its argu-
ments. Finally, one should bear in mind that the northwestern languages are
spoken on the outskirts of the IA language area, where language contact is of
considerable influence. For instance, as has been shown in the previous chapter,
Tibeto-Burman influence is noticeable on Shina. Furthermore, the genetic ties
with, and areal closeness to, Indo-Iranian may also have played a role in the
development of the language.
Indian Kashmir is situated in the high north of India, bordering the Himala-
yas. Kashmiri is spoken by more than 5,600,000 people (Ethnologue 2010). Kash-
miri has only recently been taught in primary schools (Koul 2005: 79).¹ The lan-
guage is in heavy competition with English and Urdu, which are still considered
to be of more social importance. Consequently, a high level of bilingualism exists
for Kashmiri and Urdu, particularly in the city of Srinagar (in the city of Jammu,
the competition is between Hindi and Dogri). According to Koul and Schmidt
(1984), there are two main dialects of Kashmiri, which are spoken outside the
cities in the valley, Kashtawari and Poguli (Koul 2003: 898).
The following Section 5.1 gives an overview of the alignment features of Kash-
miri, focusing on the case marking of the core arguments and verb agreement.

1 http://www.kashmirlife.net/index.php?option=com_content&view=article&catid=25%3Ap
ublic-affairs&id=519%3Aalien-mother-&Itemid=165
176 Western Indo-Aryan: Kashmiri

Section 5.2 discusses the phenomenon of “pronominal suffixes” in Western IA.

Starting with the pattern of suffixes in Kashmiri, a comparison is drawn with
other languages in the region. Section 5.3 examines Kashmiri and Marathi, two
languages in which the alignment patterns of second person pronouns are dis-
tinctly different from the alignment patterns of other personal pronouns. Finally,
Section 5.4 summarizes the Kashmiri characteristics and embeds them within a
broader overview of the non-IA languages in the region.

5.1 Kashmiri: description

5.1.1 Case marking in Kashmiri

Kashmiri has an inflectional case marking system that distinguishes four inflec-
tional cases: nominative, objective, ergative and ablative (cf. Sommer 1997; Koul
and Wali 2006).² The nouns are further declined in number and gender, cf.
Table 5.1.

m.sg m.pl f.sg f.pl

nom – – – –
erg an/palat. av i/an av
obj as/is an i an
abl i av i av

Table 5.1 Case endings in Kashmiri³

The nominative in Kashmiri is the unmarked case, without an additional ending.

Nominative plurals of masculine and feminine forms are irregular, and are
formed in various ways. Most masculine nouns do not have a distinct plural form.

2 These cases have been given different names in the literature. For instance ergative is
referred to as “agent” in Grierson (1911), and the objective case is most commonly denoted as
“dative”. The terminology preferred here is in line with the other IA languages and is discussed
in Chapter 2.
3 Kashmiri today is predominantly written in Arabic script (occasionally also in Devanagari).
The vowel system of Kashmiri is more extensive than that of the other Indo-Aryan languages
discussed. I choose to follow a common, simplified transcription of the vowels, i.e., the high
vowels i and u, the lower high vowels ī and ū, mid vowels e/e:, ǝ/ǝ: and o/o:, lower mid a, and
low central ā (cf. Koul 2003: 900). For consonants, I use the IA transcription. A sound which has
undergone palatalization is indicated by y, e.g., myǝ:l ‘fathers’.
 Kashmiri: description 177

If they do, the plural is formed by a change of vowel or a palatalization, or a

combination of both, e.g., mo:l ‘father’, myǝ:l ‘fathers’. For these words, the erga-
tive singular also takes the palatalized form, which implies that the nominative
masculine plural form is syncretic with the ergative masculine singular form. The
objective form of these particular nouns is -is instead of -as (cf. Wali & Koul 1997:
151). Feminine plurals often take the ending -i, but they can also be formed by the
same mechanisms of the masculine plural.⁴ In addition to these four inflectional
cases (cf. Masica’s layer I), other case functions, such as instrumental, posses-
sive and locative, are expressed by means of postpositions added to the ablative
or objective form of the nouns. This range of cases is similar to the wide range of
postpositional marking in other IA languages (cf. Masica’s layer II and III).
Pronouns follow the same inflectional pattern. The first two persons are
proper personal pronouns; third person pronouns are anaphorically extended
demonstrative pronouns.⁵

m.sg m.pl f.sg f.pl

nom 1 bi ǝsy bi ǝsy

2 tsi tohy tsi tohy
3 yi yim yi yim
erg 1 me asi me asi
2 tse tahi tse tahi
3 yemy yimav yemi yimav
obj 1 me asi me asi
2 tse tahi tse tahi
3 yemis yiman yemis yiman
abl 1 me asi me asi
2 tse tahi tse tahi
3 yemi yimav yemi yimav

Table 5.2 Pronouns in Kashmiri (cf. Koul and Wali 2006: 51–52)

As already noted by Kachru and Pandharipande (1979: 202), the first and second
pronominal ergative form is the same as the objective case form, while the third
person takes a specific ergative case form. The authors claim that speech act

4 A detailed overview of this complex matter is found in Koul and Wali (2006: 29–31) and
Shackle (1984).
5 There are three types of pronouns for the third person: proximate, remote within sight,
and remote out of sight. For the sake of brevity, only the forms of the proximate pronoun are
included in Table 5.2.
178 Western Indo-Aryan: Kashmiri

participants (SAP) do not have an ergative case in Kashmiri. However, the first
and second person ablative form is the same as the objective and ergative form
as well, so it seems more adequate to maintain a formal opposition between an
unmarked nominative case form and all other ‘oblique’ case forms for the SAP,
rather than the absence of an ergative case.
The marking of the subject in Kashmiri is consistent. In the imperfective, A
and S are nominative; in the perfective, A takes the ergative case form, while S is
in the nominative.⁶ The marking of O is more complex. Its position on the referen-
tial hierarchy plays a role, as in Asamiya and Nepali, but in Kashmiri the impor-
tance of the hierarchy is extended to the relation between A and O. Differences
in the levels on the referential hierarchy between two arguments of a transitive
clause lead to the attestation of inverse and direct transitions (cf. Hook and Kaul
1987). The direct transition is the more ‘natural’, or functionally unmarked situa-
tion (cf. Comrie 1989: 128), where A is higher on the referential hierarchy than O.
In these instances, the marking of O is nominative, as A, cf. tsi in ex. (199) from
Hook and Kaul (1987: 52).

(199) bi so:za-th tsi to:r

I.nom send.fut.1sg-2sg you.nom there
‘I will send you there.’

In the inverse transition, O is higher on the hierarchy than A, or both O and A

are on the same level. For logical reasons, this hierarchy only occurs when both
arguments are third person (cf. Hook and Kaul 1987). O is marked in the objective
case form in the inverse transition, and the A is marked in the nominative; cf. the
following example from Koul and Wali (2006: 117).

(200) tsi ch-u-kh me parināv-ān

you.nom aux.prs-m-2sg I.obj teach-ptcp.prs
‘You teach me.’

The distinction in case marking between direct and inverse transitions is only
applicable in the imperfective (cf. Table 5.3 for an overview). In the perfective
constructions, O always takes the nominative case form.

6 As with other IA languages, a number of intransitive verbs exists in Kashmiri that do take an
ergative marking on S. These verbs fall in the same class of similar verbs in Hindi, known under
various names as unergatives, anticausatives etc. and are discussed in Section 2.4.4, since
they are not specific for Kashmiri alone.
 Kashmiri: description 179

O 1st person O 2nd person O 3rd person

A 1st person Ø⁷ nom nom

A 2nd person obj Ø nom
A 3rd person obj obj obj

Table 5.3 Marking of O in the imperfective construction⁷

Although O is unmarked in the direct transition, it is obligatorily marked on the

verb by means of a pronominal suffix whenever it is a pronoun (cf. Section 5.2).
Person is one of the features that are part of Silverstein’s referential hierar-
chies. In the person hierarchy, first and second person are the highest ranked,
followed by third pronominal persons and ultimately by nouns. It is clear that in
the case marking and agreement pattern in Kashmiri, this particular referential
hierarchy plays an important role. In contrast to the other IA languages, in Kash-
miri the alignment is determined by the relation between the two core arguments.
In other words, the features of the person hierarchy displayed by A and O are
ranked against each other, and the alignment is based on, and varies according
to, this ranking. One could call this pattern an example of “hierarchical align-
ment”. This alignment type is not dependent on the semantic role or the syntactic
function of the core arguments, but it is fully determined by how these arguments
are ranked on the referential hierarchies (Siewierska 2004: 55–56). Hierarchical
alignment is cross-linguistically associated with person agreement. Although it is
true, however, that the occurrence of the pronominal suffixes in Kashmiri are in
part determined by referential hierarchies, it is the case marking that is hierarchi-
cally based (cf. Table 5.3). In comparison to the other IA languages, the person
hierarchy has a greater influence on the alignment pattern in Kashmiri. In perfec-
tive constructions, however, the influence of the hierarchy is restricted, given that
O cannot be marked and is necessarily in the nominative case, irrespective of its
ranking on the hierarchy.
The IO is always marked in the objective, as in the following example from
Koul and Wali (2006: 50).⁸

7 In sentences such as, e.g., ‘I see myself in the mirror’ the reflexive is expressed with a noun
pān, which means that the reflexive is treated as a third person O, not as a first person pronoun.
8 In certain causative perfective constructions, the IO can take the nominative case, as in, for
instance, kūr (‘the girl’) in asi parinǝ:v kūr hisāb (‘we taught the girl math’). In these instances,
the verb agrees with the IO. However, the use of an objective case for the IO is also possible
here (cf. Hook and Koul 1984b: 105).
180 Western Indo-Aryan: Kashmiri

(201) me di-yiv pǝ:ṃsi

I.obj give-imp.h money
‘Please give me money.’

If an O that should be marked in the objective case occurs in a ditransitive con-

struction, it may lose its objective marking. As in Nepali, a restriction exists on
two objective-marked arguments in one clause. However, this construction rarely
occurs (it is implied that there are three animate arguments in one sentence), and
if it does, it is considered to be ungrammatical by some speakers (cf. Hook and
Kaul 1987: 60).
The inverse-direct transition is not only noticeable in the case marking, it
also influences the agreement pattern and, in particular, the occurrence of pro-
nominal suffixes (cf. Section 5.2).

5.1.2 Verb agreement in Kashmiri

5.1.2.1 The verb system

There are three tenses in Kashmiri: present, past and future. The aspect can be
imperfect, perfect or conditional. Kashmiri lacks a subjunctive (Wali and Koul
1997).

a. The imperfective: present and future tenses and conditionals

The present is a periphrastic tense consisting of a present participle and a copula,
which is the verb āsun, meaning ‘to be’. The present participle in Kashmiri is
indeclinable and ends on -ān. It is also the indicator of the imperfect aspect. Com-
bined with the verb āsun in a present, past or future form, it expresses a habitual
or progressive (Koul and Wali 2006: 86). The future tense may be expressed with
a form of the participle (present or perfect), combined with a future copula, but
there is also a synthetic future tense. This synthetic future tense consists of the
root to which person-marking suffixes are added. In contrast to the conjugated
form of āsun used for the present tense, the future does not agree in gender (cf.
Bhat 1987). The conditional is formed in the same way as the future tense, i.e.,
indicated through suffixes added to the root. However, the conditional may also
be formed periphrastically, with a present or perfect participle in combination
with the copula in the conditional tense.

b. The perfective: past tenses and perfects

Kashmiri possesses a fully developed past tense and perfect aspect system. Tense
and aspect do not merge (as they do in Hindi, cf. Davison 2002). It is the past tense
 Kashmiri: description 181

that is most often used, while the perfect -mut-participle indicates the perfect
aspect (Wali & Koul 1997: 233).
Kashmiri has three kinds of past tenses, each of which are expressed with
a different formation. Koul and Wali (2006: 86–87) label them proximate past,
indefinite past and remote past, following Grierson (1911: 43). The proximate past
is used for events that happened recently, the indefinite past for events that hap-
pened at an unknown point of time in the past, and the remote past is used for
events that happened a long time ago. There is a past verb stem, in which the
stem vowel alternates according to the conjugational pattern. Grierson (1911: 42)
divides the verbs into three groups of formally different conjugations: 1/ transi-
tive and impersonal verbs, 2/ 64 exceptional intransitives, and 3/ intransitives.
The impersonal class of verbs are intransitive verbs behaving as transitives by
taking an ergative subject and a neutral agreement form. These unergative verbs,
which we also find in other IA languages, are discussed in Section 2.4.4. Group 2,
the exceptional intransitives, are intransitive verbs that take, by exception, the
same conjugation as the transitive verbs of group 1. However, unlike the imper-
sonal verbs from group 1, they show no other formal feature of transitivity, such
as the ergative marking of A. On a semantic level, Wali and Koul (2006: 99) argue
that group 2 intransitives are different from group 3 intransitives because the first
ones are more “active”. Group 3, then, includes the both semantically and for-
mally intransitive verbs.
Grierson (1911) further claims that there are four ways to form the stem of a
past tense, irrespective of the type of conjugation: I. the root + palatalization of
end consonant (this palatalization may be combined with a change of color of the
vowel), II. root + yo:v; III. root + yāv; and IV. root + ey-āv. The different stem forms
are assigned to the different conjugations to express the three past tenses:

I/II III

proximate past -y -yo:v-

indefinite past -yo:v- -yāv-
remote past -(y)eyo:v- -(y)eyāv-

Table 5.4 Grierson’s divisions of conjugational type I, II and III (updated with data from Koul
and Wali 2006: 89)

Note that these endings, and the stem, are further modified according to the
gender, number and case of the argument they agree with.
The perfect aspect is expressed with the perfect participle ending on the
inflectional suffix -mut and a form of the verb āsun. The copula may be in the
182 Western Indo-Aryan: Kashmiri

present, past or future tense, to form the present perfect, past perfect and future
perfect tense, respectively.

5.1.2.2 Imperfective constructions

The verb forms expressing present tense or imperfect aspect include the simple
present, the future, past and future progressive, and present and future con-
ditional. S and A of these verb forms are generally in the nominative case. The
marking of O is accorded to the referential hierarchy and the direct-inverse transi-
tion patterns as outlined in Section 5.1.1 (cf. Hook & Kaul 1987: 52).
The verb agrees with the nominative subject. Gender agreement is expressed
by a vowel alternation in the stem of the verb. Person and number agreement is
expressed by suffixes, which are different for the present and future tense. In the
present and progressive tenses, the finite verb – the copula – takes no ending
when the subject is a lexical NP or a third person pronoun. In Table 5.5, the past
and present conjugations of the copula are given. However, when the subject is a
first or second pronoun, so-called “pronominal suffixes” are added to the verb.
The suffixes are different for number. The same pronominal suffixes also occur in
other paradigms (cf. Section 5.2.2).

(202) bi ch-u-s kitāb par-ān

I.m.nom aux.prs-m-1sg book read-ptcp.prs
‘I am reading a book.’

(203) bi ǝ:si-s kitāb par-ān

I.f.nom aux.pst.f-1sg book read-ptcp.prs
‘I was reading a book.’

masculine feminine

present sg pl sg pl

1 chu-s chi cha-s chi

2 chu-kh chi-v cha-kh cha-vi
3 chu chi cha cha

past

1 o:su-s ǝ:sy ǝ:si-s ās-i

2 o:su-kh ǝ:si-vi ǝ:si-kh āsi-vi
3 o:s ǝ:sy ǝ:s āsi

Table 5.5 Past and present conjugations of the copula ‘to be’ in Kashmiri (Koul 2006: 33, 38)
 Kashmiri: description 183

Verbs in a future tense agree in person and number with the subject by means of
the personal endings from Tabel 5.6 (but see Section 5.2.2).

sg pl

1 -i -av
2 -akh -iv
3 -i -an

Tabel 5.6 Personal endings of the future

The verb endings of the future tense are directly attached to the main verb and not
to a copula, as exemplified in ex. (204) with the verb vuch, ‘to see’, which receives
the first person plural future ending -av.

(204) əsy vuch-av hākī mac ti tso:ri baji

we.nom watch-fut.1pl hockey match also four o’clock
‘We will also watch the hockey match at four o’clock.’ (Koul 1985: 225)

If the nominative pronominal O is lower in person than A, the verb obligatorily

expresses agreement with this argument by means of a pronominal suffix. Ex.
(205) (Wali and Koul 200: 116) shows where the pronominal suffix for O is added,
while ex. (206) shows where there is no need for a pronominal suffix, since O is
nominal.

(205) bi ch-u-s-an su parināv-ān

I.nom aux.prs-m-1sg-3sg he.nom teach-ptcp.prs
‘I am teaching him.’

(206) tsi ch-u-kh bati khev-ān

you.nom aux.prs-m-2sg rice eat-ptcp.prs.
‘You are eating food.’

Pronominal subjects and O are easily omitted without causing ambiguity, con-
sidering that the marking on the verb indicates the person of both arguments, as
exemplified in ex. (207) (Wali and Koul 2006: 116).⁹

9 Note that verb and copula change place in this example. This is due to the V2-rule in Kash-
miri, according to which the conjugated verb always occurs at the second place in a sentence.
184 Western Indo-Aryan: Kashmiri

(207) parināv-ān ch-ih-an

teach-ptcp.prs aux.prs-2sg-3sg
‘You are teaching him’

If the O is nominal, the verb may take the pronominal suffix -n referring to this
O, although the referent is not a pronoun. This pronominal suffix is optional, but
the O receives a connotation of definiteness when the suffix is present, e.g. (208)
(Koul 2006: 68). Interestingly, the suffix -n can refer to the following arguments: a
third person pronominal O in the nominative case (ex. (205)); the inanimate O of
a future verb (ex. (239)); and a third person animate O (ex. (208)). Furthermore, -n
also occurs in the perfective constructions in again other ways (cf. Section 5.1.2.3).

(208) bi an-a-n panun do:s ti

I.nom bring-fut.1sg-sg my own friend also
‘I will bring my friend also’

Aside from nominative O’s, an O in the objective case form can also be marked on
the finite verb by means of a pronominal suffix. If O is in the second person, the
suffix is obligatory, and the pronoun itself may be overt. First or third pronominal
O’s are obligatorily deleted if a suffix is added to the verb. Ex. (209) shows the
pronoun as overt, so there is no pronominal suffix on the finite auxiliary, while
ex. (210) shows the pronoun as covert, and a pronominal suffix causes a sound
alternation in the stem of the auxiliary. Note also the difference with ex. (205), in
which A and O have a different hierarchical relation.

(209) tsi ch-u-kh me parināv-ān

you.nom aux.prs-m-2sg I.obj teach-ptcp.prs
‘You teach me.’

(210) parināv-ān ch-ih-ə:m

teach-ptcp.prs aux.prs-2sg-1sg
‘You teach me.’

The presence of a pronominal IO can also be indicated on the verb. As with an

objective O, a second person IO is always indicated by a pronominal suffix and
may be covert, while first and third personal pronouns IO are always deleted
when their pronominal suffix is added to the verb. For instance, in ex. (211) and
(212), the IO is not overtly expressed, but its person and number is inferable from
the pronominal suffix that is added to the finite verb:
 Kashmiri: description 185

(211) bi ch-u-s-as kath van-ān

I.nom aux.prs-m-1sg-3sg story[m] tell-ptcp.prs
‘I tell him a story.’ (Koul and Wali 2006: 117)

(212) bi soz-u-vi cāy kamr-as+manz

I.nom send-fut.1sg-2pl tea room-obj.sg+in
‘I will send tea for you in your room.’ (Koul 1985: 220)

In ex. (213) all pronouns are covert. This example also illustrates the order of the
suffixes, viz. A-O-IO:

(213) soz-ān ch-u-s-an-av

send-ptcp.prs aux.prs-m-1sg-3sg-2pl
‘I am sending him to you.’

In Kashmiri, the pronominal suffix cross-referencing to the subject is always

closest to the verb, followed by the suffix referring to O. The suffix referring to an
IO, if present, always comes last. However, this is by no means the constituent
order one finds in every language in the area (as discussed in Section 5.2).

5.1.2.3 Ergative case marking in the perfective

In perfective constructions, A is in the ergative case form, and O is in the nomina-
tive case form – the same case as S. Contrary to the present tenses, O is never in
the objective case. In the past tense, the intransitive verb agrees with its nomina-
tive subject for person, gender and number. Agreement in gender is expressed by
alternations in the stem vowel of the verb. When S is third person, there is no verb
ending. When S is a first or second personal pronoun, the verb takes the same
personal endings as the copula verb in the present tense. Ex. (214) and (215) are
intransitive verbs in a past tense, agreeing with S.

(214) su gav vuny bāzar

he.nom go.pst.m.sg now market
‘He just went to the market.’
186 Western Indo-Aryan: Kashmiri

(215) subih-an go:-s bi sə:r-as

morning-obj.pl go.pst.m-1sg ¹⁰ I.nom walk-obj
‘In the morning, I went for a walk.’

When the verb is transitive, A takes the ergative case. The finite verb agrees in
gender and number with O, which is indicated by alternations in the stem vowel,
as in the following examples.

(216) ḍakṭar-an dyut zor zi project

doctor-erg give.pst.m.sg emphasis[m] that project
gatsh-i mukmal sapd-un
need-fut.3sg complete make-inf
‘The doctor emphasized that the project should be completed.’
(Koul 2008: 1)

(217) me por yi akhbār

I.erg read.pst.m.sg this newspaper[m]
‘I read this newspaper.’

(218) yemy līch ciṭh

he.erg write.pst.f.sg letter[f]
‘He wrote a letter.’

(219) me pəry akhbār

I.erg read.pst.m.pl newspaper[m]pl¹¹
‘I read newspapers.’

When O is a first or second pronoun, the person and number of the pronoun is
obligatorily marked on the verb by a pronominal suffix. The verb of course still
agrees in gender and number with O, cf. ex. (220):

(220) temi on-u-s bi

she.erg bring.pst-m-1sg I.nom
‘She brought me.’

10 Because of the first person singular ending -s, the stem cluster -av- becomes -o-. Sound
alternations are extremely common in Kashmiri.
11 Note that akhbār has an invariant plural.
 Kashmiri: description 187

The transitive past verb thus basically agrees with O, in the form of vowel alterna-
tion and the addition of nominative pronominal suffixes.¹² However, when the
ergative subject is a pronoun, it can be indicated on the verb. The addition of a
pronominal suffix is obligatory for an ergative second person A. First and third
person pronominal ergative A’s can also mark the verb with a pronominal suffix
and they may be deleted in this case, but the verb marking and the elision are
optional and less common.

(221) tahi sūz-av-ā yi āfsani kin

you.erg.pl send.pst.m.sg-2pl-Q this short story[m] somewhere
chāp-in+khə:tri?
print-inf.obj+for
‘Did you send this short story anywhere for printing?’ (Koul 1985: 230)

(222) tse əni-th¹³ zi mez

you.erg.sg bring.pst.m.pl-2sg two tables[m]pl
‘You brought two tables.’

The same kind of agreement pattern and pronominal suffixes applies in tenses
with a perfect aspect. The perfect participle ending on -mut and the copula āsun
(‘to be’) agree primarily with O. The participle takes over the gender and number,
and āsun follows the rules of the past tense. Ex. (223) is intransitive, ex. (224) and
(225) are transitive. In these examples, the verb displays no ergative pronominal
suffix. Ex. (226) and (227) are transitive constructions with a pronominal subject,
and this time there is a pronominal suffix, which refers to A.

12 In “tritransitive” causative constructions with more than two objects, the finite verb seems
to prefer agreement with the closest constituent in the nominative case, irrespective of the
semantic role this constituent assumes in the clause. For instance, in the following sentences,
the finite verb cāvino:v agrees with dad:
māji cāvino:v me=athi ko:ri dad
mother.erg had given.m.sg I=by girl.obj milk[m]
‘Mother had the milk given to the girl by me.’
cāvinǝ:v agrees with kūr in the following construction:
māji cāvinǝ:v me=athi kūr dad
mother.erg had given.f.sg I=by girl[f] milk[m].
‘Mother had the girl given milk by me.’ (cf. Hook and Koul 1984b: 107).
13 The history of this form is as follows: any-th > anyi-th > ǝni-th (Hook p.c.).
188 Western Indo-Aryan: Kashmiri

(223) tohy ch-i-v-ā kəśīr-i gə:-mity?

you.nom.pl aux.prs-m.pl-2pl-q Kashmir[f]-abl go-ptcp.prf.m.pl
‘Have you gone to Kashmir?’ (Koul 2006: 71)

(224) me ch-e təm’sinzi sārey kitāb-i pari-matsi

I.erg aux.prs-f.pl his all book[f]-pl read-ptcp.prf.f.pl
‘I have read all of his books.’ (Koul 1985: 249)

(225) me ch-a yi kitāb pər-mits

I.erg aux.prs-f.sg this book[f] read-ptcp.prf.f.sg
‘I have read this book.’

(226) tohi ch-u-v-ā yetī panun makān

you.erg.pl aux-m.sg-2pl-q here one’s own house[m]
banov-mut?
make-ptcp.prf.m.sg
‘Have you here constructed your own house?’ (Koul 1985: 349)

(227) tohi ch-a-v-ā ti kitāb

you.erg.pl aux.prs-f.sg.-2pl-q also book[f]
pər-mits?
read-ptcp.prf.f.sg
‘Have you also read this book?’ (Koul 1985: 249)

5.2 Pronominal suffixes in Western Indo-Aryan

5.2.1 Bound pronouns or agreement markers?

Pronominal suffixes in Western IA are known under various names, such as

bound pronouns, pronominal clitics, pronominal affixes, (en)clitic pronouns
and so forth. Although the term ‘suffix’ sounds perhaps more old fashioned than
‘clitic’, it is an adequate description for Kashmiri. According to Bynon (1989: 57)
“clitics retain mobility, whereas affixes have fixed positions within the word”.
‘Clitic’ is a term often used in Romance linguistics, as in an example from Chilean
Spanish given by Bynon (1989: 32): le di el libro a María, ‘I am giving the book to
Maria’, where le is a clitic referring to María. It is not suffixed, although it does
occupy a certain fixed position in the sentence. However, pronominal suffixes
in Kashmiri are always added to the verb in a fixed position, occasionally even
 Pronominal suffixes in Western Indo-Aryan 189

causing a phonological merge with the final sound of the verb form. Hence, the
term ‘suffix’ seems to be more appropriate in this context than ‘clitic’.¹⁴
In historical typological literature, pronominal suffixation has been consid-
ered a stage in the evolution of dependent marking to head marking (cf. Givón
1976; Nichols 1986; Bynon 1989). It is generally accepted that agreement markers
originate from pronouns. These pronouns used to be anaphoric, referring to a
topicalized lexical NP. The anaphoric pronoun becomes gradually more “bound”
to the verb and after turning into a clitic without any anaphoric meaning, it
ultimately becomes an affix that is morphologically totally bound to the verb,
i.e., a syntactic agreement marker (cf. Lehmann 1982). The overt lexical NP, to
which the pronoun had been anaphorically related, is originally topicalized but
becomes incorporated into the clause and eventually ends up as the syntactic
and semantic argument of the sentence (Givón 1976; Bynon 1989: 31). The two
stages of a ‘clitized’ pronoun and a ‘suffixed’ pronoun are not always easily dis-
tinguished (cf. Bresnan and Mchombo 1987). The most salient difference between
the construction in the languages described by Givón (1976) (Bantu), Bresnan and
Mchombo (1986) (Chichewa), and Bynon (1989) (Swabian German), and the pro-
nominal suffix system in Kashmiri is that the suffixes in Kashmiri do not refer to
a lexical NP but generally refer to a pronoun in an argument role. The topicaliza-
tion of the lexical NP does not play a role in Kashmiri because there is no lexical
NP involved. The pronominal suffix must always refer to a pronoun. The pronoun
and the suffix can occur together, but sometimes, this co-occurrence is prevented
by a combination of factors related to the argument role and the referential hier-
archy of the arguments.
A synchronic view is given by Siewierska (2004: 121–123), who distinguishes
three different types of person agreement on the verb, which she calls person
“markers”. The first of these are person markers that must always occur in combi-
nation with an overt referent. This type of person marker clearly expresses “gram-
matical agreement”, because it cross-references to the overt core argument. The
second type involves the opposite situation in which a person marker does not
occur in combination with an overt core argument. In this pattern, the person
markers do not seem to be agreement markers because no overt argument exists
for cross-referencing. However, this type of person markers might express “ana-

14 This opinion is certainly not universally accepted. Note for instance that Wali and Koul
(1994, 2002) use “clitics”. Haig (2008) uses “pronominal clitics” throughout his work on
Iranian. Pronominal clitics in Iranian, however, have a more flexible position than pronominal
suffixes in Kashmiri.
190 Western Indo-Aryan: Kashmiri

phoric agreement” (cf. Bresnan and Mchombo 1986)¹⁵ in the sense that they could
refer anaphorically to a nominal argument that is not expressed in the sentence.
On that view, person markers are anaphoric pronouns that are “bound” to the
verb. Finally, the third type either occurs in combination with overt core argu-
ments or without overt arguments.
The Kashmiri pronominal suffixes do not seem to belong to any of these three
types. The main reason for this conclusion is that the Kashmiri pronominal suffixes
refer, as a rule, to pronominal arguments – although, under certain circumstances,
some suffixes do occur referring to an overt, nominal argument. Moreover, some of
the Kashmiri pronominal suffixes are predominantly optional, whereas other ones
are obligatory, and some only occur when there is no overt argument.
In the remainder of this Section 5.2, I will make the terminological distinction
between “pronominal suffixes” and “agreement suffixes”. I use the term “pro-
nominal suffixes” when I refer to morphemes that refer exclusively to pronomi-
nal arguments. “Pronominal suffixes” are predominantly optional and they are
capable of occurring in clusters of different pronominal suffixes. “Pronominal
suffixes” come closest to being “bound pronouns”. Conversely, “agreement suf-
fixes” are obligatory, and they can cross-reference to any nominal or pronominal
argument. Therefore, “agreement suffixes” are instances of “grammatical agree-
ment”. In spite of this apparently clear-cut distinction, differentiating between
pronominal suffixes and agreement suffixes is sometimes merely a terminologi-
cal issue that does not hold in an empirical study. This is also the case with Kash-
miri. For instance, suffixes referring to a first or second person nominative subject
of a present tense (e.g., the ones added to the verb āsun in Table 5.5) could be ana-
lyzed as agreement suffixes, considering that they are obligatory. It is also pos-
sible, though, to regard them as pronominal suffixes, given that they only refer
to pronouns. A pronominal suffix referring to a second person ergative subject is
obligatory, again overruling the optionality condition of most pronominal suf-
fixes. The suffixes -n/-kh can indicate the number of the inanimate O in the future
tenses and the animate O in the past tenses, making it disputable to classify them
as purely ‘pronominal’ suffixes. Perhaps because of this complexity, in Kashmiri
and other Western IA languages, the occurrence of pronominal suffixes is widely
acknowledged and has sparked much of discussion (cf. Hook and Koul 1984;
Hook and Kaul 1987; Wali and Koul 1994; Wali and Koul 2002; Butt 2004 amongst
others). The fact that the suffixes occur in the northwestern area in particular,

15 The term “anaphoric agreement” is somehow unfortunate. It could be argued that it con-
tains a contradiction: “agreement” presupposes an argument to agree with, while “anaphoric”
entails that the person marker only refers to an argument earlier mentioned in the discourse.
 Pronominal suffixes in Western Indo-Aryan 191

and also appear in Iranian languages, has led to the conjecture that pronominal
suffixes are an area-bound feature.

5.2.2 Pronominal suffixes in Kashmiri

There are three types of pronominal suffixes in Kashmiri, all of which related to
the case marking of the pronouns they refer to.¹⁶
The pronominal suffixes referring to an argument in the nominative case are
found in Table 5.7. They may cross-reference to 1/ the nominative subject of an
imperfective construction, 2/ the nominative O of a perfective construction, and
3/ the nominative S of a perfective construction. These pronominal suffixes are
obligatory in all instances. They are very difficult to distinguish from agreement
suffixes, particularly because third persons do not have any overt pronominal
suffix. As a result, no conclusive observable difference exists between an agree-
ment suffix agreeing with a nominal third person and a pronominal suffix refer-
ring to a pronominal third person. In other words, the verb only agrees in gender
and number with a third person, nominal or pronominal, and the lack of a nomi-
native pronominal suffix indicates that the form is a third person (recall that pro-
nominal suffixes are obligatory with first and second pronouns). The nominative
suffix paradigm is illustrated with the verb āsun in Table 5.5.

sg pl

1 s /
2 kh v(i)
3 / /

Table 5.7 Nominative pronominal suffixes (Koul and Wali 2006: 113)

The objective suffixes are used to cross-reference to an objective-marked O in an

imperfective or to an IO, e.g., (210)–(213).¹⁷

16 For the sake of convenience and tradition (except for Hook and Koul 1984), I will borrow
the terminology of the cases for the suffixes, thus referring to ‘nominative’ suffixes, ‘objective’
suffixes and ‘ergative’ suffixes. Of course, these suffixes do not fulfill the role of case markers,
nor are they formally related to case forms. They simply cross-reference to arguments in that
particular case.
17 These suffixes are commonly referred to as ‘dative’ suffixes. In line with the terminology I
use for the cases, I prefer ‘objective’.
192 Western Indo-Aryan: Kashmiri

sg pl

1 m /
2 y v(i)
3 s kh

Table 5.8 Objective pronominal suffixes (Koul and Wali 2006: 114)

Their use is relatively straightforward. The second person suffixes are obliga-
tory, while the first and third suffixes only occur if the pronoun is not overtly
expressed. The overt presence of the pronoun is optional, just like the attachment
of the suffix.
The third person singular -s suffix, in particular, has a wider range of use
than simply referring to O/IO. According to Hook and Kaul (1987), -s can also refer
to a dativus commodi that is not overtly expressed, as for instance in the following
example (Hook and Kaul 1987: 66):

(228) su par-ya-s kitāb

he.nom read-fut.3sg-sg book[f]
‘He will read the book to him/her.’¹⁸

The third type of suffixes are ergative. These are used to cross-reference an erga-
tive subject in the perfective. Second person suffixes are obligatory, while first
and third person suffixes may be optionally added.

18 This sentence would normally be translated as su pari kitāb tǝmis (Koul, p.c.). Hook (p.c.)
pointed out to me that the suffix -s can refer to a wide range of arguments that are not overtly
expressed in Kashmiri, such as an (indirect) beneficiary or a non-overt possessor, similar to
the use of the pronominal suffix in Siraiki in ex. (246). While Hook and Kaul (1987) refer to this
use of -s as a dativus commodi, -s can in fact be used on much more occasions than the ones
that traditionally fall within the purview of the dativus commodi. I leave it to further research
whether the use of -s is perhaps better described in terms of a dativus ethicus, which expresses
a certain degree of emotional involvement. Note that in Eastern IA, in particular in Magahi, the
verb can also be completed with a suffix referring to a non-overt argument, which has been
called the “Addressee component”, showing involvement of a non-overt argument with respect
to the conversational act (cf. Section 3.2.1.3).
 Pronominal suffixes in Western Indo-Aryan 193

sg pl

1 m /
2 th v(i)
3 n kh

Table 5.9 Ergative pronominal suffixes (Koul and Wali 2006: 116)

In the following example, the suffix -th refers to the ergative second person
subject. Note that the verb agrees with O in gender and number, indicated in the
stem of the verb.

(229) tse əni-th zi me:z

you.erg bring.pst.m.pl-2sg two tables[m]pl
‘You brought two tables.’

In combinations of suffixes, the ergative suffixes directly follow the verb, which
agrees with O in gender and number. When the ergative subject is an inanimate
third person, -n (sg.)/-kh (pl.) is added at the end. If O is a third person, either
pronominal or animate, the number of O is additionally marked at the end of the
verb with the same suffixes, i.e. -n/ -kh (adding a meaning of definiteness to O, cf.
infra). In ex. (230), the masculine singular of O is indicated in the vowel pattern
of the stem on-, while the suffix -v- indicates that the second person ergative A
and -n is added to indicate (redundantly) that the O is a pronominal third person:

(230) tahi on-iv-on su yo:r

you.erg bring.pst.m.sg-2pl-3sg he.nom here
‘You brought him here.’

Note that -n is not the expected suffix because that would be a nominative suffix
referring to a nominative O of a perfective. However, the third person nominative
suffixes are invariably zero suffixes, which is additional evidence of a preference
for the ergative suffix to indicate the number and person of the unmarked O in
analogy with the pattern of the imperfective constructions.
Thus far, the roles of the suffixes have been described as discrete, and
they seem to correspond with the case form of the cross-referenced arguments.
However, the ergative suffixes form an exception, because they also refer to the
unmarked O of an imperfective. As outlined in Section 5.1.1, the O of an imperfec-
tive takes the nominative case in a direct transition, i.e., when it is lower on the
referential scale than the subject. In these instances, the verb takes an obligatory
194 Western Indo-Aryan: Kashmiri

suffix cross-referencing the pronominal nominative O, and this suffix shares its
form with the ergative suffixes.

(231) bi so:za-th tsi to:r

I.nom send.fut.1sg-2sg you.nom there
‘I will send you there’

In the above example, the suffix -th cross-references an unmarked O of an imper-

fective construction. It may have become clear at this point that the ergative suffix
paradigm cross-referencing an unmarked O contradicts the syntactic functions
and case forms of the argument. Recall that ergative suffixes can refer to an erga-
tive-marked A and to a nominative O. The conclusion must be that the suffixes are
determined neither by the case of the arguments (ergative and nominative), nor
by their syntactic function (A and O).
Several accounts of this unusual state of affairs have been given in the litera-
ture. Hook and Koul (1984) explain it from an argument-distinguishing perspec-
tive, which relies on the linguistic principle of economy. They define the erga-
tive suffixes in the negative, i.e., as “anti-absolutive” suffixes. In other words,
the suffixes remain related to morphological case, although this time they do not
cross-reference to a case form but are ‘in opposition’ with a case form, i.e., the
absolutive. According to Hook and Koul, the ergative suffixes always refer to the
counterpart of the argument in the absolutive case (1984: 130) and, as such, dis-
tinguish between the different arguments. In a perfective construction, the erga-
tive suffix refers to the ergative marked A, and Hook and Koul’s absolutive case
here is the unmarked (nominative) O. For the imperfective, it is less clear. The
ergative suffix refers to the unmarked (nominative) O, but A is also unmarked
nominative. Hook and Koul (1984) regard the nominative A as the “absolutive”
case, considering the nominative O as the “anti-absolutive” case to which the
suffix refers. It is assumed that the nominative A is “more” absolutive than the
nominative O, because the subject role is traditionally associated with the nomi-
native (/absolutive) case.
According to Hook (p.c.), the distribution of the pronominal suffixes indi-
cating the ergative and nominative core arguments is “chiasmatic”. By that, he
means that the two forms display a paradigmatic distribution that on the surface
resembles the form of a cross.¹⁹ In Kashmiri, this kind of distribution applies to

19 Hook (p.c.) draws an interesting parallel with Harauti, in which a similar “chiasmatic”, or
inverse, distribution of alignment features can be observed, but this distribution occurs in the
word order. As discussed in Section 6.2.2, in Harauti both A and O can be marked with the
 Pronominal suffixes in Western Indo-Aryan 195

the ergative and nominative suffixes: the ergative suffixes refer to the (ergative)
A in perfective constructions and to the (nominative) O in imperfective construc-
tions, while the distribution of the nominative suffixes is the other way around.
For example, nominative suffixes refer to the (nominative) A in imperfective con-
structions and to the (nominative) O in perfective constructions as in the follow-
ing Table:

perfective A (erg) O (nom)

ergative suffixes nominative suffixes
imperfective A (nom) O (nom)
nominative suffixes ergative suffixes

Table 5.10 “Chiasmatic” distribution of the pronominal suffixes in Kashmiri

Sometimes, the order in which the pronominal suffixes are attached to the verb
indicates a similar “chiasmatic” distribution, as in the following examples:

(232) bi ch-u-s-ath tsi vuch-ān

I.nom aux-m.sg-1sg-2sg you.nom see-ptcp.prs
‘I am seeing you.’

(233) bi ch-u-s-ath tse vuch-mut

I.nom aux-m.sg-1sg-2sg you.erg see-ptcp.prf
‘You have seen me.’

In ex. (232), the nominative suffix -s- refers to A (bi) and directly follows ch-u-
which agrees in gender and number with A. In ex. (233), on the other hand, the
same suffix -s- refers to O (bi) but it occurs in the same position in the complex
verb form. However, this nominative-ergative order in ex. (233) is exceptional in
Kashmiri. It occurs only sporadically in present perfect constructions (Hook 1987:
67). Normally, the ergative suffix precedes the nominative suffix, as for instance
in the next example (Hook 1987: 67):

same postposition naiṃ. This postposition marks the ergative A in perfective constructions,
but in an imperfective construction, the same postposition can also mark O. The second core
argument in these constructions is unmarked. Interestingly, the unmarked argument is always
the first argument in the sentence so that the word order in imperfective constructions (A-O)
may be said to be reversed vis-à-vis the word order in perfective constructions (O-A) (see Allen
1960: 10). Hook also draws attention to the case marking in Pashto and to verb agreement in
the Iranian language Vafsi-Tati and Assyrian, which display similar “chiasmatic” distributions.
196 Western Indo-Aryan: Kashmiri

(234) tse on-uth-as bi

you.erg bring.m.sg-2sg-1sg I.nom
‘You brought me.’

The concept of a “chiasmatic” distribution is a convenient tool for describing the

behavior of the ergative suffixes in Kashmiri, considering that one core argument
is cross-referenced by these suffixes in order to distinguish the arguments from
each other, irrespective of the case or the syntactic function of that argument.
Another hypothesis has been put forward by Wali and Koul (1994). They relate
the use of the ergative pronominal suffix uniquely with case roles, disregarding the
case form, and they do this from a generative perspective. For them, the use of the
ergative suffixes indicates that “the bare nonperfective object shares an identical
lexical case with the ergative” (Wali and Koul 1994: 29). In contrast to the formal
evidence, they claim that the oblique form of the ergative is only a superficial
feature and that, lexically, the ergative is neither oblique nor “absolutive”, com-
pletely similar to the case of “the bare nonperfective object” (the nominative object
in an imperfective construction). On the basis of the distribution of the pronomi-
nal suffixes, they characterize the ergative by the features [-absolutive, -oblique] in
opposition with the “absolutive” [+absolutive, -oblique], and the “dative”, a pure
oblique, by [-absolutive, +oblique]. That way, the ergative case is on a par with
the case of the unmarked O of an imperfective construction, which also has these
negative feature values. However, their arguments in favor of applying the [-abso-
lutive] and [-oblique] values to the ergative in Kashmiri stem from two other IA
languages, Marwari and Marathi. They assume that in Marwari, there exists some
kind of a “bare” ergative, i.e., a pronoun that is used as A for a perfective verb form
but which is formally not marked as an ergative. Nevertheless, they argue, there is
no agreement between the subject and the verb, which should be the case if the
unmarked pronoun were “absolutive”. Consequently, the ergative does not have
the feature value of [+absolutive]. Similarly, in Marathi, an oblique case (marking
O) triggers neutral agreement, while an ergative form has no influence whatsoever
on the agreement. The ergative, thus, also lacks the property to control neutral
agreement, and lacks [+oblique]. Wali and Koul’s search for arguments for the
feature values in languages other than Kashmiri of course raises some questions.
First, it provides the undesired space to adduce counterevidence for their claims,
again, from other languages. For instance, the ergative marked subject in Nepali is
marked by agreement on the verb, while in Marwari, the verb does agree in gender
and number with the oblique-marked O. The evidence of these languages contra-
dicts Wali and Koul’s hypothesis based on Marathi and Marwari; by all means,
however, the subject of the argument is still Kashmiri. Moreover, the formal evi-
dence in Kashmiri of a difference between the ergative case of A and the nomina-
 Pronominal suffixes in Western Indo-Aryan 197

tive case of O of an imperfective should not be ignored out of hand. This is because
only case displays a different marking associated with a different argument role.
The ergative A and O only share the same pronominal suffixes; the suffixes of every
nominative O are obligatory, while only those of the second person are obligatory
for the ergative. In this view, Wali and Koul seem to make the mistake of transfer-
ring features of the pronominal suffixes to the cases. Case marking can, however,
assume a different function and alignment from verb agreement (cf. Section 1.1.2).
It seems that, all in all, Wali and Koul’s argument boils down to what Hook and
Koul (1984) had already observed: the ergative pronominal suffixes are anti-abso-
lutive in the sense that they are used in transitive sentences to refer to the “less”
nominative (or “absolutive”) argument. In imperfective construction the identifi-
cation of A and O is entirely dependent on the verb agreement pattern (cf. Hook
and Koul 1984: 129). The pronominal suffix referring to A is obligatory, and if O is
pronominal, the use of a suffix is also obligatory. The hypothetical combination of
two nominative pronominal suffixes, one referring to A and one to O, would only
enhance the ambiguity of the case marking instead of resolving it, because it would
be impossible to distinguish the arguments even on the basis of the suffixes.²⁰ The
use of objective suffixes for O is no option because a marked O is already cross-
referenced by an objective pronominal suffix in the imperfective. Ergative suffixes
are otherwise not used in combination with an imperfective verb. Using them for
the unmarked O, therefore, is a remaining option and economical solution.
Interestingly, pronominal suffixes in Kashmiri can be “adopted” in construc-
tions with a complement clause. Hook and Kaul (1987) distinguish various forms
of adoption, which is in fact a kind of “raising” constructions. The main verb
adopts the pronominal suffixes of O or IO of the verb, finite or non-finite, of a
subordinate clause. This may lead to odd constructions, such as one in which an
intransitive main verb is marked with an objective pronominal suffix because the
subordinate construction is transitive, and the O has been adopted. The adoption
of pronominal suffixes illustrates once again the flexibility of the Kashmiri head
marking system.²¹ The application of the adoption “rules” is entirely optional.
Below are a few examples where adoption is applied, all from Hook and Kaul
(1987). The adopted suffix is marked in bold in the glosses.

20 Obviously, one could refer to the order of the suffixes, e.g., the first one refers to the sub-
ject, the second to O. From the study of the Kashmiri dialect Poguli, however, it appears that
the order of the suffixes has changed through the ages in Kashmiri, so confusion could have
persisted (cf. Hook 1987).
21 In other IA languages, the “raising” of an argument of a subordinate verb to the main verb
is not uncommon, but the raising is observable in the case marking rather than the verb agree-
ment (cf. Bickel and Yadava 2000).
198 Western Indo-Aryan: Kashmiri

(235) bi yatsha-n su vuch-un

I.nom want.fut.1sg-3sg he.nom see-inf
‘I shall want to see him.’

(236) bi ā-s-ath tsi syiṭye:šan-i pyaṭh-itul-nyi

I.nom come.pst-1sg-2sg you.nom station-abl at-abl get-inf
‘I came to get you from the station.’

(237) tsi kus chu-h-e:n yatsh-ān bi

you.nom who.nom aux-2sg-3sg want-ptcp.prs I.nom
gotshu-s-an an-un
shall.pst-1sg-3sg bring-inf
‘Who do you want I should bring?’

The future tense displays a particular pattern with respect to pronominal suffixes.
According to Koul and Wali’s grammar (2006: 102–103), a third person inanimate
O can optionally mark the verb for number. The suffix indicating a singular O is
-n and for a plural O it is -kh (the vowel linking these suffixes with the conjugated
verb alternates according to the sound laws in Kashmiri). The endings for the
future tense with an inanimate O are shown in Table 5.11 below.

A sg O pl O

1sg -an -akh

2sg -ihǝ:n -ihǝ:kh
3sg -i -yakh
1pl -iho:n -iho:kh
2pl -ihūn -ihūkh
3pl -in -inakh

Table 5.11 Person endings of a transitive future verb with an inanimate O (Koul and Wali
2006: 103)

The following examples illustrate the options. Either O is not indicated on the
verb as in ex. (238), or it is indicated, and then we find the verb form par-an as in
ex. (239).

(238) bi par-i yi kitāb

I.nom read-fut.1sg this book[f]
‘I will read this book.’
 Pronominal suffixes in Western Indo-Aryan 199

(239) bi par-an yi kitāb

I.nom read-fut.1sg.sg this book[f]
‘I will read this book.’

The constructions in ex. (238) and (239) do not have exactly the same meaning.
The indication of O in ex. (239) entails that this O, yi kitāb, is more definite/spe-
cific than the O in ex. (238). The second pattern of future conjugation, illustrated
in Table 5.11, functions in the same way as the case marking ko of a marked O in
Hindi, i.e., the -n/-kh suffixes mark the definiteness/specificity of O (cf. Masica
1986; Bhat 1987: 129; Koul and Wali 2006: 102). Furthermore, Hook and Kaul
(1987) note that these future tense -n/-kh suffixes occur much more often in refer-
ence to an animate O than to an inanimate (but definite) O-argument, as in ex.
(208) (repeated here as (240)):

(240) bi an-a-n panun do:s ti

I.nom bring-fut.1sg-sg my own friend also
‘I will bring my friend also.’

In line with the idea that -n/-kh are semantic indicators of definiteness is the fact
that they can also refer to an inanimate ergative A, because they may indicate the
definiteness of this A-argument.

5.2.3 Pronominal suffixes in Poguli

Research on pronominal suffixes in Poguli has been done by Hook (1987), who
gives an overview of the various suffixes used in the language, and in much
earlier work by Grierson (1903 VIII, II) and Bailey (1908). Poguli is closely related
to Kashmiri and is often considered to be a dialect of Kashmiri (Koul 2003). The
pronominal suffix system is very similar to that of Kashmiri. Poguli displays addi-
tional resemblances to Kashmiri in other linguistic respects. An O in Poguli is for
instance also consistently unmarked in a past tense. Furthermore, the case of O
in an imperfective is, in Poguli, also determined by the relation of O and A on the
referential hierarchy. The same conditions for a direct or inverse transition from
Kashmiri could be applied to Poguli (Hook 1987: 65).
The pronominal suffix system in Poguli is more elaborate than in Kashmiri, in
the sense that there are four groups of suffixes instead of three, although for each
person, many suffixes overlap in form. There are nominative, ergative, objective
200 Western Indo-Aryan: Kashmiri

and accusative suffixes²². The pronominal suffixes in Poguli also indicate gender
(Hook 1987: 64), although this only seems to be the case for the nominative sin-
gular suffixes and the third person suffixes. It is, however, also possible that the
agreement triggers the gender differentiation rather than the suffixes, given that
sound alternations may be caused by the combination of agreement markers in
the stem and the attached pronominal suffixes. I nonetheless follow Hook in
giving both the masculine and feminine forms of the suffixes, e.g., -us (m.), -is (f.).

nom erg

sg pl sg pl

1 -us/-is -asam -m -n
2 -us/-is -ath -t -u (-v)
3 -u/-i -uā/-ia -ni/-nye -ne/-nye

acc obj

sg pl sg pl

1 -m -n
2 -s -u (-v) -t -u (-v)
3 -n -nan/-nyen -s/-sā -nan/-nyen

Table 5.12 Pronominal suffixes in Poguli

The accusative suffixes are the fourth kind of pronominal suffixes in Poguli.
This type of suffixes are absent in Kashmiri. They refer to the nominative O of an
imperfective, an argument which is indicated by the ergative suffixes in Kashmiri.
The “anomaly” between nominative case form and ergative suffix in Kashmiri, as
discussed in the previous Section 5.2.2, is thus absent from Poguli because the
problem is resolved by an extra type of suffix. This leads to a possible hypothesis
that Kashmiri, formerly, also used to have accusative suffixes but that they even-
tually merged with another type of suffixes, ergative or objective. This claim is not
farfetched if one takes a closer look at the suffixes in Poguli; many forms are the
same and can easily merge, particularly the forms of the ergative and objective
suffixes.
Possible combinations of suffixes in Poguli are rather restricted. First and
second person ergative suffixes may be combined with objective ones, a com-

22 In Hook’s terminology these are called absolutive, ergative, dative and accusative suffixes,
respectively.
 Pronominal suffixes in Western Indo-Aryan 201

bination that we also find in Kashmiri and Sindhi. Nominative suffixes may be
combined with objective or ergative ones, but other combinations are not found.
However, it is remarkable that the order of the suffixes attached to the verb differs
from Kashmiri. In Kashmiri, the suffix cross-referencing to the subject is always
closest to the verb form, irrespective of whether it is an ergative or a nomina-
tive suffix. In Poguli on the other hand, the nominative suffix always comes first,
even when an ergative suffix is present. Hook (1987) infers that the ergative-first
order from Kashmiri must be a later evolution. However, as I will show in the
following Sections 5.2.4– 5.2.5, the ergative is also the suffix in the first position
in Sindhi and Siraiki. This suggests that there is another reason why the erga-
tive suffix occupies the first position in Kashmiri. For instance, there could be a
preference for indicating the grammatical relation of subject {A, S} in such a way
that it must appear closest to the verb. The reason for the nominative-first order
in Poguli may be that nominative suffixes are more similar to person agreement
than to the pronominal suffixes of the kind found in Kashmiri, thus implying that
the nominative suffixes in Poguli are obligatory and necessarily attached directly
to the verb. In contrast, in the ergative construction in Kashmiri, agreement with
O is indicated on the verb, and nominative suffixes are optional.
Cross-linguistically, there is no outspoken preference for one of the two
options. According to Siewierska (2004: 167), the person affix referring to O occurs
more often before the person affix referring to A than vice-versa, but she does
not consider the case marking of the arguments (i.e., she does not take into con-
sideration whether A is ergative or nominative). Nevertheless, Siewierska argues
that a possible motivation for this order is the “principle of relevance” advanced
by Bybee (1985), which posits that elements which have a stronger “semantic and
syntactic bond” are formally closer together than those which do not have such
a bond. Siewierska (2004: 167) assumes that the “semantic and syntactic bond”
between the verb and O is stronger than the bond between the verb and A, and
that the suffix referring to O is therefore normally closer to the verb than the suffix
referring to A. In Kashmiri, of course, O is already indicated on the perfective verb
form by means of gender/number agreement.

5.2.4 Pronominal suffixes in Sindhi

Another neighboring language of Kashmiri, Sindhi, also uses pronominal suf-

fixes, albeit in a slightly different manner. Pronominal suffixes in Sindhi can refer
to pronouns in the function of the ergative subject, O, IO and any object express-
ing a goal/source of motion. The latter objects, in particular, are never cross-ref-
erenced on the verb in Kashmiri. Another difference is that the suffixes in Sindhi
202 Western Indo-Aryan: Kashmiri

may also be attached to a noun or a postposition, in which case they refer to a

pronominal possessor or to a (covert) pronoun determined by the postposition.
There are two kinds of pronominal suffixes in Sindhi, ergative suffixes and
object suffixes, cf. Table 5.13.

sg. erg sg. obj pl. erg pl. obj

1 m(i)/māṃ m(i) (s)ūṃ, siṃ (s)uṃ, siṃ

2 ī/i ī/i va va
3 iṃ s(i) ūṃ n(i), na

Table 5.13 Pronominal suffixes in Sindhi

The ergative pronominal suffixes refer to the pronominal subject of a transitive

verb, but only in the perfective. In Sindhi, the simple past is expressed by parti-
cipial form, which always agrees in gender and number with O. Moreover, erga-
tive suffixes are only added to the verb when O is a third person. The following
examples from Cole (2001) and Trumpp (1872) illustrate the use of the ergative
suffix:

(241) huna=khe āndo-mi

he.obl=obj bring.pst.3sg-1sg
‘I brought him.’

(242) paṛh-iu-va
read-pst.m.sg-1sg
‘You read [something, m.].’

These suffixes are optional, and they can, but do not have to, occur in combina-
tion with an overt pronominal subject. According to Khubchandani (1962), the
overt use of the pronoun is preferred in formal language use and literature, while
the use of a pronominal suffix without an overt pronoun is colloquial. The use of
both pronoun and pronominal suffix is “substandard”.
The ‘object’ pronominal suffixes are the second type. They could be added to
any verb, and they refer to an O, IO or goal/source of motion. In brief, they refer
to any objectival, non-subject argument possible. In the following example from
Cole (2001), the use of the object suffix ni refers to an unexpressed third person
plural IO:
 Pronominal suffixes in Western Indo-Aryan 203

(243) tūṃ kitāb-a ḍīndo+hue-ni

you book[m]-pl give.ptcp.prs+aux.prs.2sg-3pl
‘You used to give the books to them.’

Object suffixes only differ formally from the ergative suffixes for the third person.
In combination with the copula hūanū, there is no formal difference at all between
an ergative and an object suffix; the copula can take one of the two in reference
to any argument role. In some constructions, combinations of an ergative and an
object suffix occur. The object suffix then always follows the ergative suffix, as in
ex. (244) from Cole (2001). Note that the first person singular ergative suffix mi
alternates with māṃ when it is followed by an object suffix.

(244) ca-yo-māṃ-si to īnd-āsīṃ

say-m.sg-1sg-3sg then come.fut-1pl
‘I said to him that we will come.’

Cole (2001) distinguishes a third category of pronominal suffixes in addition to

the object and ergative ones. She calls this the category of the nominative pro-
nominal suffixes, which are used to refer to the subject of future tenses and
intransitive past tenses (cf. 1st p.pl ending -āsīṃ from ex. (244)). Cole claims that
nominative suffixes only exist for first and second person. For the third person,
the verb agrees in gender and number with the subject, but there is zero-marking
for person. Khubchandani (1962) and Trumpp (1872), on the other hand, consider
these endings as (person) agreement on the verb, and not as pronominal suffixes.
It is true that they are not optional. They also tend to indicate gender agreement in
contrast to the ergative and object pronominal suffixes (e.g., tūṃ hal-eṃ ‘you[m]
go’ vs. the feminine form tūṃ hal-āṃ ‘you[f] go’, Cole 2001). Furthermore, there
is no restriction on the combination of “nominative suffixes” and object suffixes,
contrary to the combination of ergative and object suffixes. Moreover, nominative
suffixes only occur as attachments to verbs rather than to nouns or pronouns. For
these reasons, the suggestion made by Khubchandani and Trumpp to consider
them as person agreement seems more convincing than Cole’s proposition of a
third category of suffixes.
Sindhi also displays pronominal suffixes to indicate possessive and adverbial
relations between nouns and postpositions. Pronominal suffixes can be added to
nouns expressing human relationships or parts of the body. Postpositions can be
combined with suffixes, giving rise, for example, to saṇu-mi ‘with me’. The forms
are the same as the object suffixes, with slight variation concerning the second
person. First and second plural pronouns are never cross-referenced by pronomi-
nal suffixes with nouns or postpositions.
204 Western Indo-Aryan: Kashmiri

The pronominal suffixes in Sindhi are clearly “bound pronouns”, in the sense
that their form is almost identical to the one of overt pronouns (cf. Table 5.14).
Note that in Sindhi, the ergative is expressed with the oblique form of the noun
and pronoun.

sg.nom sg.obl pl.nom pl.obl

1 āūṃ, māṃ mūṃ asīṃ asāṃ

2 tūṃ to (t)av(āh)īṃ (t)av(āh)āṃ
3 hī, hīu, hīa, hu, hua hina, hūna hī, hu hinani, hūnan

Table 5.14 Pronouns in Sindhi

5.2.5 Pronominal suffixes in Siraiki

Siraiki/Saraiki is a language spoken in Pakistan, close to the Sindhi-area.²³

Khubchandani (1962) considers Siraiki to be a dialect of Sindhi; however, it
seems to be more closely related to Punjabi than to Sindhi. The matter is not yet
settled (cf. Shackle 1976: 2). As a Northwestern IA language, Siraiki displays a
use of pronominal suffixes comparable to Sindhi, though slightly different in
certain respects. The first source of the use of pronominal suffixes in Siraiki is
Shackle’s reference grammar (1976). Akhtar (1997), and Butt (2004) also describe
and analyze the same pronominal suffixes, but in their opinion, the language is
Punjabi. However, pronominal suffixes in Standard Punjabi are a marginal phe-
nomenon; they are, for instance, not mentioned in Bhatia’s reference grammar
(1993). It is therefore probable that the language described by Akhtar is a variant
of Punjabi, which is influenced by Siraiki. The accounts by Shackle and Akhtar
will be discussed together. Note that although the pronominal suffixes are the
same, they may occur in various degrees of obligation in the variants of Siraiki
and Punjabi.
In Siraiki, the suffixes only occur when the argument they refer to is not
overtly expressed – it is impossible to find both a pronominal suffix on the verb
and an overt pronoun together in one and the same clause. The suffixes are only
attached to finite verb forms, and not to nouns or postpositions, as in Sindhi.

23 There has been some discussion about the name of this language; I follow Shackle (1976: 1)
who prefers Siraiki above other nomenclatures such as Western Punjabi (avoiding confusion
with Standard Punjabi) or Lahndī (not frequent among native speakers).
 Pronominal suffixes in Western Indo-Aryan 205

There are two kinds of suffixes: nominative suffixes used for intransitive subjects,
which only occur in constructions with a past participle, and oblique suffixes,
which may cross-reference to possessives, marked O²⁴, IO and EXP, and to ergative
subjects. The distinction between nominative and oblique pronominal suffixes
is thus quite a deviation from the Sindhi system, in which the ergative suffixes,
though referring to an oblique case, contrast with the suffixes referring to any
object. The nominative suffixes in Siraiki do not differ significantly from person
agreement. As with the nominative suffix paradigm distinguished by Cole (2001)
in Sindhi, Siraiki has no nominative suffixes for the third person. An important
difference between the person markings in Sindhi, however, is that the nomina-
tive suffixes in Siraiki are optional. With the past tense, the use of an ergative
suffix is the most common. In particular, the third person singular oblique suffix
-s or -sū is frequently used in the Punjabi dialect as described by Akhtar (1997).

sg pl

1 s se
2 oṃ he

Table 5.15 Siraiki nominative pronominal suffixes

sg pl

1 m se
2 o/i he/je
3 s(ū) ne

Table 5.16 Siraiki oblique pronominal suffixes

It is possible to add two suffixes one after the other to the finite verb, but only in
certain combinations. The ergative suffix comes first. The pronominal suffixes
may also be added to the negational particle instead of to the verb, cf. the fol-
lowing example from Akhtar (1997). Because of their variability with respect to
the verb/particle they attach to, the pronominal suffixes in Siraiki are closer to
clitical forms.

24 Definite and animate O’s are marked in Siraiki. Since unmarked O’s are necessarily inani-
mate and/or indefinite, they are not likely to be pronominal, which may explain why they are
not cross-referenced to by a pronominal suffix.
206 Western Indo-Aryan: Kashmiri

(245) kam nāī-s kī-tā

work not-3sg do-pst.m.sg
‘He/She did not do the work.’

There is, in fact, only one important condition related to pronominal suffixes
in Siraiki, which is that the suffix never occurs in combination with the overt
pronoun to which it refers. The suffix -s(ū) seems not to appear in combination
with a habitual tense (cf. Akthar 1997), which is constructed with the auxiliary ai.
Akhtar (1997) further mentions that the third person suffix -s(ū) can also be
attached to the verb while referring to a possessive pronoun. He provides the fol-
lowing example:

(246) muṛ-e kam kar-d-e-s

boy-nom.pl work do-prs-3pl-3sg
‘His boys work.’

This use is unique for Siraiki.

5.3 Second person importance: hierarchical relations and

Marathi

As discussed above, special importance is attributed to the second speech act

participant in Kashmiri. Second person arguments are obligatorily cross-refer-
enced by a pronominal suffix on the finite verb, while the pronominal suffixes of
a first or third person are in most cases optional.²⁵ This suggests that the person
hierarchy in Kashmiri is dominated by the second person (which is cross-linguis-
tically not uncommon, as observed by Nichols 1992). However, with regard to case
marking, the hierarchy seems to be “first person > second person > third person”.
A first person A is related to the nominative marking of a second or third person
O, while a second person A is related only to the nominative marking of a third
person O (a first person O being objective). If the second person were the highest
in the person hierarchy, the first and the third person O would be marked in the
same way. Nevertheless, there is no other IA language in which the indication of
a second person argument on the verb is always obligatory, except for Marathi.
Marathi is one of southernmost variants of IA, spoken in and around the area of

25 The exception is the pronominal suffixes of a nominative O of an imperfective, which are

always obligatory, irrespective of the person.
 Second person importance: hierarchical relations and Marathi 207

Mumbai and the state of Maharashtra. Although quite far away from the north-
western area, Marathi also displays the use of pronominal suffixes, but only for
second person ergative subjects.²⁶
In Marathi, third person A-arguments are marked with the -ne postposition
when the verb is in the past tense. The finite verb then agrees in gender and number
with O if it is unmarked. First and second pronouns, on the other hand, do not
take a postposition or a different form for A. However, even with an unmarked A,
the verb retains its agreement with the O. Remarkably, when A is second person,
there is double agreement: in this case, the verb not only agrees with the O, but
also with A for number. In particular, a suffix -s is added when A is second person
singular, and -t when it is plural. In Western Marathi, these suffixes are obliga-
tory, while they are optional in Eastern Marathi (cf. Pandharipande 1997: 412). Ex.
(247) (from Wali 2006: 10) is an example of the marking with -s.

(247) tu kāvitā vāc-l-i-s

you.m poem[f] read-pst-f.sg-2sg
‘You read the poem.’

The suffixes -s/-t, however, are not specifically associated with an argument in
the ergative case. They primarily refer to a subject in the second person. Intransi-
tive verbs in the past tense, for instance, receive the -s/-t when S in the nominative
is a second person, as in bas-l-ā-s ‘you (m.sg) sat’ (example taken from Pand-
haripande 1997: 412). The suffixes -s/-t are also the endings of the second person
in the present tense, irrespective of whether the verb is transitive or intransitive
(cf. Wali 2006: 11; Pandharipande 1997: 409). The following example is from Wali
(2006: 11).

(248) tu ghari jāc-t-e-s

you.f home go-prs-f.sg-2sg
‘You are going home.’

When the perfect participle in combination with a copula is used to express the
present perfect, -s is added to the copula (which, if it is in the present tense, is
the normal ending), but the participle itself displays no marking. The resem-
blance with the treatment of second person pronouns in Kashmiri is remarkable,
with the only difference being that in Marathi the suffix always refers to A/S,

26 There are reasons to believe that Marathi also had other pronominal suffixes in former
stages of its history, which have, however, disappeared in the course of time (cf. Hook 1987).
208 Western Indo-Aryan: Kashmiri

irrespective of the case (nominative or ergative), and whether the verb is transi-
tive or not.²⁷
Rather surprisingly, however, Hook (2011) notes that in Old Marathi this
-s/-t suffix cross-referenced to O, not to an (ergative) subject, as in the following
example from (Master 1964: 130–131):

(249) joḍunu tumhīṃ dīdhal-e-ti

join you.nom give-pst.m.pl-2pl
‘He joined you.’

This illustrates, once more, the variation and flexibility of pronominal suffixes.
They occur in reference to A, O, and IO, and they are clearly not tied to only one
type of alignment.

5.4 Conclusion

The Northwestern IA languages have always been considered distinct from the
central IA languages, to the extent that Kashmiri was said not to belong to the IA
language family at all. In alignment structure, however, Kashmiri shows similari-
ties to Central IA, particularly with respect to case marking. There is an ergative
case marking that only shows up in perfective constructions. The same situa-
tion can be found in Hindi, with the difference being that in Kashmiri the case
marking proceeds by means of an inflectional morpheme, not a postposition as in
Hindi. The marking of O in Kashmiri, on the other hand, is exceptional for the IA
languages. In contrast to Central IA, O is not marked in Kashmiri when it is defi-
nite or animate, and in imperfective constructions it is marked when it is higher
in person than A. Although this produces an entirely different result of O-marking
in Kashmiri and Hindi, the basic principle of the marking of O is nevertheless the
same: O is marked when it is ranked high on the referential hierarchies. In the
Rajasthani language Harauti, the rule that O is only marked when it is definite/
animate is also modified because of the marking of A: O can only be marked if A
is not already (cf. Section 6.2.2). The marking of O in Kashmiri is not governed
by entirely different principles, and it remains determined by the semantics of

27 Although the second person pronoun is undoubtedly the most powerful controller of pro-
nominal agreement, other pronouns may also cause the addition of a pronominal suffix, at least
in certain dialects. Wali (2006: 10) mentions dialectal agreement with a third person ergative
subject, cf. ex. tyā=ne (he.erg) lili=lā (Lili[f]obj) mār-l-i-na (beat-pst-f.sg-3sg), ‘he beat Lili’.
 Conclusion 209

the referential hierarchy. This conclusion is also corroborated by the finding that
in Kashmiri the objective case is also used, as in all other IA languages, to mark
the IO. Moreover, it seems as if Kashmiri somehow “compensates” for the fact
that definiteness/animacy are not formally indicated on a nominal O by suffix-
ing -n/–kh to the verb. When the suffixes -n/–kh are added to the verb, in the last
position, they refer to a core argument that must be interpreted as more definite/
animate than when no verbal suffix is referring to it. Again, the same semantic
properties of definiteness and animacy play a similar role in Kashmiri as in other
IA languages, although they are indicated in different ways in both languages.
With regard to the perfective constructions, Kashmiri displays a strict ban
on the marking of O. Therefore, the perfective constructions in Kashmiri may be
considered to come closest to the definition of ergativity, which says that O is
treated in the same way as S. The lack of marking of O in the perfective construc-
tions can be attributed to several factors. One of the most telling findings is that,
already in the imperfective constructions, O is not marked in the same, semanti-
cally based way as in Hindi. The objective marker in Hindi, Asamiya, and Nepali
is a semantic marker, and formally a postposition. The objective in Kashmiri, on
the other hand, is an inflectional case. If the O in a Kashmiri perfective construc-
tion would be marked in the objective case, Kashmiri would display a tripartite
pattern, with an ergative A, a nominative S, and an objective O. In the typologi-
cal literature, such a pattern is considered uneconomical (cf., e.g., Comrie 1978),
which may explain why it is avoided in Kashmiri. However, the tripartite pattern
occurs in Hindi, Asamiya, and Nepali, but in these languages, the marking of O
depends additionally on semantic conditions and differs from the purely syntac-
tic marking of A and S.
Although Kashmiri case marking is almost a textbook example of ergative
case marking, it is important to point out that the Kashmiri perfective construc-
tion deviates from a “true” ergative construction as far as its agreement pattern
is concerned. The perfective verb forms are, just like in Central IA, based on
participial forms in combination with a copula. The participial form on -ān that
in Kashmiri is used to form the past tense is invariant, and the perfect partici-
ple agrees in gender and number with S or O. The copula, however, is always
expressed in Kashmiri (in languages such as Hindi and Rajasthani this is not
always the case), and it occurs according to the verb-second (V2) rule in Kashmiri,
which stipulates that the conjugated verb always occurs in the second position
of the clause. Because the copula is always present, there is person agreement
in every construction in Kashmiri – unlike in Hindi, where the perfect participle
expressing the past occurs alone and only agrees in gender and number with S/O.
In this respect, Kashmiri is more similar to Nepali than to Hindi. In general, in
the perfective constructions, the verb agrees in gender, number, and person with
210 Western Indo-Aryan: Kashmiri

the unmarked O. The ergative A and IO can also be marked on the verb, by means
of pronominal suffixes. Pronominal suffixes are somewhere between anaphoric
bound pronouns and grammatical agreement. The nominative suffixes that refer
to an imperfective A, S and a perfective O are obligatory, and they come closest
to morphemes expressing grammatical agreement. Moreover, the nominative
suffixes can occur with or without an overt core argument. The same situation
holds for the ergative suffixes that refer to an unmarked O in an imperfective con-
struction. They are obligatory, and therefore similar, again, to grammatical agree-
ment. The ergative suffixes that refer to an ergative pronominal A are optional.
The objective suffixes, on the other hand, only occur when the pronoun is not
overt. Both the ergative A and the objective suffixes referring to a second person
pronoun are obligatory, irrespective of the presence of the overt pronoun.
In the neighboring languages of Poguli, Sindhi, and Siraiki, one encounters
considerable variation in the use of pronominal suffixes. Poguli has a system that
resembles Kashmiri, apart from being a little more extensive. The pronominal suf-
fixes in Sindhi and Siraiki can be attached to nominal arguments and not only to
verbs. In Siraiki, pronominal suffixes are only expressed on the verb when the
argument is not overtly expressed, so this is a clear case of bound pronouns. In
Sindhi, there is a preference for a pattern similar to the one found in Siraiki, but
exceptions occur when the argument is overtly expressed. The exception to this
‘areal’ pattern is Shina; Gilgiti Shina and Kohistani Shina do not use pronominal
suffixes. However, Bailey (1924: 218) mentions the occurrence of pronominal suf-
fixes in the Guresi dialect of Shina. The versatility of the phenomenon of pronomi-
nal suffixes is confirmed by their occurrence in Indo-Iranian. In a number of Indo-
Iranian languages, pronominal suffixes also occur, and again, there is no trace of a
similar pattern between the languages, the variation is extensive (for an overview,
see Emeneau [1965] 1980: 139). Pronominal suffixes in Iranian often function as
clitics rather than suffixes because they do not have fixed positions in the sentence
(cf. Jahani and Korn 2009; Kieffer 2009; Robson and Tegey 2009; Windfuhr 2009;
Edelman and Dodykhudoeva 2009: 782; Bashir 2009). According to Haig (2008:
107), pronominal suffixes in Persian primarily refer to the “Indirect Participant”,
which is his term for arguments expressed by the dative/genitive case in Iranian,
e.g. IO, EXP, possessor, and goal/source of motion. Moreover, the O of a present
tense verb and the A of a past tense verb can also be referred to by pronominal
clitics, although this happens in a less consistent way, according to Haig (2008:
112). Pronominal clitics can be attached to nouns and prepositions, for instance,
in order to cross-reference to a possessor argument or a complement in modern
Persian, thus resembling the pronominal suffix system in Sindhi (Haig 2008: 337).
Windfuhr and Perry (2009: 487) maintain that clitics in Persian may also refer to
an argument that conveys an “indirect experience” expressed in the clause.
 Conclusion 211

The presence of pronominal suffixes similar to the ones in contemporary Iranian

has been attested to since the Middle Iranian stage (cf. Haig 2008: 105). From
the Vedic/Avestic stage, however, enclitic pronouns occurred frequently in the
Indo-Aryan and Iranian languages (cf. Masica 1991: 254; Emeneau [1965] 1980:
154). These pronouns were morphologically shortened versions of the normal
pronouns, but they were not ‘suffixed’ to the verb. It is a plausible hypothesis
that these old Vedic/Avestic clitics survived in the form of pronominal suffixes
attached to the verb, or to any other part of the sentence in Iranian, and that
under the influence of Iranian, the suffix system in Indo-Aryan became rein-
forced (cf. Emeneau [1965] 1980: 155). This diachronic hypothesis, however, has
yet to have been confirmed by conclusive historical evidence.²⁸

28 A comparison with the Indo-Iranian languages should not be limited to the phenomenon of
pronominal suffixes, a broader comparison between the alignment patterns in Indo-Aryan and
Indo-Iranian may be very fruitful for historical and synchronic research as well, in particular
with regard to the so-called “loss” of ergativity in Indo-Iranian. See Hewson and Bubenik (1997)
on ergativity in Persian, Bynon (1979) on ergativity in Kurdish, Payne (1980) on Pamir languages
and Matras (1997) on Kurmanji, among others.
6 Central Indo-Aryan: Rajasthani
The Central IA languages are spoken in the Indian states of Uttar Pradesh,
Madhya Pradesh, Uttarakhand, Himachal Pradesh, Jharkhand, Bihar, Haryana,
Chhattisgarh, Delhi and Rajasthan. This central area, which is intersected by the
Ganges and Yamuna rivers, is known as the ‘Hindi belt’ where Hindi is the offi-
cial language (and the language of primary education), and where the local lan-
guages are considered by many of the inhabitants merely as dialects of Hindi. The
Rajasthani languages are spoken in the state of Rajasthan, a rural and pastoral
area in the west of the Hindi belt. “Rajasthani” (a concept introduced by G.A.G.
Grierson in his Linguistic Survey (1903 IX, II)) includes all of the related “dia-
lects” spoken in the state of Rajasthan. Rajasthani thus is not a specific language;
rather, it refers to a cluster of languages, which, although intrinsically closely
related, do display different features.
As the standard urban means of communication in the area, Hindi is the
dominant language of the Central Indo-Aryan region. A recurring comparison
between Hindi and Rajasthani is unavoidable in this Chapter because most of
the literature on Central Indo-Aryan is based on Hindi. However, the alignment
pattern of Hindi is not representative of all of the central Indo-Aryan languages,
including Rajasthani. In addition to the Rajasthani language group, there are other
groups of closely related dialectal varieties of Central Indo-Aryan. First, there are
the “Hindi varieties”, some well-known dialects of Hindi, of which Braj, Bundeli
and Awadhi are the most important. Hindi is only one variety of this group, and
such a classification gives a fairly different perspective from approaches where
Hindi is the central language. Second, the Bihari languages, which are discussed
in Section 3.2.1, are also often included in Central IA because they are spoken
between the two core areas of Central IA and Eastern IA. However, as was shown
in Chapter 3, the alignment of the Bihari languages is closer to Eastern IA. Third,
two important Indo-Aryan languages, Punjabi and Gujarati, are also included in
Central IA. Note that Gujarati is in many respects similar to Rajasthani because
they share the same ancestor (Old Western Rajasthani; cf. Tessitori 1914), whereas
Punjabi displays more similarities with Hindi. For some other features, in particu-
lar the use of pronominal suffixes, Punjabi is similar to the Northwestern IA lan-
guages (cf. Section 5.2.5). The southwestern language Marathi has been partly dis-
cussed in Section 5.3, but other parts of its grammar are discussed in Section 6.2.4.¹

1 The examples in this Chapter are taken from several short stories in Marwari and Harauti
included in Saraswat and Prem (1984), Sharma (1992) and from published linguistic sources. A
native speaker has been consulted to verify the data.
 Description of Marwari 213

6.1 Description of Marwari

Marwari and Harauti are two Rajasthani languages. Marwari is the Rajasthani
language with the highest number of speakers. Harauti is a smaller Rajasthani
language, but it displays some typical case marking features which are unlike any
other IA language.
In general, if one refers to the “language” Rajasthani, the description applies
most of the times to Marwari, the most widespread and best known Rajasthani
dialect. Marwari is spoken by app. 13,023,000 people (Gusain 2004, Ethnologue
2010) in Western Rajasthan. Marwari has spread to other parts of India outside
Rajasthan by emigration. It is also attested in Nepal and Pakistan. Marwari has
the highest number of speakers, but its status as best known Rajasthani language
is also due to its rich literary history (Kothari 1991: 9).

6.1.1 Case marking

The main case functions in Marwari, and in Rajasthani in general, are indicated
by postpositions. There are two inflectional case forms (layer I), the unmarked
nominative form and the oblique. With consonant stems, the nominative and
the oblique have merged in the singular, but not in the plural, cf. Table 6.1. The
oblique form of the (pro)noun occurs only in combination with a postposition.
Some of the layer II postpositions expressing semantic cases are syu for the instru-
mental function, meṃ and par for indications of place, amongst others. Layer III
case markings are found in combinations of an oblique + genitive postposition
+ lexical postposition. The genitive postposition agrees in case (nominative or
oblique), gender and number with its head like an adjective. In Marwari it takes
the forms m.sg. ro, m.pl. ra, f.sg. and pl. ri.
The postposition related to the marking of the core arguments is ne, or pho-
nological variants of ne, e.g., nai, naiṃ, neṃ, depending on the dialect. ne is a
multifunctional postposition in the Rajasthani languages, but most often it func-
tions as the marker of O and IO, as in Marwari. Note that there is no ergative case
marker in Marwari.
In the paradigm of the pronouns, the nominative-oblique distinction is
absent for first and second person, although there is a wide range of variation of
pronominal forms. Third person pronouns are anaphoric extensions of demon-
stratives.
214 Central Indo-Aryan: Rajasthani

m.sg f.sg m.pl f.pl

nom vowel -o/-au -ī -ā -ī

nom consonant Ø Ø
obl vowel -e/-ai -ī -āṃ -ī
obl consonant Ø -āṃ

Table 6.1 Nominative and oblique case forms of nouns ending on a vowel and nouns ending on
a consonant in Marwari

nom.sg obl.sg nom.pl obl.pl

1 me, meṃ, m(h)ai(ṃ), māiṃ me, mai mhe mhe

hum, mhu(ṃ), mhū mhā
2 taiṃ taiṃ the, tame the, tame
tūṃ thā, tamā
3 (u)vo, o uṇ (u)vai vāṃ

Table 6.2 Pronouns in Marwari

The variation in forms of the pronouns given in Table 6.2 is due to the differ-
ent pronunciations in different regions. However, there is a clear distinction
between two kinds of phonological tendencies in the range of first person sin-
gular pronouns, i.e. pronouns with an -e- or -ai-sound, and pronouns with an
-u- sound. This goes back to a historical merging of cases, to which we come back
in Section 6.2.1. Apart from these pronominal forms, there is also a second person
honorific form in Marwari, i.e. āpām. Bahl (1989: 37) further distinguishes an
inclusive and an exclusive first person plural pronoun, of which the inclusive is
āṃpai, and the exclusive mhe. Like with the nouns, postpositions are added after
the oblique forms of the pronouns.

6.1.2 Verb system and agreement in Marwari

The verb system in Rajasthani is based on a combination of participial forms and

a copula, conjugated for tense and mood. In the different languages, there are dif-
ferent forms of the copula ‘to be’, but two are prevalent: stems in hV- and stems
in chV-. Subjunctive and future forms are synthetic, with a conjugational ending.
In Marwari, there are two kinds of passives. One is constructed with the auxil-
iary verb meaning ‘to go’, (as also occurs in other Central IA languages such as
Hindi). The second passive is a synthetic passive, formed with the infix -īj-. The
 Description of Marwari 215

two passives are used in different contexts. For an overview of the verb paradigm
in Marwari, see Grierson (1903 IX, II), Bahl (1991: 94), Gusain (2004).
Marwari displays split agreement: in the imperfective constructions, the verb
always agrees with A/S, but in the tenses with a perfective verb form, the intran-
sitive verb agrees with S and the transitive verb with O. In participle-based con-
structions, there is agreement in gender and number marked both on the partici-
ple and the copula in a past tense, cf. ex. (250). In constructions with a participle
and a past copula, there is no person agreement. Agreement in person is marked
only on the present tense copula, cf. ex. (251); the participle agrees in gender and
number. One exception noted by Khokhlova (1992: 89) is that a feminine parti-
ciple does not differentiate for number, and neither does a third person present
tense copula.

(250) … jakai din mhārai=sūṃ ā bāt

which day I.gen=abl this thing[f]
ka-yī+h-ī
say-pst.f.sg+aux-pst.f.sg
‘…on which day he had said this to me.’ (Barhat 1984: 41)

(251) kāīṃ bhāī, kasy-o jādū kar-yo+h-ai

what brother what-m.sg magic[m] do-pst.m.sg+aux-prs.3sg
‘What, brother, what magic have you done?’ (Hamsa 1984: 172)

Note that the agreement with O is consistent in any transitive perfective tense
construction. This implies that the verb always agrees with O, irrespective of
whether that O is case marked or not, cf. ex. (252), where ma=naiṃ is marked, but
still controls agreement.

(252) mammī ma=naiṃ god=meṃ uṭhā+li-yo

Mama[f] I=obj lap=loc heave+take-pst.m.sg
‘Mammy took me up in her lap.’ (Barhat 1984: 42)

A deviating pattern of agreement has been reported for the Marwari spoken in the
area of the city of Jodhpur (cf. Magier 1983; Khokhlova 2006). The pattern is cross-
linguistically unusual, because it implies a split of agreement pattern between
the participle and the copula. The example from Magier (1983) that is commonly
cited is the following:
216 Central Indo-Aryan: Rajasthani

(253) mhaiṃ sītā=ne dekh-ī+h-ūṃ

I.m Sita[f]=obj see-pst.f.sg+aux-prs.1sg
‘I saw Sita.’

This agreement pattern is marginal, but is not in the process of disappearing,

since it is still accepted by native speakers and used in recent literature (cf. ex.
(254) from Stronski 2010: 86, Māṇak 2008, Jan. 44):

(254) mhaiṃ aiḍi ḍukanaṃ ḍekh-i+h-uṃ

I.m such shop[f]pl see-pst.f.pl+aux-prs.1sg
‘I have seen this kind of shops.’

In Gujarati as well, there are examples of such an agreement pattern. Gujarati

normally displays the same pattern as Marwari, i.e., the perfective verb agrees
with O in gender, number and person, as illustrated in ex. (255). Note that Guja-
rati does not have a different case form for a pronominal A or S – third persons
and nominal A-arguments take an ergative case marking -e.

(255) me kam kar-yuṃ+ch-e

I work[n] do-pst.n.sg+aux.prs-3sg
‘I have done the work.’

However, occasionally a construction that is similar to the Marwari construction

in ex. (254) appears, e.g.,

(256) me baghā menejament=ne vinantī

I whole management=obj request[f]
karī+ch-uṃ²
do-pst.f.sg+aux.prs-1sg
‘I have requested the complete management [if…].’

In this pattern, the agreement of the participial form is as expected, i.e., it agrees
in gender and number with O. However, the agreement of the copula is with A
(1st person), which is unexpected, given that the agreement pattern of the aux-

2 http://www.divyabhaskar.co.in/2008/09/23/0809231925_ceo_death.html. Thanks to Peter

Hook for drawing my attention to this example.
 Description of Marwari 217

iliary normally corresponds to the agreement pattern of the participial.³ With

respect to Marwari, Allen (1960), Bahl (1989) and Khokhlova (2002) claim that
this agreement occurs only when A is a SAP, in particular a first person, so the
referential hierarchy and possibly also the relation between A and O could play a
role in establishing the conditions of this agreement. If A is high on the referential
hierarchy, or higher than O, the copula agrees with A. Arguably, if both O and A
are third person, it is impossible to recognize which argument the copula agrees
with, since the copula only agrees in person, not in gender. Moreover, many forms
of the copula are identical for number and person; for instance, in Gujarati, the
second singular and the third singular and plural form of the copular verb are the
same (che; Marwari and Hindi second and third person singular hai). Therefore it
may seem as if the pattern only occurs with a first/second person A.
In contrast to Allen (1960), Magier (1983) does not report on any conditions
for A-agreement, according to him it may show up any time. He argues that it is
exemplary of how ergative agreement is gradually disappearing. However, this
split agreement pattern remains marginal and largely unattested in texts. Against
the assumption that such pattern will spread and lead to pure subject agreement
in Marwari, I would argue that agreement with O in Marwari is much more perva-
sive than in other Central Indo-Aryan languages such as Hindi.⁴

3 According to Klaiman (1987: 84), the pattern in which the agreement of the participle is
different from that of the copula is also found in Mewari, though in a less complete form. In
Gujarati and Mewari, it is possible to find examples in which the participle agrees with the O,
but the copula occurs in the default form (3m.sg.), showing no agreement at all. Such a split be-
tween the agreement of the copula and the the agreement of the main verb has been reported
in Marathi as well.
4 Khokhlova (1992: 90) gives the following example, which is perhaps indicative for the influ-
ences which led to the auxiliary-main verb split in agreement:
mhaim saugan=lai lī+h-ūṃ
I.m oath[f]=obj take.pst.f.sg+aux-prs.1sg
‘I swore an oath.’
She notes the split in agreement between the auxiliary and the main verb, and reports that
this example is taken from a Jodhpur based Marwari speaking author. However, this example is
exceptional because of the lai form of the objective postposition, which is typical for Northern
IA languages, such as Nepali. In Nepali, agreement with A is the rule and not the exception, and
it is possible that the lai-marking in this example indicates influence from Northern Indo-Aryan.
218 Central Indo-Aryan: Rajasthani

6.2 Differential marking

In the Rajasthani languages, postpositions marking core arguments display a

high degree of multifunctionality. The use of most postpositions in Rajasthani
is based on principles of differential marking, in the sense that their occurrence
is not only syntactically determined but also subject to semantic conditions (cf.
Section 1.1.3). The postpositions in the different Rajasthani languages, as well as
in Hindi and Punjabi, share a great amount of formal similarity, but they can
function in the most divergent ways across the individual languages. The follow-
ing sections 6.2.1–6.2.4 deal with differential marking in a range of IA languages,
and more specifically: the variation in subject marking in Marwari from a histori-
cal perspective, the multifunctional postposition naiṃ in Harauti, the differen-
tial marking by means of the Gujarati and Marwari postposition ne, and the case
marking in Punjabi, Gujarati and Marathi.⁵

6.2.1 Marwari subject marking: a historical perspective

In contrast to Harauti, Hindi and other Central IA languages, there is no ergative

postposition in Marwari. Consider the examples below. Ex. (257) is an intransitive
sentence with a form for S mhaiṃ, of which I will assume that it is a nominative.
A in ex. (258) is in the same form as the one of S in ex. (257). In ex. (258), the O
śaraṇ=naiṃ is marked according to the criteria of DOM, i.e. it is an animate and
definite argument.

(257) mhaiṃ bego hī vāṃ=rai=kanaiṃ jā-y pūg-yo

I fast emph he=gen=close go-cvb arrive-pst.m.sg
‘I got to him in a hurry.’ (Hamsa 1984: 170)

(258) mhaiṃ śaraṇ=naiṃ dekh-ī

I.m Sharan[f]=obj see-pst.f.sg
‘I saw Sharan.’ (Hamsa 1984: 172)

The absence of an ergative case marking in the constructions with a perfective

verb in a Central IA language is quite remarkable, considering that the agreement
of transitive perfective verbs is with O, hence displaying an ergative agreement

5 Parts of the following sections are based on a revision of Verbeke (2010).

 Differential marking 219

pattern.⁶ Historically however, there are traces that there used to be a separate
form to indicate A in earlier stages of Marwari. First, there is an extensive variety
of pronominal forms (cf. Table 6.2, Gusain 2004: 23–24; Drocco 2008: 98). It has
been established by Khokhlova (1992) that the forms with the phoneme -u- come
from the former nominatives of Old Rajasthani, whereas the forms with -e- or -ai-
were formerly used only as obliques. In Old Rajasthani, the oblique was not only
used as the inflectional form before a postposition. It also functioned on its own
as an ergative case; compare, for instance, the oblique form rājāi ‘king’ in the fol-
lowing example from Old Rajasthani (Khokhlova 1992: 76):

(259) mujh=nai rājā-i ādeś dīdh-au+ch-ai

I.obl=obj king[m]-obl.sg order[m] give.pst-m.sg+aux-prs.3sg
‘The king has given the order to me.’

In contemporary Marwari, these forms of the pronouns (mhai-, etc.) occur not
only as obliques with a postposition, but also as unmarked pronouns designat-
ing S and A without a postposition. Consider the following examples (260)–(263).
They illustrate that in contemporary Marwari, the hum, mhu(ṃ) forms and the
me(ṃ), mhai(ṃ) forms are both used for A and S, despite their different his-
torical background. Although ex. (260) shows an intransitive construction that
always takes a nominative S, the first person form mhaiṃ is used. In ex. (261), the
same pronoun (the variant māiṃ) is used as A of the verb ‘to say’. Ex. (262) has
the pronoun mhūṃ as A. In the second part of the example, the second person
pronoun tūṃ is A, although historically this was a nominative form restricted to
intransitive sentences. Ex. (263) combines both forms mhūṃ and maiṃ in exactly
the same construction, thus proving that there is no apparent difference in func-
tion between the two forms of the pronoun in contemporary Marwari.

(260) mhaiṃ baḍā bābū=rī ṭebal=rai=sāmaiṃ ūbho h-o

I big boss=gen table=gen=in front standing be-pst.m.sg
‘I was standing in front of the table of the big boss.’ (Hamsa 1984: 170)

6 Note that Kutchi Gujarati apparently has the same alignment: no ergative case marking but
O-agreement (cf. Patel 2007: 36).
220 Central Indo-Aryan: Rajasthani

(261) lārlā dināṃ sagaḷ-ā pataṃg uḍāvai+h-ā, māiṃ

passed days all-nom.m.pl kite[f] fly.prs+aux-pst.m.pl I
mamma=naiṃ ka-yo
ma=obj say-pst.m.sg
‘In the past days, everyone was flying a kite – I said to mamma.’ (Barhat
1984: 39)

(262) mhūṃ pappū=naiṃ pūch-yo – tūṃ mā’ṭ sahib=naiṃ kyūṃ

I Pappu=obj ask-pst.m.sg you master sir=obj why
konī bulā-yo?
not invite-pst.m.sg
‘I asked Pappu – Why did you not invite master?’ (Barhat 1984: 41)

(263) mhūṃ mammī=naiṃ kai-yo, paṇ mammī isai kām=meṃ

I ma=obj tell-pst.m.sg but ma such work=loc
phaṃsar-ī h-ī ke mammī=naiṃ yād konī
sunk-f.sg be-pst.f.sg that ma=obj memory[m] not
rai-yo, maiṃ pāpā=naiṃ ka-yo – paṇ pāpā
stay-pst.m.sg I pa=obj tell-pst.m.sg. but pa
bhī me=rī bāt konī suṇ-ī
also I=gen word[f] not hear-pst.f.sg
‘I told mamma, but mamma was sunk in so much work that mamma did
not remember, I told papa – but papa also did not listen to my words.’
(Barhat 1984: 41)

The oblique form of the third person singular uṇ rarely occurs as A or S. However,
if it is used, there is free variation between the use of the oblique and the nomina-
tive pronoun vo for A (cf. Magier 1983: 245).⁷
According to Khokhlova (1992), the strict opposition between the ergative
and the nominative is only maintained in the m.pl. paradigm. The form vāṃ is
the oblique form of the third person plural pronoun, nominative vai, in combi-
nation with a postpostion. In some constructions, vāṃ is also used for A, as in
ex. (264) and (265). Confirming Khokhlova’s research, I have found no examples
where vāṃ occurs as S.

7 Of course, it is possible that there is a certain preference for a form dependent on the area
where the language is spoken, and it is conceivable that the degree of contact with other Ra-
jasthani languages or Hindi may influence the frequency of the occurrence of uṇ or vo.
 Differential marking 221

(264) vāṃ pūch-yo – kām=meṃ koī dikkat to nīṃ

he.erg?pl ask-pst.m.sg work=loc some problem[f] then not
ā-ī?
come-pst.f.sg
‘He asked – hasn’t any problem come up in the job?’ (Hamsa 1984: 171)

(265) bāṃ savāl gaḷat batā-yo

he.erg?pl question[m] wrong speak-pst.m.sg
‘He said, the question is wrong.’ (Barhat 1984: 40)

A third person pronominal A is not exclusively expressed with the pronoun vāṃ.
The originally nominative form vai in the following ex. (266) is widely attested,
and can be used either as A or as S:

(266) vai baṛā het=sūṃ mhārī āḍī sirkā-ī

he.pl great care=with I.gen problem[f] solve-pst.f.sg
‘He solved my problems with great care.’ (Hamsa 1984: 170)

If case is defined as a system of formal and functional oppositions (cf. Dik 1978;
Masica 1991), vāṃ is not the structurally ergative case of the third person pronoun,
given that the variant vai is used for A as well. The use of vai shows that there is
no systematic difference between the marking of the S and A in Marwari, just as
the other pronouns show.
The concept of a systematic ergative case seems not to exist in contemporary
Marwari. However, as the evidence showed for the first person singular, there
used to be an ergative pronominal case. Moreover, remnants of this case can still
be found in the third person plural paradigm, in which the use of the oblique case
of the pronoun is restricted to A. An explanation for this fact may be found in
the closeness of the third person pronoun and the nominal declension. The third
person pronouns are anaphoric extensions of demonstrative adjectives, and as
such they display more similarities to nouns than to the first and second personal
pronouns.
Indeed, nouns in Marwari generally possess a nominative and an oblique
case form, but often there is no overt difference between the two cases. The
oblique case is only used and recognized when it is followed by postpositions.
However, in some paradigms we can again notice residual forms of an older erga-
tive case, especially for the masculine plural paradigm; not coincidentally, these
forms resemble the paradigm of the third person pronoun. It is only in the mascu-
line paradigm ending in a vowel that the different cases take distinct inflections
(cf. Table 6.1). Nouns ending in a consonant remain unchanged in the nominative
222 Central Indo-Aryan: Rajasthani

plural and oblique singular and take -āṃ for the oblique plural. The nominative
is the case used for all subjects. Consider example (267); bhāyalā, a noun ending
in a vowel, is in the nominative plural case.

(267) sagaḷ-ā bhāyal-ā pūch-yo – āj mhā=nai

all-nom.m.pl friend[m]-nom.pl ask-pst.m.sg today I=obj
sagaḷ-āṃ=nai kiyāṃ yād kari-yā?
all-obl.m.pl=obj why memory[m] do-pst.m.pl
‘All the friends asked – Why do you remember us all today?’ (Sharma 1992:
106)

As in the third person plural pronominal paradigm, the paradigm of masculine

plural nouns is most susceptible to traces of the older division between a nomi-
native and an oblique without postposition (= ergative). According to Khokhlova
(1992: 87), “in m.sg. paradigm of nominal stems in -o/-au, there exist free vari-
ation of nominative and ergative markers for A while S is generally marked by
Nominative”. She gives the following examples (268) and (269), in which the stem
of the noun, however, does not end in -o/-au, but in a consonant. (Khokhlova
1992: 87–88):

(268) anek vidvān likha-‘r nahiṃ bhej+sak-yā

many scholar.pl write-cvb not send+can-pst.m.pl
‘Many scholars could not send [their] articles.’ (Chandrasimh 1966:2)
< after they had written [them], many scholars could not send [them]

(269) bidvān-āṃ thor-au mārg dikhā-yau

scholar-pl some-m.sg way[m] show-pst.m.sg
‘The scholars showed some way.’ (Bhandari 1966:20)

For Khokhlova, the difference between the form vidvān (vidva:n) as S and
bidvānāṃ (bidva:na:M) as A is illustrative of the discrepancy between the nomi-
native case and the ergative case. However, in the feminine paradigm, the ending
-āṃ is often used for S, as is shown in ex. (270), although the expected form is the
invariant form āṃkhyā. The ending in -āṃ can be considered to be a mere variant
of -ā for the nominative feminine plural, since it is not confined to marking A.

(270) abkai vāṃ=rī āṃkhyāṃ mhārā=pai jamyoḍ-ī h-ī

now he=gen eye[f]nom.pl I.gen=loc fixed-f.pl be-iprf.f.pl
‘This time his eyes were fixed on me.’ (Hamsa 1984: 170)
 Differential marking 223

In conclusion, the case marking of pronouns and nouns functioning as A is not

ergative in Marwari because there is no systematic differentiation between the
case marking of A and S. There is only one restriction on this general conclusion:
the masculine plural third person pronoun vāṃ and some forms of masculine
plural nouns are rarely, if ever, used for S. In Marwari, although both the histori-
cally ergative and nominative form of the first and second personal pronoun are
still in use, it is only in the third person plural that a functional difference is still
made. Third person plural A’s are restricted to the oblique form.
At first sight, the case marking in Marwari seems to be in line with the cross-
linguistic generalization that nouns, and by extension demonstratives, tend to
be more marked than pronouns if they function as A: third person plural nouns
and pronouns are more marked than first/second pronouns (cf. Silverstein 1976;
Dixon 1994: 86, Chapter 2). However, as Filiminova (2005: 87) rightly remarks, if
pronouns tend to pattern accusatively and nouns ergatively, then the develop-
ment of case alignment in Marwari is rather unexpected. Following Magier (1983),
she points out that in Marwari, the ergative case of the pronouns has “extended”
to the case marking S, e.g., the old ergative form for the first person singular mhai
is in contemporary Marwari also used for S (and in Gujarati, Hindi, Marathi). Fil-
iminova further argues that, historically, the pronouns retained the distinction
between ergative and nominative over a longer period of time than did the nouns
(cf. Khokhlova 1995, 2001). Proof of this can still be found in modern Marwari: the
pronominal forms mhai and hum are still formally clearly distinct from each other,
whereas in the nominal paradigm, the difference between an oblique (formerly
ergative) and nominative form has almost completely disappeared. This stands
in contradiction with Silverstein’s prediction that the pronominal paradigm will
display a greater tendency towards an accusative pattern than the nominal para-
digm, given that, historically, the pronouns had different forms for marking A
and S. Filiminova (2005: 109) therefore concludes that the case marking of pro-
nouns is more “resistant to realignments and other changes than that of nouns”,
and that the influence of the person hierarchy on ergative case marking is, at the
least, modified by historical transitions and language-particular preferences.

6.2.2 Differential marking in Harauti

Harauti is a Rajasthani language related to Marwari but spoken on a smaller scale

in and around the twin cities Kota-Bundi (4,730,000 speakers according to the
2001 Census). In Harauti, as in the other Rajasthani languages, sentences with a
perfective finite verb form display a different form of alignment than present tense
sentences. Imperfective verb forms agree with the subject. The case marking is as
224 Central Indo-Aryan: Rajasthani

follows: A and S are in the unmarked nominative, and O is marked with the post-
position naiṃ depending on to its place on the referential hierarchy.
In constructions with a perfective verb form, the alignment is markedly dif-
ferent. First, agreement is with O whenever the verb is transitive. Secondly, A
receives an ergative case marking. The Marwari pattern without ergative case
marking is thus not followed in Harauti. Another peculiarity of Harauti lies in the
fact that the same postposition used to indicate a marked O is also the marker of A
in a perfective construction. There is a rule specifying that the postposition occurs
only once in the sentence, preferably marking A (Allen 1960: 10). Though occa-
sionally the double occurrence of the postposition naiṃ occurs, it is in general
considered to be ungrammatical by most Harauti speakers (cf. Stronski 2010: 85).
Grierson (IX, II: 203) attributes the use of naiṃ as ergative marker to the influence
of the neighboring language Bundeli, which displays the same pattern as Hindi.
The following examples illustrate naiṃ functioning as marker of A. In ex.
(271) the transitive verb is in a past tense, and A is constructed with naiṃ. The
verb agrees with O in gender and number.

(271) mhaiṃ=naiṃ hāṃk-ā pāṛ-yā ar pāchai bhī

I.f=erg scream[m]-pl scream-pst.m.pl and after also
bhāg-ī paṇ moṭar nhaṃ ḍhab-ī
run-pst.f.sg but bus[f] not stop-pst.f.sg
‘I screamed and ran after it, but the bus did not stop.’ (Prem 1984: 91)

In ex. (272), A is again marked with the postposition naiṃ. Both elements of the
present perfect construction, i.e., the past participle dekhī and the auxiliary chī,
agree with the f.sg. O jhalak, which is referred to by the relative pronoun jyo.

(272) ḍokarī=kī āṃkhyāṃ=meṃ yācanā=kī asī

old woman=gen eye[f]obl.pl=loc begging[f]=gen this
jhalak ch-ī jyo mhaṃ=naiṃ iṃ=sūṃ phalī
glimpse[f] be.pst-f.sg which I=naiṃ she.obl=from first
nhaṃ dekh-ī+ch-ī
not see-pst.f.sg+aux-pst.f.sg
‘In the eyes of the old woman there was a glimpse of begging which I had
not seen from her before.’ (Prem 1984: 92)

In contrast with Hindi and most other central IA languages, an IO in Harauti

is considerably more marked with a postposition other than naiṃ, the choice
depending on lexical criteria of the verb. According to Allen (1960: 10), naiṃ is
deliberately avoided as marker of the IO in perfective constructions because of
 Differential marking 225

the rule that the postposition naiṃ cannot be used more than once in a clause.
Hence, a definite/animate O is not marked in the same way as IO.
Nevertheless, naiṃ does appear as the marker of an experiencer. Below are
examples in which the postposition naiṃ marks the experiencer role. Experiencer
constructions are common in the IA languages (cf. Section 2.4.4). In Hindi and
Marwari, the experiencer is marked with the same postposition used for O and
IO. Conversely, in Harauti, the experiencer is marked with the same postposi-
tion as an ergative A. The identical marking of the experiencer and an ergative A
cannot possibly conflict or lead to ambiguity in any of the examples because in
an experiencer construction, there is no transitive subject. Ex. (273) is a sentence
constructed with the verb lāg ‘to attach’, a typical experiencer verb. Note that the
stimulus role ḍokarī is in the unmarked nominative case.

(273) ḍokarī mhaṃ=naiṃ asī lāg-ī jasyāṃ sāṃcyāīṃ

old woman[f] I=obj such attach-pst.f.sg such just
ūṃ=kai tāīṃ khajāno mala+g-yo+ch-o
she.obl=gen that treasure[m] find+go-pst.m.sg+aux-pst.m.sg
‘The woman seemed to us as if that treasure of hers had just been found.’
(Prem 1984: 94)

The recipient function of IO with verbs like ‘to bring to’ or ‘to say to,’ however, is
generally expressed with other postpositions. For instance, in ex. (274), the abla-
tive postposition sūṃ, and in ex. (275), kai ‘for’ is used.⁸

(274) tīn-cyār jaṇāṃ=naiṃ klīnar=sūṃ bāt kah+dī

some people=erg cleaner=to thing[f] say+aux.pst.f.sg
‘Some people said to the cleaner.’

(275) potā-potyāṃ=kai beī āmbā

grandchild[m]obl.pl=for this.pl mango[f]pl
lā-ī+ch-ī
bring-pst.f.pl+aux-pst.f.pl
‘I had brought those mangoes for my grandchildren.’ (Prem 1984: 92)

8 In Hindi, the IO of verbs meaning ‘to say’ often occur with the ablative postposition se, but
the objective ko is also possible, in particular with the verb kahnā. In the other constructions as
well, one finds a variety in Hindi on ko; for instance, in a construction such as ex. (275) in Hindi,
the postposition ke lie can be used instead of ko, and in ex. (276), the locative postposition par
can be used.
226 Central Indo-Aryan: Rajasthani

Another example of a ditransitive construction is with the prototypical verb ‘to

give’. Ex. (276) is an idiomatic expression meaning ‘to give an order’, and naiṃ is
used as the marker of A. The IO takes a different morpheme, pai, a locative post-
position with the meaning ‘on’.

(276) ek mard=naiṃ ḍalevar=pai hukam calā+dyo,

one man=erg driver=on order[m] go+give.pst.m.sg
ḍalevar sāb!
driver sir
‘A man ordered the driver: driver sir!’ (Prem 1984: 92)

In the following example from Grierson (1903 IX, II: 211), the IO mhāiṃ is even
unmarked.

(277) thā=ne mhāiṃ ek uraṇo bhī nhī di-yo

you=erg I one young goat[m] also not give-pst.m.sg
‘You have never given me a young goat.’

In sum, naiṃ is also the postposition of the experiencer, whereas in typical ditran-
sitive constructions such as with the verbs ‘to bring’, ‘to say’ and ‘to give’, the IO
is predominantly marked with a different postposition.⁹
O is the third possible argument for which the multifunctional postposition
is used, but the occurrence of naiṃ is dependent on the features that determine
DOM. When O is human, it almost always takes naiṃ. When it is inanimate, it
is only marked when there is some kind of definiteness associated with it (cf.
Masica 1991: 365). naiṃ is only used once in a clause, so O is not marked with the
postposition when A is present in the sentence (cf. Allen 1960). In ex. (278), there
is no overt subject, the main verb pūchajyo being an imperative. The inanimate
but definite O moṭar of ruk- ‘to stop’ takes naiṃ. Thus, in this sentence, naiṃ is
used to indicate an argument in the sentence, but not to indicate A. Ex. (279) is a
similar example with an animate O.

9 If the conditions are right, i.e., if there is no argument marked with naiṃ in the sentence, it
is possible to find a naiṃ-marked IO, for instance, the following example from Grierson (1903,
IX, II: 203):
koi ūṃ=ne kāīṃ nhai de-to
someone he.obl=obj something not give-iprf.m.sg
‘No one used to give anything to him.’
 Differential marking 227

(278) āva-tī moṭar=naiṃ rukā-’r pūcha+j-yo ke ek thelo

driving-f.sg bus[f]=naiṃ stop-cvb ask+go-imp.2sg that one bag
to nhaṃ mal-yo
then not be found-pst.m.pl
‘After stopping the driving bus, ask once if there hasn’t been found a bag.’
(Prem 1984: 92)

(279) dekh-o īṃ ḍokarī=naiṃ!

see-imp.2sg this old woman[f]=naiṃ
‘Look at this old woman!’ (Prem 1984: 92)

Ex. (280) is in the present tense. A is the interrogative pronoun ‘who’. The O
marked by naiṃ is inanimate but definite and determined by a genitive phrase.

(280) khābā=kī cīj=naiṃ kuṇ choḍ-ai+ch-ai?

eat=gen thing=naiṃ who.nom.m.sg lose-prs+aux.prs.3sg
‘Who loses a thing to eat?’ (Prem 1984: 92)

From the above examples it is clear that compared to the other Central Indo-
Aryan languages, the multifunctionality of the postposition naiṃ is raised to
extreme levels in Harauti. Because the postposition shows up as a marker of A,
experiencer, both typically animate roles, and animate/definite O, it seems safe
to say that naiṃ is associated with a high level of animacy and definiteness (pos-
sibly extended to humanness/specificity as in Hindi, see Section 2.2). This is in
line with the idea, which is cross-linguistically well-established, that differential
object marking either hinges on the animacy or on the definiteness of O (Malchu-
kov 2008; Klein and de Swart 2011, Chapter 1). On the other hand, it is cross-lin-
guistically unusual that the same postposition used in DOM is also able to mark
an ergative A.
The finding that the postposition naiṃ in Harauti is to a certain extent multi-
functional does not necessarily contradict the principles of DOM. In Harauti, an
A in combination with a perfective construction and an experiencer argument
are ranked highest on the scale of animacy/definiteness, because the agent and
the experiencer are semantic roles which are typically either animate or defi-
nite. The A and the experiencer are always marked. Animacy and definiteness
are, however, not the only factors that determine marking in Harauti, because
these parameters cannot explain why in a perfective construction, the ergative A
is marked, and the animate O is not. Even inanimate A’s are marked, cf. ex. (281):
228 Central Indo-Aryan: Rajasthani

(281) moṭar=naiṃ pāṃ-pāṃ=kī āvāj kar-ī ar

bus=naiṃ paw-paw=gen sound[f] make-pst.f.sg and
ek jhaṭako de’r ṭhaharag-ī
one shake[m] give.cvb stop-pst.f.sg
‘The bus made the sound of paw-paw and after giving a shake it stopped.’
(Prem 1984: 93)

Consequently, the marking with naiṃ cannot only be due to the animacy/definite-
ness of the marked argument. It is therefore reasonable to assume that other fea-
tures such as discourse salience and topicality are also involved in the marking
pattern of Harauti. Although these latter kinds of properties are only rarely men-
tioned in the literature on DOM, it seems plausible that the level of salience of
an argument is inherently related to its level of animacy/definiteness. Topical-
ity determines as well the salience of an argument. The topicalized argument is
the argument about which something is being said (cf. Dik 1989: 266), and pre-
dominantly concurs with the A-argument. Therefore, a topicalized argument is
necessarily also a salient argument in the clause. Topicality constitutes a third
factor, besides animacy and definiteness, that can control differential marking
(cf. Bossong 2006).
Definiteness and animacy are two criteria, topicality being another one,
that contribute to a high level of salience (cf. Croft 1988: 165). Animate and
definite arguments tend to be more salient in discourse than inanimate and
indefinite arguments (cf. Hopper and Thompson 1980; Heine and König 2010:
94). In Harauti, it is true that O, for instance, is only marked when it is definite
and animate, i.e., when it is a discourse-prominent argument. On this view, the
main semantic purpose of the naiṃ-marking is to characterize the argument as
the main discursive argument of the sentence, and, as such, to highlight it vis-
à-vis the other arguments. If the argument is only animate or only definite, the
marking with naiṃ is not certain. This explains why the IO of typical ditransitive
verbs such as ‘to give’ or ‘to tell’ does not generally take naiṃ, given that the IO
in Harauti is not the most prominent argument in a sentence in which this role is
assumed by A. Accordingly, the IO is not marked by the postposition for discourse
prominence, but by another “adjunctive” kind of postposition. Still, in the rather
exceptional situations where the IO has a great discursive importance, it may be
marked with naiṃ. In ex. (282) for instance, the animate IO parār ‘another’ takes
naiṃ. Although this is unusual, it is the emphasized argument of the sentence: of
all people, the buffalos are sold to an utter stranger. The prominence of the IO is
moreover stressed by its position at the beginning of the sentence:
 Differential marking 229

(282) parār=naiṃ syāḷā=meṃ mhūṃ bhaiṃsy-āṃ bec’r pācho

other=naiṃ cold season=in I buffalo-nom.pl sell.cvb back
jā+r-yo+ch-o
go+prog-pst.m.pl+aux-pst.m.pl
‘It was to a stranger that I had first sold my buffalos in the cold season, and
then I was going back.’ (Prem 1984: 93)

The opposite example also shows up; in ex. (283), mhaiṃ, while being A of a
perfective verb, does not take the naiṃ postposition. It appears that neither the
syntactic obligatoriness of marking A nor the animacy of A triggers a marking.
Because the emphasis is on the action, and in particular on the hands perform-
ing the action, mhaiṃ remains unmarked. Note that there is no other argument
that qualifies to take naiṃ in ex. (283). The O poṭalī ‘luggage’ could technically be
marked with naiṃ if it were definite, but its non-specific nature renders marking
improbable.

(283) adha-khulī phāṭak=ko kuṃdo ḍāvā hath=sūṃ pakaṛ’ r mhaiṃ

half-open door=gen part[m] left hand=abl take.cvb I
jīvaṇā hath=sūṃ poṭalī bhītar phāṃk +dī
right hand=abl luggage[f] inside throw +give.pst.f.sg
‘While taking a part of the half open door with my left hand, I threw my
luggage inside with my right hand.’ (Prem 1984: 91)

In sum, the distribution of the postposition naiṃ in Harauti poses a problem,

because naiṃ cannot be considered as a structural case marker in this language.
Though the postposition occurs predominantly marking the A-argument of a per-
fective construction, it can also mark an O, IO and experiencer. Its occurrence
seems to be determined by semantic criteria of definiteness, animacy and, to a
certain extent, topicality. Though naiṃ in Harauti still has the function of dis-
tinguishing the A and O-arguments, by occurring only once in a sentence, this
seems to be only a secondary function – its primary function being the semantic
indexing of the arguments.¹⁰

10 Note that in certain Bhili languages (Rajasthani tribal languages), naiṃ can occur more
than once in a clause, marking every animate/definite/topical argument (Max Phillips, p.c.). In
these languages, naiṃ has completely lost its distinguishing function of case marker.
230 Central Indo-Aryan: Rajasthani

6.2.3 Object marking in Marwari and Gujarati

In contrast to the multifunctionality of the postposition naiṃ in Harauti, the

pattern of DOM in Marwari is of a more “traditional” kind. Definiteness and
animacy determine the differential object marking in Marwari (cf. Masica 1991:
365). O is obligatorily naiṃ-marked when it is animate and definite, e.g., the
proper name Śaraṇ in ex. (284).

(284) mhaiṃ śaraṇ=naiṃ dekh-ī

I Sharan.f.sg=obj see-pst.f.sg
‘I saw Sharan.’ (Hamsa 1984: 172)

When the O is inanimate but definite, it is optionally naiṃ-marked. In ex. (285),

the O is construed with naiṃ, and the postposition excludes any possibility of
interpreting this O as indefinite.

(285) mhaiṃ khāmosī=rī sīṃv=naiṃ pār

I silence=gen limit[m]sg=obj pass
kar+g-yo+h-o
do+go-pst.m.sg+aux-pst.m.sg
‘I passed the limit of silence.’ (Hamsa 1984: 172)

The O is never marked when it is inanimate and/or indefinite. On the other

hand, the IO in Marwari is marked with naiṃ on a regular basis; as in Hindi (see
Section 2.2, 1.1.3.2).

(286) bābū mha=naiṃ baiṭh+jāv-ṇai=ro isāro kar-yo

boss I=obj sit+go-inf=gen sign[m] make-pst.m.sg
‘The boss made me a sign to sit down.’ (Hamsa 1984: 170)

It has been argued above for Harauti that its multifunctional postposition naiṃ
functions primarily on a semantic basis of marking highly animate, definite or
topical arguments. The postposition naiṃ “indexes” the argument it marks for
these semantic features. On the other hand, the agreement pattern in Marwari
seems to function as a distinguishing mechanism. The agreement pattern is dif-
ferent from Hindi, where there is no agreement when O is marked (cf. Chapter 2).
In Marwari, there is always agreement with O (cf. ex. (284)). Thus, the occurrence
of naiṃ as a marker of a noun has no influence on the agreement of the verb; naiṃ
in Marwari does not function as an indicator of a relationship between the noun
and the verb, like typical case markers do (cf. Blake 1994: 1) and like ko does in
 Differential marking 231

Hindi. In fact, naiṃ in Marwari only indicates the definiteness or animacy of O/

IO. Therefore, like in Harauti, it is primarily a semantic marker. The fact that naiṃ
can mark O as well as IO also points to the conclusion that it is not a distinguish-
ing “accusative” case marker of which the only function is to distinguish the O of
the A-argument.¹¹
It is well known that Rajasthani and Gujarati share a common ancestry
through the ancestral “Old Western Rajasthani” (cf. Tessitori 1914). It is therefore
interesting to take a look at the Gujarati DOM. In his study of verb agreement
and object marking in Gujarati, Mistry (1998) proposes an approach in which he
considers two Gujarati postpositions as homophonous. For agreement, Mistry
(1998) observes that the verb in Gujarati agrees with the unmarked and marked
O; this is similar to Marwari and different from Hindi. In addition, he argues that
this agreement pattern forms a problem for any theoretical approach which sug-
gests that verb agreement is restricted to the nominative (unmarked) case. With
regard to Hindi, this approach has been advocated by Mohanan (1994: 105) who
argued that there is a constraint in Hindi which says that the “verb agrees with
the highest ARG [argument] associated with NOM [nominative] case”. The follow-
ing examples from Hindi illustrate this constraint:

(287) maiṃ laṛkī=ko dekh-tī+h-ūṃ

I girl[f]=obj see-prs.f+aux.prs-1sg
‘I see the girl.’

(288) tum=ne saikil bec-ī+h-ai

you=erg cycle[f] sell-pst.f+aux.prs-3sg
‘You’ve sold cycles.’

In both these examples, it is the nominative which controls agreement, although

maiṃ is the subject in ex. (287) and saikil is O in ex. (288). If there is no nomi-
native argument at all, the verb takes the “default” form, i.e., 3rd person m.sg.
Mohanan’s argument draws on Saxena (1981) who also considered agreement in
Hindi as a purely formal matter, i.e., as being exclusively restricted to unmarked
arguments. Saxena was however criticized by Comrie (1984). Comrie argued,
convincingly, that one also needed to look at the grammatical relations to evalu-
ate the agreement pattern in Hindi. He based this argument on the agreement in
Indo-Aryan languages, in particular on Gujarati and certain dialects of Marathi.

11 This is also reflected in my labeling of this postposition as “objective” instead of using the
more common terms “accusative” or “dative”.
232 Central Indo-Aryan: Rajasthani

Indeed, Gujarati (and Marwari) displays a different verb agreement, because the
verb agrees with the marked O (cf. Mistry 1998: 429). To solve this contradiction
between verb agreement in Hindi, which is apparently well explained by Mohan-
an’s constraint, and the “irregular” agreement in Gujarati, Mistry argues that the
Gujarati postpositional marker of the O, ne, is not a case marker, but a “specific-
ity” marker that functions on semantic grounds. As in Marwari, this marker of O
has no bearing on the agreement pattern; that is, it is not able to restrict or block
agreement (Mistry 1998: 433–437). Compare the following example from Mistry
(1998: 435) in which the perfective verb baraḍyā agrees in gender and number
with the masculine plural O mag=ne:¹²

(289) kiśor-e mag=ne baraḍ-yā

Kishor-erg mung beans[m]=obj grind-pst.m.pl
‘Kishor ground mung beans.’

Thus far, Mistry’s approach is essentially the same as the one supported here for
Marwari: naiṃ in Marwari is considered to be a marker of animacy/definiteness
instead of a structural accusative case marker. However, Mistry notes that some
Gujarati verbs take objects obligatorily marked with ne (NE in his article), and
they agree by default (i.e., they do not agree with any argument in the construc-
tions, but take the neuter singular form). Examples of these verbs include maḷ- ‘to
meet’, vaḍh- ‘to rebuke’, karaḍ- ‘to bite’, coṭ- ‘to cling’, bheṭ- ‘to embrace’, cum- ‘to
kiss’, nam- ‘to bow’, and paraṇ- ‘to marry’ (Mistry 1998: 429). The examples that
Mistry (1998: 428) offers, are all constructions of obligation (cf. ex. (290)), and
in the only example in which he uses a perfective verb, the verb turns out to be
exceptional in that it does not take an ergative-marked A and rather seems to be
intransitive¹³ (cf. ex. (291)).

12 Note that Gujarati has retained three grammatical genders, i.e. masculine, feminine and
neuter, and that the ergative is expressed by the inflectional oblique case -e.
13 Note that the verb law ‘to bring’ in Gujarati also does not take an ergative marked A in the
perfective, similar to Hindi lānā ‘to bring’ (cf. Mistry 1998: 247, Chapter 2). As in Hindi, it is dif-
ficult to consider law as an intransitive verb, or as an experiencer verb. However, it is interest-
ing that there seem to be more verbs in Gujarati than in Hindi that pattern intransitive on the
syntactic level although they are transitive on the semantic level. Moreover, the use of ne as the
marker of the experiencer in a construction of obligation appears to be restricted to Northern
Gujarati. In Southern Gujarati, the ergative suffix -e is used instead (cf. Mistry 1976: 267).
 Differential marking 233

(290) kiśor=ne kāgaḷ=ne aḍ-v-uṃ+ha-t-uṃ

Kishor=obj letter[m]=obj touch-des-n.sg+aux-pst-n.sg
‘Kishor wished to touch the letter.’

(291) kiśor kāgaḷ=ne aḍ-y-o

Kishor letter[m]=obj touch-pst-m.sg
‘Kishor touched the letter.’

Mistry argues that in constructions with these particular verbs (of which aḍ, ‘to
touch’, is an example in (290) and (291)), NE functions as a case marker, i.e., as
marker of O. This “NE” is homophonous with the aforementioned ne postposi-
tion. According to Mistry, NE is different from ne because the latter one serves
as a specificity marker and does not prevent agreement. In other words, Mistry
distinguishes the case marker NE from the “specificity” marker ne, on the basis of
the argument that NE can prevent agreement, whereas ne cannot.
A number of objections can be raised against Mistry’s analysis. First, the
above-mentioned verbs do not take an O as Mistry assumes (and as their transla-
tions do in English). On the contrary, they are verbs that take the semantic role of
recipients.¹⁴ The argument marked by ne in these constructions is considered to
be an IO, not an O. The obligatoriness of NE indicates this, as IOs and experiencer
arguments (i.e., non-O’s) are always marked in Marwari and Gujarati. Moreover,
the use of cognate verbs in Hindi points in the same direction. For instance, the
Hindi verb milnā ‘to meet’ is not transitive in the ordinary meaning of the term,
and in expressions such as ex. (292) (Montaut 2004: 195), the agent-experiencer
ham takes the postposition ko; obviously, the verb does not agree with this
marked argument, but with the patient-stimulus.

(292) hamār-ī cāzī ham=ko mil+ga-ī+h-ai

we.gen-f Chazi[f] we=obj get+go-pst.f.sg+aux-prs.3sg
‘We have found (gotten back) our Chazi.’

The proposed homophony between the case marker NE and the specificity marker
ne is in fact reducible to another instance of the multifunctionality of the object
marker in the IA languages: it marks O under certain semantic conditions, the IO
according to an extension of DOM and experiencer arguments (cf. Kittilä 2006).

14 More literal, though perhaps grammatically incorrect, English translations of the afore-
mentioned verbs would therefore be ‘to join with’, ‘to rebuke to’, ‘to bite in’, ‘to cling to’, ‘to
embrace with’, ‘to kiss with’, ‘to bow to’, and ‘to marry with’, respectively.
234 Central Indo-Aryan: Rajasthani

The verb in Gujarati and Marwari always agrees with O, whether it is marked or
not. However, agreement with an IO or an experiencer, marked with the same
postposition, is out of the question.¹⁵
In sum, whereas Marwari naiṃ and Gujarati ne fulfill the “indexing” role
of case markers, being primarily semantic markers, the agreement in these
languages has the role of “distinguishing” the O- and A-arguments by consist-
ently marking O. It is the verb agreement in Marwari and Gujarati that fulfils the
argument-distinguishing and relational function that is attributed to ko in Hindi
(cf. Magier 1990: 213, but see Spencer 2005). From the point of view of iconic-
ity, it is interesting to observe that ko is also formally very distinct from ne/naiṃ.
Although ko is also a postposition that indicates a definite/animate O, it does not
allow agreement. On the other hand, compared to the Rajasthani languages and
Gujarati, the ergative postposition ne in Hindi is reduced in multifunctionality,
because the use of ne restricted to the specific syntactic functions of marking A
and does not extend to any other argument.

6.2.4 The marking of subjects and experiencers in Gujarati and Punjabi

The postposition ne shows up for different functions in both Gujarati and Punjabi.
Another alignment feature that these two languages have in common is that SAP
pronouns show different marking from third person pronouns and nouns.
In contrast with Marwari, Gujarati does mark A with a specific marker in per-
fective constructions. This marker is -e, an oblique form, which is also used to
mark the locative. The marker -e can also be added to an attributive adjective that
modifies a nominal form marked by -e, which shows that it is a true inflectional
form, not a postposition (cf. Cardona 1965: 62, 70). In Gujarati, ne is the objective
postposition, which is added only to a (pro)nominal form. The conjugation of
the pronouns is given in Table 6.3. First and second person singular pronouns
take a separate ergative form when they are used as A, but they do not have the
typical -e ending of the nominals; for their plurals, the ergative has merged with
the nominative, giving one form that ends in -e for both A and S.

15 Experiencer verbs in Hindi and Marwari generally agree with the stimulus role, not with the
experiencer, which is marked, respectively, with ko or naiṃ. Agreement between the experi-
encer verb and the experiencer is possible outside the IA languages, as shown by Bickel (2003)
in a study of the Tibeto-Burman language Bilhari. The IA languages included in Bickel’s study,
Maithili and Nepali, do not display experiencer agreement and therefore follow the general IA
pattern (cf. Chapter 3–4, Section 2.4.3).
 Differential marking 235

sg 1 2 3 pl 1 2 3

nom huṃ tuṃ e apṇe ta(h)me e

erg meṃ teṃ e(h)ṇe apṇe ta(h)me e(h)mṇe
obj ma=ne ta=ne e(h)=ne apṇ=ne ta(h)m=ne e(h)m=ne

Table 6.3 Personal pronouns in Gujarati (cf. Cardona 1965: 92–93)¹⁶

In Punjabi, ne is the postposition that marks the A of a verb in the perfective.

Whereas in Gujarati, the first and second person plural forms do not have a sep-
arate form for A and S, in Punjabi none of the SAP pronouns have a different
marking for A.¹⁷

sg 1 2 3 pl 1 2 3

nom mai tū é asī tusī é

erg mai tū é=ne asī tusī énā=ne
obj mai=nū tai=nū é=nū sā=nū tuā=nū énā=nū

Table 6.4 Personal pronouns in Punjabi (cf. Bhatia 1993: 229, the diacritics are simplified)

First and second person pronouns are unmarked for A and do not take the ne
postposition. Verb agreement, however, is with O, not with the unmarked A.
Compare the following examples from Bhatia (1993: 86).

(293) maiṃ kamm kītā

I work[m] do.pst.m.sg
‘I worked.’, lit.: ‘I did the work.’

(294) kuṛī=ne kamm kītā

girl[f]=erg work[m] do.pst.m.sg
‘The girl worked.’, lit.: ‘The girl did the work.’

16 Certain forms, such as the exclusive and formal plural and the remote third person pronoun,
are not included in this and the following tables for reasons of space. These forms are all
declined in a similar way.
17 Punjabi has the same alignment pattern as Marathi, except for the use of pronominal suf-
fixes (cf. Chapter 6). The Marathi ergative marker ne becomes nī if the argument is plural (cf.
Ashwini and Deo 2006: 379).
236 Central Indo-Aryan: Rajasthani

Note that variation in the ergative marking exists. Bhatia (1993: xxxi), for instance,
observes that the different marking for first/second versus third person occurs
only in the Punjabi spoken around Powadhi, Ambala. According to Bhatia, the
Punjabi spoken around Majhi and Amritsar always marks A with ne, including
SAP pronouns functioning as A. Other variants of Punjabi follow yet another
strategy by not marking A at all.
Case marking in Punjabi seems to be a classic example of a “split conditioned
by the semantic nature of NPs” (cf. Dixon 1994: 83, Chapter 2). Unlike Marwari,
there is a clear difference in Punjabi between first/second pronominal arguments
with accusative case marking and third person pronouns and nouns with erga-
tive marking. However, it is important to note that with respect to agreement, all
arguments (pronominal and nominal arguments alike) pattern in the same way.
There is ergative agreement with S/ O in a perfective verb construction. Hence, the
difference in alignment between pronouns and nouns is more a historical devel-
opment than a fundamental difference of pronouns “preferring” a different align-
ment pattern than nouns (cf. Filiminova 2005, Section 6.2.1).
Pronominal case marking in Gujarati is highly irregular. All singular pro-
nouns and the third person plural pronoun display a distinction between ergative
and nominative case. Only the first and second person plural have nominative
case for both A and S. The opposition between the first/second person plural and
the third person plural is consonant with the idea that nouns and third person
pronouns pattern more as ergative than first/second pronouns. The question still
remains as to why this split is absent in the singular paradigm. Typologically,
splits based on number are diverse, and according to Dixon (1994) the split can
be in either direction because (i) number is a category that can be expressed by
nouns as well as pronouns;¹⁸ (ii) number alone cannot determine whether an
argument is more likely to function as A than as O. Whether singular or plural
pronouns pattern as ergatives is thus merely a matter of preference in a particu-
lar language. However, in splits, “there are likely to be more distinctions in the
singular than in the plural” for argument marking (Dixon 1994: 92), and this is
where Gujarati differs, since we find an opposition in first/second versus third
in the plural, but not in the singular. Arguably, historical developments in the
morphological paradigm have led to this pattern in Gujarati.

18 According to Dixon (1994: 92–94), number contrasts with person, which is more associated
with pronouns (because only pronouns distinguish between first, second, and third person).
However, number also contrasts with gender, which is more associated with nouns, because
nouns are typically distinguished for being masculine, feminine, and neuter (in contrast to first/
second pronouns, which, in many languages, have only one form for all genders).
 Differential marking 237

For other alignment features, Punjabi is in general closer to Hindi than to the Raj-
asthani languages. The Punjabi postposition for O and IO is nū(ṃ). Only a definite
or animate O is marked, and agreement in Punjabi is with the subject, except in
perfective transitive verb constructions, where there is an unmarked O. The expe-
riencer in a construction of obligation in Punjabi always takes the objective form,
as muṇḍe=nūṃ in the following example from Bhatia (1993: 37).

(295) muṇḍe=nūṃ katāb paṛ-nī pav-egī

boy=obj book[f] read-inf.f compel-fut.f.3sg
‘The boy will have to read a book.’

It has been reported that in Hindi, these constructions also occur with the post-
position ne, in particular in the Urdu spoken around Lahore and Delhi (cf. Butt
and King 1991, 2004: 6; Bashir 1999). Butt and King (2004: 6) argue that a subject
marked with ne “is interpreted as wanting to perform the action”, whereas if it is
marked with ko (the usual construction), it “must perform the action”. They give
the following examples to illustrate this.

(296) nadya=ne zu jā-nā h-ai

N=erg zoo go-inf be.prs-3sg
‘Nadya wants to go to the zoo.’

(297) nadya=ko zu jā-nā h-ai

N=obj zoo go-inf be.prs-3sg
‘Nadya must go to the zoo.’

In standard Hindi/Urdu, this particular use of ne is unusual. It is possible that

it is an extension of the ergative use of ne, which occurred under the influence
of Punjabi, where ko is non-existent. However, as pointed out by Masica (1990:
335), when it marks the experiencer of a construction of obligation, ne is used in
Punjabi to express only “weak compulsion”. Moreover, ne never occurs with first/
second pronouns; compare the following Punjabi examples taken from Masica
(1990: 335):

(298) muṇḍe=ne jā-ṇā ai

boy=erg go-inf be.prs.3sg
‘The boy has to go.’ (vs.: the boy should/ought to go)
238 Central Indo-Aryan: Rajasthani

(299) maiṃ jā-ṇā ai

I go-inf be.prs.3sg
‘I have to go.’

In a construction expressing “strong compulsion”, the objective postposition

nūṃ marks the experiencer. Although Masica’s intention is to argue against the
borrowing of ne for constructions of obligation from Punjabi to Hindi, his exam-
ples actually strengthen Butt’s view that in Hindi and Punjabi, objective postposi-
tions are used for strong obligations and that the ergative postposition denotes a
lesser degree of compulsion.¹⁹
In Marathi, the postposition of the ergative, ne (ne is the singular form, nī the
plural form), assumes a number of additional functions that go beyond marking
the ergative. Ne is not only the marker of the nominal A of a perfective construc-
tion but also the instrumental marker of inanimate agents in constructions with
intransitive verbs (cf. ex. (300) from Wali 2005: 45). Moreover, ne is also used –
and much more frequently than in Punjabi and Hindi – in combination with a
construction of obligation, which can be expressed by means of a subjunctive in
Marathi (cf. ex. (301) from Wali 2005: 46). This latter construction can either be
transitive or intransitive. However, the subjunctive verb never agrees with the ne-
marked argument (Pandharipande 2003: 711). If the agent of such a construction
is in the nominative case, then the meaning is optative rather than obligative (cf.
Wali 2004; Montaut 2009b).

(300) varya=ne arsa phuṭ-l-a

wind=ins mirror be broken-pst-3m.sg
‘The mirror was broken due to the wind.’

(301) lili=ne dhav-av-e

Lili=erg run-sbjv-3n.sg
‘Lili should run.’

However, just as in Hindi and Punjabi, an infinitive construction combined with

an objective-marked argument can also express an obligation in Marathi, as in
the following example from Pandharipande (1990: 163):

19 In Gujarati, the objective is used to mark experiencers of modal constructions with ‘want’
and ‘must’ verbs. It may be replaced by a “pronominal adjective” (a genitive form of the pro-
noun) in constructions of obligation without a modal auxiliary (Cardona 1965: 95).
 Differential marking 239

(302) tyā=lā śaḷe=lā zāyca āhe

he=obj school=loc go.inf be.prs.3sg
‘He has to go to school.’

Again, the pattern with the ergative postposition and the subjunctive expresses a
less strong compulsion than the infinitival construction with an objective-marked
experiencer argument.
Finally, the pattern found in Hindi, Punjabi, and Marathi in which the erga-
tive postposition marks the experiencer of a construction of obligation is far from
exceptional in Indo-Aryan. For instance, it also occurs in Nepali and the Pahari
languages Garhwali and Kumaoni (cf. Section 4.1.2, Montaut 2009b: 311).

6.2.5 Differential marking: summary

The extension of ne in Hindi to modal constructions of obligation and permis-

sion once more confirms the multifunctionality of the postposition ne in the IA
languages. The main function of ne in the IA languages may be best described
as ‘marking a core argument’. Whether this core argument functions as A or O is
subject to language-specific variation, which occurs to a considerable extent. If
we take into account that most of the speakers of IA languages are multilingual,
speaking more than one IA language, historical changes due to language contact
are more than likely.
Yet, in the central Indo-Aryan languages discussed in Section 6.2 – namely,
Hindi, Marwari, Harauti, Gujarati, and Punjabi– there is one recurring trend
with regard to postpositional marking. The postposition with the form ne/naiṃ
marks an argument high on the scale of definiteness/animacy, or in other words,
an argument that is highly “discourse prominent”. For Hindi and Harauti, this
means that ne marks A. In Hindi, ne contrasts with the objective postposition
ko, which marks the other core arguments (i.e., a definite/animate O, IO, and the
experiencer argument). In Marwari, ne does not mark ergative A. In this language,
the formal distinction between ergative and nominative has largely disappeared.
Gujarati is similar, but still has an oblique marking of A in certain paradigms. In
Punjabi, the same merge of the ergative and nominative form has taken place in
the pronominal paradigm, but ne is the ergative marker of third person pronouns
and nouns. As in Hindi, Punjabi has an extra postposition, nūṃ, which marks O,
IO, and the experiencer. The use of this postposition seems to result in its argu-
ment not controlling verb agreement.
The postposition ne does occur in Marwari and Gujarati as a semantic marker
of O without influencing verb agreement. It also occurs as a marker of IO and
240 Central Indo-Aryan: Rajasthani

the experiencer. Eventually, in Harauti, ne can mark any prominent argument,

although there are certain restrictions on its occurrence. Whereas A marked with
ne in Harauti does not control agreement, an O marked with ne can control agree-
ment if the clause is perfective. Except in Hindi, Punjabi and Marathi (Section 5.3),
then, it looks as though verb agreement is not influenced in any way by case
marking. In all of the languages, there are the following two general principles:
(1) the A of a transitive perfective verb never controls agreement and (2) agree-
ment is controlled by O. The exception is that in Hindi, Punjabi and Marathi, a
marked O cannot control agreement.
Hindi, with its strict use of ne for A and ko for O, and with the restriction on
agreement with any overtly marked argument, seems to be exceptional among
the aforementioned languages. As will become clear in Section 6.3 below, this
conclusion is confirmed when we look at the alignment types found in the dia-
lects of Hindi, which pattern much less consistently than the alignment patterns
in Hindi itself.

6.3 Central Indo-Aryan and Hindi: variation on a micro-scale

6.3.1 Hindi and Braj: O-marking and agreement

This degree of standardization in Hindi is best illustrated with a comparison

between Hindi and a variant of Hindi, such as Braj. Braj is a language spoken
around the city of Agra in the state Uttar Pradesh, where Hindi is the main lan-
guage. Braj used to be a language with a flourishing literary tradition in the 14th-
16th centuries but is now merely a spoken “dialect” of Hindi. The alignment in Braj
is, contrary to what one might expect, not the same as in Hindi. Some tendencies
seem to be more closely related to Rajasthani; others are typical of Braj only.
In Braj, ergative case is expressed with the postposition naiṃ (cf. Varma 1935:
96). In its distribution, the use of naiṃ is similar to that of ne in Hindi. In con-
trast to the imperfective, A in the perfective construction takes the ergative suffix
naiṃ/ne. S remains in the nominative case. There are two cases with a layer I form
in Braj; the unmarked nominative case and the oblique. The pronouns take an
oblique form before the postposition naiṃ. However, as in Hindi, some pronomi-
nal forms display a different oblique before naiṃ than before other postpositions
(e.g., taiṃ=naiṃ; haman=naiṃ). The first person singular maiṃ remains in the
nominative before the postposition naiṃ (cf. Masica 1991: 252). The conjugation
of the pronouns in Braj is shown in Table 6.5.
 Central Indo-Aryan and Hindi: variation on a micro-scale 241

nom.sg nom.pl obl.sg obl.pl

1 maiṃ, meṃ, hauṃ, hoṃ, hūṃ ham mo (maiṃ=naiṃ) ham (haman=naiṃ)

2 tū, tūṃ tum to (taiṃ=naiṃ) tum
3 bau, bu, bo; bā (f) be, bai, we bā, wā, gwā un, bin, win, gwanu

Table 6.5 Pronominal paradigm Braj

In modern Braj, a second form of the first person singular is still found: hauṃ.
Hauṃ is only used for S of an intransitive verb or for the subject of an imperfec-
tive verb. It never occurs in combination with the postposition naiṃ. In older Braj,
there was a clear distinction between the two forms of the first person pronoun
maiṃ and hauṃ. The latter was used strictly for a nominative subject, whereas
maiṃ was used as the ergative subject without a postposition, as is exemplified
in ex. (303), from a 17th century Braj text (Snell 1991: 9).

(303) tāte ā-i śaraṇa maiṃ gah-ī

then come-cvb refuge[f] I.obl grasp-pst.f.sg
‘So I came and took refuge.’

The two ways of forming a first person in Braj (mai- and hu-) recalls the situa-
tion in Marwari (and Gujarati), where there is also a varying use of an ai-pro-
noun and an u-pronoun, in transitive and intransitive sentences (cf. Khokhlova
1992). In Marwari, the pronominal forms (m)hūṃ and maiṃ are used for the same
argument, regardless of whether the verb is transitive or intransitive. Marwari (a
Rajasthani language) is different from Braj in that there is no postposition for A.
As Braj is now spoken in an area where there is a strong Hindi influence in
the media and education, the language is more influenced by Hindi than earlier
forms of the language were. Despite the fact that the ergative postposition naiṃ
is used consistently in Braj, the language is positioned somewhere between
Marwari and Hindi with respect to pronominal case marking. For instance, the
existence of two morphologically different forms of the first person pronoun is
unlike Hindi, but similar to Marwari. In modern Braj, maiṃ can be followed by
an ergative postposition, and hauṃ is used in all other instances (Verma 1935). In
contrast to Marwari, where both forms are used without functional differences,
the occurrence of maiṃ and hauṃ in Braj is determined by grammatical relations:
maiṃ is associated with an ergative A, hauṃ is preferred for S and A is used in
an imperfective construction. Furthermore, the use of maiṃ and hauṃ is partly
subject to which dialectal variant of Braj is used. Varma (1935: 70) states the fol-
lowing: “Dans l’ensemble du pays où les formes en h- sont employées, on trouve
aussi maẽ, mais là où maẽ est normal, les formes en h- manquent absolument.” It
242 Central Indo-Aryan: Rajasthani

is clear, then, that maiṃ is the functionally unmarked form in Braj, whereas the
use of hauṃ is limited to S and to the subject of the non-perfective verb forms.
Additionally, Liperovsky (2007: 148) notes that nouns in the oblique plural can
function as A without the naiṃ postposition, something that does not occur in
Hindi. Compare, e.g., the following example:

(304) coranu sabu mālu-asbābu nikāri+lī-yau

thief.obl.pl all luggage[m] go out+take-pst.m.sg
‘The thieves took out the whole luggage.’

Nevertheless, in Marwari, there is a tendency to differentiate between A and S in

the masculine plural: S never takes the oblique case, while A may take oblique
or nominative case in free variation (cf. Khokhlova 1992). The plural paradigm of
nouns functioning as subject is exceptional in both Braj and Marwari.
In Braj, the postposition kau or kūṃ is used exclusively for a marked O.
However, the conditions licensing the use of this postposition are not as rigid
as in Hindi. In Hindi, all pronouns must take a ko-postposition, because of
their animacy and definiteness. Yet, in Braj, pronouns can appear without a
postposition,²⁰ as exemplified by tū in ex. (305) (Liperovsky 2007: 151).²¹

(305) tū mere mālik=neṃ yādi kī-yau+h-ai

you I.gen master=erg remembrance do-pst.m.sg+aux.prs
‘My master has recalled thee.’

20 The question of whether the use of a postposition with pronouns in Braj is related to a
semantic difference is a contentious issue and is the subject of current research. From the
examples in Liperovsky (2007: 150–151), it is clear that A is either lower in person than the
unmarked pronominal O or is in third person. Furthermore, in cases in which there is agree-
ment with a marked O in Braj O is always animate in the examples given in Liperovsky (2007).
However, to draw definite conclusions about the role of referential hierarchies in the alignment
of Braj, more material on Braj is needed.
21 I asked a native speaker of Braj to assess this example as well as the following ones cited
by Liperovsky (2007). My informant altered all examples in order to obtain an alignment pattern
that better conformed to the alignment of Hindi. For instance, he changed to in example (305)
to to=kū, maiṃ in ex. (306) to mo=kū. He was in doubt about the agreement in ex. (306) and he
hesitated to change mārī to the expected m.sg. form in Hindi māro. On the basis of this brief
inquiry, it is of course impossible to give a definitive judgment of Liperovsky’s examples (which
are from a collection of folk tales) or to assume that among speakers who know Braj and Hindi
there is a tendency to prefer the Hindi pattern of alignment. I leave this question to further
research.
 Central Indo-Aryan and Hindi: variation on a micro-scale 243

The agreement of transitive verbs in the perfective is either with an unmarked or

marked O. In Hindi, O-agreement occurs only if O is nominative. Agreement in
perfective constructions in Hindi is always in gender and number. This occurrence
is related to the fact that the perfective verb was a participle historically, which,
because of its deverbal nature, can agree only in gender and number, regardless
whether it is used as a finite verb or as an adjective. The present perfect construc-
tion consists of a participle and a conjugated verb. One might expect, then, that
the conjugated verb would agree in person with O. This agreement is impossible,
not because there is a restriction on agreement in person, but because first and
second person O’s are always marked with ko, and formal marking of O rules
out agreement in Hindi. Agreement with a pronominal O is thus automatically
prevented in Hindi. However, in Braj, the use of ko is not as restricted. We find
occurrences where a pronominal O does not take the postposition and agreement
in person occurs, as for instance in the following ex. (306) from Liperovsky (2007:
150), where the copula h-ūṃ is first person singular, agreeing with the first person
maiṃ, an unmarked O.

(306) maiṃ rājā=neṃ sāt tār-e=bhītar mūṃ-di

I.f king=erg seven lock[m]-pl=in confine-ptcp.prs.f.sg
rākh-ī+h-ūṃ
keep-pst.f.sg+aux-prs.1sg
‘The king has been confining me under seven locks.’

In general, agreement is freer in Braj because the verb can also agree with a
marked O. In ex. (307) the verb mār-ī displays a feminine ending, agreeing with
the marked O vā=kū ‘she’ (Liperovsky 2007: 149).

(307) yā bāt=kū sun vā=kai khasam=ne vā=kū bauhut

this thing=obj hear.cvb she=gen husband=erg she=obj much
mār-ī+ī
beat-pst.f.sg+aux.pst.f.sg
‘On having heard this, her husband had beaten her severely.’

In conclusion, Braj is somewhere between Hindi and Marwari with respect to its
agreement patterning. Marwari never differentiates agreement between a marked
or unmarked O, and there is always agreement between the perfective verb and
O, as in the ex. (307) from Braj. However, in Braj, this kind of agreement is excep-
tional. More frequently, the same agreement as in Hindi occurs: the marking of
O does not appear in combination with O-agreement. Still, the postpositional
marking and agreement in Braj is less strict than in Hindi, in the sense that in
244 Central Indo-Aryan: Rajasthani

Braj, sometimes there is agreement with a marked O; such agreement is prohib-

ited in Hindi. Furthermore, unlike in Hindi, there is no obligatory rule for the
postpositional marking of plural nouns in functioning as A; nor is there one for
pronouns in the function of O. Agreement in Braj is a new challenge to the con-
straint in Mohanan (1994: 105), which says that in Hindi, the verb agrees with
the nominative argument, irrespective of its grammatical role. This constraint is
clearly violated in more distant IA languages, and it turns out to be too strong
for Braj, a language assumed to be very closely related to Hindi, because agree-
ment between a marked O and the verb is possible – and perhaps required (cf.
Section 6.2.3).

6.3.2 Bundeli and Bagheli

Another variety of Hindi is the language Bundeli, spoken in the region of Bun-
delkhand, a historically important area in Central India that extends over the
states of Uttar Pradesh and Madhya Pradesh. Bundeli is predominantly a spoken
language and is not known to have much literature. It is in competition with
Hindi, while also bordering the area where Braj is spoken. Its features are mainly
related to Hindi and Braj, and it is considered a Western dialect of Hindi. Its align-
ment is similar to these two languages. There is an ergative postposition neṃ,
which is used as the marker for A with a verb in the perfective (cf. Jaiswal 1962:
99). O is marked only when animate and definite, in which case its marker is
the postposition khoṃ or its variant khaṃ. This is also the postposition of the
IO, except with beneficiaries, of which Jaiswal (1962: 102) says that the layer III
postposition ke lāneṃ is also often used. The verb always agrees with the subject,
except in the perfective, where it agrees with O if the latter is in the nominative
case, e.g. ex. (308) from Jaiswal (1962: 146).

(308) tum=neṃ ek ām khā-o

you=erg one mango[m] eat-pst.m.sg
‘You ate a mango.’

Jaiswal (1962) describes Standard Bundeli (if it is at all possible to speak of a

standard variety of this language). In the introduction to his work, some varie-
ties of Bundeli are discussed. Kundri (Kunḍrī) is one particular dialect of Bundeli.
Kundri is spoken in the area around Banda and Hamirpur in the state Uttar
 Central Indo-Aryan and Hindi: variation on a micro-scale 245

Pradesh.²² This region is close to the Eastern Hindi area. Awadhi is a well-known
representative of Eastern Hindi (cf. Section 3.2.2), but another variety is Bagheli.
Kundri is spoken close to where Bagheli is spoken. Kundri is a mix of both the
Eastern Hindi Bagheli and the Western Hindi Bundeli, as can also be seen from
the alignment of the language. Consider for instance the following examples.

(309) mumtāj kah-is abai maiṃ mītiṃg=mā

Mumtaj say-pst.3sg now I meeting=loc
jā-t+h-auṃ
go-ptcp.prs+aux.prs-1sg
‘Mumtaj said, “Now I’m going to a meeting”.’

(310) chah sāl pahile gāṃv=kā mumtāj grām+sabhā=ke jamīn

six year early village=gen Mumtaj village+council=gen land
aür tālāb=kā ghervā=ke bāunḍrī
and pond=gen mortgage=gen boundary
ban-vā+dih-is+ha-vai
make-caus+give-pst.3sg+aux.prs-3sg
‘Six years ago Mumtaj of the village mortgaged the land and the pond of
the village council.’

(311) ham kaünau māripīṭ nahīṃ kī-n

we nobody beating not do.pst-1pl
‘We didn’t give anyone a trashing.’

22 To give an illustration of the complex situation of the varieties of Hindi, consider the follow-
ing: the language Kundri is described by Jaiswal as a variant of Bundeli, “a kind of mixed Bun-
deli” (Jaiswal 1962: 11). He argues that there are two types of Kundri, “[t]he Kunḍrī of Banda is
somewhat different from the Kunḍrī of Hamirpur and is based on Bagheli, but possesses many
characteristic features of Bundeli, whereas Hamirpuri Kunḍrī is based on Bundeli with some
Bagheli elements.” This division is taken over by Masica (1999: 433), for whom Kundri-2 is a
dialect of Bundeli, spoken in Hamirpur, and Kundri-1 “a subdialect of the jūṛar dial. of Bagheli”,
spoken in the Banda district. Both classifications are based on Grierson’s LSI. There are about
11,000 to 15,000 speakers of Kundri.
The examples are from the website http://www.nirantar.net/khabar_slides.htm. The language
is called Bundeli, but from a comparison of the language features and the area in which the lan-
guage is spoken (Chitrakoot and Banda districts), it is clear that it is not the standard Bundeli
and is most probably Kundri.
246 Central Indo-Aryan: Rajasthani

In these examples it is clear that Kundri does not follow the alignment pattern of
the Standard Bundeli discussed by Jaiswal (1962). First, there is no ergative post-
position or any other marker for A of a perfective verb. Second, the perfective verb
does not agree with O, but with the subject. As for alignment, Kundri is certainly
closer to the Eastern Hindi varieties than to Bundeli. Additionally, note the simi-
larities in conjugation with Awadhi. The sound -i- is important for the ending of
the verb in the past tense, and we also find it in the t2 conjugation in Awadhi (cf.
Chapter 3). In contrast to Kundri, in Awadhi, the main verb takes an additional -i,
and the form of the copula is slightly different from the Kundri copula. Compare
Kundri dihis havai ‘he gave’ and Awadhi dekhisi hai ‘he saw’. The Awadhi first
plural form ends in -en ( e.g. dekhen) whereas in Kundri, the form ends in -in,
(e.g. kīn) without sandhi resulting in ki-in. Finally, the Kundri lexicon has words
typical for Eastern Hindi.

6.3.3 Varieties of Hindi: conclusion

In comparing both Bagheli-influenced Kundri with Bundeli, and Braj with Hindi
above, an attempt was made to present and discuss the range of variation that
can be found in alignment types across Hindi dialects. A number of syntactic
similarities to the other Central IA languages have been observed. For instance,
there is never agreement with an ergative-marked A, and the agreement with O
in a perfective construction varies. In Braj, the verb may or may not agree with
O. In Kundri, Bagheli, and Awadhi, the verb always agrees with the subject, and
ergative marking of A is in general absent. Hindi is the only IA language in which
there is considerably less variation, which is perhaps due to its high degree of
standardization and its use in primary schools and media.

6.4 Central Indo-Aryan: Conclusion

The area of the Hindi belt, which is a fertile stretch of land because of the presence
of two great rivers, is a densely populated region that has often played a key role
in the course of Indian history. It is where the Moghul emperors established their
empire and introduced Arabic and Persian and their literary traditions and where
the Rajput princes held court. It is also the region where the local languages Braj
and Awadhi were instrumental in establishing a tradition of devotional litera-
ture in the “colloquial” languages at the time of the Moghuls. These are perhaps
some of the reasons for the immense variation in languages one finds there. Even
though at first sight the Central Indo-Aryan languages display a number of simi-
 Central Indo-Aryan: Conclusion 247

larities in alignment (for example, they all have the form ne), a closer investiga-
tion shows that the alignment is different in every language.
Central IA languages display a preference for gender and number agreement,
which is clearly based on the participial origin of the finite verb form. Apparently
related with the gender/number agreement is the fact that this agreement is often
with O. This agreement seems to be much stronger in the languages toward the
west of the area (Rajasthani and Gujarati). Hindi and its dialects do not display
consistent agreement with a marked O.
The purpose of postpositions in Central IA seems more of an “indication” of
the semantic properties of their arguments (as illustrated by the DOM pattern,
which the O-postposition follows) than of a syntactic, “distinguishing” func-
tion. In many languages, ne is multifunctional, Hindi, again, being the excep-
tion. The postposition ko in Hindi and the postposition nūṃ in Punjabi are dif-
ferent from the multifunctional naiṃ in Gujarati in that the argument they mark
can no longer control verb agreement. Finally, standardized Hindi may be said,
in general, to constitute an exception among the Central IA languages. Its clear
“rules” of agreement and case marking are not adequate to account for the varia-
tion one finds in Rajasthani and the Hindi dialects.
7 Conclusion

7.1 The definition of ergativity: summary

Dixon (1979; Comrie 1978) defined an ergative construction as a construction in

which O is treated the same way as S and different from A. With respect to mor-
phological ergativity, this “treatment” refers to verb agreement and case marking.
Syntactic ergativity, in which S and O are syntactically treated the same way in
coordination and conjunction patterns, does not occur in the Indo-Aryan lan-
guages discussed in this study. On the contrary, ellipsis of every core argument is
possible and is determined not only by syntactic constraints but also by semantic
and discourse-related factors (cf. Section 2.4.3).
The pattern of morphological ergativity defined by Comrie and Dixon is
attested in Indo-Aryan, but before a construction can be called ergative, several
conditions must be fulfilled. Of the Indo-Aryan languages discussed in this study,
Hindi, Kashmiri, Poguli, Sindhi, Siraiki, Marathi, Harauti, Gujarati, Punjabi,
Braj, and Bundeli display a “perfectly ergative” construction with ergative case
marking of A and perfective verb agreement with S and O. This observation is not
tantamount to classifying these languages as ergative, because they also display
other (and more) constructions that clearly do not conform to the definition of
ergativity.
In the above-mentioned languages, ergative constructions occur only under
specific circumstances. These circumstances are determined by the tense/aspect
of the verb and the place of the core arguments in the referential hierarchies. The
present study has revealed that these two factors directly or indirectly influence
the alignment patterns of constructions in Indo-Aryan. The following table pro-
vides an overview of the constructions in IA that are entirely ergative. The condi-
tions under which a particular type of alignment appears are not strict but must
be interpreted for each language separately. That is, the category of “perfective”
may differ slightly across the languages (for instance, compare the tense system
in Nepali, Section 4.1.3, with the tense system in Kashmiri, Section 5.1.2.1). This
type of difference also applies to the condition of “referentiality”, which includes
different aspects in different languages. Hence, although I use the general term
“referentiality” (ref-) in Tables 7.1 through 7.6, “referentiality” may refer to differ-
ent features of animacy, definiteness, person, and topicality. Moreover, the term
“referentiality” does not necessarily assume a relation between the referential
features of A and those of O. If there is such a relation, the alignment displays a
pattern in which the marking of A determines the marking of O and vice versa.
 The definition of ergativity: summary 249

The constructions in Table 7.1 are extracted from an inventory of 200 Indo-Aryan
constructions drawn from 20 languages (cf. appendix 1). For each language, two
intransitive and eight transitive constructions have been included, which differ
with respect to one of the following three conditions: tense/aspect of the verb,
the referentiality of O, and the referentiality of A. The case marking and the verb
agreement in each of the constructions were subsequently determined for the
transitive constructions; the intransitive constructions are only relevant for com-
parison between the form of A, O and S, and are not included in the following
analyses.

language tense/aspect ref-O ref-A case marking verb agreement

Hindi-Urdu perfective low high S=O≠A S, O

Hindi-Urdu perfective low low S=O≠A S, O
Kashmiri perfective high low S=O≠A S, O
Kashmiri perfective low low S=O≠A S, O
Poguli perfective high low S=O≠A S, O
Poguli perfective low low S=O≠A S, O
Sindhi perfective low low S=O≠A S, O
Siraiki perfective low low S=O≠A S, O
Marathi perfective low low S=O≠A S, O
Harauti perfective high high S=O≠A S, O
Harauti perfective high low S=O≠A S, O
Harauti perfective low high S=O≠A S, O
Harauti perfective low low S=O≠A S, O
Gujarati perfective low high S=O≠A S, O
Gujarati perfective low low S=O≠A S, O
Punjabi perfective low low S=O≠A S, O
Braj perfective low high S=O≠A S, O
Braj perfective low low S=O≠A S, O
Bundeli perfective low high S=O≠A S, O
Bundeli perfective low low S=O≠A S, O

Table 7.1 Ergative constructions in Indo-Aryan

The Rajasthani language Harauti is the Indo-Aryan language with the highest
number of ergative constructions (all four perfective constructions); that is,
Harauti is the Indo-Aryan language with the lowest number of conditions that
must be fulfilled for a construction to qualify as ergative. All ergative construc-
tions in Harauti require that the verb be perfective; this requirement is the sole
condition. The place of the arguments in the referential hierarchies does not
directly interfere with the case marking or the verb agreement because Harauti
restricts the occurrence of the postposition naiṃ, which can mark any core
250 Conclusion

argument. In perfective transitive constructions, naiṃ already marks A; thus, a

“second” naiṃ-marking in the sentence, the marking of O, is straightforwardly
impossible. Furthermore, the verb in perfective constructions in Harauti always
agrees with O. As a result, O has the same nominative case marking as S, and the
verb agrees with S and O.
Hindi displays the same alignment under the same conditions as Braj and
Bundeli. In contrast with Harauti, a high-ranked O is marked and prevents
ergative agreement from occurring. Hence, O is treated entirely differently from
S in Hindi, Braj and Bundeli. In Braj, constructions with a high-ranked O may
pattern ergatively in instances in which O is unmarked. In contrast with Hindi,
an animate and even pronominal O in Braj does not always take an objective
marking. However, these constructions are rather exceptional and seem to be
gradually disappearing from the language, perhaps under the influence of Hindi
(Section 6.3.1). Thus, these constructions are not included in the table. Gujarati
patterns in the same way as Hindi and its Western dialects in this table. However,
in Gujarati, the first and second plural pronouns do not display a distinctive erga-
tive case; this characteristic distinguishes Gujarati from the more regular pro-
nominal paradigm of Hindi.
Kashmiri displays the same pattern as the closely related language Poguli. All
arguments that are low on the referential hierarchy (i.e., non-pronominal argu-
ments) cannot control pronominal suffixes, and the verb in the perfective con-
structions in this table consistently agrees with the nominative O. The pronomi-
nal suffixes in Kashmiri that can refer to an ergative inanimate A are optional; this
construction is not included in the table. However, in constructions in which the
A-argument is high on the referential hierarchy, pronominal suffixes are much
more common and sometimes obligatory. In these instances, the ergative O-agree-
ment is accompanied by an additional A-agreement. Sindhi, Siraiki, Marathi, and
Punjabi only occasionally display ergative constructions because all three condi-
tions, the tense/aspect of the verb and the places on the referential hierarchies of
the A-argument and the O-argument influence the case marking of the arguments
and the verb agreement. However, there is a clear tendency in these languages
that the pronominal arguments pattern differently than the nominal arguments.
In summary, “perfectly ergative” constructions are rather limited in number
in Indo-Aryan. Twenty of the 160 transitive constructions included in the inven-
tory display a “perfectly ergative” construction. This low number is due to the
many so-called “splits” in alignment in Indo-Aryan languages, which I will
discuss in the next section.
 Splits 251

7.2 Splits

The term “split” is rather unfortunate when applied to Indo-Aryan languages

because it presupposes that an “abnormal” pattern is split from a normal pattern.
It is generally assumed that Indo-Aryan is split ergative, in the sense that it dis-
plays ergative constructions as exceptions to a predominantly accusative pattern.
However, the dominance of the accusative construction in Indo-Aryan is not con-
firmed by the data contained in Table 7.2. There are indeed more “perfectly accu-
sative” constructions than ergative constructions (i.e., there are 31 accusative

language tense/aspect ref-O ref-A case marking verb agreement

Hindi-Urdu imperfective high high S=A≠O S, A

Hindi-Urdu imperfective high low S=A≠O S, A
Asamiya perfective high high S=A≠O S, A
Asamiya imperfective high high S=A≠O S, A
Bangla perfective high high S=A≠O S, A
Bangla perfective high low S=A≠O S, A
Bangla imperfective high high S=A≠O S, A
Bangla imperfective high low S=A≠O S, A
Bhojpuri perfective high high S=A≠O S, A
Bhojpuri imperfective high high S=A≠O S, A
Bhojpuri imperfective high low S=A≠O S, A
Magahi imperfective high high S=A≠O S, A
Magahi imperfective high low S=A≠O S, A
Awadhi imperfective high high S=A≠O S, A
Awadhi imperfective high low S=A≠O S, A
Nepali imperfective high low S=A≠O S, A
Marathi perfective high high S=A≠O S, A
Marathi imperfective high high S=A≠O S, A
Marathi imperfective high low S=A≠O S, A
Marwari imperfective high high S=A≠O S, A
Marwari imperfective high low S=A≠O S, A
Harauti imperfective high high S=A≠O S, A
Harauti imperfective high low S=A≠O S, A
Gujarati imperfective high high S=A≠O S, A
Gujarati imperfective high low S=A≠O S, A
Punjabi imperfective high high S=A≠O S, A
Punjabi imperfective high low S=A≠O S, A
Braj imperfective high high S=A≠O S, A
Braj imperfective high low S=A≠O S, A
Bundeli imperfective high high S=A≠O S, A
Bundeli imperfective high low S=A≠O S, A

Table 7.2 Accusative constructions in Indo-Aryan

252 Conclusion

constructions and 20 ergative constructions, but overall, 31 is still rather low com-
pared to the total sum of 160 examined transitive constructions). Nevertheless,
the range of languages displaying an accusative pattern is wider than that dis-
playing an ergative pattern. One finds accusative constructions in the languages
Hindi, Asamiya, Bangla, Bhojpuri, Magahi, Awadhi, Nepali, Marathi, Marwari,
Harauti, Gujarati, Punjabi, Braj, and Bundeli. Whereas the ergative construction
is clearly limited to Central and Western Indo-Aryan, accusative constructions are
found across the entire region in which the Indo-Aryan languages are spoken.
The single accusative construction in Nepali must be put between brackets,
since A-arguments may optionally take an ergative marking. The reasons for this
marking cannot be directly linked to any referential hierarchy (cf. Section 4.2).
The ergative and accusative constructions are syntactically stable patterns.
They are able to distinguish A-arguments from O-arguments in an equal measure,
and in an economical and unambiguous way, by morphosyntactic mechanisms
such as case marking and verb agreement. However, some constructions in Indo-
Aryan are neither ergative nor accusative. This observation indicates that align-
ment in Indo-Aryan is not a purely syntactic issue. While the arguments are “dis-
tinguished” from each other by case marking and verb agreement, case marking
and verb agreement are also mechanisms that have “indexing” functions.

7.2.1 The influence of referential hierarchies

7.2.1.1 Referentiality of O
The objective marking of O in Indo-Aryan languages is primarily semantically
determined. As a form of case marking, the function of the objective is certainly
more directed toward indexing referential features of O than toward distinguish-
ing O from A. The marking of O is influenced neither by the marking of A nor by
the need to mark O differently from A. In most Indo-Aryan languages, the concept
of a “hierarchy” is applicable only in the sense that one can compare the place of
a marked O with the place of an unmarked O on the hierarchy, and then draw con-
clusions about the correlation between markedness and the referential features.
Only Kashmiri and Poguli show a correlation between the marking of the A- and
O-arguments within one construction. In both of these Western Indo-Aryan lan-
guages, the features of A and O are ranked against each other on the referential
hierarchies, and their respective ranking (i.e., A < O or O < A) determine their
marking.
The influence of the referential hierarchies leads to alignment patterns that
differ from the accusative and ergative alignment. In addition to these patterns,
Comrie (1978) distinguished the neutral type of alignment, in which all arguments
 Splits 253

are coded in the same way (A=S=O); the tripartite alignment pattern, in which all
arguments are coded differently (A≠S≠O); and the last type, in which A is coded
in the same way as O and differently from S, which is extremely uncommon.
In Indo-Aryan, if O is high in animacy/definiteness, the expected case marking
pattern is tripartite (A is ergative) or accusative (A is nominative) because O will
be marked with the objective marker and will thus differ from A and S. Indeed, a
number of constructions in which O is high in animacy/definiteness display this
type of tripartite case marking (22/80), and accusative case marking is even more
frequent (52/80). A few constructions in Table 7.3 combine a highly referential O
with an ergative case marking pattern. It is not surprising that these construc-
tions occur only in Kashmiri and Poguli, in which the O in perfective construc-
tions cannot be marked, irrespective of its place on the referential hierarchies.
Harauti displays the same pattern due to the aforementioned restriction on two
naiṃ markings in a single clause.

language tense/aspect ref-O ref-A case marking verb agreement

Kashmiri perfective high high S=O≠A A, S, O

Kashmiri perfective high low S=O≠A S, O
Poguli perfective high high S=O≠A A, S, O
Poguli perfective high low S=O≠A S, O
Harauti perfective high high S=O≠A S, O
Harauti perfective high low S=O≠A S, O

Table 7.3 Ergative case marking in constructions with a high referentiality of O

Conversely, if the referentiality of O is low, the preferred case marking patterns

are ergative (28/80) (if A is ergative), and neutral (52/80) (if A is nominative).
Remarkably, the present investigation has found that the referentiality of
O has no particular influence on the agreement pattern. In general, accusative
agreement is the most common type of agreement among the IA constructions
(88/160). In constructions with a high-ranked O, 39/80 constructions display
accusative agreement, and 21/80 display tripartite agreement. In constructions
with a low-ranked O, 49/80 constructions display accusative agreement, and
12/80 display tripartite agreement. This difference is caused by the languages in
which a high-ranked O controls the occurrence of a pronominal suffix.
The dissociation of verb agreement and the referential properties of O is
consistent with Bickel’s findings (2008) that although features from referential
hierarchies obviously influence the case marking, particularly the marking of O,
alignment in general is not influenced by these hierarchies. That is, there is no
254 Conclusion

observable tendency toward ergative or accusative agreement associated with a

high- or low-ranked O.
The exceptional languages are Hindi, Punjabi, and Marathi, in which the
marking of O does not co-occur with O-agreement. However, in the other lan-
guages with ergative agreement and differential object marking (DOM), viz.
Marwari, Harauti, and Gujarati, the marking of O has no influence on the agree-
ment pattern. Remarkably, there is a clear morphological difference between
the markers of O in these languages, and this difference parallels the distinction
between O-agreement and no O-agreement. As shown in the following table, the
case marker of a high-ranked O in Hindi and Punjabi has a distinctive form, in
contrast with the O-marker in the Rajasthani languages (Marwari and Harauti)
and in Gujarati. Marathi may be considered an exception, because the origin
and relation of the lā-postposition seems closer to the Nepali objective marker
lāī than to the Hindi marker ko (see Pandharipande 1990: 173). The Sindhi objec-
tive marker khe is reminiscent of the Hindi form, and like Hindi, Sindhi has no
gender/number agreement with a marked O. However, personal agreement is
possible in Sindhi in the form of a pronominal suffix referring to the marked O.

Hindi-Urdu ko
Punjabi nūṃ
Marwari naiṃ
Harauti naiṃ
Gujarati ne
Marathi lā
Sindhi khe

Table 7.4 Objective postpositions in Central and Western Indo-Aryan

7.2.1.2 Referentiality of A
In the typological literature, it has been argued that the differential marking of A
is related to the person hierarchy. Animacy and definiteness have a much smaller
role in determining the marking of A. In Indo-Aryan, animacy and definiteness
do not influence the marking of A at all. The person hierarchy ranks the pronoun
with respect to the nouns, yet the cut-off point for differential marking often lies
between the first/second person and the third person. Research has suggested
that there is a cross-linguistic pattern according to which a pronominal A tends
to pattern accusatively and a nominal A ergatively. However, Bickel (2008) argues
that this correlation is merely incidental and that the motivation behind the vari-
ation differs for each particular language.
 Splits 255

The overview of the Indo-Aryan languages provided in Appendix 1 shows that an

A that is high on the referential hierarchy displays an accusative case marking
more often than an ergative or tripartite marking (different from O and from S).
However, the difference is negligible (28/80 vs. 24/80). Furthermore, if we closely
examine the forms of the pronouns that do not display any difference between A
and S in languages in which nouns differ formally when used as A or S, then we
observe that most of these pronominal forms were originally oblique forms that
functioned as ergative markers. For instance, the first person singular pronouns
in Asamiya (maï), Marwari and Punjabi (maiṃ), and Marathi (mī) are all old
oblique forms. It is understandable that these forms are not additionally marked
with an ergative postposition because they are already considered to be case
marked. With respect to the factor of the referentiality of A, Gujarati is an interest-
ing case in point, given that in that language only first and second person plural
pronouns do not display a different marking for A and S. However, the forms of
the first and second pronoun are apṇe and tahme, respectively; both pronouns
end with -e, which is also the ergative ending of nouns in general in Gujarati.
In summary, the different case marking of pronouns and nouns is caused not
by a difference in the referentiality of A but rather by the tendency of pronouns
to phonologically change and the greater tendency of case forms to merge more
easily with each other into one form compared with nouns (cf. Filiminova 2005).
In Marwari, the development of the merging of ergative and nominative marking
has even spread to the nominal paradigm insofar that there is no significant dif-
ference in use between the nominative and oblique forms as markers of A (except
in the third person plural, Section 6.2.1).
Considering that the apparent accusativity of pronouns is due to historical
developments in the case paradigm and not to specific referential properties of A,
it may be expected that verb agreement is not influenced by the different marking
of nouns and pronouns. The inventory of alignment patterns in Appendix 1 shows
that this expectation is correct. In Punjabi and Marathi perfective verb construc-
tions, the verb agreement remains the same throughout the nominal and pro-
nominal paradigms; however, pronouns do not have an overt ergative marking,
whereas nouns do.
Nevertheless, there is a slight variation in the verb agreement patterns that is
caused by the referentiality of A, but this variation is due to another factor, viz.,
the presence of pronominal suffixes. Pronominal suffixes occur only referring to
pronominal arguments, which are high on the person hierarchy. However, pro-
nominal suffixes do not “prefer” a certain type of alignment. Their occurrence is
rather determined by the discourse prominence of the argument to which they
refer in a relationship of coreference (i.e., anaphorically). In some languages,
256 Conclusion

such as Siraiki and Sindhi, the coreferential antecedent of a pronominal suffix

need not be a core argument.
There is a second type of pronominal suffixes in Kashmiri. A pronominal
suffix of this second type refers to a particular argument and conveys a certain
level of definiteness to the argument to which it refers. Whereas agreement suf-
fixes have a distinguishing function and cross-reference the core arguments of a
construction, these pronominal suffixes indicate the discourse-prominence of the
argument to which they refer, even when that argument is not overtly expressed.

7.2.2 Tense/Aspect/Mood splits

The Indo-Aryan languages are typical examples of languages in which the so-
called Tense/Aspect/Mood split occurs. Ergative alignment is associated with
perfective constructions, and accusative alignment is associated with imperfec-
tives. Except for Asamiya and, to a certain extent Nepali, this split appears in
every IA language that possesses an ergative pattern. However, following Ander-
son (1992), I contend that the tense/aspect split in Indo-Aryan is not functionally
motivated by a semantic similarity between the ergative and perfective construc-
tions, but that the split is rather the result of a specific historical development,
viz., the simplification of the verb paradigm in Old and Middle Indo-Aryan. In
the Eastern languages, the pattern of gender/number agreement with O was lost
when the perfective verb changed to a non-participial form displaying personal
agreement with A. However, in Asamiya and the Bihari languages, traces of the
older participial agreement pattern can still be found, particularly in the transi-
tive marking of the third person of a past tense verb.
If we assume that the tense/aspect split is a consequence of a development in
the tense paradigm of verbs, then we may expect that the split is primarily related
to features concerning the verb, particularly, verb agreement. The verb agreement
is often different from the case marking pattern in a single language. Verb agree-
ment appears to be primarily associated with a particular tense/aspect and lacks
the semantic distinctions that are often conveyed by case marking. For instance,
in the few imperfective constructions in Asamiya, Shina, and Nepali that have
ergative case marking, the verb agreement is consistently accusative; this pattern
is expected for imperfective constructions.
 Splits 257

language tense/aspect ref-O ref-A case marking verb agreement

Asamiya imperfective low low S=O≠A A, S

Nepali imperfective low high S=O≠A A, S
Shina imperfective low high S=O≠A A, S
Shina imperfective low low S=O≠A A, S

Table 7.5 Imperfective constructions with ergative case marking

It is reasonable to assume that the ergative case marking in imperfective con-

structions is the result of a relatively recent development. Asamiya, Nepali, and
Shina are all spoken at the borders of the IA language area. It is possible, and
even likely, that the spread of ergative case marking from the perfective to the
imperfective constructions has occurred under the influence of neighboring, non-
Indo-Aryan languages; however, based on this study, it is not clear whether erga-
tive case marking was absent in the imperfective constructions in older stages of
Asamiya. With respect to Shina, this influence on the basis of language contact
is almost certain because this language displays a type of the ergative marking
in imperfective constructions that is quite different from the ergative marker in
perfective constructions, and this feature is most likely borrowed from Tibetan
languages spoken in the area. In Table 7.5, Nepali has only one imperfective con-
struction with an ergative-marked A. Ergative case marking in Nepali is related to
a high degree of topicality of the argument and the perfective aspect conveyed in
the construction, whereas in Asamiya, the marking of a pronominal A does not
differ from the marking of nominative A.
Conversely, ergative verb agreement often occurs in combination with accusa-
tive case marking in perfective constructions. Ten out of eighteen perfective con-
structions with accusative case marking display the possibility of O-agreement,
and many of these constructions are from the Eastern, “accusative” languages (cf.
Table 7.6). In these languages, gender and number have been lost as grammatical
categories, and person agreement seems to have replaced the “adjectival” agree-
ment of Western and Central Indo-Aryan.
The phenomenon of pronominal suffixes in Western Indo-Aryan and the
person agreement in Eastern Indo-Aryan have often been considered to be related.
For instance, Chatterji (1926: 974) refers to the person marking on the verb as
“pronominal affixation”. He argues that the addition of pronominal suffixes to
verb forms has led to the development of person agreement on the verb. Person
agreement in Eastern Indo-Aryan is equally well the result of the main verb form
merging with the copula. The simple present and past in Nepali are in that regard
illustrative because they clearly show that the verb form, although no longer peri-
phrastic, consists of a combination of the main verb and a copular form.
258 Conclusion

If there is a similarity between pronominal suffixes and agreement, it is between

the Western pronominal suffixes and the additional suffixes in Eastern Indo-
Aryan, particularly the Bihari languages, which refer to a marked object and are
called “secondary agreement”. Similar to the pronominal suffixes, which are
often optional, O-agreement is not obligatory in the Eastern languages and is
partially determined by the high referentiality of O. Bickel, Bisang and Yadava
(1999) argue that pronominal suffixes provided the basis of the personal agree-
ment system in Maithili, an Eastern Indo-Aryan language. With respect to specific
uses, the similarities between the Western pronominal suffixes and the “second-
ary agreement” of the Eastern languages are evident. For instance, in Magahi, the
“Speech Addressee”, which is an unexpressed argument that is not part of the
transitive or intransitive construction, can be referred to with a suffix attached
to the verb. This use is highly reminiscent of the use of the objective pronomi-
nal suffixes in Kashmiri and Siraiki because these suffixes can also refer to an
unexpressed argument that, although not overtly expressed in the construction,
is involved in the speech act.
However, irrespective of these similarities between Western and Eastern pro-
nominal suffixes, it is revealing that O-agreement in Eastern Indo-Aryan often
occurs only in perfective constructions. This pattern may constitute evidence of a
relationship between the original participial form of the verb and the possibility
that this verb agrees with O.

language tense/aspect ref-O ref-A case marking verb agreement

Asamiya perfective high high S=A≠O A,S

Awadhi perfective high high S=A≠O A,S,O
Awadhi perfective high low S=A≠O A,S,O
Bangla perfective high low S=A≠O A,S
Bangla perfective high high S=A≠O A,S
Bhojpuri perfective high high S=A≠O A,S
Bhojpuri perfective high low S=A≠O A,S,O
Magahi perfective high high S=A≠O A,S,O
Magahi perfective high low S=A≠O A,S,O
Maithili perfective high high S=A≠O A,S,O
Maithili perfective high high S=A≠O A,S,O
Marathi perfective high low S=A≠O A,S
Marwari perfective high high S=A≠O S,O
Marwari perfective high low S=A≠O S,O
Punjabi perfective high high S=A≠O S
Siraiki perfective high high S=A≠O A,S,O

Table 7.6 Perfective constructions with accusative case marking

 Splits 259

7.2.3 Classifying the Indo-Aryan languages

More than 100 years ago, Grierson (1903) classified the Indo-Aryan languages
according to an inner-outer division. The “outer” languages are the languages
spoken at the eastern and western borders of the Indo-Aryan region. The inner
languages are the languages spoken in the “Hindi belt”. Grierson’s classification
was based on the theory of dialectology and isoglosses (which was popular at
the time). This theory assumes that the group of outer languages has developed
further from the proto-language, whereas the inner languages are the most “con-
servative” languages. The features he selected for this classification were phono-
logical properties and the occurrence of person agreement.
Grierson’s “inner-outer” model was refuted by Chatterji (1926: 169; Masica
1991: 450). Alternatively, Chatterji suggested a line of separation between Eastern
and Western Indo-Aryan and subdivided the Central Indo-Aryan languages on the
eastern or western side of this line. Whereas Grierson relied on phonological cri-
teria, Chatterji primarily based his division on the presence or absence of a con-
jugated past tense. Chatterji argued that Grierson’s argument, according to which
the inner languages had gender/number agreement and the outer languages had
person agreement, severely underestimated the differences between the Western
and Eastern Indo-Aryan verbs and their respective agreement patterns. Chatterji
conjectured that the Western verbs were essentially similar to the Central Indo-
Aryan verbs and that person agreement in Western Indo-Aryan was optional. He
noted that Eastern languages have a transitive past tense verb that is active and
conjugated, whereas the Western languages retain the “nominal” system of Old
Indo-Aryan. Thus, Chatterji concluded that Eastern languages have progressively
changed, whereas the Western languages still retain more features of older stages
of Indo-Aryan.
Today, Grierson’s idea of an outer and inner circle of language “types” has
been refuted. However, the alternative view, according to which there is a straight
line between the Eastern and Western languages, is also an oversimplification
with respect to the attested language patterns. Alignment in Indo-Aryan is much
more varied than these accounts suggest. For instance, it would be implausible to
consider the alignment of Nepali and the Bihari languages as strictly “Eastern” or
“Western”. Montaut (2009) suggests a division into three systems: the “Bangla”
system, the “Hindi-Urdu” system, and the “Sindhi” system. These three systems
are parallel to the geographical distribution of the languages. According to
Montaut, the “Bangla” system illustrates the end of a diachronic circle that pro-
gresses from an accusative to an ergative and back to an accusative construction.
Therefore, Bangla seems to follow the alignment of languages such as French or
Persian. Because person markers have been attached to a Bangla verb, the con-
260 Conclusion

struction has lost all of its ergative features. However, although this pattern might
be observable in Bangla, not all of the other Eastern languages are strictly accusa-
tive. For instance, I refer to the Asamiya alignment discussed in Chapter 3.
The “Hindi-Urdu” system is the most “ergative” system and represents a
straight evolution from the OIA pattern, which is considered “le noyau dur de
l’héritage sanscrit” (Montaut 2009: 10). In that sense, Montaut concurs with Gri-
erson by arguing that Hindi is the language that has the most conservative fea-
tures, although she posits this argument on different grounds. Montaut refers to
the participial construction that Hindi inherited from OIA rather than focusing on
phonological features and person agreement.
The “Sindhi” system is similar to the “Hindi-Urdu” system, but pronominal
suffixes occur only in the Northwestern Indo-Aryan languages; this difference is
probably due to contact with one or more non-Indo-Aryan languages. However,
crucially, and as Montaut (2009) shows, each language displays a range of varia-
tion; some constructions tend to be closer to one system, and some are closer to
another system. Nepali, Bihari, and even the Hindi varieties are languages that
fall somewhere in between the different systems. Nevertheless, Montaut’s divi-
sion is very convenient and much more fine-grained than the inner-outer or the
West-East theory. It is therefore better suited to describe the variation of align-
ment patterns attested in Indo-Aryan.
Finally, another classification has been provided by Klaiman (1987: 96), who
ranks the IA languages on a scale ranging from very ergative to least ergative.
Klaiman ranks Kashmiri and Pashto (Indo-Iranian) among the “most” ergative
languages, mainly because Kashmiri lacks DOM in the perfective constructions.
Marwari is slightly less ergative. Gujarati and Mewari, Hindi, Bundeli, Panjabi,
Marathi, Baluchi (Indo-Iranian), Lahnda, and Sindhi are still less ergative. Nepali
and Burushashki (which is a language isolate) are more accusative, and Shina,
Ladakhi, Balti, Asamiya, and nonstandard Bangla are the least ergative of all.
Although Klaiman’s criteria parallel the criteria used in this study, I believe that
it makes little sense to claim that one language is “more ergative” than another
language, because there are different alignment features that are independently
motivated. Not all features indicate “ergativity”; for instance, the presence of pro-
nominal suffixes is almost orthogonal to whether or not a language is “ergative”.

7.2.4 Case marking versus verb agreement

As illustrated in Sections 7.2.1–7.2.2, in many Indo-Aryan languages, case marking

and verb agreement present different alignment patterns. It was argued that
agreement is associated with the tense/aspect split, whereas case marking is more
 Core arguments and grammatical relations 261

closely associated with the referential hierarchies of the arguments. Pronominal

suffixes confirm their middle position by showing a greater association with the
semantic properties of case marking than with the properties of agreement, such
as indexing definiteness/animacy on an argument. Although one may speak of
a “mismatch” between case marking and verb agreement (cf. Woolford 2006),
there is no historical or functional reason to assume that case marking and verb
agreement must show the same type of alignment. On the contrary, case marking
and verb agreement are complementary in the Indo-Aryan languages, and they
do not conflict. Typically, agreement fulfils the function of “distinguishing” the
arguments, whereas case marking is primarily a mechanism for “indexing” the
arguments (cf. Section 1.1.2). However, in the languages in which case marking
does not indicate the definiteness or animacy of a core argument, agreement may
indicate these features (e.g., the suffix -n/-kh in Kashmiri). Moreover, we found
that the case marking patterns may change or eventually disappear in languages
with consistent agreement patterns, such as Harauti, Marwari, and Nepali.

7.3 Core arguments and grammatical relations

The term “subject” can be used with respect to Indo-Aryan as a term that is similar
to its traditional definition of the grouping of A and S. In most imperfective con-
structions in Indo-Aryan, there is a clear morphological and syntactic basis for
treating A and S the same way. In perfective constructions, the morphosyntactic
properties of case marking and agreement show a different treatment of A and S,
but on the level of the syntactic properties of coordination and conjunction, they
act the same – to the extent that these syntactic properties are the criteria that
distinguish grammatical relations in Indo-Aryan. In Section 2.4.3, I argued that
certain measures that are commonly believed to identify subjects are not effective
with respect to Indo-Aryan. For instance, a reflexive pronoun can refer to every
argument in most Indo-Aryan languages. Word order is also a doubtful criterion;
although every language has a specific dominant order in texts, specific discourse
motivations of word order have a much more important role in spoken language. In
patterns of coordination, the tendency to group A and S together is stronger than
for other arguments. However, this tendency is not an absolute “rule”. Montaut
(2001: 338) accurately summarizes the subject properties of Hindi as follows:

“La catégorie de sujet n’est donc pas obligatoire en hindi/ourdou, et pour cette raison on
peut considérer qu’elle n’est pas pertinente dans la langue, ne concernant que certaines con-
structions, bien qu’un nombre non négligeable d’énoncés (…) puissent s’analyser comme
ayant sujet, c’est-à-dire un terme auquel s’attachent l’ensemble des propriétés subjectales”.
262 Conclusion

A similar conclusion can be drawn with respect to the Indo-Aryan languages

in general. Therefore, “A” and “S” are certainly the more convenient terms to
describe the grammatical relations on the basis of their morphosyntactic coding
properties. Nevertheless, I concur with Montaut that the term “subject” can aptly
be used to describe a number of constructions in Indo-Aryan, particularly con-
structions with imperfective active verbs.
The ergative marking is similar throughout the Indo-Aryan languages that I
have analyzed. There exist the forms ending in -e, the oblique marking, and the
forms based on the postposition ne or le. Table 7.7 provides an overview of the
forms of the ergative markers in the languages discussed in this study.

Hindi-Urdu =ne
Kashmiri (Poguli) -an/-y
Sindhi -e
Siraiki -e
Marathi =ne/=nī
Harauti =naiṃ
Gujarati -e
Punjabi =ne
Braj =nai
Bundeli =ne
Nepali =le
Asamiya -e

Table 7.7 Ergative case markers in Indo-Aryan

The distinction between O and IO is a more complicated matter. Indo-Aryan does

not possess secundative nor indirective alignment; in Indo-Aryan, only a particu-
lar type of O and IO are treated in the same way. Kittilä (2006) explains this type
of marking as extended differential object marking (EDOM) by considering it as
a pattern in which the marking of O and IO is determined by animacy/definite-
ness. However, because an IO is typically animate/definite, Kittilä claims that the
marking has become generalized for every IO. Thus, I have chosen to retain the
term “IO” to emphasize the similarity in marking with an O despite the differ-
ing semantic roles of O and IO. The semantic role of the IO is traditionally the
recipient/beneficiary of an action. Asamiya makes a distinction in the marking
of the semantic roles of recipient and beneficiary: recipients are marked with the
objective suffix -k, and beneficiaries are marked with lai. In Hindi (ko vs. ke lie),
Bundeli (ko vs. ke lāneṃ) and the other languages, this distinction is also present,
but the use of the objective marker is often predominant. It is hardly a coinci-
dence that Asamiya -k and lai are two forms that are also attested in the other
 The motivation of the ergative pattern in Indo-Aryan 263

Indo-Aryan languages. Forms with -k are generally used to mark the objective
case (for instance -k in Asamiya, ko in Hindi, and, perhaps, the related form nūṃ
in Punjabi). Asamiya’s lai is related to lāī in Nepali (and perhaps to le in Nepali),
ne in Hindi and Punjabi, and naiṃ in Marwari and Harauti. Whereas ko is clearly
an objective that marks a definite/animate O, lai and ne are multifunctional mor-
phemes. In most languages, lai/ne has one dominant function (i.e., serving as an
ergative marker in Hindi and Punjabi, and an objective marker in Marwari and
Gujarati). However, ne in Harauti can mark the argument in the ergative and the
objective cases, whereas Nepali has lāī for the objective and le for the ergative
case. Other languages, such as Sindhi, Kashmiri, and Gujarati use the oblique
form to mark the ergative case.

7.4 The motivation of the ergative pattern in Indo-Aryan

Considering the extensive variation in case marking and verb agreement, and in
alignment in general, that emerges from the present study, it does not make much
sense to attribute the existence of the ergative pattern to some cognitive motiva-
tion, according to which the ergative construction primarily resembles an intran-
sitive construction on conceptual grounds. It seems likely that a convergence of
various different factors has led to this pattern in the Indo-Aryan languages and
that cognitive or discursive motivations, although they may have been a factor,
are not the main reason for the existence of the ergative pattern.
It must be borne in mind that the tense/aspect split in alignment is histori-
cally motivated. This historical dimension places an important constraint on any
direct motivation of ergativity, particularly on the assumed similarity between the
conceptualizations of perfective and ergative constructions. There is no semantic
similarity establishing a direct link between the two constructions or between a
perfective construction and ergative case marking. If such a semantic similarity
did exist, then the Asamiya ergative case marking with imperfective verbs and the
Marwari accusative case marking with perfective verbs would be incomprehensi-
ble, because both languages would break the link of perfectivity with ergativity.
In the typological literature, it is customary to consider ergative construc-
tions as exemplary transitive constructions with two arguments that are maxi-
mally distinct from each other (cf., e.g., Næss 2007). However, this interpretation
conflicts with the cognitive motivation that the ergative construction is similar to
the intransitive construction in that both share the same theme-relation (i.e., the
S is the theme-argument of the intransitive verb and the O is the theme-argument
of an “ergative” verb). In contrast with this cognitive motivation, I think that the
present study has demonstrated that the ergative argument is not syntactically
264 Conclusion

“outside” the construction. Although one could argue that this finding applies
to Hindi to a certain extent, the observed agreement with ergative-marked argu-
ments in Asamiya, Shina, Nepali, and the Northwestern Languages (at least to
some degree) definitively shows that the ergative A is, as a rule, a core argument
that is firmly entrenched in the transitive construction.
The idea that there is a resemblance between the ergative and intransitive
constructions in the Indo-Aryan languages seems to be derived primarily from
the historical development of the perfective constructions in these languages.
As explained in Chapter 2, a participial form, i.e., a deverbal form that had
undergone a valency-reducing operation, became the general way to express
a perfect aspect/past tense in Indo-Aryan. Valency-changing mechanisms are
very common in the Indo-Aryan languages, even in modern Indo-Aryan. The
morphology of Indo-Aryan verbs is built upon valency-increasing mechanisms:
divalent verbs are often morphologically causative, in the sense that they derive
from modified monovalent verbs. Therefore, Indo-Aryan languages belong to the
“transitivizing” class of languages, which “treat intransitives as basic and transi-
tives as derived” (Nichols et al. 2004:149; Haspelmath 1993)²³.
Because the OIA ta-participle was passive-like, it was initially used only in
intransitive constructions. However, when the participial periphrasis became
the regular form of the finite verb, the participle occurred in both transitive and
intransitive constructions. In fact, the idea that the ergative construction is based
on a theme-relation is simply another variant of the passive-as-ergative idea. In
contemporary Indo-Aryan, there is no “reduced” transitivity of ergative construc-
tions. On the contrary, an ergative construction contains an agentive agent, an
affected patient, and a verb expressing the completion of an action. These three
criteria arguably indicate a high level of transitivity. Actually, perfective construc-
tions seem to “remediate” their participial origin by structurally emphasizing the

23 Note that Haspelmath (1993: 98) argues that the causative (divalent) form of verbs in Hindi
is the basic form. For example, in the pair phirnā ‘to be turned’ and phernā ‘to turn’, phernā is
the basic verb and phirnā the derived verb, according to Haspelmath. He maintains that causa-
tive verbs may have either a high vowel (ī, ū) or a mid-vowel (o, e), whereas the monovalent
verb can have only a high vowel (i, u). However, this account is inaccurate. The greater variation
of vowels in causative verbs in Hindi is caused by a phonological modification of the vowel of
monovalent verbs that occurs in the formation of the causative. For example, the vowel of a
monovalent verb can be substituted by the stem vowel in guṇa form (i.e., prefixing a short a to
a simple vowel sound) or vṛddhi form (i.e., prefixing a long ā to a simple vowel sound) (Kellogg
1938: 28, 255). Contrary to Haspelmath’s claim, the number of possible stem vowels is unre-
lated to the distinction between the basic and derived verbs. In Hindi, phirnā, which is the form
of the monovalent verb, is undoubtedly the basic verb form, not the derived form.
 The motivation of the ergative pattern in Indo-Aryan 265

transitivity of the construction. This tendency is apparent in the different conju-

gations of transitive and intransitive verbs in Awadhi, Maithili, and Magahi and in
the possibility of O-agreement in these languages. With these mechanisms, these
languages emphasize the difference between an intransitive and a transitive con-
struction. In addition to the Eastern Indo-Aryan languages, Kashmiri, Shina and
Sindhi also have different conjugations for transitive and intransitive verbs.
Furthermore, Kittilä (2011: 365), following Hopper and Thompson (1980: 271),
suggests that there is a difference in transitivity in perfective and imperfective
constructions. Considering the historical developments that led to the TAM split
in Indo-Aryan, there is no particular need to assume this difference. However,
from a synchronic point of view, the Indo-Aryan languages show differing align-
ment patterns that always follow the line between perfective and imperfective
and never the line between past and non-past (e.g., a present perfect construction
is typically ergative, whereas a past imperfect construction is not ergative). In that
respect, if the hypothesis regarding the extension of the marker le in Nepali, as
suggested in Section 4.2, proves to be correct, then the indication of A in imper-
fective constructions in Nepali is another way of indicating the high level of
transitivity of the construction. Both the telicity of the action and topicality of A,
which are features indicated by the presence of the postposition le, are assumed
to be typical of highly transitive constructions.
Indo-Aryan languages also emphasize the indication of “de-transitivity”.
There are constructions in Indo-Aryan that display a type of “reduced” transi-
tivity. These constructions contain experiencer verbs that express psychological
states (e.g., Hindi yād ānā ‘to remember’, Bangla bhaloi lag, ‘to like’). The expe-
riencer and the unergative constructions can be considered opposites. Unerga-
tive constructions are construed with intransitive verbs that show the transitive
pattern of argument coding (i.e., in perfective constructions, the sole argument is
coded as an A). Conversely, experiencer constructions are construed with transi-
tive verbs that show a type of intransitive coding of the arguments with the expe-
riencer in an oblique case and the stimulus argument coded as an intransitive S.
The diverging alignment of the two constructions is caused by the verb semantics
assuming a higher priority than the syntax. Both constructions display a type
of “semantic alignment”. However, the experiencer and unergative construction
are completely different from the ergative alignment, and the non-nominative
marking of an experiencer is entirely different in origin from the ergative marking
of an A.
There is ample evidence suggesting that the ergative agreement goes straight
back to the reanalysis of the participial construction. However, the ergative case
is a later development; this case appeared only after different inflectional case
endings had merged. The ergative case is one of many cases in the paradigm of
266 Conclusion

Indo-Aryan. The ergative case is a purely syntactic case with distinctive proper-
ties that do not greatly differ from the accusative case in, for example, Latin and
Greek, in the sense that both the ergative and the accusative indicate that there is
a second, non-nominative core argument in the clause and that the construction
is transitive. The ergative is a factor that determines alignment, as all syntactic
cases do, but the presence of an ergative case does not imply ergative alignment.
The ergative case in Indo-Aryan differs from O-agreement and O-marking. This
case functions independently from agreement. However, because the ergative
case is successful in distinguishing the transitive A from the second core argu-
ment O, this case marker has a well-defined function and may be said to have
become a proper member of the case paradigm in many Indo-Aryan languages.

7.5 Final observations

In the literature, ergativity has been considered as a determining factor in the

Indo-Aryan languages, setting them apart from Indo-European in general.
Because the ergative pattern is highly uncommon among the Indo-European lan-
guages and because it occurs only under certain conditions in the Indo-Aryan
languages, it has been argued that the ergative pattern is a pattern “at odds”
with Indo-Aryan and is likely to disappear (e.g., it has apparently disappeared in
Bangla). However, the cross-linguistic comparison of Indo-Aryan languages pre-
sented in this study has shown that this view does not do justice to our findings.
The ergative construction is less widespread in Indo-Aryan than may be expected
and is not likely to disappear in the languages in which it occurs.
There is a strong tendency in the Indo-Aryan languages toward one par-
ticular type of agreement, either with O or with A. In a number of languages,
this “primary” agreement is complemented with pronominal suffixes or a “sec-
ondary” type of agreement. The ergative case, in turn, is a morphological case
marking that does not provide information about verb agreement. The agree-
ment with O, ergative agreement, seems to be the result of a historical develop-
ment in the tense/aspect paradigm. Therefore, there is no direct motivation that
“explains” the ergative agreement pattern; on the contrary, the motivation for this
pattern is rather “indirect” in the sense that the transition in the tense paradigm
of a participial form becoming the finite verb was “accompanied” by the emer-
gence of an ergative agreement pattern. Nevertheless, because of this historical
development, the indication of the transitivity of the construction is crucial in the
Indo-Aryan languages.
Ergative case marking is functionally motivated: it is an appropriate device
to distinguish the core arguments from each other and thus, to indicate that a
 Final observations 267

construction is transitive. Although the ergative case marking fulfils this twofold
function of indicating A and indicating the transitivity of the verb, there is
space for the marking of O to have a different function. The marking of O was
previously considered as the marking of the accusative case, but in my opinion,
O-marking functions on an entirely different level and rather indicates semantic
factors related to the referential hierarchies, such as animacy, definiteness, and
topicality. Therefore, the analysis of the marking of O supports the hypothesis
that coding transitivity is crucial in Indo-Aryan, considering that the semantic
features associated with O-marking all indicate a high level of affectedness and
individuation of the O-argument.
Further research regarding these issues remains to be conducted. This study
was designed to be a comparative study of ergativity in the main Indo-Aryan
languages, but the variation in alignment in Indo-Aryan is expected to be even
greater than the variation that has emerged from the present study. Many smaller
Indo-Aryan languages, which have not yet been fully described and analyzed,
display entirely different patterns. Although we are fortunate that various schol-
ars have embarked on the study of several of these endangered languages in
recent years, we have only begun to scratch the surface.
8 Appendix

8.1 Appendix 1

Language Verb form Ref O Ref A/S Case marking Agreement

1 Hindi perfective itr itr

2 Hindi imperfective itr itr
3 Hindi perfective high high S≠A≠O S
4 Hindi perfective high low S≠A≠O S
5 Hindi imperfective high high S=A≠O A,S
6 Hindi imperfective high low S=A≠O A,S
7 Hindi perfective low high S=O≠A S,O
8 Hindi perfective low low S=O≠A S,O
9 Hindi imperfective low high S=A=O A,S
10 Hindi imperfective low low S=A=O A,S
11 Asamiya perfective itr itr
12 Asamiya imperfective itr itr
13 Asamiya perfective high high S=A≠O A,S
14 Asamiya perfective high low S≠A≠O A,S,O
15 Asamiya imperfective high high S=A≠O A,S
16 Asamiya imperfective high low S≠A≠O A,S
17 Asamiya perfective low high S=A=O A,S
18 Asamiya perfective low low S=O≠A A,S,O
19 Asamiya imperfective low high S=A=O A,S
20 Asamiya imperfective low low S=O≠A A,S
21 Bangla perfective itr itr
22 Bangla imperfective itr itr
23 Bangla perfective high high S=A≠O A,S
24 Bangla perfective high low S=A≠O A,S
25 Bangla imperfective high high S=A≠O A,S
26 Bangla imperfective high low S=A≠O A,S
27 Bangla perfective low high S=A=O A,S
28 Bangla perfective low low S=A=O A,S
 Appendix 1 269

Language Verb form Ref O Ref A/S Case marking Agreement

29 Bangla imperfective low high S=A=O A,S

30 Bangla imperfective low low S=A=O A,S
31 Bhojpuri perfective itr itr
32 Bhojpuri imperfective itr itr
33 Bhojpuri perfective high high S=A≠O A,S
34 Bhojpuri perfective high low S=A≠O A,S,O
35 Bhojpuri imperfective high high S=A≠O A,S
36 Bhojpuri imperfective high low S=A≠O A,S
37 Bhojpuri perfective low high S=A=O A,S
38 Bhojpuri perfective low low S=A=O A,S
39 Bhojpuri imperfective low high S=A=O A,S
40 Bhojpuri imperfective low low S=A=O A,S
41 Maithili perfective itr itr
42 Maithili imperfective itr itr
43 Maithili perfective high high S=A≠O A,S,O
44 Maithili perfective high low S=A≠O A,S,O
45 Maithili imperfective high high S=A≠O A,S,O
46 Maithili imperfective high low S=A≠O A,S,O
47 Maithili perfective low high S=A=O A,S
48 Maithili perfective low low S=A=O A,S
49 Maithili imperfective low high S=A=O A,S
50 Maithili imperfective low low S=A=O A,S
51 Magahi perfective itr itr
52 Magahi imperfective itr itr
53 Magahi perfective high high S=A≠O A,S,O
54 Magahi perfective high low S=A≠O A,S,O
55 Magahi imperfective high high S=A≠O A,S
56 Magahi imperfective high low S=A≠O A,S
57 Magahi perfective low high S=A=O A,S
58 Magahi perfective low low S=A=O A,S
59 Magahi imperfective low high S=A=O A,S
60 Magahi imperfective low low S=A=O A,S
61 Awadhi perfective itr itr
270 Appendix

Language Verb form Ref O Ref A/S Case marking Agreement

62 Awadhi imperfective itr itr

63 Awadhi perfective high high S=A≠O A,S,O
64 Awadhi perfective high low S=A≠O A,S,O
65 Awadhi imperfective high high S=A≠O A,S
66 Awadhi imperfective high low S=A≠O A,S
67 Awadhi perfective low high S=A=O A,S,O
68 Awadhi perfective low low S=A=O A,S,O
69 Awadhi imperfective low high S=A=O A,S
70 Awadhi imperfective low low S=A=O A,S
71 Nepali perfective itr itr
72 Nepali imperfective itr itr
73 Nepali perfective high high S≠A≠O A,S
74 Nepali perfective high low S≠A≠O A,S
75 Nepali imperfective high high S≠A≠O A,S
76 Nepali imperfective high low S=A≠O A,S
77 Nepali perfective low high S=O≠A A,S
78 Nepali perfective low low S=O≠A A,S
79 Nepali imperfective low high S=O≠A A,S
80 Nepali imperfective low low S=A=O A,S
81 Kashmiri perfective itr itr
82 Kashmiri imperfective itr itr
83 Kashmiri perfective high high S=O≠A A,S,O
84 Kashmiri perfective high low S=O≠A S,O
85 Kashmiri imperfective high high S=A≠O/S=A=O A,S,O
86 Kashmiri imperfective high low S=A≠O A,S,O
87 Kashmiri perfective low high S=O≠A A,S,O
88 Kashmiri perfective low low S=O≠A S,O
89 Kashmiri imperfective low high S=A=O A,S,O
90 Kashmiri imperfective low low S=A=O A,S
91 Poguli perfective itr itr
92 Poguli imperfective itr itr
93 Poguli perfective high high S=O≠A A,S,O
94 Poguli perfective high low S=O≠A S,O
 Appendix 1 271

Language Verb form Ref O Ref A/S Case marking Agreement

95 Poguli imperfective high high S=A≠O A,S,O

96 Poguli imperfective high low S=A≠O A,S,O
97 Poguli perfective low high S=O≠A A,S,O
98 Poguli perfective low low S=O≠A S,O
99 Poguli imperfective low high S=A=O A,S,O
100 Poguli imperfective low low S=A=O A,S,O
101 Sindhi perfective itr itr
102 Sindhi imperfective itr itr
103 Sindhi perfective high high S≠A≠O S
104 Sindhi perfective high low S≠A≠O S
105 Sindhi imperfective high high S=A≠O A,S,O
106 Sindhi imperfective high low S=A≠O A,S,O
107 Sindhi perfective low high S=O≠A A,S,O
108 Sindhi perfective low low S=O≠A S,O
109 Sindhi imperfective low high S=A=O A,S
110 Sindhi imperfective low low S=A=O A,S
111 Siraiki perfective itr itr
112 Siraiki imperfective itr itr
113 Siraiki perfective high high S=A≠O A,S,O
114 Siraiki perfective high low S≠A≠O S,O
115 Siraiki imperfective high high S=A≠O A,S,O
116 Siraiki imperfective high low S=A≠O A,S,O
117 Siraiki perfective low high S=A=O A,S,O
118 Siraiki perfective low low S=O≠A S,O
119 Siraiki imperfective low high S=A=O A,S
120 Siraiki imperfective low low S=A=O A,S
121 Marathi perfective itr itr
122 Marathi imperfective itr itr
123 Marathi perfective high high S=A≠O A,S
124 Marathi perfective high low S≠A≠O S
125 Marathi imperfective high high S=A≠O A,S
126 Marathi imperfective high low S=A≠O A,S
127 Marathi perfective low high S=A=O A,S,O
272 Appendix

Language Verb form Ref O Ref A/S Case marking Agreement

128 Marathi perfective low low S=O≠A S,O

129 Marathi imperfective low high S=A=O A,S
130 Marathi imperfective low low S=A=O A,S
131 Shina perfective itr itr
132 Shina imperfective itr itr
133 Shina perfective high high S≠A≠O A,S
134 Shina perfective high low S≠A≠O A,S
135 Shina imperfective high high S≠A≠O A,S
136 Shina imperfective high low S≠A≠O A,S
137 Shina perfective low high S=O≠A A,S
138 Shina perfective low low S=O≠A A,S
139 Shina imperfective low high S=O≠A A,S
140 Shina imperfective low low S=O≠A A,S
141 Marwari perfective itr itr
142 Marwari imperfective itr itr
143 Marwari perfective high high S=A≠O S,O
144 Marwari perfective high low S=A≠O S,O
145 Marwari imperfective high high S=A≠O A,S
146 Marwari imperfective high low S=A≠O A,S
147 Marwari perfective low high S=A=O S,O
148 Marwari perfective low low S=A=O S,O
149 Marwari imperfective low high S=A=O A,S
150 Marwari imperfective low low S=A=O A,S
151 Harauti perfective itr itr
152 Harauti imperfective itr itr
153 Harauti perfective high high S=O≠A S,O
154 Harauti perfective high low S=O≠A S,O
155 Harauti imperfective high high S=A≠O A,S
156 Harauti imperfective high low S=A≠O A,S
157 Harauti perfective low high S=O≠A S,O
158 Harauti perfective low low S=O≠A S,O
159 Harauti imperfective low high S=A=O A,S
160 Harauti imperfective low low S=A=O A,S
 Appendix 1 273

Language Verb form Ref O Ref A/S Case marking Agreement

161 Gujarati perfective itr itr

162 Gujarati imperfective itr itr
163 Gujarati perfective high high S≠A≠O S,O
164 Gujarati perfective high low S≠A≠O S,O
165 Gujarati imperfective high high S=A≠O A,S
166 Gujarati imperfective high low S=A≠O A,S
167 Gujarati perfective low high S=O≠A S,O
168 Gujarati perfective low low S=O≠A S,O
169 Gujarati imperfective low high S=A=O A,S
170 Gujarati imperfective low low S=A=O A,S
171 Punjabi perfective itr itr
172 Punjabi imperfective itr itr
173 Punjabi perfective high high S=A≠O S
174 Punjabi perfective high low S≠A≠O S
175 Punjabi imperfective high high S=A≠O A,S
176 Punjabi imperfective high low S=A≠O A,S
177 Punjabi perfective low high S=A=O S,O
178 Punjabi perfective low low S=O≠A S,O
179 Punjabi imperfective low high S=A=O A,S
180 Punjabi imperfective low low S=A=O A,S
181 Braj perfective itr itr
182 Braj imperfective itr itr
183 Braj perfective high high S≠A≠O S,O
184 Braj perfective high low S≠A≠O S,O
185 Braj imperfective high high S=A≠O A,S
186 Braj imperfective high low S=A≠O A,S
187 Braj perfective low high S=O≠A S,O
188 Braj perfective low low S=O≠A S,O
189 Braj imperfective low high S=A=O A,S
190 Braj imperfective low low S=A=O A,S
191 Bundeli perfective itr itr
192 Bundeli imperfective itr itr
193 Bundeli perfective high high S≠A≠O S
274 Appendix

Language Verb form Ref O Ref A/S Case marking Agreement

194 Bundeli perfective high low S≠A≠O S

195 Bundeli imperfective high high S=A≠O A,S
196 Bundeli imperfective high low S=A≠O A,S
197 Bundeli perfective low high S=O≠A S,O
198 Bundeli perfective low low S=O≠A S,O
199 Bundeli imperfective low high S=A=O A,S
200 Bundeli imperfective low low S=A=O A,S

8.2 Appendix 2

The following examples are illustrations of each abstract construction from

Appendix 1. The numbering of the sentences parallels the constructions in Appen-
dix 1, i.e., example (1) is an illustration of an intransitive construction in Hindi in
a perfective form, etc.. The examples taken from the running text are indicated
with their reference number to the text; the source of not previously mentioned
examples is explicitly mentioned. For the constructions with the value of Ref-O or
Ref-A as “high”, the preferred examples were constructions with a first or second
person. However, in instances when there is no difference in coding between pro-
nouns and animate arguments, a third person pronoun or an animate argument
is sometimes used instead. Similarly, A’s with a low referentiality value are often
animate arguments, since inanimate A-arguments are not very common. Prefer-
ence is given to a finite verb form; sometimes, however, an infinite construction
is included. When no example is provided, I refer to examples with an identi-
cal construction in the same language. Both the table in Appendix 1 and the fol-
lowing sentences should only be consulted in combination with the information
found in the running text.

(1) maiṃ so ga-yā+h-ūṃ (26)

I.m sleep go-pst.m.sg+aux.prs-1sg
‘I have gone to sleep.’

(2) maiṃ so jā-ūṃ-g-ī (22)

I.f sleep go-1sg-fut-f
‘I will go to sleep.’
 Appendix 2 275

(3) maiṃ=ne us=ko dekh-ā

I.f=erg she.obl=obj see-pst.m.sg
‘I saw her.’

(4) rām=ne mujh-e pahle dekh-ā+th-ā

Ram=erg I-obl.obj before see-pst.m.sg+aux.pst-m.sg
‘Ram had seen me before.’

(5) maiṃ us=ko dekh-tā+h-ūṃ

I.m she.obl=obj see-prs.m.sg+aux.prs-1sg
‘I see her.’

(6) rām mujh-e dekh+rahā+h-ai

Ram I-obl.obj see+prog+aux-3sg
‘Ram is watching me.’

(7) maiṃ=ne us-e tumhār-ī gand-ī kamiz-eṃ

I=erg he.obl-obj your-f dirty-f shirt[f]-nom.pl
de+dīṃ (1)
give+give.pst.f.pl
‘I gave him your dirty shirts back.’

(8) rām=ne wah ghar banā-yā (31)

Ram=erg that house[m] make-pst.m.sg
‘Ram constructed that house.’

(9) tum kitāb rām=ko kyoṃ de-tā+h-o? (23)

you.m book[f] ram=obj why give-prs+aux.prs-2pl
‘Why are you giving Ram the book?’

(10) rām kitāb paṛh-tā+th-ā (59)

Ram book[f] read-prs+aux.pst-m.sg
‘Ram was reading a book.’

(11) tekhet awase’ sesa-lai na-r-’l (112)

she of course end=to not-wait-pst.3
‘She, of course, did not wait for the end [to come].’
276 Appendix

(12) āmi jagannāth mandira-t somā-m (87)

we Jagannath temple-loc go-fut.1
‘All of us will go into the Jagannath temple.’

(13) maī tā-k kaṛā najara-t rākh-il-oṃ+heṃten

I he-obj strict watch-loc keep-pst-1+sbjv
‘I would have kept him in strict watch.’ (Baruah 1980: 922)

(14) māhīmāk-e tejīmalā-k biyā-lai pathā-l-e (114)

stepmother-erg Tejimala-obj marriage-to send-pst-3tr
‘The stepmother sent Tejimala to the marriage.’

(15) mai to-k bajāb pār-im

I you-obj save can-fut.1
‘I could save you.’ (Baruah 1980: 1060)

(16) mor la’rā-i mo-k sahāy+kar-ib

my boy-erg I-obj help+do-fut.3
‘My son will help me.’ (Kakoti 2011: 7)

(17) kāli āpuni cakā meja-bor ān-il-e ne? (72)

yesterday you.h chair table-pl bring-pst-2h Q
‘Did you bring chairs and tables yesterday?’

(18) kitāb-khan rām-e likh-ich-e (3.1.1)

book-def Ram-erg write-prf-3tr
‘Ram has written the book.’

(19) taï etiyā inginiyāriḥ na-paṛ-a ne? (97)

you now engineering not-do-prs.2 q
‘Don’t you do engineering now?’

(20) kakāyer-e okālati na-kar-e na?

elder brother=erg law not-do-prs.3 no
Doesn’t your elder brother practice law? (Baruah 1980: 156)

(21) biṛālanī gel

cat go.pst.3
‘The cat went.’ (Radice 1994: 221)
 Appendix 2 277

(22) tara ki ashbe?

they Q come.fut.3
‘Are they coming?’ (Nasrin and van der Wurff 2009: 47)

(23) tini ama-ke jete dilen na

he I-obj go give.pst.3 not
‘He didn’t allow me to go.’ (Nasrin and van der Wurff 2009: 165)

(24) ora amader-ke cear-er=songe doṛi=diye beṃdhe+fele-ch-il-o.

they we.obl-obj chair-gen=with rope=with bind+throw-prf-pst-3
‘They had bound us with a rope to a chair.’ (Nasrin and van der Wurff
2009: 153)

(25) ami ona-ke cin-i

I she-obj know-prs.1
‘I know her.’ (Nasrin and van der Wurff 2009: 50)

(26) amader kaki amader-ke khub bhalobash-en

we.obl aunt we.obl-obj much love-prs.3
‘Our aunt loves us a lot.’ (Nasrin and van der Wurff 2009: 44)

(27) ami-o saṛi pore-ch-i

I-emph saree wear-prf-1
‘I have put a sari on.’ (Nasrin and van der Wurff 2009: 199)

(28) pāk bāhinī rājaśāhī śahar-er puro kantrol niye+nil

Pak army Rajshahi town-gen full control take+take.pst.3
‘The Pak military took full control of Rajshahi town.’ (Radice 1994: 251)

(29) apni ki onek praktis koren?

you how much practice do.prs.3
‘How much do you practice?’ (Nasrin and van der Wurff 2009: 73)

(30) sangbadrikra amar obostha bojh-e na

journalist my situation understand-prs.3 not
‘The journalists don’t understand my situation.’ (Nasrin and van der Wurff
2009: 70)
278 Appendix

(31) laṛikā ā+gail

boy come+go.pst.3
‘The boy has arrived.’ (Verma 2003: 528)

(32) tu sutat+bāṛa
you sleep.prs+aux.prs.2
‘You are sleeping.’ (Verma 2003: 532)

(33) tu i phul=ke gulāb bujha-la

you this flower=obj rose think-pst.2
‘You thought this flower (definite object) a rose.’ (Verma 2003: 534)

(34) u hamrā=(ke) dekhue

he we.obl=obj see.pst.3
‘He saw us.’ (Tiwari 1960: 170)

(35) ham torā laṛki=ke ṭhik=se dekh-ab

I your daughter=obj well=with see-fut.1
‘I will look after your daughter well.’ (Verma 2003: 533)

(36) karikhahavā mohi mārat+bā

black-faced I.obl (poetry) beat.prs+aux.prs.3
‘The black-faced one beats me.’ (Grierson 1885: 136)

(37) u admī dekh-al-as-i

he man see-pst-tr-3
‘He saw men.’ (Tiwari 1960: 161)

(38) u phul mahak-al-an

he flower smell-pst-3
‘He smelled the flowers.’ (Verma 2003: 534)

(39) ham ām-o khāeb

I mango-incl eat.fut.1
‘I will eat mangoes too.’ (Verma 2003: 528)

(40) isar sajāi de-su

god punishment give-sbjv.3
‘Let god give punishment.’ (Tiwari 1960: 185)
 Appendix 2 279

(41) ram ge-l+ch-aith (137)

Ram go-pst.itr+aux.prs-3h
‘Ram has gone.’

(42) ram jait+ch-aith

Ram go.prs+aux.prs-3h
‘Ram goes.’ (Yadav 2003: 487)

(43) ham to=rā kaniya=keṃ dekh-au-l-iaunh (140)

I you=obj bride=obj show-caus-pst-1.2nh.3h
‘I showed you the bride.’

(44) majisṭar sāheb ham=rā dekh-al-ainh

Magistrate sir I=obj see-pst-3h.1
‘His Honor, the Magistrate, saw me.’ (Southern Maithili)
(Grierson 1885: 36)

(45) ham to=rā choṛi+deb

I you=obj go+give.fut.1.2h
‘I will let you go.’ (Northern Maithili) (Grierson 1885: 37)

(46) ohinā ham=rā sabha=keṃ bhaṛkabaiachi.

they I=obj all=obj quarrel.prs.3.1
‘They make us quarrel.’ (Northern Maithili) (Grierson 1885: 34)

(47) toṃ ok=ra kitab de-l-hak (136)

you he=obj book give-pst-2mh.3nh
‘You gave him a book.’

(48) ram khai-ne+ch-aith

Ram eat-pst.tr+aux.prs-3h
‘Ram has eaten.’ (Yadav 2003: 487)

(49) ham akhbar paiṛh+rahal+chi

I newspaper read+prog+aux.prs.1
‘I am reading a newspaper.’ (Yadav 2003: 483)

(50) hari-ji bhajan gab-ait+ch-aith

hari-h song sing-prs+aux.prs-3h
‘Hari is singing a song.’ (Bickel et al. 1999: 496)
280 Appendix

(51) ham sut-l-i+he (130)

I sleep-ptcp-1+aux.prs
‘I have slept.’

(52) ham ghar-e jā+hi

I house-loc go+aux.prs
‘I am going home.’ (Verma S. 2003: 508)

(53) ham okrā dekh-al-iai (141)

I he.nh.obj see-pst-1.3nh
‘I saw him.’

(54) bhagavān ham=rā daliddhar kai-lk-an

god I=obj beggar make-pst-3h
‘The Giver has made me a beggar.’ (Gaya) (Grierson 1885: 81)

(55) ham-hūṃ apanā laṛikā=ke skūl=me bheje+lā+cāha+hī

I-emph my boy=obj school=to send+bring+want+aux.prs
‘I also want to send my boy to school.’ (South Patna) (Grierson 1885: 80)

(56) aur ham=rā to=rā laṛāva+hath

and we=obj you=obj quarrel.caus+aux.prs.3
‘[They] make you and us quarrel.’ (Gaya) (Grierson 1885: 82)

(57) ham okar intajār kai-l-iai

I his waiting make-pst-1.3nh
‘I waited for him.’ (Verma 1991: 140)

(58) jādugarwā rājā=ke nokar banā+de-lk-ai

magician king=obj servant make+give-pst-3
‘The magician made the servant to a king.’ (Verma 1991: 139)

(59) toṃ uṭhamāṃ jāe na paib

you there go not can.fut.2
‘You cannot go there.’ (Grierson 1885: 77)

(60) rām-jī sab=ke hisāb kara+hathin

Ram-h all=obj cognizance do+aux.prs.3h
‘Ram takes cognizance of all.’ (Gaya) (Grierson 1885: 81)
 Appendix 2 281

(61) jale ui lauṭa-īṃ (144)

when she.pl return-pst.f.pl
‘When she returned.’

(62) gopal=ke cāriu larikā jia-ti+haiṃ

Gopal=gen four boy live-prs+aux.prs.3pl
‘Gopal’s four sons are living.’ (Saksena [1937] 1971: 220)

(63) ū hamai mār-isi

he I.obl beat-pst.tr.3sg
‘He beat me.’ (Saksena [1937] 1971: 162)

(64) ui=ke baraijā kāṭi+khāini

he=obj wasp.pl sting+eat.pst.tr.3pl
‘The wasps stung him.’ (Saksena [1937] 1971: 442)

(65) tui=kā mai tīni pāni=se na dhoi+deuṃ

you=obj I three water=with not wash+give.fut.1sg
‘I will not wash you away with three waters.’ (Saksena [1937] 1971: 432)

(66) id. 65

(67) kā ī=ma ke gulgul-ā tum lih-e+hau? (143)

q this=from some sweet-pl you.f take-pst.tr+aux.prs.2pl
‘Have you taken some sweets from this?’

(68) bahū tale tīni gulgul-ā nikāri lih-isi (144)

daughter-in-law then three sweet-pl take out.inf take-pst.tr.3sg
‘The daughter-in-law took out three sweets.’

(69) mai miṭhāī khā-ti+hoihauṃ

I sweets eat-prs+aux.fut.1sg
‘I shall be eating sweetmeats.’ (Saksena [1937] 1971: 265)

(70) sāheb log tau pāṃc-pāṃc chaāhā dāuṃ ṭipan

sahib people then five-five six time tiffin
uḍāwa-ti+haiṃ
indulge-prs+aux.prs.3pl
‘While the sahebs indulge in tiffin meals five or six times.’ (Saksena [1937]
1971: 154)
282 Appendix

(71) poi=ko man=mā pani kehī ciso pas-na ga-yo (153)

husband=gen mind=loc too little chill enter-inf go-pst.3sg
‘A little chill entered the husband’s mind too.’

(72) tinī-harū āṃkhā band gar-era bas-dai-nan (187)

he-pl eye close do-cvb sit-prs-neg.3pl
‘They don’t sit with their eyes closed [anymore].’

(73) mai=le Prembahādur=lāi dekh-eṃ

I=erg Prembahadur=obj see-pst.1sg
‘I saw Prembahadur.’ (Bickel 2011: 404)

(74) baigun=ko kāṃḍā=le tyas=lāī ghoc-yo (148)

aubergine=gen thorn[f]=erg she=obj prick-pst.3sg
‘A thorn of the aubergine-plant pricked her.’

(75) kina hāmī=lāī kasai(=le) samjh-os? (164)

why we=obj someone=erg remember-sbjv.3sg
‘Why would anyone remember us?’

(76) svara-harū(=le) mānauṃ us=lāī bolā-i-rah-ekā+ch-an (169)

voice-pl=erg as if he=obj call-lnk-prog-prf.ptcp.pl+prs-3pl
‘[it was] as if voices were calling him.’

(77) lathi=le maiṃ=le sisi phuṭa-eṃ (152)

stick=ins I=erg bottle break-pst.1sg
‘I broke the bottle with the stick.’

(78) rājā=le mukh bharī pāṇī hāl-e (147)

king=erg mouth full water put-pst.3sg.h
‘The king put his mouth full [with] water.’

(79) tarsā-eko ra na-tarsā-eko hajūr(=le) thāhā

terrify-prf.ptcp or neg-terrify-prf.ptcp you.h=erg knowledge
pā-i-bakse-lā! (167)
get-lnk-royal.h-fut.3sg
‘Terrified or not terrified, you will find out.’
 Appendix 2 283

(80) tapāīṃ=ko hajām(=le) mero ghāu kina

you.h=gen barber=erg my wound why
kāṭ-i-diṃ-dai-na (161)
stitch-lnk-give-prs-neg.3sg
‘Why doesn’t your barber stitch my wound?

(81) su gav vany bāzar (214)

he.nom go.pst.m.sg now market
‘He just went to the market.’

(82) toh’ gatsh-iv dili

you.nom. go-fut.2pl Delhi
‘You will go to Delhi.’ (Koul and Wali 2006: 17)

(83) tahi on-iv-on su yo:r (230)

you.erg bring.pst.m.sg-2pl-3sg he.nom here
‘You brought him here.’

(84) temi on-u-s bi (220)

she.erg bring.pst-m-1sg I.nom
‘She brought me.’

(85) tsi ch-u-kh me parināv-ān (200)

you.nom aux.prs-m-2sg I.obj teach-ptcp.prs
‘You teach me.’

bi so:za-th tsi to:r (199)

I.nom send.fut.1sg-2sg you.nom there
‘I will send you there.’

(86) su parinā-vi me
he.nom teach-fut.3sg I.obj
‘He will teach me.’ (Koul and Wali 2006: 104)

(87) tahi por-v-ā akhbār? (73)

you.erg read.pst.m.sg-2pl-q newspaper[m]
‘Did you read the newspaper?’
284 Appendix

(88) ḍakṭar-an dyut zavar zi project

doctor-erg give.pst.m.sg emphasis[m] that project
gatsh-i mukmal sapd-un (216)
need-fut.3sg complete make-inf
‘The doctor emphasized that the project should be completed.’

(89) bi ch-u-s kitāb par-ān (202)

I.m.nom aux.prs-m-1sg book[f] read-ptcp.prs
‘I am reading a book.’

(90) m’on dos ch-u ārām-i=sān kām kar-ān

my friend aux.prs-3m.sg rest-abl=with work do.ptcp.prs
‘My friend works leisurely.’ (Koul and Wali 2006: 44)

(91) āu bi go-s
I.nom fear go.pst-1sg
‘I was afraid.’ (Hook 1987: 64)

(92) yaw laal korah warha dhaw+cha

these ruby where from going+aux.prs.3pl
‘Where are these rubies (going) from?’ (Hook: http://www-personal.
umich.edu/~pehook/poguli.html - Accessed on 10 May 2012)

(93) yīs ti āu teri pent-i-s

yesterday you.erg I.f.nom there send.pst-f.sg-1sg
‘Yesterday, you sent me there.’ (Hook: http://www-personal.umich.
edu/~pehook/poguli.html - Accessed on 10 May 2012)

(94) teni āu ān-t-us-nan-e

he.erg I.nom bring-pst-1pl-3sg-prf
‘He has brought me for them.’ (Hook 1987: 67)

(95) kinaer ān-ma-n

how bring-fut.1sg-3sg
‘How shall we bring her?’ (Hook: http://www-personal.umich.edu/~pehook/
poguli.html – Accessed on 10 May 2012)

(96) sā ti teru peni +lev-i-(t)

she.nom you.obj there send+pour-fut.3sg-(2sg)
‘She will send you there.’ (Hook 1987: 66)
 Appendix 2 285

(97) āu bi go-s ti tes ciṭhi daī+le

I.nom fear go.pst-1sg you.erg he.obj letter give+pour.pst.2sg
‘I was afraid that you might have given him the letter.’ (Hook 1987: 64)

(98) cīn=sun rāza dweb bāpor-us pyon-c-in

China=gen king.erg woman merchant-obj send-pst-3sg
‘The king of China sent the woman to the merchant.’ (Hook: http://www-
personal.umich.edu/~pehook/poguli.html – Accessed on 10 May 2012)

(99) as yu sweecker bāpor tyurhu pyon-ma-n

we it thinking merchant there send-fut.1pl-3sg
‘With this in mind we will send the merchant there.’ (Hook: http://www-
personal.umich.edu/~pehook/poguli.html – Accessed on 10 May 2012)

(100) bāpor lāl cīn=sani rāz-as wātal+chu

merchant ruby China=gen king-obj arrive.caus+aux.prs.3sg
‘The merchant delivers the rubies to the king of China.’ (Hook: http://www-
personal.umich.edu/~pehook/poguli.html – Accessed on 10 May 2012)

(101) a-y-ūsī
come-pst-1sg
‘I came.’ (Khubchandani 2003: 644)

(102) tūṃ hal-eṃ

you go-prs.2sg
‘You go.’ (Cole 2001)

(103) mūṃ huna=khe āndo

I.erg he.obl=obj bring.pst.m.sg
‘I brought him.’ (Cole 2001)

(104) id. 103

(105) maṃ ḍ’īs-aṃ-ī+tho

I look.prs-1sg-2sg+aux.prs.m.sg
‘I look at you.’ (Khubchandani 1962: 77)

(106) id. 105

286 Appendix

(107) muṃ kītab-a paṛh-y-a/ paṛh-iu-va

I.erg book[m]-pl read-pst-m.pl read-pst-1sg
‘I read books.’ (Khubchandani 2003: 647)

(108) chokria kuto di-ṭho

girl[f]erg dog[m] see-pst.m.sg
‘The girl saw the dog.’ (Cole 2001)

(109) tūṃ kitāb-a ḍīndo+hue-ni (243)

you. book-[m]-pl give.prs+aux.prs.2sg-3pl
‘You used to give the books to them.’

(110) hu chokīr-a amba khai-nī+tha

those boy-pl mango.pl eat-3pl+aux.prs.m.pl
‘Those boys are eating mangoes.’ (Khubchandani 2003: 651)

(111) o gea
he go.pst.m.sg
‘He went.’ (Shackle 1976: 144)

(112) hik aorat bai=de ghar vai-si

one woman she=gen house go-fut.3sg
‘One woman will go to another’s house.’ (Shackle 1976: 168)

(113) sumbal=nū mariā-s

sumbal=obj beat.pst-3sg
‘S/he beat Sumbal.’ (Akhtar 1997: 6)

(114) fawad=ne mariā-s

fawad=erg beat.pst-3sg
‘Fawad beat him.’ (Akhtar 1997: 6)

(115) o mai-kuṃ pāgal pyā baṇedai

he I-obj mad emph make.prs.3sg
‘He is making me mad.’ (Shackle 1976: 146)

(116) fawad mare-ga-s

Fawad beat-fut-3sg
‘Fawad will beat him.’ (Akhtar 1997 :8)
 Appendix 2 287

(117) kam ki-tā-s

work do-pst.m.sg-3sg
‘S/he did the work.’ (Akhtar 1997: 7)

(118) fawad=ne sumbal=nū katāb di-tī

Fawad=erg Sumbal=obj book[f] give-pst.f.sg
‘Fawad gave the book to Sumbal.’ (Akhtar 1997: 2)

(119) asaṃ amb kha-suṃ

we mango eat-fut.1pl
‘We shall eat mangoes.’ (Shackle 1976: 144)

(120) muṛ-e kam kar-d-e-s (246)

boy-nom.pl work do-prs-3pl-3sg
‘His boys work.’

(121) mādzhī badlī dzhālī

my transfer[f] happen.pst.f.sg
‘My transfer happened.’ (Pandharipande 1997: 5)

(122) tu ghari jāc-t-e-s (248)

you.f home go-prs-f.sg-2sg
‘You are going home.’

(123) mī bāī=lā pāhila

I woman=obj see.pst.3sg.n
‘I saw the woman.’ (Pandharipande 1997: 135)

(124) presiḍenṭ=ne tyā=lā sekreṭarī tsyā padāwar nem-le

president=erg he=obj secretary his position appoint-pst.3sg.n.
‘The president appointed him in the position of secretary.’ (Pandhari-
pande 1997: 323)

(125) tumhī ma=lā sāmbhāḷāwe¹

you I.obl=obj take care.opt.3sg.n
‘You may protect me.’ (Pandharipande 1997: 45)

1 Because the mood is optative, there is no agreement with A.

288 Appendix

(126) tyātsa na bolṇa ma=lā botsta

his not talk[n]inf I.obl=obj hurt.prs.3sg.n
‘His not talking hurts me.’ (Pandharipande 1997: 189)

(127) tu kāvitā vāc-l-i-s (247)

you.m poem[f] read-pst-f.sg-2sg
‘You read the poem.’

(128) tyā=ne don āmbe khālle

he=erg two mango[m]pl eat.pst.3pl.m
‘He ate two mangoes.’ (Pandharipande 1997: 15)

(129) tī hindī śikel

she Hindi learn.fut.3sg
‘She will learn Hindi.’ (Pandharipande 1997: 2)

(130) mohan āmbe khā-to

Mohan mango[m]pl eat-prs.3sg.m
‘Mohan eats mangoes.’ (Pandharipande 1997: 12)

(131) ātho mosfar ā-l-o

this way traveller.nom come-prf-3m.sg
‘A traveller came from this way.’ (Schmidt and Kohistani 2008: 69)

(132) ye toomo bubā-di boj-tha

he his father-to go-inj
‘He should go to his own father.’ (Schmidt and Kohistani 2008: 85)

(133) moṃ se-s gooz-a uch-y-ā-s

I.erg he-obj home-obl take-caus-prf.m.sg-1
‘I took him home.’ (Schmidt and Kohistani 2008: 82)

(134) ich-i se-ṇo gooz-a uch-y-ā-o (198)

bear-erg he-obj.pl home-obl take-caus-prf.m.sg-3
‘The bear took them home.’

(135) ciyee mo ye-s kudi cak-ū-s?

now I.erg he-obj where look-fut-1sg
‘Now where should I look for him?’ (Schmidt and Kohistani 2008: 78)
 Appendix 2 289

(136) id. 134

(137) ye-s-i so ṭak har-yā-o

he-obl-erg that button take away-prf.m.sg-3
‘He took away the button.’ (Schmidt and Kohistani 2008: 76)

(138) iman toolo-ga dodook-i aṭ-eeg-a-an

faith toolo-and dodooko-erg bring-prf-3m.pl-aux.prs
‘Toolo and Dodooko have brought the faith.’ (Schmidt and Kohistani
2008: 50)

(139) be-s jo th-ooṇ-as kapar-y-o-gi? (197)

we-erg what do-1pl-aux.prs potsherd-pl-obj-with
‘What are we to do with potsherds?’

(140) cayāṃiṃ-s cīṃ=cīṃ th-ī-n-i

sparrow[f]-erg chirping do-impf-aux.prs-3f.sg
‘The sparrow is chirping.’ (Schmidt and Kohistani 2008: 52)

(141) mhaiṃ bego hī vāṃ=rai=kanaiṃ jā-y pūg-yo (257)

I fast emph he=gen=close go-cvb arrive-pst.m.sg
‘I got to him in a hurry.’

(142) rojīnā ghare bhī to jāvo+ho

daily house.loc too you go.ptcp.prs+aux.prs.m.sg
‘Do you still go daily to the house?’ (Hamsa 1984: 172)

(143) mhaiṃ śaraṇ=naiṃ dekh-ī (253)

I.m Sharan[f]=obj see-pst.f.sg
‘I saw Sharan.’

(144) mammī ma=naiṃ god=meṃ uṭhā+li-yo (252)

Mama[f] I=m.obj lap=loc heave+take-pst.m.sg
‘Mammy took me up in her lap.’

(145) vai mha=naiṃ cor-nijar=sūṃ dekhai+hā

he I.m=obj thief-look=with see.prs+aux.pst.m.sg
‘He was looking at me with the look of a thief.’ (Hamsa 1984: 170)
290 Appendix

(146) ek aṇajāṇī-sī cuppī mha=naiṃ gher+lī+hī

a unknown-like silence I=obj spread+take+aux.pst.f.sg
‘An unknown silence was spreading over me.’ (Hamsa 1984: 173)

(147) vai baṛā het=sūṃ mhārī āḍī sirkā-ī (266)

he great care=with I.gen problem[f] solve-pst.f.sg
‘He solved my problems with great care.’

(148) paṇ pāpā bhī merī bāt konī suṇ-ī (263)

but pa also I.gen word[f] not hear-pst.f.sg
‘But papa also did not listen to my words.’

(149) āp kyūṃ taklīph dekh-o

you why trouble see-prs.2sg
‘Why do you take the trouble?’ (Hamsa 1984: 171)

(150) pāpā ma=naiṃ pataṃg nīṃ lyā’r+de-tā

papa I=obj kite not bring+give-prs.m.sg
‘Papa wouldn’t bring me a kite.’ (Barhat 1984: 39)

(151) paṇ moṭar nhaṃ ḍhab-ī (271)

but bus[f] not stop-pst.f.sg
‘But the bus did not stop.’

(152) ghaṃṭā=lār ek moṭar aṭhī=sūṃ ar ek uṭhī=sūṃ

hour=by one bus here=from and one there=from
chūṭai+chai
leave.prs+aux.prs.3sg
‘Each hour, one bus leaves from here, and one from there.’ (Prem 1984: 91)

(153) u uṃ=ne dekhi+chai

she he.obl=obj see.pst.f+aux.prs.3sg
‘He has seen her.’ (Allen 1960: 10)

(154) ḍokri=ne ya-i kari

old woman=erg this-f.sg do.pst.f
‘The old woman did this [thing, f.].’ (Allen 1960: 11)
 Appendix 2 291

(155) u uṃ=ne dekhai+chai

he he=obj see.prs+aux.prs.3sg
‘He sees him.’ (Allen 1960: 10)

(156) khābā=kī cīj=naiṃ kuṇ choḍai+chai? (280)

eat=gen thing=naiṃ who.nom.m.sg lose.prs+aux.prs.3sg
‘Who loses a thing to eat (definite object)?’

(157) mhaiṃ=naiṃ hāṃk-ā pāṛ-yā (271)

I.f=erg scream[m]-pl scream-pst.m.pl
‘I screamed.’

(158) ek mard=naiṃ ḍalevar=pai hukam calā+dyo,

one man=erg driver=on order[m] go+give.pst.m.sg
ḍalevar sāb! (276)
driver sir
‘A man ordered the driver: driver sir!’

(159) muṃ chora=ne photi duṃgo

I boy=obj book[f] give.fut.1sg
‘I will give the boy a book.’ (Allen 1964: 343)

(160) choro kam kare+che

boy[m] work do+aux.prs.3sg
‘The boy does the work.’ (Allen 1964: 343)

(161) te kāl paho’nch-yo

he yesterday reach-pst.m.sg
‘He reached yesterday.’ (Ganathe 2005: 88)

(162) huṃ niśāle ja-to+nathī

I school go-prs+neg.aux
‘I do not go to school.’ (Ganathe 2005: 77)

(163) me ehṇe kale jo-yo

I.erg he.obj yesterday see-pst.m.sg
‘I saw him yesterday.’ (Cardona 1965: 75)
292 Appendix

(164) kiśor-e mag=ne baraḍ-yā (289)

Kishor-erg mung beans[m]=obj grind-pst.m.pl
‘Kishor ground the mung beans (definite object).’

(165) te tam=ne bolāve+che

he you=obj call+aux.prs.3sg
‘He is calling you.’ (Ganathe 2005: 105)

(166) kiśor rāj=ne pajav-śe

Kishor Raj=obj harass-fut.3sg
‘Kishor will harass Raj.’ (Mistry 1997: 426)

(167) me kam kar-yuṃ+che (255)

I.erg work[n] do-pst.n.sg+aux.prs.3sg
‘I have done the work.’

(168) ramesh-e traṇ capḍ-i kharid-i

Ramesh-erg three book[f]-pl buy-pst.f.sg
‘Ramesh bought three books.’ (Cardona 2003: 670)

(169) huṃ paysa ap-uṃ

I money give-prs.1sg
‘I will pay.’ (Cardona 2003: 680)

(170) ramesh shuṃ kam kare+che

Ramesh what work do+aux.prs.3sg
‘What does Ramesh do?’ (Cardona 2003: 680)

(171) maiṃ baiṭh-iā+hāṃ

I sit-pst.m.sg+aux.prs.1sg
‘I sat.’ (Tolstaya 1981: 44)

(172) maiṃ jāṃ-dā+hāṃ

I go-prs+aux.prs.1sg
‘I go.’ (Gill 21)

(173) maiṃ tarī bhaiṃ=nūṃ viāh toṃ pahilāṃ vekhi-a

I your sister=obj marriage then before see-pst.m.sg
‘I saw your sister before the marriage.’ (Tolstaya 1981: 67)
 Appendix 2 293

(174) us=ne maiṃ=nūṃ āpṇā dost baṇā-iā

he=erg I=obj his friend make-pst.m.sg
‘He made me his friend.’ (Tolstaya 1981: 50)

(175) uh maiṃ=nūṃ vekh-dā+haiṃ

he I=obj see-prs+aux.prs.3sg
‘He sees me.’ (Tolstaya 1981: 68)

(176) rāj rattan=nūṃ apṇā vairī man-dā+hai

Raj Rattan=obj his enemy consider-prs+aux.prs.3sg
‘Raj considers Ratan his enemy.’ (Bhatia 1993: 134)

(177) maiṃ kamm kītā (293)

I work[m] do.pst.m.sg
‘I worked.’, lit.: ‘I did the work.’

(178) kuṛī=ne kamm kītā (294)

girl[f]=erg work[m] do.pst.m.sg
‘The girl worked.’, lit.: ‘the girl did the work.’

(179) maiṃ cāh nahīṃ pīṃ-da

I tea not take-prs.m.sg
‘I do not take tea.’ (Gill 22)

(180) doveṃ tiaṃ roṭi tīar kar+raiaṃ+san

both daughter[f]pl food ready do+prog.f.pl+aux.pst.3pl
‘The two daughters were preparing the meal.’ (Shackle 2003: 611)

(181) tumhāre caccā gāṃo ga-e

your uncle village go-pst.m.pl
‘Your uncle has gone to the village.’ (Varma 1935: 76)

(182) tum kahāṃ jāl+hao?

you where go+aux.prs.2sg
‘Where are you going?’ (Varma 1935: 75)

(183) gaurī=naiṃ syām=kūṃ dekh-yau

Gauri=erg Syam=obj see-pst.m.sg
‘Gauri saw Syam.’ (Liperovsky 2007: 145)
294 Appendix

(184) tū mere mālik=neṃ yādi kī-yau+hai (305)

you I.gen master=erg remembrance did-pst.m.sg+aux.prs
‘My master has recalled thee.’

(185) bao laṛkā=kao naokar=sae calbāut+hae

he boy=obj worker=abl carry.caus+aux.prs.3sg
‘He makes the boy carried by the worker.’ (Varma 1935: 101)

(186) birāmman binneṃ bulāibe āy-au

Brahman they.obj call.inf come-pst.m.sg
‘The Brahman came to call them.’ (Liperovsky 2007: 148)

(187) taiṃ=naiṃ sac kah-yo

you=erg truth tell-pst.m.sg
‘You told the truth.’ (Varma 1935: 74)

(188) coranu sabu mālu-asbābu nikāri+lī-yau (304)

thief.obl.pl all luggage[m] go out+take-pst.m.sg
‘The thieves took out the whole luggage.

(189) bao laoṃṛā=kao ām det+hae

he boy=obj mango give.prs+aux.prs.3sg
‘He gives a mango to the boy.’ (Varma 1935: 94)

(190) jo koī kām kar-ae

who those work do-sbjv.3sg
‘They who would work…’ (Varma 1935: 92)

(191) bo chata=se kūd-o

he roof=from jump-pst.m.sg
‘He jumped from the roof.’ (Jaiswal 1962: 59)

(192) bo āu-ta+hue
he come-prs+aux.sbjv
‘He may be coming.’ (Jaiswal 1962: 141)

(193) ū=ne machariyana=kho khā+ḍāl-o

he=erg fish=obj eat+throw-pst.m.sg
‘He ate the fish (definite object) up.’ (Jaiswal 1962: 66)
 Appendix 2 295

(194) dhobī=ne gadhana=kho khūba pīṭ-o

washerman=erg donkey=obj very beat-pst.m.sg
‘The washerman beat the donkeys bitterly.’ (Jaiswal 1962: 65)

(195) maiṃ ū=kho jo ute baiṭho+hai jāna-ta+hauṃ

I he=obj who here sit+aux.3sg know-prs+aux.prs.1sg
‘I know him who is sitting here.’ (Jaiswal 1962: 78)

(196) ī=kho sab cīna-ta+haiṃ

he=obj all know-prs+aux.prs.3pl
‘Everybody knows him.’ (Jaiswal 1962: 75)

(197) tum=neṃ ek ām khā-o (308)

you=erg one mango[m] eat-pst.m.sg
‘You ate a mango.’

(198) dukāndārana=ne dukāne banda kara+dai

shopkeeper=erg shop[f]pl close do+give.pst.f.pl
‘The shopkeepers have closed the shops.’ (Jaiswal 1962: 65)

(199) ham garmī=kī dupariyā=me kām naiṃ kara-ta

we summer=gen afternoon=in work not do-prs.m.pl
‘We do not work in the hot summer afternoon.’ (Jaiswal 1962: 71)

(200) jamidāra hamesa ḍākuana=kho baṛhāvā

zamindar always dacoit=obj encouragement
de-ta+h-aiṃ
give-prs+aux.prs-3pl
‘Zamindars always give encouragement to the dacoits.’ (Jaiswal 1962: 65)
9 References
Abadie, Peggy. 1974. Nepali as an ergative language. Linguistics of the Tibeto-Burman area 1.
156–177.
Abraham, Werner. 2000. The aspect-case typology correlation: perfectivity triggering (split)
ergativity. Burzio’s generalization explained. In Erik Reuland (ed.), Arguments and case:
explaining Burzio’s generalization, 129–190. Amsterdam: Benjamins.
Acharya, Jayaraj. 1991. A descriptive grammar of Nepali and an analyzed corpus. Washington,
D.C.: Georgetown University Press.
Aissen, Judith. 1999. Markedness and subject choice in optimality theory. Natural language
and linguistic theory 17. 673–711.
Aissen, Judith. 2003. Differential object marking. Iconicity versus economy. Natural language
and linguistic theory 21. 435–483.
Akhtar, Raja N. 1997. Affix -s(uu) constructions in Punjabi. Essex Graduate Working Papers in
Linguistics 1. 1–18.
Allen, William. 1960. Notes on the Rajasthani verb. Indian Linguistics 21. 1–13.
Amritavalli, R. and Partha P. Sarma. 2002. A case distinction between unaccusative and
unergative subjects in Assamese. Snippets 5. 6–7.
Andersen, Paul K. 1990. Review of “Passive and voice” M. Shibatani. Journal of Linguistics
26(1). 189–202.
Anderson, Stephen R. 1976. On the notion of subject in ergative languages. In Charles N. Li
(ed.), Subject and topic, 1–25. New York: Academic press.
Anderson, Stephen R. 1977. On mechanisms by which languages become ergative. In Charles
N. Li (ed.), Mechanisms of syntactic change, 317–363. Austin & London: University of Texas
Press.
Anderson, Stephen R. 1992. A-morphous morphology. Cambridge: Cambridge University Press.
Andrews, Avery D. 2001. Non-canonical A/S marking in Icelandic. In Alexandra Aikhenvald,
Robert M. W. Dixon and Masayuki Onishi (eds.), Non-canonical marking of subjects and
objects, 85–111. Amsterdam: Benjamins.
Bahl, Kali C. 1989. Adhunik rajasthani ka samracanatmak vyakarana [A structural grammar of
Modern Rajasthani]. Jodhpur: Rajasthani Sahitya Samsthan.
Bailey, T. Grahame. 1908. The languages of the northern Himalayas. London: Royal Asiatic
Society.
Bailey, T. Grahame. 1924. Grammar of the Shina language. London: Royal Asiatic Society.
Barđdal, Johanna. 2011. Lexical vs. structural case: a false dichotomy. Morphology 21. 619–654.
Barđdal, Johanna. 2011. The rise of dative substitution in the history of Icelandic: A diachronic
construction grammar account. Lingua 121(1). 60–79.
Barđdal, Johanna and Thorhallur Eythórsson. 2012. “Hungering and lusting for women and
fleshly delicacies”: Reconstructing grammatical relations for Proto-Germanic. Transactions
of the Philological Society 110(3): 363–393.
Barhat. 1984. Masterjī. R. Saraswat and P. Prem (eds.), Aaj ri Rajasthani kahaniyan
[Contemporary Rajasthani stories]. New Delhi: Sahitya Akademi.
Barua, Hemachandra. 1859. Asamiya vyakarana [Assamese grammar]. Sibsagar.
Baruah, P. N. Dutta 1980. An intensive course in Assamese. Mysore: Central institute of Indian
languages.
 References 297

Bashir, Elena 1999. The Urdu and Hindi postposition ne: its changing role in the grammar. In
Rajendra Singh (ed.), The yearbook of South Asian languages and linguistics, 11–36. New
Delhi: Sage Publications.
Bashir, Elena. 2003. Dardic. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan
languages, 818–894. London: Routledge.
Bashir, Elena. 2009. Wakhi. In Gernot Windfuhr (ed.), The Iranian languages, 825–858. London:
Routledge.
Bayer, Joseph. 2004. Non-nominative subjects in comparison. In Peri Bhaskararao and
K. V. Subbarao (eds.), Non-nominative subjects. Vol I, 49–76. Amsterdam: Benjamins.
Beames, John. 1872–1879. A comparative grammar of the Modern Aryan languages of India.
London: Trubner & Co.
Benveniste, Emile. 1952. La construction passive du parfait transitif. Bulletin de la Société
Linguistique de Paris 48. 52–62.
Bhat, D. N. S. 1991. Grammatical relations. The evidence against their necessity and
universality. London: Routledge.
Bhat, Rajesh. 1987. A descriptive study of Kashmiri. Delhi: Amar Prakashan.
Bhatia, Motia. 1985. An intensive course in Punjabi. Mysore: CIIL.
Bhatia, Tej K. 1993. Punjabi. A cognitive-descriptive grammar. London: Routledge.
Bhattacharyya, B. K. 1997. Janajatiya Sadhu: A collection of tribal folktales of the North-East.
New Delhi: Sahitya Akademi.
Bickel, Balthasar. 2003. Referential density in discourse and syntactic typology. Language
79(4). 708–736.
Bickel, Balthasar. 2004. The syntax of experiencers in the Himalayas. In Peri Bhaskararao and
K. V. Subbarao (eds.), Non-nominative subjects. Vol I, 77–112. Amsterdam: Benjamins.
Bickel, Balthasar. 2008. On the scope of referential hierarchy in the typology of grammatical
relations. In Greville G. Corbett and Michael Noonan (eds.), Case and grammatical
relations, 191–210. Amsterdam: Benjamins.
Bickel, Balthasar. 2011. Grammatical relations typology. In Jae J. Song (ed.), The Oxford
handbook of language typology, 399–445. Oxford: Oxford University Press.
Bickel, Balthasar, Walter Bisang and Yogendra P. Yadava. 1999. Face vs. empathy: the social
foundation of Maithili verb agreement. Linguistics 37(3). 481–518.
Bickel, Balthasar and Johanna Nichols. 2009. Case marking and alignment. In Andrej
Malchukov and Andrew Spencer (eds.), The Oxford handbook of case, 304–321. Oxford:
Oxford University Press.
Bickel, Balthasar and Alena Witzlack-Makarevich. 2008. Referential scales and case alignment:
reviewing the typological evidence. In Marc Richards and Andrej Malchukov (eds.), Scales,
1–37. Leipzig: Linguistische Arbeitsberichte.
Bickel, Balthasar and Yogendra P. Yadava. 2000. A fresh look at grammatical relations in
Indo-Aryan. Lingua 110. 343–373.
Blake, Barry J. 2001. Case. Cambridge: Cambridge University Press.
Bokarev, E. A., (ed.). 1950. Ergativnaja konstrukcija predlozenija. Moskva: Izd-vo inostrannoj
literatury.
Bornkessel-Schlesewsky, Ina, Kamal K. Choudhary, Alena Witzlack-Makarevich and Balthasar
Bickel. 2008. Bridging the gap between processing preferences and typological distri-
butions: Initial evidence from the online comprehension of control constructions in Hindi.
In Marc Richards and Andrej Malchukov (eds.), Scales, 397–436. Leipzig: Linguistische
Arbeitsberichte.
298 References

Bossong, Georg. 1983. Animacy and markedness in universal grammar. Glossologia 2–3. 7–20.
Bossong, Georg. 1984. Ergativity in Basque. Linguistics 22. 341–392.
Bossong, Georg. 1985. Empirische Universalienforschung: Differentielle Objektmarkierung in
Neuiranischen Sprachen. Tübingen: Narr.
Bossong, Georg. 2006. Meaning, form and function in basic roles. In Ina Bornkessel, Matthias
Schlesewsky, Bernard Comrie, Angela D. Friederici (eds.), Semantic role universals and
argument linking, 237–262. Berlin: Mouton.
Bresnan, Joan and Sam A. McHombo. 1987. Topic, pronoun, and agreement in Chichewa.
Language 63(4). 741–782.
Brown, N. 1848. Grammatical Notices of the Assamese Language. Sibsagor: American Baptist
Missionary Press.
Brown, Penelope and Stephen C. Levinson. [1978] 1987. Politeness: Some universals in
language use. Cambridge: Cambridge University Press.
Bubenik, Vit. 1989. On the origin and elimination of ergativity in Indo-Aryan languages.
Canadian journal of linguistics 34. 377–398.
Bubenik, Vit. 1998. A historical syntax of late Middle Indo-Aryan (Apabhramsa). Amsterdam:
Benjamins.
Burrow, Thomas. 1965. The Sanskrit language. London: Faber & Faber.
Burzio, Liugi. 1986. Italian syntax. A government-binding approach. Dordrecht: Reidel.
Butt, Miriam. 2004. The role of pronominal suffixes in Punjabi. Paper presented at the DGFS
workshop Evolution syntaktischer Relationen, University of Mainz, 25–27 February.
Butt, Miriam. 2006. Theories of case. Cambridge: Cambridge University Press.
Butt, Miriam. 2007. The dative-ergative connection. Invited paper presented at CIEFL,
Hyderabad.
Butt, Miriam. and Tracy H. King. 2004. The status of case. In Anoop Mahajan and Vineeta Dayal
(eds.), Clause structure in South Asian languages, 153–198. Dordrecht: Kluwer.
Butt, Miriam. and Tikaram Poudel. 2007. Distribution of the ergative in Nepali. Paper presented
at the University of Leipzig.
Bybee, Joan. 1985. Morphology. Amsterdam: Benjamins.
Bybee, Joan and Osten Dahl. 1989. The creation of tense and aspect systems in the languages
of the world. Studies in Language 13(1). 51–103.
Bynon, Theodora. 1979. The ergative construction in Kurdish. Bulletin of the School of Oriental
and African studies 42(2). 211–224.
Bynon, Theodora. 1989. Pronominal attrition, clitic doubling and typological change. Folia
Linguistica Historica 13(1–2). 27–63.
Bynon, Theodora. 2005. Evidential, raised possessor, and the historical source of the ergative
construction in Indo-Iranian. Transactions of the philological society 103(1). 1–72.
Cardona, George. 1965. A Gujarati reference grammar. Philadelphia: University of Pennsylvania
Press.
Cardona, George. 1970. The Indo-Iranian construction mana (mama) krtam. Language 46(1).
1–12.
Cardona, George. 1974. The Indo-Aryan languages. Encyclopedia Britannica (15th ed.) 9.
439–450.
Cardona, George and Dhanesh Jain, (eds.). 2003. The Indo-Aryan languages. London:
Routledge.
Chafe, Wallace L. 1994. Discourse, consciousness, and time: the flow and displacement of
conscious experience in speaking and writing. Chicago: Chicago University Press.
 References 299

Chatterjee, Jayati. 2008. Lexical motivation for passive forms of some Bangla verbs. South
Asian language review 8(1). 100–106.
Chatterji, Suniti K. 1926. The origin and development of the Bengali language. London: George
Allen & Unwin.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.
Clark, T. W. 1963. Introduction to Nepali. Cambridge: W. Heffer & Sons.
Cole, Jennifer. 2001. Sindhi. In Jane Garry and Carlo Rubino (eds.), Facts about the world’s
languages: An encyclopedia of the world’s major languages, past and present, 647–653.
New York: The H.W. Wilson Company.
Cole, Peter, Wayne Harbert, Gabriella Hermon and S. N. Sridhar. 1980. The acquisition of
subjecthood. Language 56(4). 719–743.
Cole, Peter and Jerrold M. Sadock. 1977. Grammatical relations. New York: Academic Press.
Comrie, Bernard. 1973. The ergative, variations on a theme. Lingua 32. 239–253.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
Comrie, Bernard. 1978. Ergativity. In Winfred P. Lehmann (ed.), Syntactic typology. Studies in the
phenomenology of language, 329–393. Sussex: The Harvester Press.
Comrie, Bernard. 1984. Reflections on verb agreement in Hindi and related languages.
Linguistics 22. 857–864.
Comrie, Bernard. 1989. Language universals and linguistic typology. Oxford: Basil Blackwell
Publishing.
Comrie, Bernard. 2008. Alignment of case marking of full noun phrases. In Matthew Dryer and
Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Munich: Max
Planck Digital Library. http://wals.info/chapter/98 (Accessed 4 April 2011).
Cooreman, Ann, Barbara Fox and Talmy Givon. 1984. Discourse definition of ergativity. Studies
in Language 8(1). 1–34.
Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge University Press.
Coseriu, Eugenio. 1987. Formen und Funktionen. Studien zur Grammatik. Tübingen: Niemeyer.
Creissels, Dennis. 2008. Direct and indirect explanations of typological regularities: The case of
alignment variations. Folia Linguistica 42(1). 1–38.
Creissels, Dennis. 2008b. Remarks on split intransitivity and fluid intransitivity. In Olivier
Bonami and Patricia Cabredo-Hofherr (eds.), Empirical issues in syntax and semantics 7,
139–168. Paris: Colloque de syntaxe et sémantique.
Creissels, Dennis. 2009. Uncommon patterns of core term marking and case terminology.
Lingua 119(3). 445–459.
Croft, William. 1988. Agreement versus case marking and direct objects. In Michael Barlow
and Charles Ferguson (eds.), Agreement in natural language: Approaches, theories,
descriptions, 159–179. Stanford: CSLI.
Croft, William. 2003. Typology and universals. Cambridge: Cambridge University Press.
Cysouw, Michael and Diane Forker. 2009. Reconstruction of morphosyntactic function:
nonspatial usage of spatial case marking in Tsezic. Language 85(3). 588–617.
Das Gupta, Bidhu B. 1981. Assamese self-taught. Calcutta: Das Gupta Prakashan.
Das, Pradeep K. 2006. Grammatical agreement in Hindi-Urdu and its major varieties. München:
Lincom.
Dasgupta, Probal 2004. Some non-nominative subjects in Bangla. In Peri Bhaskararao and K. V.
Subbarao (eds.), Non-nominative subjects. Vol I, 129–140. Amsterdam: Benjamins.
300 References

Davidse, Kristin. 1992. Transitivity/ergativity. The Janus-headed grammar of actions and events.
In Martin Davies and Louise Ravelli (eds.), Advances in systemic linguistics, 105–135.
London: Pinter.
Davison, Alice. 1990. Comments on the papers by Elena Bashir, Peter Hook, Colin P. Masica, and
Arlene Zide. In Manindra K. Verma and K. P. Mohanan (eds.), Experiencer subjects in South
Asian languages, 355–364. Stanford: CSLI.
Davison, Alice. 1999. Ergativity: functional and formal issues. In Michael Darnell, Edith
Moravcsik, Frederick J. Newmeyer, Michael Noonan and Kathleen Wheatley (eds.),
Functionalism and formalism in linguistics, 177–210. Amsterdam: Benjamins.
Davison, Alice. 2002. Agreement features and projections of TENSE and ASPECT. In Rajendra
Singh (ed.), The yearbook of South Asian Languages and linguistics, 12–27. New Delhi:
Sage.
De Clercq, Eva. 2003. Een kritische studie van Svayambhudeva’s Paümacariu. Gent: University
of Ghent dissertation.
De Clercq, Eva. 2010. The Apabhraṃśa of Svayambhūdeva’s Paümcariu. Mumbai: Hindi Granth
Karyalay.
de Swart, Peter. 2006. Case markedness. In Leonid Kulikov, Andrej Malchukov and Peter de
Swart (eds.), Case, valency and transitivity, 249–267. Amsterdam: John Benjamins.
de Swart, Peter, Monique Lamers and Sander Lestrade. 2008. Animacy, argument structure and
argument encoding. Lingua 118. 131–140.
Debrunner, Albert. 1954. Jacob Wackernagel Altindische Grammatik, Vol II. Göttingen:
Vandenhoeck und Ruprecht.
Degener, Almuth. 2008. Shina-Texte aus Gilgit (Nord-Pakistan) : Sprichwörter und Materialien
zum Volksglauben. Wiesbaden: Harrassowitz.
DeLancey, Scott. 1981. An interpretation of split ergativity and related patterns. Language 57.
626–657.
DeLancey, Scott. 1985. Lhasa Tibetan evidentials and the semantics of causation. Proceedings
of the annual Berkeley Linguistics Society 11. 65–72.
DeLancey, Scott. 1986. Evidentiality and volitionality in Tibetan. In Wallace L. Chafe and Johanna
Nichols (eds.), Evidentiality: The linguistic coding of epistemology, 201–213. Norwood:
Ablex.
DeLancey, Scott. 2004. The blue bird of ergativity. In Francesco Queixalós (ed.), Ergativity in
Amazonia III, 1–37. Paris: CELIA-CNRS.
Deo, Ashwini and Devyani Sharma. 2006. Typological variation in the ergative morphology in
Indo-Aryan. Linguistic typology 10. 369–418.
Dik, Simon C. 1978. Functional grammar. Amsterdam: North-Holland Publishing Company.
Dik, Simon C. 1980. Studies in functional grammar. London/New York: Academic Press.
Dik, Simon C. 1989. The theory of functional grammar. Part I: The structure of the clause.
Dordrecht: Foris.
Dirr, Albert. 1912. Rutulskij jazyk [The Rutul language]. Sbornik Materialov dlya Opisaniya y
Plemn Kavkaza 42(1). 1–204.
Dixon, Robert M. W. 1972. The Dyirbal language of North Queensland. Cambridge: Cambridge
University Press.
Dixon, Robert M. W. 1979b. Is Dyirbal ergative? Corrections and comments. Linguistics 17(11–1).
1003–1015.
Dixon, Robert M. W. 1979. Ergativity. Language 55(1). 59–138.
Dixon, Robert M. W. 1994. Ergativity. Cambridge: Cambridge University Press.
 References 301

Dixon, Robert M. W. 1995. Complement clauses and complementation strategies. In Frank R.

Palmer (ed.), Grammar and meaning: essays in honour of Sir John Lyons, 175–220.
Cambridge: Cambridge University Press.
Dixon, Robert M. W. 2010. Basic linguistic theory. Grammatical topics. Oxford: Oxford University
Press.
Dixon, Robert M. W. and A. Aikhenvald. 2000. Introduction. In Robert M. W. Dixon and
Alexandra Aikhenvald (eds.), Changing valency: Case studies in transitivity, 1–29.
Cambridge: Cambridge University Press.
Donohue, Marc and Søren Wichmann (eds). 2008. The typology of semantic alignment. Oxford:
Oxford University Press.
Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–619.
Drocco, Andrea. 2008. L’ergatività in hindi. Alessandria: Edizione dell’Orso.
Dryer, Matthew. S. 1986. Primary objects, secondary objects, and antidative. Language 62(4).
808–845.
Dryer, Matthew. S. 1997. Are grammatical relations universal? In Joan Bybee, John Haiman and
Sandra Thompson (eds.), Essays on language function and language type: dedicated to
T. Givón, 115–143. Amsterdam: Benjamins.
Du Bois, John W. 1987. The discourse basis of ergativity. Language 63. 805–855.
Du Bois, John W. 2003. Argument structure: Grammar in use. In John W. Du Bois, Lorraine E.
Kumpf and William J. Ashby (eds.), Preferred argument structure, 11–60. Amsterdam:
Benjamins.
Du Bois, John W., Lorraine E. Kumpf and William J. Ashby, (eds.). 2003. Preferred Argument
Structure. Grammar as architecture for function. Amsterdam: Benjamins.
Durie, Marc. 1988. Preferred argument structure in an active language: Arguments against the
category ‘intransitive subject’. Lingua 74. 1–25.
Edelman, Dzhoy J. I. and Leila R. Dodykhudoeva. 2009. The Pamir languages. In Gernot
Windfuhr (ed.), The Iranian languages, 773–786. London: Routledge.
Emeneau, Murray B. [1965] 1980. India and linguistic areas. In A. S. Dil (ed.), Language and
linguistic area. Essays by Murray B. Emeneau, 126–166. Stanford: Stanford University
Press.
Estival, Dominique and John Myhill. 1988. Formal and functional aspects of the development
from passive to ergative systems. In Masayoshi Shibatani (ed.). Passive and voice,
441–492. Amsterdam: Benjamins.
Evans, Nicholas. 1999. Why argument affixes in polysynthetic languages are not pronouns.
Evidence from Bininj Gun-wok. Sprachtypologie und Universalienforschung 52. 255–281.
Everaert, Christine. 2004. Basisgrammatica van het Hindi. Gent: Academia Press.
Farrell, Patrick. 2005. Grammatical relations. Oxford: Oxford University Press.
Filiminova, Elena. 2005. The noun phrase hierarchy and relational marking: problems and
counterevidence. Linguistic typology 9. 77–113.
Fillmore, Charles. 1968. The case for case. In E. Bach and R. T. Harms (eds.), Universals in
linguistic theory, 1–91. Holt: Rinehart and Winston.
Foley, William A. and Robert D. J. Van Valin. 1980. Role and reference grammar. In E. Moravcsik
and J. R. Wirth (eds.), Current approaches to syntax, 329–352. New York: Academic Press.
Foley, William A. and Robert D. J. Van Valin. 1984. Functional syntax and universal grammar.
Cambridge: Cambridge University Press.
Fortescue, Michael D. 1984. West Greenlandic. London: Croom Helm.
302 References

Gabelentz, Hans C. v. d. 1861. Ueber das Passivum: Eine sprachvergleichende Abhandlung.

Abhandlungen der Königlich Sächsischen Gesellschaft der Wissenschaften 8. 451–546.
Ganathe, N. S. R. 2005. Learn Gujarati in 30 days. Chennai: Balaji Publications.
García-Miguel, José M. 2007. Clause structure and transitivity. In Dirk Geeraerts and Hubert
Cuyckens (eds.), The Oxford handbook of cognitive linguistics, 753–781. Oxford: Oxford
University Press.
Genetti, Carol. 1988. A syntactic correlate of topicality in Newari narrative. In John Haiman
and Sandra Thompson (eds.), Clause combining in grammar and discourse, 29–48.
Amsterdam: Benjamins.
Genetti, Carol. 1999. Variation in agreement in the Nepali finite verb. In Yogendra P. Yadava and
William W. Glover (eds.), Topics in Nepalese linguistics, 542–555. Kathmandu: Royal Nepal
Academy.
Genetti, Carol. 2007. A grammar of Dolakha Newar. Berlin: GmbH & Co.
Gildea, Steve. 2004. Are there universal cognitive motivations for ergativity? In Francesco
Queixalós (ed.), Ergatividade na Amazonia II, 1–54. Brasilia: CNRS, IRD and the
Laboratório de Línguas Indígenas, UnB.
Gill, T. S. (no publication date mentioned). Learn Punjabi. New Delhi: Star Publications.
Givón, Talmy. 1976. Topic, pronoun and grammatical agreement. In Charles N. Li (ed.), Subject
and topic, 149–188. New York: Academic Press.
Givón, Talmy. 1983. Topic continuity in discourse: A quantitative cross-language study.
Amsterdam: Benjamins.
Givón, Talmy. 1984. Syntax. A functional-typological introduction. Amsterdam: Benjamins.
Givón, Talmy. 1995. Functionalism and grammar. Amsterdam: Benjamins.
Givón, Talmy. 2001. Syntax: an introduction. Amsterdam: Benjamins.
Goddard, Cliff. 1982. Case systems and case marking in Australian languages: a new
interpretation. Australian journal of linguistics 2. 167–196.
Gonda, Jan. 1951. Remarks on the Sanskrit passive. Leiden: Brill.
Goswami, Golokchandra. 1982. The structure of Assamese. Gauhati: Department of Publication
Gauhati University.
Goswami, Golokchandra and Jyotiprakash Tamuli. 2003. Asamiya. In George Cardona and
Dhanesh Jain (eds.), The Indo-Aryan languages, 429–522. London: Routledge.
Goswami, Upendranath. 1978. An introduction to Assamese. Gauhati: Mani-Manik Prakash.
Greenberg, Joseph. 1966. Language Universals. In Thomas Sebeok (ed.), Current trends in
linguistics: Theoretical foundations, 61–112. The Hague/Paris: Mouton & Co.
Grierson, George A. 1883. Seven grammars of the dialects and subdialects of the Bihari
language. Calcutta: Bengal Secretariat Press.
Grierson, George A. 1903–1926. Linguistic survey of India. Calcutta: Office of the Superin-
tendent, Government Printing.
Grierson, George A. 1911. A manual of the Kashmiri language comprising grammar,
phrase-book and vocabularies. Oxford: Clarendon Press.
Gusain, Lakhan. 2004. Marwari. München: Lincom Europa.
Haig, Geoffrey. 2008. Alignment change in Iranian languages : a construction grammar
approach. Berlin: Mouton de Gruyter.
Hale, Austin. 1973. Clause, sentence and discourse patterns in selected languages of Nepal.
Norman: Summer Institute of linguistics of the University of Oklahoma.
Halliday, M. A. K. 1967. Notes on transitivity and theme in English, Parts 1 and 2. Journal of
Linguistics 3. 37–81, 199–244.
 References 303

Halliday, M. A. K. [1966] 2003. Grammar, society and the noun. In M. A. K. Halliday and
J. Webster (eds.), On language and linguistics, 50–73. London: Continuum international
publishing.
Hamsa. 1984. Cuppī [Silence]. In Rawat Saraswat and Prem (eds.), Aaj ri Rajasthani kahaniyan
[Contemporary Rajasthani stories], 170–174. New Delhi: Sahitya Akademi.
Hargreaves, David. 2005. Agency and Intentional Action in Kathmandu Newar. Himalayan
linguistics 5. 1–48.
Harris, Alice C. 1990. Alignment typology and diachronic change. In Winfred P. Lehmann (ed.),
Language typology 1987. Systematic balance in language, 67–91. Amsterdam: Benjamins.
Harris, Alice C. 1997. Review Ergativity by R.M.W. Dixon. Language 73(2). 359–374.
Harris, Alice C. and Lyle Campbell. 1995. Historical syntax in cross-linguistic perspective.
Cambridge: Cambridge University Press.
Harris, Richard M. and Rama N. Sharma. 1968. A basic Hindi reader. New York: Rochester
University.
Haspelmath, Martin. 1990. The grammaticization of passive morphology. Studies in language
14(1). 25–72.
Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations.
In Bernard Comrie and Maria Polinsky (eds.), Causatives and transitivity, 87–120.
Amsterdam: John Benjamins.
Haspelmath, Martin. 2005. Argument marking in ditransitive alignment types. Linguistic
discovery 3(1). 1–21.
Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal of
Linguistics 42(1). 25–70.
Haspelmath, Martin. 2009. Terminology of case. In Andrej Malchukov and Andrew Spencer. The
Oxford handbook of case, 505–517. Oxford: Oxford University Press.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic
studies. Language 86(3). 663–687.
Heath, Jeffrey. 1979. Is Dyirbal ergative? Linguistics 17(5–6). 401–463.
Heath, Jeffrey. 1980. Dyirbal ergativity. Counter-rejoinder. Linguistics 18(5–6). 505–521.
Heine, Bernd and Christa Konig. 2010. On the linear order of ditransitive objects. Language
Sciences 32(1). 87–131.
Helmbrecht, Johannes. 2001. Head-marking versus dependent-marking languages. In Martin
Haspelmath (ed.), Language typology and language universals, 1424–1432. Berlin:
Mouton de Gruyter.
Hewson, John and Vit Bubenik. 1997. Tense and aspect in Indo-European languages.
Amsterdam: Benjamins.
Hjelmslev, Louis. 1972. La catégorie des cas. Munchen: Wilhelm Fink Verlag.
Hock, Hans H. 1986. P-oriented constructions in Sanskrit. In B. Krishnamurti (ed.), South Asian
languages: Structure, convergence and diglossia, 15–26. New Delhi: Motilal Banarsidass.
Hoernle, August F. R. 1880. A comparative grammar of the Gaudian (Aryo-Indian) languages.
Amsterdam: Philo Press.
Hook, Peter E. 1978. The Hindi compound verb: What it is and what it does. In K. S. Singh (ed.),
Readings in Hindi-Urdu linguistics, 109–124. Delhi: National Publishing House.
Hook, Peter E. 1987. Poguli syntax in the light of Kashmiri: A preliminary report. Studies in the
linguistic sciences 17(1). 63–71.
304 References

Hook, Peter E. 1988. Paradigmaticization: A case study from South Asia. In Shelley Axmaker,
Fourteenth Annual meeting. General session and parasession on grammaticalization,
293–303. Berkeley: Berkeley Linguistics Society.
Hook, Peter E. 1990. Experiencers in South Asian languages: A gallery. In Manindra K. Verma
and K. P. Mohanan (eds.), Experiencer subjects in South Asian languages, 319–334.
Stanford: CSLI.
Hook, Peter E. 1991. On identifying the conceptual restructuring of passive as ergative in
Indo-Aryan. In Madhav M. Deshpande and Saroje Bhate (eds.), Paninian studies. Professor
S. D. Joshi felicitation volume, 177–200. University of Michigan: Center for South and
Southeast Asian Studies.
Hook, Peter E. 1996. Kesar of Layul: A Central Asian epic in the Shina of Gultari. In William
Hanaway and Wilma Heston (eds.), Studies in Pakistani popular culture, 121–183. Lahore:
Sang-e-Meel and Lok Virsa.
Hook, Peter E. 2011. Alignment, realignment and paths of syntactic change. A case study from
Indo-Aryan. Presentation at the Linguistic Anthropology Seminar, University of Virginia,
7 October. http://artsandsciences.virginia.edu/linganthseminar/archive/2011_Fall.html
Hook, Peter E. and Vijay K. Kaul. 1987. Case alternation, transitionality and the adoption of
direct objects in Kashmiri. Indian Linguistics 48. 52–69.
Hook, Peter E. and Ashok K. Koul. 2002. Under the surface of the South Asian linguistic area:
More on the syntax of derived transitives and causatives in Kashmiri. In Omkar N. Koul and
Kashi Wali (eds.), Topics in Kashmiri linguistics, 103–112. New Delhi: Creative.
Hook, Peter E. and Omkar N. Koul. 1984. Pronominal suffixes and split ergativity in Kashmiri.
In Omkar N. Koul (ed.), Aspects of Kashmiri Linguistics, 123–135. New Delhi: Bahri
Publications.
Hook, Peter E. and Omkar N. Koul. 1984b. On the grammar of derived transitives and causatives
in Kashmiri. In Omkar N. Koul and Peter E. Hook (eds.), 62–90. Aspects of Kashmiri
linguistics. New Delhi: Bahri Publications.
Hook, Peter E. and Omkar N. Koul. 2004. Case as agreement. In Peri Baskararao and K.V.
Subbarao (eds.), Non-nominative subjects, Vol. I, 213–225. Amsterdam: Benjamins.
Hook, Peter E. and Omkar N. Koul. 2006. Valency sets in Kashmiri. In Tasaku Tsunoda and Taro
Kageyama (eds.), Voice and grammatical relations. In honor of M. Shibatani, 43–84.
Amsterdam: Benjamins.
Hopper, Paul and Sandra Thompson. 1980. Transitivity, grammar and discourse. Language 56.
251–299.
Hopper, Paul and Elisabeth C. Traugott. 1993. Grammaticalization. Cambridge: Cambridge
University Press.
Hutt, Michael. 1997. Modern literary Nepali. An introductory reader. New Delhi: Oxford
University Press.
Hutt, Michael and A. Subedi. 1999. Nepali. London: Teach yourself books.
Iggesen, Oliver. 2005. Case-asymmetry. München: Lincom Europa.
Jackendoff, Ray. 1972. Semantic interpretation in generative grammar. Cambridge: MIT Press.
Jahani, Carina and Agnes Korn. 2009. Balochi. In Gernot Windfuhr (ed.), The Iranian languages,
634–693. London: Routledge.
Jaiswal, Mahesh P. 1962. A linguistic study of Bundeli. Leiden: E.J. Brill.
Jeffers, Robert J. 1976. The position of the Bihari dialects in Indo-Aryan. Indo-Iranian Journal 18.
215–225.
 References 305

Jelinek, Eloise. 1984. Empty categories, case and configurationality. Natural Language &
Linguistic Theory 2(1). 39–76.
Jespersen, Otto. 1909–1949. A modern English grammar on historical principles. Copenhagen:
Ejnar Munksgaard.
Jha, Subhakanta. 1958. The formation of the Maithili language. London: Luzac and Company.
Kachru, Yamuna and R. Pandharipande. 1979. On ergativity in selected South Asian languages.
South Asian languages analysis 1. 193–209.
Kakati, Bani. 1941. Assamese, its formation and development. Guwahati: Lawyer’s Book Stall.
Kakoti, S. 2011. Amrityu Amrit [Amrit’s life potion]. In Pradip Acharya (ed.), Winners all, 1–32.
Guwahati: Munin Barkotoki Memorial Trust.
Karki, Tika B. and Chij K. Shrestha. 1974. Basic course in spoken Nepali. Kathmandu: The Nepal
Press.
Keenan, Edward L. 1976. Towards a universal definition of “subject”. In Charles N. Li (ed.),
Subject and topic, 303–333. New York: Academic Press.
Keine, Stefan. 2007. Reanalysing Hindi split-ergativity as a morphological phenomenon. In
Jochen Trommer and Andreas Opitz (eds.), 1–2-many, 73–127. Leipzig: Linguistische
Arbeitsberichte.
Kellogg, Samuel H. 1938. A grammar of the Hindi language. London: Kegan Paul, Trench,
Trubner and Co.
Khokhlova, Liudmilla V. 1992. Trends in the development of ergativity in New Indo-Aryan.
Osmania Papers in Linguistics 18. 71–98.
Khokhlova, Liudmilla V. 1995. The development of patient-oriented constructions in Late
Western NIA languages. Osmania Papers in Linguistics 21. 15–54.
Khokhlova, Liudmilla V. 2001. Ergativity attrition in the history of Western Indo-Aryan
languages. In Rajendra Singh (ed.), The yearbook of South Asian languages and
linguistics, 158–184. New Delhi: Sage Publications.
Khokhlova, Liudmilla V. 2002. Syntactic peculiarities of Rajasthani. Paper presented at the
7th European Conference on Modern South Asian Studies. Heidelberg: University of
Heidelberg.
Khubchandani, Lachman. 1962. Pronominal suffixes in Sindhi. Indian Linguistics 23. 72–81.
Kibrik, Alexander E. 1979. Canonical ergativity and Daghestan languages. In Frans Plank (ed.),
Ergativity. Towards a theory of grammatical relations, 61–78. New York: Academic Press.
Kibrik, Alexander E. 1997. Beyond subject and object: toward a comprehensive relational
typology. Linguistic typology 1(3). 279–346.
Kieffer, Charles M. 2009. Parachi. In Gernot Windfuhr (ed.), The Iranian languages, 693–721.
London: Routledge.
Kimball, Geoffrey. 1991. Koasati Grammar. Lincoln: University of Nebraska Press.
Kittilä, Seppo. 2002. Remarks on the basic transitive sentence. Language Sciences 24(2).
107–130.
Kittilä, Seppo. 2005. Optional marking of arguments. Language Sciences 27(5). 483–514.
Kittilä, Seppo. 2005b. Recipient-prominence vs. beneficiary-prominence. Linguistic typology 9.
269–297.
Kittilä, Seppo. 2006. Object-, animacy- and role-based strategies - A typology of object
marking. Studies in Language 30(1). 1–32.
Kittilä, Seppo. 2007. A typology of tritransitives: alignment types and motivations. Linguistics
45(3). 453–508.
306 References

Kittilä, Seppo. 2007b. On the encoding of transitivity-related features on the indirect object.
Functions of Language 14(1). 149–164.
Kittilä, Seppo. 2008. Defining prototypical transitivity. Lingua 118(12). 2012–2020.
Kittilä, Seppo. 2009. Case and the typology of transitivity. In Andrej Malchukov and Andrew
Spencer (eds.), The Oxford handbook of case, 356–365. Oxford: Oxford University Press.
Kittilä, Seppo. 2011. Transitivity typology. In Jae J. Song (ed.), The Oxford handbook of typology,
346–367. Oxford: Oxford University Press.
Klaiman, Miriam H. 1976. Subjecthood, reflexivization, emphatic pronominalization and clause
matedness in Bengali (with some side glances at Hindi). In Manindra K. Verma (ed.), The
notion of subject in South Asian languages, 137–151. Madison: University of Wisconsin.
Klaiman, Miriam H. 1978. Arguments against a passive origin of the IA ergative. Papers from the
annual regional meeting of the Chicago Linguistic Society 14. 204–216.
Klaiman, Miriam H. 1980. Bengali dative subjects. Lingua 51. 275–295.
Klaiman, Miriam H. 1981. Volitionality and subject in Bengali: a study of semantic parameters in
grammatical processes. Bloomington: Indiana University.
Klaiman, Miriam H. 1987. Mechanisms of ergativity in South Asia. Lingua 71. 61–102.
Klaiman, Miriam H. 1990. The prehistory of noun-incorporation in Hindi. Lingua 81(4). 327–350.
Klein, Udo and Peter de Swart. 2011. Case and referential properties. Lingua 121(1). 3–19.
Klimov, Georgij A. 1973. Ocerk obscej teorii ergativnosti. Moskva: Nauka.
Klimov, Georgij A. 1974. On the character of “active” languages. Linguistics 131. 11–23.
Klimov, Georgij A. 1979. On the position of the ergative type in typological classification. In
Frans Plank (ed.), Ergativity. Towards a theory of grammatical relations, 327–332. New
York: Academic Press.
Korn, Agnes. 2003. Balochi and the concept of Northwestern Iranian. In Carina Jahani and
Agnes Korn (eds.), The Baloch and their neighbours. Ethnic and linguistic contact in
Balochistan in historical and modern times, 49–60. Wiesbaden: Reichert.
Kothari, Deva. 1991. Rajasthani bhasa aur uski boliyam [The Rajasthani language and its
dialects]. Udaipur: Sahitya Samsthan.
Koul, Omkar N. 1985. An intensive course in Kashmiri. Mysore: Central Institute of Indian
Languages.
Koul, Omkar N. 2003. Kashmiri. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan
languages, 895–952. London: Routledge.
Koul, Omkar N. 2005. Studies in Kashmiri linguistics. New Delhi: IILS.
Koul, Omkar N. 2006. Spoken Kashmiri: A language course. Delhi: IILS.
Koul, Omkar N. 2008. Kashmiri Newspaper Reader. Hyattsville: Dunwoody Press.
Koul, Omkar N. and Ruth L. Schmidt. 1984. Dardistan revisited: An examination of relationship
between Kashmiri and Shina. In Omkar N. Koul and Peter E. Hook (eds.), Aspects of
Kashmiri linguistics, 1–26. New Delhi: Bahri Publications.
Koul, Omkar N. and Kashi Wali. 2006. Modern Kashmiri grammar. Springfield: Dunwoody
Press.
Kumakhov, Mukhadin, Karina Vamling and Zara Kumakhova. 1996. Ergative case in the
Circassian languages. Lund University Working Papers in Linguistics 45. 93–111.
Kuno, S. 1987. Functional syntax: Anaphora, discourse and empathy. Chicago: University of
Chicago Press.
Langacker, Ronald. 1991. Foundations of Cognitive Grammar, Vol II, Descriptive Application.
Stanford: Stanford University Press.
Langacker, Ronald. 1999. Grammar and conceptualization. Berlin: Mouton de Gruyter.
 References 307

Lazard, George. 1998. Actancy. Berlin: Mouton de Gruyter.

Lazard, George. 1998b. De la transitivité restreinte à la transitivité généralisée. In André
Rousseau (ed.), La transitivité, 55–84. Paris: Presses Universitaires du Septentrion.
Lazard, George. 2001. Le marquage différentiel de l’objet. In Martin Haspelmath (ed.),
Language Typology and Language Universals, 873–885. Berlin: Walter de Gruyter.
Lazard, George. 2002. Transitivity revisited as an example of a more strict approach in
typological research. Folia Linguistica 36(3–4). 141–190.
Lehmann, Christian. 1982. Universal and typological aspects of agreement. In Hansjakob Seiler
and Franz J. Stachowiak (eds.) Apprehension: das sprachliche Erfassen von Gegenstünden,
Teil II, 201–267. Tübingen: Narr.
Lehmann, Christian. 2002. Thoughts on grammaticalization. Erfurt: Seminar für Sprachwis-
senschaft der Universität.
Lemmens, Maarten. 1998. Lexical perspectives on transitivity and ergativity: causative
constructions in English. Amsterdam: Benjamins.
Levin, Magnus. 2001. Agreement with collective nouns in English. Stockholm: Almqvist and
Wiksell.
Li, Chao. 2007. Split ergativity and split intransitivity in Nepali. Lingua 117. 1462–1482.
Li, Charles. N. 1976. Subject and topic. New York: Academic Press.
Li, Charles N. 1977. Mechanisms of syntactic change. Texas: University of Texas Press.
Lidz, Jeffrey. 2006. The grammar of accusative case in Kannada. Language 82(1). 10–32.
Liperovsky, V. P. 2007. Notes on the marking of actants in Braj (in comparison with Modern
Standard Hindi). In Colin Masica (ed.), Old and new perspectives on South Asian
languages. Grammar and semantics, 144–152. New Delhi: Motilal Banarsidass.
Magier, David. 1983. Components of ergativity in Marwari. Papers from the annual regional
meeting of the Chicago Linguistic Society 19(1). 244–255.
Magier, David. 1985. Case and transitivity in Marwari. In Arlene Zide, David Magier and Eric
Schiller (eds.), Proceedings of the conference on participant roles (CLS 1984 meeting),
149–159. Bloomington: Indiana University Linguistics Club.
Magier, David. 1990. Dative/Accusative subjects in Marwari. In Manindra K. Verma and K. P.
Mohanan (eds.), Experiencer subjects in South Asian languages, 213–220. Stanford: CSLI.
Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Massachussets, MIT
dissertation.
Malchukov, Andrej. 2006. Transitivity parameters and transitivity alternations: Constraining
co-variation. In Leonid Kulikov, Andrej Malchukov and Peter de Swart (eds.), Case, valency
and transitivity, 329–358. Amsterdam: Benjamins.
Malchukov, Andrej, Martin Haspelmath and Bernard Comrie. 2010. Ditransitive constructions: A
typological overview. In Andrej Malchukov, Martin Haspelmath and Bernard Comrie (eds.),
Studies in ditransitive constructions, 1–59. Berlin: Mouton de Gruyter.
Malchukov, Andrej 2008. Animacy and assymetries in differential case marking. Lingua 118.
203–221.
Malchukov, Andrej and Helen de Hoop. 2011. Tense, aspect, and mood based differential case
marking. Lingua 121(1). 35–47.
Mallinson, G. and Barry J. Blake. 1981. Language typology. Amsterdam: North-Holland
Publishing Company.
Manaster-Ramer, Alexis. 1994. The origin of the term ergative. Sprachtypologie und Universali-
enforschung 47. 211–214.
Manto, S. H. 1984. Thanda Gosht [Cold flesh]. New Delhi: Saqi Book Depot.
308 References

Masica, Colin. 1976. Defining a linguistic area. South Asia. Chicago: University of Chicago
Press.
Masica, Colin. 1986. Definiteness-marking in South Asia. In Badhriraju Krishnamurti (ed.),
South Asian languages: Structure, convergence and diglossia, 123–146. New Delhi:
Motilal Banarsidass.
Masica, Colin. 1990. Varied case markings in obligational constructions. In Manindra K. Verma
and K. P. Mohanan (eds.), Experiencer subjects in South Asian languages, 335–342.
Stanford: CSLI.
Masica, Colin. 1991. The Indo-Aryan languages. Cambridge: Cambridge University Press.
Master, Alfred. 1964. A grammar of Old Marathi. Oxford: Clarendon Press.
Matras, Yaron. 1997. Clause combining, ergativity, and coreferent deletion in Kurmanji. Studies
in language 21. 613–653.
Matthews, Dave. 1984. A course in Nepali. New Delhi: Heritage Publishers.
McGregor, William B. 2009. Typology of ergativity. Language and linguistics compass 3(1).
480–508.
McGregor, William B. 2010. Optional ergative case marking systems in a typological-semiotic
perspective. Lingua 120(7). 1610–1636.
Meakins, Felicity and Carmel O’Shannessy. 2010. Ordering arguments about: Word order and
discourse motivations in the development and use of the ergative marker in two Australian
mixed languages. Lingua 120(7). 1693–1713.
Medhi, Kaliram. 1936. Assamese grammar and origin of the Assamese language. Calcutta: P.C.
Ray, Sri Gouranga Press.
Merlan, Francesca. 1985. Split intransitivity: functional oppositions in intransitive inflection. In
Johanna Nichols and Anthony C. Woodbury (eds.), Grammar inside and outside the clause:
Approaches to theory from the field, 324–362. Cambridge: Cambridge University Press.
Mescaninov, I. I. 1967. Ergativnaja konstrukcija v jazykax razlicnyx tipov. Leningrad: Nauka.
Michailovsky, Boyd. 1996. L’inférentiel du népalais. In Zlatka Guentchéva (ed.), L’énonciation
médiatisée, 109–124. Leuven: Peeters.
Mishra, Mithilesh K. 1990. Dative/Experiencer subjects in Maithili. In Manindra K. Verma and
K. P. Mohanan (eds.), Experiencer Subjects in South Asian Languages, 105–117. Stanford:
CSLI.
Mistry, P. J. 1976. Subject in Gujarati: An examination of verb agreement phenomenon. In
Manindra K. Verma (ed.), The notion of subject in South Asian languages, 240–268.
Madison: University of Wisconsin.
Mistry, P. J. 1997. Objecthood and specificity in Gujarati. In Jane H. Hill, P. J. Mistry and Lyle
Campbell. The life of language: Papers in linguistics in honor of William Bright, 425–442.
Berlin: Mouton de Gruyter.
Mithun, Marianne. 1991. Active/Agentive case marking and its motivations. Language 67(3).
510–546.
Mithun, Marianne. 1999. The languages of native North America. Cambridge: Cambridge
University Press.
Mithun, Marianne. and Wallace Chafe. 1999. What are S, A, and O? (Ergative-absolutive and
nominative-accusative languages). Studies in Language 23(3). 569–596.
Mohanan, Tara. 1994. Argument structure in Hindi. Stanford: CSLI Publications.
Monier-Williams, Monier. 1858. Rudiments of Hindustani Grammar. London: Cheltenham
College.
 References 309

Montaut, Annie. 1991. Voix, aspect, diathèses en hindi modern. Syntaxe, sémantique,
énonciation. Leuven: Peeters.
Montaut, Annie. 1996. La genèse des systèmes accompli et futur en Indo-Aryan. Journal
Asiatique 284(2). 325–360.
Montaut, Annie. 1998. L’ergativité en indo-aryen: un procès actif contrôlé par son agent ou une
prédication (stative) de localisation. Recherches linguistiques de Vincennes 27. 139–156.
Montaut, Annie. 1999. Le renouvellement des formes verbales en Hindi/Ourdou Moderne.
Journal Asiatique 287(2). 587–628.
Montaut, Annie. 2001. La notion de sujet en Hindi-Ourdou: Actance, diathèse et localisation.
Bulletin de la société de linguistique de Paris XCVI(1). 311–348.
Montaut, Annie. 2004. A grammar of Hindi. München: Lincom.
Montaut, Annie. 2004b. Oblique main arguments in Hindi as localizing predications:
Questioning the category of subject. In Peri Bhaskararao and K. V. Subbarao (eds.),
Non-nominative subjects. Vol II., 33–56. Amsterdam: Benjamins.
Montaut, Annie. 2006. The evolution of the tense-aspect system in Hindi/Urdu: The status of
the ergative alignment. In Miriam Butt and Tracy H. King (eds.), Proceedings of the LFG06
Conference, 365–385. Stanford: CSLI.
Montaut, Annie. 2006b. Mirative meanings as extensions of aorist in Hindi/Urdu. In Rajendra
Singh (ed.), The yearbook of South Asian languages and linguistics, 71–86. Berlin: Walter
de Gruyter.
Montaut, Annie. 2007. Réfléchi, transitif et intransitif en hindi/ourdou. In André Rousseau,
Didier Bottineau and Daniel Roulland (eds.), L’Enoncé réfléchi, 229–248. Rennes: Presses
Universitaires de Rennes.
Montaut, Annie 2009. Ergativité et classification des langues indo-aryennes: évolutions
internes ou contacts externes. In Jean-Marie Durand and Antoine Jacquet (eds.), Acte du
colloque Centre et périphérie: Approches nouvelles des orientalistes, 347–361. Paris: J.
Maisonneuve.
Montaut, Annie. 2009b. Ergative and pre-ergative patterns in Indo-Aryan as predications of
localization. A diachronic view of past and future systems. In Ali R. Fatihi (ed.), Language
Vitality in South Asia, 295–325. Aligarh: Aligarh Muslim University.
Montaut, Annie and Christiane Pilot-Raichoor. 1994. Sémantique actancielle et cas
morphologiques en hindi et en badaga. Persée 28. 90–112.
Moravcsik, Edith and Jessica R. Wirth. 1980. Current approaches to syntax. New York: Academic
Press.
Morgenstierne, Georg. 1961. Dardic and Kafir Languages. The Encyclopedia of Islam, Vol II,
facs. 25. Leiden: E.J. Brill.
Morland-Hughes, W. R. J. 1947. A grammar of the Nepali language in the Roman and Nagri
script. London: Luzac & Co.
Næss, Åshild. 2002. Transitivity and accusativity in Pileni. Rongorongo Studies 12(2). 6–76.
Næss, Åshild. 2004. What markedness marks: the markedness problem with direct objects.
Lingua 114. 1186–1212.
Næss, Åshild. 2006. Case semantics and the agent-patient opposition. In Leonid Kulikov,
Andrej Malchukov and Peter de Swart (eds.), Case, valency and transitivity, 309–327.
Amsterdam: Benjamins.
Næss, Åshild. 2007. Prototypical transitivity. Amsterdam: Benjamins.
Nasrin, Mithun B. and Wim A. M. van der Wurff. 2009. Colloquial Bengali. London: Routledge.
310 References

Nespital, Helmut. 1981. Das Futursystem im Hindi und Urdu. Ein Beitrag zur semantischen
Analyse der Kategorien Tempus, Aspekt und Modus und ihrer Grammeme. Wiesbaden:
Steiner.
Newmeyer, Frederick J. 1998. Language form and language function. Cambridge: MIT Press.
Newmeyer, Frederick J. 2010. On comparative concepts and descriptive categories: A reply to
Haspelmath. Language 86(3). 688–695.
Nicholl, George F. 1885. A Bengali grammar also an Asamese grammar. London: W.H. Allen &
Co.
Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 62(1).
56–119.
Nichols, Johanna. 1990. Some preconditions and typical traits of the stative-active language
type. In Winfred P. Lehmann (ed.), Language typology 1987. Systematic balance in
language, 95–114. Amsterdam: Benjamins.
Nichols, Johanna. 1992. Linguistic diversity in space and time. Chicago: Chicago University
Press.
Nichols, Johanna. 2008. Why are stative-active languages rare in Eurasia? Typological
perspective on split subject marking. In Marc Donohue and Søren Wichmann (eds.), The
typology of semantic alignment systems, 121–139. Oxford: Oxford University Press.
Nichols, Johanna, David A. Peterson and Jonathan Barnes. 2004. Transitivizing and detransi-
tivizing languages. Linguistic typology 8. 149–2011.
Palmer, Frank R. 1994. Grammatical roles and relations. Cambridge: Cambridge University
Press.
Pandharipande, Rajeshwari. 1997. Marathi. London: Routledge.
Pandharipande, Rajeshwari. 2000. Experiencer (Dative) NPs in Marathi. In Manindra K. Verma
and K. P. Mohanan (eds.), Experiencer subjects in South Asian languages, 161–179.
Stanford: CSLI.
Pandharipande, Rajeshwari. 2003. Marathi. In George Cardona and Dhanesh Jain (eds.), The
Indo-Aryan languages, 698–728. London: Routledge.
Pandharipande, Rajeshwari and Yamuna Kachru. 1977. Relational grammar, ergativity, and
Hindi-Urdu. Lingua 41. 217–238.
Patel, Pritty. 2007. Split ergativity and the nature of agreement. London: University College
London dissertation.
Payne, John. 1980. The decay of ergativity in Pamir languages. Lingua 51. 147–186.
Peeters, Bert. 2000. Book review M. Lemmens, Lexical perspectives on transitivity and
ergativity: causative constructions in English. Studies in Language 24(3). 683–694.
Perlmutter, David M. 1983. Studies in relational grammar. Chicago: University of Chicago Press.
Peterson, John M. 1998. Grammatical relations in Pali and the emergence of ergativity in
Indo-Aryan. München: Lincom.
Peterson, John M. 1999. The Nepali subordinated verb. In Yogendra P. Yadava and William W.
Glover (eds.), Topics in Nepalese linguistics, 337–370. Kathmandu: Royal Nepal Academy.
Peterson, John M. 2002. The Nepali converbs: a holistic approach. In Rajendra Singh (ed.), The
yearbook of South Asian languages and linguistics, 93–135. New Delhi: Sage.
Plank, Frans (ed.). 1979. Ergativity. Towards a theory a theory of grammatical relations. London:
Academic Press.
Pokharal, Madhav P. 1998. Nepali vakya vyakaran [Nepali syntax]. Kathmandu: Ekta Books.
Polinsky, Maria and Bernard Comrie. 1999. Agreement in Tsez (Nakh-Daghestanian language
family). Folia Linguistica 33(1–2). 109–130.
 References 311

Post, Mark W. 2008. Grammaticalization and the discourse distribution of serial verbs in
Assamese. In Wilaiwan Khanittana and Paul Sidwell (eds.), SEALSXIV(2): papers from
the 14th meeting of the Southeast Asian Linguistics Society, 63–74. Canberra: Pacific
Linguistics.
Pott, August F. 1873. Unterschied eines transitiven und intransitiven Nominativs. Beiträge zur
vergleichenden sprachforschung auf dem Gebiete der arischen, celtischen und slawischen
Sprachen 7. 71–94.
Pray, Bruce R. 1976. From passive to ergative in Indo-Aryan. In Manindra K. Verma. The notion of
subject in South Asian languages, 195–211. Madison: University of Wisconsin.
Prem. 1984. Khajano [Treasure]. In Rawat Saraswat and Prem (eds.), Aaj ri Rajasthani
kahaniyan [Contemporary Rajasthani stories], 91–94. New Delhi: Sahitya Akademi.
Primus, Beatrice. 1999. Cases and thematic roles: ergative, accusative and active. Tübingen:
Niemeyer.
Primus, Beatrice. 2009. Case, grammatical relations, and semantic roles. In Andrej Malchukov
and Andrew Spencer (eds.), The Oxford handbook of case, 261–275. Oxford: Oxford
University Press.
Primus, Beatrice. 2011. Case-marking typology. In Jae J. Song (ed.), The Oxford handbook of
linguistic typology, 303–321. Oxford: Oxford University Press.
Radice, William. 2007. Teach yourself Bengali. London: Hodder Education.
Ray, Sidney H. and Alfred C. Haddon. 1893. A study of the languages of Torres Straits with
vocabularies and grammatical notes, Part I. Proceedings of the Royal Irish Academy, Third
Series II. 463–616.
Ray, Tapas. 2003. Oriya. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan
languages, 444–476. London: Routledge.
Regamey, Constantin. 1954. A propos de la construction ergative en indo-aryan modern Sprach-
geschichte und Wortbedeutung. In G. Redard, Festschrift Albert Debrunner, 363–381.
Bern: Francke.
Riccardi, Theodore. 2003. Nepali. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan
languages, 538–580. London: Routledge.
Robinson, William. 1839. A grammar of the Assamese language. Serampore.
Robson, Barbara and Habibullah Tegey. 2009. Pashto. In Gernot Windfuhr (ed.), The Iranian
languages, 721–772. London: Routledge.
Rumsey, Alan. 1987. Was Proto-Indo-European an ergative language? Journal of Indo-European
Studies 15. 19–37.
Saksena, Baburam. [1937] 1971. Evolution of Awadhi. New Delhi: Motilal Banarsidass.
Sapir, Edward. 1917. Review of Uhlenbeck 1916. International Journal of American linguistics
1(1). 82–86.
Saxena, Anuradha. 1981. Verb agreement in Hindi. Linguistics 19(5–6). 467–474.
Scancarelli, Janine. 1985. Referential strategies in Chamorro narratives: Preferred clause
structure and ergativity. Studies in language 9(3). 335–362.
Schmidt, Ruth L. 2004. A grammatical comparison of Shina dialects. In Anju Saxena. Himalayan
languages past and present, 33–55. Berlin: Mouton de Gruyter.
Schmidt, Ruth L. and Razwal Kohistani. 2008. A grammar of the Shina language of Indus
Kohistan. Wiesbaden: Harrassowitz Verlag.
Schmidt, Wilhelm. 1902. Die sprachlichen Verhältnisse von Deutsch-Neuguinea. Zeitschrift für
afrikanische, ozeanische und ostasiatische Sprachen 6. 1–99.
312 References

Schuchardt, Hugo. 1896. Ueber den passiven Charakter des Transitivs in den kaukasischen
Sprachen. Sitzungsberichte der kaiserlichen Akademie der Wissenschaften 133.
Seely, Jonathan. 1977. An ergative historiography. Historiographia Linguistica 4(2). 191–206.
Shackle, Christopher. 1976. The Siraiki language of central Pakistan: A reference grammar.
London: School of Oriental and African studies.
Shackle, Christopher. 1984. Towards a morphological classification of Kashmiri monosyllabic
nouns. In Omkar N. Koul and Peter E. Hook. Aspects of Kashmiri Linguistics, 46–61. New
Delhi: Bahri Publications.
Sharma, Mukunda M. 1963. Assamese for all. Or, Assamese self-taught. Jorhat: Asam Sahitya
Sabha.
Sharma, Ramnivas. 1992. Rajasthani Lokagathavan [Rajasthani folktales]. New Delhi: Sahitya
Akademi.
Shibatani, Masayoshi, (ed.). 1988. Passive and voice. Amsterdam/Philadelphia: John
Benjamins.
Siewierska, Anna. 1984. The passive. A comparative linguistic analysis. Kent: Croom Helm.
Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press.
Siewierska, Anna. and Dik Bakker. 2009. Case and alternative strategies: word order and
agreement marking. In Andrej Malchukov and Andrew Spencer (eds.), The Oxford
handbook of case, 290–303. Oxford: Oxford University Press.
Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Robert M. W. Dixon (ed.),
Grammatical categories in Australian languages, 112–171. Canberra: Australian Institute of
Aboriginal studies; Atlantic Highlands, NJ: Humanities Press.
Smith, Richard R. 1974. Awadhi/Kannauji transition phenomena and their correlates: a study in
dialect geography. New York: Cornell University dissertation.
Snell, Rupert. 1991. The Hindi classical tradition. A Braj Bhasa reader. London: SOAS.
Sommer, Anton F. W. 1997. Elementargrammatik des Kaschmiri. Wien: Parkcopydruck.
Song, Jae J. 2011. The Oxford handbook of linguistic typology. Oxford: Oxford University Press.
Speijer, Jakob S. [1886] 1998. Sanskrit syntax. New Delhi: Motilal Banarsidass.
Spencer, Andrew. 2005. Case in Hindi. In Miriam Butt and Tracy H. King (eds.), Proceedings of
the LFG05 Conference, Bergen, 429–446. Stanford: CSLI.
Spencer, Andrew. 2006. Syntactic vs. morphological case. Implications for morphosyntax. In
Leonid Kulikov, Andrej Malchukov and Peter de Swart (eds.), Case, valency and transitivity,
3–22. Amsterdam: Benjamins.
Srivastava, Dayanand. 1962. Nepali language. Its history and development. Calcutta: Calcutta
University Press.
Strand, Richard F. 1973. Notes on the Nuristani and Dardic languages. Journal of the African and
Oriental Studies 93(3). 297–305.
Stronski, Krzysztof. 2009. Approaches to ergativity in Indo-Aryan. Lingua Posnaniensis LI.
77–118.
Stronski, Krzysztof. 2010. Non-nominative subjects in Rajasthani and Central Pahari. Lingua
Posnaniensis LII(1). 81–97.
Stump, Gregory T. and Ramawatar Yadav. 1988. Maithili verb agreement and the control
agreement principle. Papers from the annual regional meeting of the Chicago Linguistic
Society 24(2). 304–321.
Subbarao, K. V. 2000. Syntactic typology and South Asian languages. In Rajendra Singh
(ed.), The yearbook of South Asian languages and linguistics, 93–108. New Delhi: Sage
Publications.
 References 313

Tchekhoff, Claude. 1978. Aux fondements de la syntaxe: l’ergatif. Paris: Press Universitaire de
France.
Tessitori, Liugi P. 1914–1916. Notes on the grammar of Old Western Rajasthani. Indian
Antiquary 43–45. 1–106.
Thapa, Suryakala. 2001. Tuntuni ko [About a bird]. New Delhi: Sahitya Akademi.
Thomsen, Ole N. 1994. Dyirbal ergativity and embedding. A functional-pragmatic approach.
Studies in language 18(2). 411–488.
Tiwari, Udai N. 1960. The origin and development of Bhojpuri. Calcutta: The Asiatic Society.
Tournadre, Nicholas. 1996. L’ergativité en Tibétain. Louvain: Editions Peeters.
Trask, Robert 1979. On the origins of ergativity. Frans Plank (ed.), Ergativity. Towards a theory of
grammatical relations, 385–404. London: Academic Press.
Trask, Robert. 2002. Ergativity and accusativity in Basque. In Kristin Davidse and Beatrice
Lamiroy (eds.), The nominative and accusative and their counterparts, 265–284.
Amsterdam: John Benjamins.
Trombetti, Alfredo. 1903. Delle relazioni delle lingue caucasiche, Part II. Giornale delle Società
asiatica italiana 16. 145–175.
Trubetzkoy, Nikolai S. 1939. Le rapport entre le déterminé, le déterminant et le défini. In Charles
Sechehaye and Charles Bally (eds.), Mélanges de linguistique, offerts à Charles Bally,
75–82. Genève : Georg et Cie.
Trumpp, Ernest. 1872. Grammar of the Sindhi language. Osnabrück: Biblio Verlag.
Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21(2). 385–396.
Tuite, Kevin J., Asif Agha and Randolph Graczyk. 1985. Agentivity, transitivity and the question
of active typology. Papers from the annual regional meeting of the Chicago Linguistic
Society 21(2). 252–270.
Turnbull, Archibald. 1982. Nepali grammar and vocabulary. New Delhi: Asian educational
services.
Uhlenbeck, Christianus C. 1901. Agens und Patiens im Kasussystem der indogermanischen
Sprachen. Indogermanische Forschungen 12. 170–171.
Uhlenbeck, Christianus C. 1916. Het passieve karakter van het verbum transitivum of van het
verbum actionis in taalen van Noord-Amerika. Verslagen en Mededeelingen der Koninklijke
Akademie van Wetenschappen 5(2). 187–216.
Ura, Hiroyuki. 2006. A parametric syntax of aspectually conditioned split-ergativity. In Alana
Johns, Diane Massam and Juvenal Ndayiragije (eds.), Ergativity: Emerging issues, 111–142.
Dordrecht: Springer.
Vaillant, André. 1936. L’ergatif Indo-Européen. Bulletin de la société de linguistique de Paris 37.
93–108.
Van Valin, Robert D. J. 1990. Semantic parameters of split ergativity. Language 66. 221–260.
Van Valin, Robert D. J. and Randy J. LaPolla. 1997. Syntax. Structure, meaning and function.
Cambridge: Cambridge University Press.
Verbeke, Saartje. 2010. Postpositions and agreement in the Rajasthani languages. Indoger-
manische Forschungen 115. 35–64.
Verbeke Saartje and Ludovic De Cuypere. 2009. The rise of ergativity in Hindi. Assessing the
role of grammaticalization. Folia Linguistica Historica 30(1). 367–389.
Varma, Dhirendra. 1935. La langue Braj (Dialecte de Mathura). Paris: Librairie d’Amérique et
d’Orient Adrien-Maisonneuve.
314 References

Verma, Manindra K. 1976. The notion of subject and the data from Nepali. In Manindra K. Verma
(ed.), The notion of subject in South Asian languages, 270–285. Madison: University of
Wisconsin.
Verma, Manindra K. and K. P. Mohanan, (eds.). 1990. Experiencer subjects in South Asian
languages. Stanford: CSLI.
Verma, Manindra K. and T. N. Sharma. 1979. Intermediate Nepali Structure. New Delhi:
Manohar.
Verma, Sheela. 2003. Magahi. In George Cardona and Dhanesh Jain (eds.), The Indo-Aryan
languages, 498–514. London: Routledge.
Villar, Francisco. 1984. Ergativity and animate/inanimate gender in Indo-European. Kuhns
Zeitschrift 97(2). 167–197.
Von Heusinger, Klaus. 2011. Specificity. In Klaus von Heusinger, Claudia Maienborn and Paul
Portner. Semantics: An international handbook of natural language meaning. Vol 2,
1025–1058. Berlin: Mouton de Gruyter.
Wali, Kashi. 2004. Nonnominative subjects in Marathi. In Peri Bhaskararao and K. V. Subbarao
(eds.), Non-nominative subjects. Vol I, 245–274. Amsterdam: Benjamins.
Wali, Kashi. 2005. Marathi. Muenchen: Lincom Europa.
Wali, Kashi. 2006. Marathi. New Delhi: IILS.
Wali, Kashi. and Ashok K. Koul. 1994. Kashmiri clitics and ergative case. Indian Linguistics 55.
77–96.
Wali, Kashi and Ashok K. Koul. 2002. Kashmiri clitics: The role of case and CASE. In Omkar
N. Koul and Kashi Wali (eds.), Topics in Kashmiri linguistics, 17–42. New Delhi: Creative
Books.
Wali, Kashi and Omkar N. Koul. 1997. Kashmiri: A cognitive-descriptive grammar. London:
Routledge.
Wallace, William. 1981. Object marking in the history of Nepali: A case of syntactic diffusion.
Studies in the linguistic sciences 11(2). 107–128.
Wallace, William. 1982. The evolution of ergative syntax in Nepali. Studies in the linguistic
sciences 12(2). 147–211.
Wallace, William. 1985. Subjects and subjecthood in Nepali: An analysis of Nepali clause
structure and its challenges to RG and GB. Illinois: University of Illinois, Urbana.
Whitney, William Dwight. 1896. A Sanskrit grammar, including both the classical language, and
the older dialects, of Veda and Brahmana. Leipzig: Breitkopf und Härtel.
Wichmann, Søren. 2008. The study of semantic alignment: retrospect and state of the art. In
Marc Donohue and Søren Wichmann (eds.), The typology of semantic alignment, 3–23.
Oxford: Oxford University Press.
Wierzbicka, Anna. 1980. Case marking and human nature. Australian journal of linguistics 1.
43–81.
Willems, Klaas. 1997. Kasus, grammatische Bedeutung und kognitive Linguistik: ein Beitrag zur
allgemeinen Sprachwissenschaft. Tübingen: Tübinger Beiträge zur Linguistik.
Willems, Klaas. 2000. Form, meaning and reference in natural language: A phenomenological
account of proper names. Onoma 35. 85–119
Windfuhr, Gernot (ed.). 2009. The Iranian languages. London: Routledge.
Windfuhr, Gernot and John R. Perry. 2009. Persian and Tajik. In Gernot Windfuhr (ed.), The
Iranian languages, 416–544. London: Routledge.
Woolford, Ellen. 2006. Case-agreement mismatches. In Cedric Boeckx (ed.), Agreement
systems, 299–316. Amsterdam: Benjamins.
 References 315

Wunderlich, Dieter. 1985. Über die Argumente des Verbs. Linguistische Berichte 97. 183–227.
Yadava, Yogendra P. 1999. The complexity of Maithili verb agreement. In Rajendra Singh
(ed.), Yearbook of South Asian languages and linguistics, 139–152. New Delhi: Sage
Publications.
Yip, Moira, Ray Jackendoff and Joan Maling. 1987. Case in tiers. Language 63(2). 217–250.
Zaenen, Annie, Joan Maling and Höskuldur Thráinsson. 1985. Case and grammatical functions:
The Icelandic passive. Natural Language and Linguistic Theory 3. 441–483.
Zwicky, Arnold M. 1985. Clitics and particles. Language 61(2). 283–305.
Index of subjects
A see S, A, O (P) case marking 5–7, 9–14, 16, 22–23, 26–36,
ablative 79, 83, 96, 136, 176–178, 225 38, 40–44, 49, 51–52, 55, 61, 64, 106,
absolutive 10, 13, 22, 24, 27, 29–31, 45, 50, 109, 156, 197, 248–263
52–54, 58, 79, 194, 196–197 – Asamiya 113–114, 141
accusative agreement 26, 43, 156, 172, – Kashmiri 176–180
253–254 – Marwari 213, 223
accusative alignment 13, 22, 27, 29, 34, – Nepali 147–148, 153–156
41–43 , 49, 135, 137, 256 case syncretism 30, 78–80
accusative case 10, 14, 30, 57, 73–74, causative 25, 60, 97–98, 104, 149, 264
231–232, 266–267 causativization 60, 98
accusative case marking 236, 253, 255, 257, clitic 30, 33, 42, 188–189, 205, 210–211
258, 263 coding properties 13–14, 16, 19, 22, 29, 43,
active alignment 12 50, 52, 262
active language 40, 45, 109, cognate object 108, 118, 148
adjectival participle 33, 87, 133, 140, 145 competing motivations 46–47
adjunct 28–29, 31 conjunction reduction 16, 100, 150
adposition 30, 42 construction, definition 26–27
affectedness 37–38, 62–63, 267 contentive typology 12
agency 21–22, 32, 36, 39, 43, 59, 62–63 control 36, 107–108, 171
agent 1, 12, 18–21, 36–37, 39–44, 46–49, 53, converb 84–85, 91–92, 100, 102, 116–118,
56–63, 76–80, 83–88, 92–99, 102, 107, 149–151, 174
109–110, 114, 148–149, 161, 172, 227, coordination 16–17, 22, 52, 102, 150, 248,
238, 248, 252, 259–261, 264 261
Aktivus 56–57, 59 copula 68, 91, 126, 129, 132–133, 139–140,
alignment 2, 5–7, 9–16, 22, 25–27, 29, 153–155, 163, 173–174, 180–185, 203,
34–36, 40– 47, 52–55, 63–64, 263 207, 209, 214–217, 246, 257
ambitransitive verbs 51, 98 coreference 100–101, 255
animacy 35–39, 70–74, 93, 137, 145, 148, cross-referencing see verb agreement
209, 227–232, 239, 242, 248, 253–254,
261–262, 267 Dardic 175
anti-absolutive 194, 197 dative 27–28, 40, 63, 68, 73–74, 79, 85, 98,
aspect 41, 49, 94–96, 126–127, 129, 152, 121, 196, 210
166, 180, 248–250, 256, 260, 263, 266 definiteness 35–39, 70–74, 93, 119,
attention flow 48–49 142–145, 184, 193, 199, 209, 226–232,
autonomy/dependence 52 239, 242, 248, 253–256, 261–262, 267
differential object marking 36–40, 70, 74,
behaviour-and-control properties 22–23 119, 162, 218, 227, 230–233, 247, 254,
beneficiary 28, 74, 121–122, 262 260
borrowing 91, 173, 238, 257 differential subject marking 36, 38–40, 162
bound pronouns 33–34, 43, 145, 156, direct object 1, 9–10, 14–15, 18–19, 22, 27,
188–190, 204, 210 50–51, 57, 79
‘bound’ ‘free’ split 43 discourse 34, 40, 44–48, 53, 102, 145, 228,
239, 248, 255–256, 261
 Index of subjects 317

discriminatory approach see distinguishing grammatical roles 20, 22, 25, 28, 100, 244
function of case grammaticalization 46–47, 55, 88, 91
distinguishing function of case 31–32, 34,
36–37, 55, 61, 74, 161–162, 229–230, habitual 94, 164–165, 180, 206
234, 247, 252, 256, 261 head- and dependent-marking 13, 41–44,
ditransitive construction 24, 28, 180, 226 189, 197
double object construction 25 hierarchical alignment 179
Hindi belt 67, 212, 246, 259
ellipsis 21, 55, 100–102, 248 honorificity 116, 126, 134–138, 148
empathy 136–137 humanness 32, 35, 39, 46, 71–72, 114, 142,
ergative agreement see O–agreement 226
ergative alignment 12, 19–20, 30, 34, 41, 43,
49, 59, 63, 135, 252, 256, 265–266 immediacy of involvement 53
ergative case 1–2, 13, 30, 36, 39, 42–43, imperfective, definition 41, 63, 96
48–53, 56–60, 63, 80–86 , 92, 94, 100, Inaktivus 59
103, 109, 115, 117, 142–144, 152, 156, incorporation 93
166, 169–173, 177–178, 181, 185–186, indefinite-NP deletion 17
196, 207–208, 213, 218–224, 236, 240, indefinite past 181
246, 248, 250, 252, 255, 257, 262–266 indexing/flagging 32
ergative case marking (pattern) 6, 26, 43, indexing function of case 31–32, 34, 37, 74,
143, 156, 185, 208–209, 223, 253, 234, 252, 261
256–257, 263, 266–267 indirect object 22–25, 27, 85,
ergative language 8, 12, 31, 45, 53, 56, 59, indirective alignment 24, 27, 74, 262
63, 81, 86, 260 individuation 37–38, 62, 267
ergative marking see ergative case Indo-Aryan language area 66–67
ergativity, definition 12, 22, 31, 45, 56 infinitive 92, 118, 123, 148–149, 152, 238
experiencer 29, 63, 99–104, 106–108, instrumental 24, 30, 56, 63, 76–79, 83–88,
124–126, 134, 149, 151–152, 171, 174, 91, 103, 113–114, 147–149, 152, 177, 213,
225–227, 229, 233–234, 237–240, 265 238
extended differential object marking 74, 262 inverse 175, 178, 180, 182, 199
IO, definition 28, 262
face 136–137
fluid S-marking 19 ko Hindi 68–74, 90, 96, 100–101, 106–107,
focalization 159, 161–162 155, 199, 230, 233–234, 237, 239–240,
focus 35, 41, 85, 159–162, 165–166, 168 243, 247, 254, 263

gender/number agreement 33, 60, 92, language contact 6–7, 111, 143–144,
132, 134, 140, 144–145, 154–156, 170–173, 175, 239, 257
169–173, 182, 185, 201, 203, 247, 254, le Nepali 147–152, 156–171, 174, 262–263,
256, 259 265
genitive 31, 58, 79, 85–88, 90, 92, 98, lexical compound verb 92–94, 108, 118
113–114, 122, 124–125, 147, 210, 213 lexical–semantic ergativity 50–53
givenness 49 light verbs 54–55, 64, 92–97, 110
goal 28, 48–49, 52, 123, 201, 210 locative 28, 31, 79, 90, 103, 113–114, 123,
grammatical relations 13–14, 21–29, 34, 43, 133, 145, 177, 226, 234
48–50, 59, 62–64, 73, 90, 100, 231, 241,
261–262
318 Index of subjects

macroroles 20–21 P see S, A, O (P)

marked nominative 115, 119 palatalization 177, 181
markedness 32, 36–40, 81–83, 162, 252 passive 1, 47–48, 51, 55–57, 60, 76–88,
markedness reversal 38–39 96–97, 103, 114, 136, 149, 214–215, 264
(Masica’s) layer system 89–90, 113, 177, 213, Passivus 57, 59
240, 244 patient 12, 18–21, 24–25, 28, 30, 32, 36–37,
mixed ergativity 42 41, 44, 48, 55, 61–63, 76–80, 85, 96–99,
morphological ergativity 12, 16, 51–52, 248 104, 106, 233, 264
patient-oriented construction 41, 85
ne perfect participle 69, 75–80, 83–84, 87, 91,
– Hindi 1, 68–69, 83, 90–91, 103–107, 144, 95–96, 104, 114, 149,
158, 170, 174, 237–240, 247, 262–263 174, 180–181, 187, 207, 209, 243
– Harauti 218, 224–228, 239–240, 249–250, perfective (definition) 41, 49, 63, 95–96,
253, 263 264
– Marwari 213, 218, 230–232, 234, 239, 263 perfectivity 96, 166, 169–170, 174, 263
neutral alignment 74 peripheral argument 48, 53
nominal ergativity 42, 170 periphrastic construction 76, 87
nominality (hierarchy) 35, 37, 39 person agreement 32–33, 91, 128–129, 132,
nominalization 51 139, 145, 155–156, 169, 173, 179, 189,
nominative case 1, 14, 18, 30, 36, 45, 52, 201–205, 209, 215, 243, 257, 259–260
57–61, 73, 79–80, 85, 100, 115, 142–143, person hierarchy 35, 39, 179, 206, 223,
151, 172–173, 178–179, 182–185, 191, 254–255
193, 200, 222, 225, 231, 236–244, 250 pivot 17, 21, 27
non-canonical argument 29, 98, 111, 124–125 possessor 31, 88, 136, 202, 210
postposition (definition) 89–90
O see S, A, O (P) preferred argument structure 46–47
O-agreement 43, 56, 86, 134, 137–138, primary object 24, 27
144–145, 153–155, 170, 186, 217–218, pronominal affixes 33, 155–156, 188
236, 240, 243–247, 250, 254, 257–258, pronominal suffixes 6–7, 100, 175–176,
265–266 179–180, 182–212, 250, 253–261, 266
object 6, 9, 11, 13–14, 19, 20, 22–24, 27–29, protoroles 20–21
43, 45, 49–52, 63, 74, 78, 86, 100, 109, proximate past 181
134–138, 148, 196, 201–205, 258
objective 68, 70, 73–74, 90–91, 96, 98–100, rah 92–93, 95, 158
103, 106–107, 113, 119, 123–125, 135, raising 197
149, 155, 174–180, 184–185, 191–192, reanalysis 55, 80–82, 155–156, 166, 169, 265
197–201, 206, 209–210, 234, 237–239, recipient 18, 24–25, 28, 74, 121–122, 225,
250–254, 258, 262–263 233, 262
obligation 99, 106, 125, 151–152, 174, 204, referential hierarchies 34–41, 49, 137,
232, 237–239 178–179, 189, 208–209, 217, 224,
oblique 40, 80, 89, 99, 103, 113–114, 147, 248–255, 261, 266
172–173, 178, 196, 204–205, 213–214, reflexivization 22, 100–101, 261
219–223, 234, 239–240, 242, 255, relative clause 52
262–263, 265 remote past 181
optional ergative marking 40 resultative 55, 87–88,
 Index of subjects 319

S see S, A, O (P)
S, A, O (P) 9–10, 15–16, 19–28, 261–262 ta-participle 76–78, 81–88, 264
sample, typological 3–4 tense 38, 41, 49, 53, 56, 64, 75–76, 82,
secondary agreement 134–138, 145, 258 94–96, 248–250, 256
secondary object 24, 27 tense/aspect/mood split 41–42, 47–49, 63,
secundative alignment 24, 27, 262 248–250, 256, 260, 263, 265–266
semantic alignment 19, 21, 30, 39–40, 44, theme 18, 28, 51–53, 104, 263–264
265 topic 46–49, 52, 61, 162
semantic case 29, 31, 58, 94 topicality 22–23, 36–37, 39, 56, 61, 101–102,
semantic roles 14, 18–20, 31, 43, 46, 61, 63 162, 228–229, 248, 257, 265–267
semi-ergative 79, 84 topicalization 189
semi-passive 79 transitive conjugation 118, 127–131, 133,
Silverstein’s hierarchy 35–36, 39, 47, 135, 140, 181
137, 179 transitivity 37, 40, 51, 56, 60–63, 108, 131,
source 48, 59, 103, 201–202, 210 166, 181, 264–267
specificity 71, 199, 227, 232–233 transitivizing 98, 264
speech act participants 135–137, 178, 206, tripartite alignment 10, 42, 44, 209, 253, 255
217, 234–236 typological ergativity 50–52
split agreement 215–217
split alignment 10, 34–35, 41, 43, 47, 54, unaccusative 108
248 unergative 108–110, 117–118, 144, 148, 181,
split ergativity 68, 251 265
split intransitivity 19
split S-marking 19 V2 175, 209
stage- and individual-level verb agreement 5, 7, 10–15, 22, 26–29,
predicates 163–166, 169 32–34, 36, 40, 42–44, 60, 106, 135,
stative 19, 44, 104–106, 132–133 138, 155, 197, 231–232, 234, 240,
stimulus 99–100, 103, 104, 125, 225, 233, 248–266
265 – Asamiya 126–127
structural case see syntactic case – Nepali 152–153
subject 8, 10, 13–16, 18–23, 25, 27, 29, verb stapling 92
45–46, 50, 52, 57–59, 61, 63, 68, 84–85, verbal ergativity 42, 170–171, 173
98–100, 102–103, 261–262 viewpoint 48–49, 103
syntactic case 29, 31, 94, 109–110, 229, 266 volitionality 62, 107, 109, 161
syntactic ergativity 12, 16–18, 51, 248
syntactic relations see grammatical relations word order 11, 22–23, 38, 84, 175, 261
Index of languages
Apabhramsha 78–79 Magahi 67, 111–112, 131–133, 137–140, 145,
Asamiya 6, 66–67, 90, 93, 102, 111–129, 131, 252, 258, 265, 269, 280
142–145, 147–149, 156, 158, 169–174, Maithili 66–67, 100, 111–112, 131–140, 145,
178, 209, 252, 255–257, 260, 262–263, 258, 265, 269, 279
268, 275–276 Marathi 66–67, 92, 176, 196, 206–208, 212,
Awadhi 67, 112, 139–141, 212, 245–246, 252, 223, 231, 238–240, 248, 250-255, 260,
265, 269–270, 281 271–272, 287–288
Marwari 67, 196, 213–225, 230–234, 236,
Bangla 65–67, 111–112, 125, 129–131, 130, 239, 241–243, 252–255, 260–263, 272,
142–143, 155, 169, 172, 252, 259–260, 289–290
265–266, 268–269, 276–277
Bhojpuri 67, 111–112, 131, 133, 139–140, 252 Nepali 6, 66–67, 92, 100, 138, 146–174,
269, 278 178, 180, 196, 209, 239, 252–265, 270,
Bihari 6, 66–67, 112, 131, 139, 145, 212, 256, 282–283
258–260
Braj 67, 212, 240–243, 246, 248, 250, 252, Oriya 66–67, 111–112, 125, 129, 142–143, 172
273, 293–294
Bundeli 67, 212, 224, 244–246, 248, 250, Pali 87–88
252, 260, 262, 273–274, 294–295 Poguli 175, 199–201, 210, 248–253,
270–271, 284–285
Dolakha Newar 51, 127, 170–171 Prakrit 78, 82
Dyirbal 16–18, 52 Punjabi 1, 66, 92, 106, 204–205, 212, 218,
235, 234–240, 247–255, 263, 273,
Early Hindi 80, 90 292–293

German 14–15, 24, 58, 88, 98 Rajasthani 7, 66–67, 82, 91, 97, 146, 155–156,
Gujarati 1, 66–67, 212, 216–218, 223, 208–209, 212–214, 218–219, 223, 231,
230–236, 239, 241, 247–250, 252, 234, 237, 240–241, 247, 249, 254,
254–255, 260, 263, 273, 291–292
Sanskrit 2, 65–66, 75–78, 84–87, 91
Harauti 76, 155, 208, 213, 218, 223–231, Shina 67, 100, 170, 172–175, 210, 256–257,
239–240, 248–250, 252–254, 261, 263, 260, 264–265, 272, 288–289
272, 290–291 Sindhi 66–67, 201–205, 210, 248, 250, 254,
Hindi 65–83, 88–110, 112, 124, 139, 141, 256, 259–260, 263, 265, 271, 285–286
144, 151–158, 170, 174–175, 180, 199, Siraiki 67, 201, 204–206, 210, 248, 250, 256,
208–209, 212, 214, 217–218, 223–227, 258, 271, 286–287
230–234, 237–248, 250, 252, 254,
260–265, 268, 274–275 Tibetan 143–144, 170, 173, 257
Hindi – Urdu 1–7, 65–68, 259–260 Tsez 15, 18
Turkish 54
Kabardian 58
Kashmiri 6, 66–67, 92, 102, 172, 175–176, Urdu 65–66, 175, 237
178–201, 206–210, 248–253, 256–265,
270, 283–284 West Greenlandic 24
Koasati 60
Kundri 244–246