An Investigation of the Syntactic, Semantic and Pragmatic Interfaces

Reciprocity in Russian: An investigation of the syntactic, semantic and pragmatic interfaces

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Lauren Ressue

Graduate Program in Slavic and East European Languages and Cultures

The Ohio State University

2015

Dissertation Committee:

Andrea Sims, Advisor

Judith Tonhauser, Advisor

Daniel Collins

Lauren Ressue

2015

Abstract

This dissertation explores two reciprocal expressions in Russian, drug druga and reciprocal -sja

verbs to determine their distribution, semantics and pragmatics. I argue that while these two

expressions are similar in many ways, they also differ in subtle ways not before discussed. While

the empirical foundations of this dissertation are data from Russian, my findings have empirical

and theoretical consequences for both the formal semantic and the typological literature on

reciprocity.

In English, the reciprocal expression each other has been studied in detail to explore the

relations between participants it is compatible with (Langendoen 1978, Dalrymple et al. 1998b,

Beck 2001). For example, the sentence in (i) is compatible with either of the relations depicted in

(a) and (b), where the arrows represent a binary painting relation.

(i) The four artists painted each other.

(a) (b)

In this dissertation, I explore the meaning of reciprocal expressions in Russian to discover

whether these expressions have the same meaning as each other and whether they contribute any ii meaning other than relations between individuals. My data comes from both a corpus study utilizing the Russian National Corpus and elicitation with native speakers of Russian.

My results suggest that while both drug druga and the -sja verbs, like each other, also convey more than one relation between individuals they also contribute other semantic content to a sentence. For example, they both introduce a temporal restriction on events. Some of the -sja verbs semantically restrict events to simultaneity, and I argue that drug druga gives rise to an implicature that the events are simultaneous. I also explore two different syntactic constructions the reciprocal -sja verbs occur in and find a number of semantic differences between the two constructions. I conclude that the syntactic environment of a reciprocal expression can affect its semantics. I furthermore find that drug druga is restricted against occurring with the preposition s when it introduces a relation of accompaniment to a sentence.

My findings suggest that while reciprocal expressions convey a set of relations between individuals, they also contribute other semantic and pragmatic content that is sensitive to a number of factors. Furthermore, I argue that while reciprocal expressions exhibit some uniformity in their meanings, these meanings also differ. I provide a formal analysis based on those proposed for English each other by Dalrymple et al. (1998b), Sabato and Winter (2005) and

Dotlačil and Nilsen (2008). My analysis extends this account to the Russian expressions and captures other semantic and pragmatic properties beyond relations between individuals. The core of the semantic analysis consists of proposed lexical entries for drug druga and the reciprocal -sja verbs, whose truth conditions allow the right predictions about the behavior of these expressions in comparison to each other (and each other) and other reciprocal expressions.

iii

Dedication

This document is dedicated to my mother, for teaching me how to love every step of the way.

Acknowledgements

I must start my acknowledgements by expressing my great appreciation to my two advisors, Drs. Andrea Sims and Judith Tonhauser, for their patience and extreme dedication to me and my project. They have taught me so much about linguistics, writing and the investigative process. What’s more, they have devoted a vast amount of time to giving me detailed and insightful comments on my work and to meeting with me to instruct and guide. This dissertation is theirs as much as it is mine.

I also thank Dr. Daniel Collins, my third committee member, for his instruction and support through my graduate career, as well as Dr. Brian Joseph for his insightful comments and suggestions, especially in the beginning stages of this dissertation.

Thank you to my four consultants, Yuliia Aloshycheva, Helen Myers, Marina Pashkova and Evgenija Tretiakova. I am indebted to you for your patience with the absurd contexts and questions I have sent you over the years and for the time you have devoted to helping me.

I am forever grateful to the Hilandar Research Library for taking me on as a GA, and to

Dr. Predrag Matejic, M.A. Pasha Johnson, Helene Senecal and Lyubamira Gribble for being wonderful to me for two years. I learned so much from you and don’t know how I was so lucky to be able to work with you. Please include this in your “works supported” list - this dissertation was certainly supported by all of you. And further thanks to Pasha for the friendship, for taking me in when I was homeless, and for all the book recommendations.

As with many graduate students, I was greatly influenced in everything by the cohort I started with: Yuliia Aloshycheva, Maggie Harrison, Kirk Jorgensen, Michael Nichols, Ceilidh

Orr and Daria Safronova. Without these fellow students, I might not have gained and retained the motivation to continue through everything. We were a great group. My thanks most especially to

Maggie Harrison and Yuliia Aloshycheva for all the laughs, adventures and challenges we experienced together.

A special thank you also to Ceilidh Orr for being there for and through everything. You were a great sounding board throughout the dissertation for everything and anything. I owe a part of Chapter 3 to you. But also, what would the dissertation have looked like without all that

Shakespeare, Casanova and kofe?

Thanks to my Apropos group that helped me find a place at OSU. I miss you guys.

I am further grateful for my colleagues in the Department of Slavic and East European

Languages and Cultures for their comments on various presentations and other less tangible support. Thanks to Maria Alley, Brian Brookes, Matt Curtis, Michael Furman, Nina

Haviernikova, Marcela Michalkova, Jeff Parker, Robert Reynolds, Anastasia Smirnova, Larysa

Stepanova, Susan Vdovichenko, and Kate White. Thanks too to Mario Rodriguez Polo and Isa

Arranz del Riego for their friendship and constant inspiration.

My gratitude to Jessie Labov, who whisked me out of Columbus at the perfect time and gifted me with as much free time as possible to write. I owe Chapter 5 to you.

Thank you to the crew at 701. You distracted me at the right times and reminded me that there are things outside of reciprocity. I was lucky to land with you. Thanks especially to

Elizabeth Custodio and Meron Bekele for keeping me afloat evening after evening.

Through everything, I’m sure that what kept me sane was the early morning skating. So thanks to Tony Habart, Lindsay Martin, Dayna Jalkanen and Kristen Spickard for all being part of it and being generally wonderful.

As for my family, thank you first to Robert, who always asked the right questions and had the right pieces of advice throughout this process. You’ve always been an inspiration for me.

Karen: thank you for your perpetual support and cheerleading through this process.

And of course my eternal thankfulness to my mother, to whom I dedicate this dissertation. None of this would have been possible without your constant kindness and wisdom.

When I’m stuck in life, you always suggest just the right, unique, positive perspective. Such guidance has been immensely useful in developing this dissertation and beyond.

And John. What can I say? Thank you for the plane tickets, weather reports, conversion assistance and careful proof-reading, all of which I couldn’t have done without. Thank you for bearing with me throughout all of this. Za!

vii

Vita

2006 ...... B.A. Modern languages, Beloit College

2007-2008 ...... University Fellowship

2008-2012 ...... Graduate Teaching Associate, Slavic and East ...... European Languages and Cultures, The Ohio ...... State University

2009 ...... M.A., Russian Linguistics, The Ohio State ...... University

2012-2014 ...... Graduate Research Associate, Resource Center ...... for Medieval Slavic Studies

Publications

2013 Review of The Semantics of Genitive Objects in Russian: A Study of Genitive of Negation and Intensional Genitive Case, by Olga Kagan. Springer, 2012. LinguistList.

2010 Formal Studies in Slavic Linguistics, co-edited with Anastasia Smirnova and Vedrana Mihaliček. Newcastle upon Tyne: Cambridge Scholars Publishing.

Fields of Study

Major Fields: Slavic and East European Languages and Cultures

viii

Table of Contents Abstract ...... ii

Dedication ...... iv

Acknowledgements ...... v

Vita...... viii

List of Tables ...... xiv

List of Figures ...... xv

Abbreviations in the Glosses ...... xvii

1 Introduction ...... 1

1.1 Motivation and goals ...... 4

1.2 Dissertation outline ...... 8

1.3 Terminology ...... 10

1.3.1 Proposed terminology and definitions ...... 11

1.4 Background on the syntax and semantics of drug druga and the -sja verbs ...... 22

1.4.1 Drug druga ...... 22

1.4.2 Verbs suffixed with -sja ...... 26

1.1 Methodology ...... 43

1.2 Looking forward ...... 47

2 Formal foundations ...... 48

2.1 Formal framework ...... 48

2.1.1 Krifka (1998) ...... 53

2.1.2 Analysis of a fragment of Russian ...... 58

2.2 Formal accounts of reciprocity ...... 62 ix

2.3 Each other conveys four relations ...... 80

2.3.1 The Cover mechanism of Schwarzschild (1996) ...... 86

2.4 Analysis of each other ...... 93

2.5 Preliminary formal definition of drug druga ...... 98

2.6 Preliminary formal definition of the -sja verbs ...... 99

2.7 Summary ...... 101

3 Reciprocal expressions and temporality ...... 103

3.1 Previous literature regarding the interaction temporality and reciprocity ...... 104

3.2 Definitions of terms ...... 112

3.3 Temporality and drug druga ...... 114

3.3.1 Temporal entailments and drug druga ...... 114

3.3.2 Temporal implicatures and drug druga ...... 115

3.3.3 Formal analysis of drug druga ...... 126

3.4 Temporality and the -sja verbs ...... 127

3.4.1 Temporal entailments and the -sja verbs ...... 128

3.4.2 Previous analyses of reciprocal verbs and events ...... 138

3.4.3 Formal analysis of -sja verbs ...... 146

3.4.4 Discussion of the findings on the -sja verbs ...... 150

3.5 Conclusions ...... 151

4 The -sja verbs and the syntax-semantics interface ...... 155

4.1 Basic properties of the two constructions ...... 159

4.1.1 Basic properties of the simple reciprocal construction ...... 159

4.1.2 Basic properties of the discontinuous reciprocal construction ...... 163

4.2 Previous literature on the discontinuous construction ...... 167

4.3 Methodological preliminaries about the corpus study ...... 182

4.4 The meaning of the simple reciprocal construction ...... 184

4.4.1 Strong Reciprocity (SR) ...... 184

4.4.2 Intermediate Reciprocity (IR) ...... 185

4.4.3 One-way Weak Reciprocity (OWR) ...... 187

4.4.4 Revised Intermediate Alternative Reciprocity ...... 189

4.4.5 Intermediate Alternative Reciprocity (IAR) ...... 191

4.4.6 Summary ...... 193

4.5 The meaning of the discontinuous reciprocal construction ...... 193

4.5.1 Semantics of the discontinuous construction in Russian ...... 194

4.5.2 Consequences of semantic generalizations for the discontinuous construction ... 205

4.5.3 The discontinuous construction and two-set relations ...... 213

4.5.4 Summary ...... 221

4.6 A semantic analysis of the -sja verbs ...... 221

4.6.1 An analysis of the simple construction ...... 222

4.6.2 An analysis of the discontinuous construction ...... 227

4.7 Theoretical conclusions ...... 258

4.7.1 Reciprocal relations ...... 258

4.7.2 Significance of syntactic construction ...... 260

5 Accompaniment and reciprocity: the meaning of drug druga ...... 262

5.1 Introduction ...... 262

5.2 Reciprocal relations and drug druga ...... 266

5.2.1 Strong Reciprocity ...... 267

5.2.2 Intermediate Reciprocity (IR) ...... 268

5.2.3 One-Way Weak Reciprocity ...... 269

5.2.4 Revised Intermediate Alternative Reciprocity ...... 271

5.2.5 Interim summary ...... 272

5.3 Drug druga and accompaniment ...... 273

5.3.1 Background on Accompaniment ...... 273

5.3.2 Definitions and a diagnostic for accompaniment ...... 280

5.3.3 Evidence that drug s drugom does not express accompaniment ...... 286

5.3.4 Analysis of accompaniment phrases ...... 291

5.3.5 Interim summary ...... 296

5.4 Drug druga and discontinuous constructions ...... 296

5.4.1 Naturally reciprocal verbs and drug s drugom ...... 302

5.4.2 Analysis of the discontinuous construction...... 307

5.5 Possible counterexamples ...... 308

5.6 Implications and conclusion...... 315

5.6.1 Drug druga is compatible with the same relations as each other ...... 317

5.6.2 Drug druga can be used productively as the object of prepositional phrases ...... 318

5.6.3 Comitativity and reciprocity ...... 319

6 Conclusions ...... 321

6.1 Summary of main empirical and theoretical conclusions of the dissertation ...... 321

6.2 Implications for theoretical semantics ...... 326

6.2.1 Reciprocal relations ...... 326

6.2.2 Implications for other linguistic areas ...... 328

6.2.3 The structure of Russian and cross-Slavic implications ...... 329

6.3 Implications for typological work ...... 330

6.3.1 Argument versus non-argument ...... 331

6.3.2 Temporality ...... 336

6.3.3 Temporality and relations between participants ...... 345

xii

6.3.4 Accompaniment ...... 348

6.3.5 A preliminary taxonomy ...... 349

Bibliography ...... 352

xiii

List of Tables

Table 1: Details on the corpus study of drug druga ...... 44

Table 2: Summary of proposed parameters for a taxonomy of reciprocity ...... 350

xiv

List of Figures

Figure 1: Depiction of the sentence John and Mary stared at each other...... 1

Figure 2: Depiction of The four girls sat next to each other on a bench...... 2

Figure 3: Possible depiction of The children went to the Halloween party disguised as each other...... 2

Figure 4: Strongly Reciprocal relation between four individuals ...... 12

Figure 5: Depiction of the sentence Mary hugged herself...... 13

Figure 6: Depiction of the sentence Pjat' mal'čikov šli drug za drugom v komnatu. ‘Five boys followed each other into the room.’ ...... 20

Figure 7: Derivation of the sentence Marija tancevala...... 59

Figure 8: Derivation of the sentence Ivan i Marija tancevali ‘Ivan and Marija danced.’ ...... 61

Figure 9: Dalrymple et al.’s (1998b) implicational hierarchy of reciprocal relations ...... 68

Figure 10: Depiction of the sentence The third-grade students in Mrs. Smith’s class gave each other measles...... 82

Figure 11: Depiction of two books ...... 83

Figure 12: Depiction of three books ...... 84

Figure 13: A possible depiction of the sentence The boys kicked each other...... 86

Figure 14: Depiction of the sentence He and scores of inmates slept on foot-wide wooden planks stacked atop each other – like sardines in a can – in garage-sized holes in the ground...... 92

Figure 15: Derivation of the sentence John and Mary hugged each other...... 97

Figure 16: Reproduction of Langendoen’s (1978: 188) depiction of reciprocal relations for two sets ...... 178

Figure 17: Reproduction of Dalrymple et al.’s (1998b: 181) depiction of the relations between the Earth, Sun and Moon...... 181

Figure 18: Depiction of an Intermediate Reciprocal Relation ...... 186

Figure 19: Depiction of Revised Intermediate Alternative Reciprocity for two sets ...... 210

Figure 20: Visualization of a Revised Intermediate Alternative Reciprocity for two sets relation ...... 218

Figure 21: Visualization of an Intermediate Alternative Reciprocity for two sets relation ...... 220

Figure 22: Derivation of a sentence that realizes the discontinuous reciprocal construction ...... 245

Figure 23: Final derivation of a sentence that realizes the discontinuous reciprocal construction ...... 254

Figure 24: Depiction of an Intermediate Reciprocal relation ...... 268

Figure 25: Evans’ (2008: 45) typology of reciprocal expressions ...... 333

xvi

Abbreviations in the Glosses

1PL 1st person plural

1SG 1st person singular

2PL 2nd person plural

2SG 2nd person singular

3PL 3rd person plural

3SG 3rd person singular

ACC Accusative

ACTL Actual

ADV Adverb

DAT Dative

F Feminine

GEN Genitive

IMPF Imperfective

INF Infinitive

INST Instrumental

LOC Locative

M Masculine

NEG Negation

NOM Nominative

NPST Non-past

xvii

PFV Perfective

PL Plural

PST Past

PTCP Participle

REFL Reflexive

RECIP Reciprocal

SG Singular

xviii

1 Introduction

In English, the meaning of the expression each other has proven challenging to characterize. As a case in point, consider the following three sentences which realize each other:

(1) John and Mary stared at each other.

(2) The four girls sat next to each other on a bench.

(3) The four children went to the Halloween party disguised as each other.

In each sentence, each other conveys a particular relation between the individuals denoted by the subject. Note, however, that the relation between individuals conveyed differs in each sentence.

The relation between John and Mary in (1) can be depicted as in Figure 1, where each arrow represents one staring-at relation.

Figure 1: Depiction of the sentence John and Mary stared at each other.

The relation between the six girls expressed by (2) can be depicted as in Figure 2, where each double-headed arrow represents a simultaneous two-way relation, such that each girl stands in a symmetric sitting-next-to relation with two other girls, such that the relations occur simultaneously.

Figure 2: Depiction of The four girls sat next to each other on a bench.

The relation between the children in sentence (3) can be depicted as in Figure 3:

Figure 3: Possible depiction of The children went to the Halloween party disguised as each other.

This sentence conveys a relation such that each child stands in the disguised-as relation to exactly one other child, and the relations are not symmetric - i.e. it is not (necessarily) the case that two

children are disguised as each other (though such an interpretation is possible). The expression each other can therefore convey (at least) three different relations between entities.

Since each other can convey different relations, it poses a puzzle as to how best to characterize its semantics. Furthermore, it begs the question for those interested in other

languages and theories that are cross-linguistically adequate as to whether other reciprocal

expressions are also compatible with different relations between participants.

In this dissertation, I explore two reciprocal expressions in Russian to contribute to our

understanding of reciprocity cross-linguistically. One of these expressions, drug druga, is much

like each other in that it occupies an argument position and refers back to the subject of the

sentence:

(4) Context: Ivan and Marija were dating but recently broke up. They’re furious at each other and have been avoiding each other. You’ve recently heard, however, that they ran into each other and had a huge fight when they saw each other. You tell a mutual friend:

Ivan i Marija videli drug druga v parke. Ivan.NOM and Marija.NOM see.PST.PL each other.ACC in park.LOC ‘Ivan and Marija saw each other in the park.’

The second reciprocal expression I call the reciprocal -sja verbs; these are verbs in Russian that realize the suffix -sja and that contribute a reciprocal meaning to the sentence.1

1 Not all verbs in Russian suffixed with -sja are reciprocal; see Section 1.4.2 for an overview of the syntax and semantics of these verbs. 3

(5) Context: Ivan and Marija have liked each other for a long time. Finally, you see them kissing in the park. You excitedly tell a mutual friend:

Ivan i Marija pocelovali-s'! Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed!’

See Section 1.4 for more details about both of these expressions.

This dissertation is dedicated to exploring the syntax, semantics and pragmatics of these

two expressions to determine their distribution and to provide an analysis that accounts for the

meaning of each expression. The results of my exploration will contribute to our understanding

of reciprocity cross-linguistically.

1.1 Motivation and goals

Scholars have been exploring the semantics of English each other since the 1970’s. They

have been fascinated by the puzzle presented above, that this one expression can convey different

relations between individuals. Many authors have explored which relations the expression is

compatible with (e.g. Fiengo and Lasnik 1973, Langendoen 1978, Dalrymple et al. 1998b, Beck

2001) and have furthermore suggested analyses for how these different meanings arise. In this

way, the semantic contribution of each other has been studied in great detail, and theories for

accounting for its distribution and meaning are well developed (see especially my summary of

Dalrymple et al. 1998b and Beck 2001 in Section 2.2 of Chapter 2).

In addition to the literature on each other, many studies have been performed on

reciprocal expressions in other languages. Many of these studies have been typological, in that

they study reciprocal expressions in many different languages (e.g. Maslova 2007, Nedjalkov

2007a, Evans et al. 2011a). Such studies focus on a wide range of phenomena related to 4

reciprocity, but notably, most do not discuss the relations between individuals like the literature on English. Some examine reciprocal relations in related languages to compare the basic syntactic and semantic properties of the different expressions (Lichtenberk 2000, Moyse-Faurie 2008).

Others look at many different reciprocal markers in one language (e.g. Nedjalkov 2007a and

Evans et al. 2011a and most of the chapters therein) to discuss the syntactic and semantic

variation within a language. If these authors discuss the semantics, they typically do not discuss

the relations between individuals that the expressions are compatible but rather note restrictions

on the distribution of expressions. For example, Kruspe (2011) claims that a reciprocal

expression in Mah Meri (Austro-Asiatic, Malaysia) cannot be used when the individuals involved

in the relation are not volitional, i.e. they do not intentionally stand in the relation. Similarly,

Burenhult (2011) argues that the acceptability of a reciprocal expression in Jahai is sensitive to

the aspect of the verb. Therefore, the literature on expressions outside of English has identified a

number of factors that affect the distribution of reciprocal expressions.

There is a tension between these two bodies of literature. The studies of English are very

detailed, yet they examine only the relations that each other conveys, to the detriment of asking

whether each other has other semantic content. The literature on other expressions, on the other

hand, has attempted to find patterns across many languages, to the detriment of exploring

particular expressions in detail. Because of this lack of detailed exploration, assumptions are

made on the basis of a few examples. Therefore, a number of questions emerge from the

intersection of the two bodies of literature. First, are other expressions compatible with the same

relations as each other? There is a gap in the literature concerning this question, since the

relations each other conveys have been thoroughly studied, but not the relations other expressions

convey. Second, what other meanings can reciprocal expressions convey? And similarly, what

syntactic, semantic and/or pragmatic restrictions exist on the distribution of reciprocal expressions?

In this dissertation, I take as a goal to expand the detailed information we have on reciprocal expressions cross-linguistically by exploring the two Russian reciprocal expressions

drug druga and the -sja verbs. They prove to be interesting because both expressions

semantically and pragmatically differ from English each other in subtle ways. For example,

although drug druga and each other are in many ways syntactically and semantically similar,

they exhibit some differences in distribution. To illustrate, consider the sentence in (6) that is

unacceptable (see Section 1.1 for an overview of my methodology):

(6) Context: Ivan and Marija went to see a movie together last night.

#Ivan i Marija pošli v kino drug s drugom. Ivan.NOM and Marija.NOM go.PST.PL to movies.ACC each with other.INST (Intended: ‘Ivan and Marija went to the movies with each other.’)

The unacceptability of this sentence is interesting because its closest English equivalent is

acceptable. This example suggests that although each other and drug druga are alike in many

ways, their semantics subtly differ. Because they differ, analyses of reciprocity that do not

consider Russian data are missing a part of the picture.

From a different perspective, a detailed description of Russian reciprocal expressions

seems reasonable in view of the very limited attention drawn to these expressions in the Russian

descriptive literature. While verbs suffixed with -sja have been discussed extensively, their

reciprocal function has received little attention. Many prominent grammars of Russian do not

even mention drug druga (e.g. Vinogradov 1947, Ivanov 1990, Timberlake 2004, Wade 2010),

and many others that do simply say that it is a reciprocal pronoun (Kamynina 1999). Therefore, a detailed study of these expressions will contribute to the grammatical coverage of Russian.

I use the tools and questions of formal semantics to investigate parameters suggested to be relevant from the typological literature to further our understanding of these two expressions and reciprocity more broadly. Based on the gaps in the literature, I pose the following research questions:

(7) (i) What relations between individuals are drug druga and the reciprocal -sja verbs compatible with?

(ii) Do either of these expressions introduce semantic or pragmatic content other than restrictions on relations?

(iii) Does the distribution of drug druga differ from that of the reciprocal -sja verbs? Do the distributions of the Russian expressions differ from that of each other?

The exploration of these questions should be useful for those interested in (i) the structure of

Russian and (ii) developing a cross-linguistically valid characterization of reciprocity.

In the next section I provide an overview of the chapters of this dissertation. After this

overview, I will present the basic terminology I will use throughout this dissertation. Section 1.4

contains background on the basic properties of drug druga and the -sja verbs. In Section 1.1, I

outline the methodology to be used through the dissertation, and in Section 1.2 I set the stage for

Chapter 2.

1.2 Dissertation outline

In Chapter 2, I outline the formal framework I couch my analyses in throughout the

dissertation. I provide background on the formal tools I will use to describe and analyze the

empirical generalizations to be made concerning the meaning of drug druga and the -sja verbs. I especially introduce algebraic structures for individuals and events from Krifka (1998). I then introduce the two main approaches to analyzing the reciprocal expression each other, one which is championed by linguists such as Dalrymple et al. (1998b) and Sabato and Winter (2005), and the other which is argued for by Langendoen (1978), Heim et al. (1991) and Beck (2001). I motivate my decision to adopt an analysis in the first tradition, similar to that of Dalrymple et al.

(1998b). I also introduce the relations between individuals that I will argue each other is compatible. I provide a derivation of a sentence that realizes each other as an example of how the many formal pieces can be used together to result in a coherent analysis. Finally, I propose basic definitions of drug druga and the reciprocal -sja verbs that I will adjust throughout this dissertation.

In Chapter 3 I ask whether drug druga and/or the -sja verbs introduce a restriction of simultaneity or sequentiality on the events expressed by the verb. I take as a preliminary goal to define these terms, since they have not been precisely defined in the literature to this point. I present evidence suggesting that although drug druga does not impose a semantic restriction, it does impose a pragmatic restriction of simultaneity on the events. I propose that drug druga introduces a generalized conversational implicature that the events expressed by the verb happen simultaneously. Although it has been argued that the -sja verbs restrict events to simultaneity

(e.g. Wierzbicka 2009, Siloni 2012), I find that this is in fact only true for a subset of these verbs.

The remainder of the verbs are indeed compatible with contexts in which the events occur

sequentially. I provide formal analyses of these reciprocal expressions and discuss what implications the temporal restrictions may have on a theory of reciprocity.

In Chapter 4 I turn my attention solely to the -sja verbs. It has been noted in the literature that the -sja verbs can participate in two distinct syntactic constructions; the simple reciprocal construction and the discontinuous reciprocal construction. These constructions have been argued to contribute different meanings to a sentence. In particular, the discontinuous construction is said to be able to convey an asymmetric relation between entities. I systematically explore the relations each construction is compatible with to better understand the meaning of each, and I indeed find differences in their semantics. I propose an analysis in which these differences partially fall out from the difference in the syntax of both constructions and partially because the -sja verbs in each construction have different lexical entries. Thus, I show that the syntactic construction a reciprocal expression is in can have consequences for its semantics.

In Chapter 5 I explore the meaning of drug druga. I find that drug druga cannot subcategorize for the preposition s ‘with’ when it introduces a relation of accompaniment to a sentence. I provide a precise definition of accompaniment and formalize this restriction. I then discuss the meaning of the preposition s ‘with’ that drug druga does subcategorize for and

suggest that it introduces a relation of accompaniment between the denotation of the subject of

the verb and that of the object of the preposition. I propose that when s does co-occur with drug

druga, that this s is the same as that in the discontinuous construction.

I conclude in Chapter 6 by summarizing my findings and discuss the implications my

empirical and theoretical claims have for the study of reciprocity and for those interested in the

structure of Russian. I also explore how my work informs a taxonomy of reciprocal expressions.

1.3 Terminology

Although, as Wierzbicka (2009: 103) notes, “reciprocity is in”, it is not “in” enough for

there to be a standard set of terminology to discuss reciprocity. This lack of standardization

suggests that finding common terms is a challenge. Linguists that treat the semantics of each

other typically simply assume that this expression is reciprocal without providing a reason for this

assumption or characterization (e.g. Heim et al. 1991, Dalrymple et al. 1998b, Beck 2001). If it is

clear that the discussion is about a particular expression, then having no definition of the term

reciprocal it is not much of a problem; it only becomes problematic when attempting to make

claims about each other that are relevant for cross-linguistic analysis.

In order to make a study of reciprocity cross-linguistically relevant, it must be made clear

that research into reciprocity is about the grammatical means by which languages encode a

certain meaning and possibly about what other meanings the same formal expression can convey.

Ideally the literature would be in agreement as to what this meaning is. As is often the case,

however, there is no such agreement. Some state that they investigate how languages encode a

symmetric relation between entities (e.g. Lichtenberk 1985, Nedjalkov 2007b, Heine and

Miyashita 2008). By a symmetric relation they mean that they study expressions that convey a

relation R such that if Rab, then Rba. Other assumptions exist, however; for example, Evans et

al. (2011a) write that the meaning by which a reciprocal expression is defined is “interaction and

projected shared intention” (24), and Wierzbicka agrees that “shared intention” is the common

meaning of reciprocal expressions. Frajzinger (2000) argues that reciprocal expressions encode a

situation when two or more entities share two semantics roles, such as AGENT and PATIENT, or

AGENT and BENEFACTOR. The literature is therefore not in agreement as to what meaning is definitional for the study of reciprocity and its expression.

Because there is no agreement as to what meaning should determine what expressions are under discussion in a study of reciprocity, it is vital that linguists make clear the definition they assume when exploring the topic. It is not uncommon, however, for authors to simply assume that a certain expression is “reciprocal” without any explanation of what makes it reciprocal. For example, Filip and Carlson (2001) begin their article on reciprocal expressions on Czech by saying that “The meanings of sentence with reciprocal expressions such as English each other . .

.” (417) and later simply assume that the expression se in Czech is a “reciprocal particle,” without

ever defining what “reciprocal” is. Similarly, Siloni (2012) discusses “reciprocal constructions”

in multiple languages without ever defining the term or discussing how one could determine

whether a particular construction is reciprocal. Without such definitions, it is not clear why we

should consider the expressions under discussion as related, and why any similarities between the

meanings of the expressions discussed are relevant to a study of reciprocity.

In this dissertation, I take as a goal to define my terminology so that others can make use

of my data and analyses for future work. At minimum we need (i) a meaning that is taken to be

the basic reciprocal meaning, by which expressions can be defined, (ii) a term to describe any

grammatical means by which reciprocal meaning is conveyed, and (iii) terms to describe any of

the other meanings that a reciprocal expression can convey. In the following section, I will

outline my definitions of a reciprocal expression and a reciprocal relation and other more minor

terms to be used throughout the dissertation.

1.3.1 Proposed terminology and definitions

I first define a relation of Strong Reciprocity (SR). This term is adopted from the formal

semantic literature (e.g. Langendoen 1978, Dalrymple et al. 1998b, Beck 2001, etc.). This

relation will be definitional for me; if a formal expression conveys this relation between entities, then I consider it to be reciprocal (see below for a more precise definition of a reciprocal expression). The set A corresponds to an expression in a sentence that denotes an NP, and R is a binary relation that will be supplied by the predicate of the sentence.

(8) Strong reciprocity:

(∀x,y ∈A)(x≠y→xRy), where A is a set with two or more members, and R is a subset of A × A.

The definition of Strong Reciprocity states that every member of set A stands in the R relation to every other member of the set. For example, consider the relation between entities depicted in Figure 4, where each number represents an individual, each arrow represents one binary relation, and each arrow represents the same type of R relation:

1 3

2 4

Figure 4: Strongly Reciprocal relation between four individuals

In this depiction, each individual stands in a relation to each other individual as the first and second argument of the R relation. If each number represents a girl, the sentence in

(9) could describe the situation in Figure 4. In this case, A corresponds to the NP the four

girls, and the R relation is a hugging relation.

(9) The four girls hugged each other.

The stipulation in (8) that A has more than one member is important, as it is agreed that

reciprocity at its heart involves more than one entity. The notation A × A means that the

R relation takes both its first and second arguments from the set A. In the sentence in (9),

each relation involves an entity denoted by the NP the four girls as its first and second

argument.

Finally, the distinctness clause (x≠y) is necessary because I want to define reciprocity as involving two distinct entities. I want to ensure that the relation that the sentence in (10) expresses, depicted in Figure 5, is not included in the study of reciprocity. The distinctness clause excludes such relations, as this sentence does not convey any relations which involve two distinct individuals.

(10) Mary hugged herself.

Figure 5: Depiction of the sentence Mary hugged herself.

I choose to use the relation of Strong Reciprocity as the definitional relation for two main reasons. First, the linguistic discussion of reciprocity has started with English each other and as is often the case, the meaning of this English expression is typically assumed to be the prototypical reciprocal meaning. Given that each other is compatible with a number of different relations between entities, I choose the most restrictive of these relations to define other expressions by. By choosing the most restrictive, I make my definition of a reciprocal expression conservative.

Second, I choose Strong Reciprocity as the way to define reciprocal expressions because there has been a dispute in the literature as to how to define a reciprocal relation. Most, if not all, linguists would agree that ‘reciprocal expressions’ are compatible with Strong Reciprocity, even if they do not consider this relation definitional. Therefore most other expressions that have been labeled reciprocal should meet my definition as well. My findings will therefore be relevant for these other studies and expressions.

An expression E is a reciprocal expression iff

(11) (i) there exists at least one sentence A which realizes the expression E and that conveys a Strongly Reciprocal relation

AND

(ii) there exists at least one grammatical sentence B that is the same as sentence A, except that the expression E is removed or replaced with an expression of the same syntactic category to the expression E, such that sentence B does not entail a Strongly Reciprocal relation

To illustrate how this definition works, I show why drug druga is a reciprocal expression according to the definition in (11).

First, it must be determined whether drug druga is realized in at least one sentence that conveys SR. Consider the following context that establishes a SR relation between the denotations of Ivan and Marija, and the sentence that is judged to be acceptable when uttered in this context.

(12) Context: You are good friends with both Ivan and Marija. You know Ivan loves Marija and Marija loves Ivan. A friend asks whether they are getting married. You respond that you don’t know, but you follow up with:

Ivan i Marija očen' ljubjat drug druga. Ivan.NOM and Marija.NOM really love.NPST.3PL each other.ACC ‘Ivan and Marija really love each other.’

Since the sentence in (12) is acceptable in the context, drug druga is realized in a sentence that

conveys SR. Therefore, drug druga passes the first test in the definition in (11) above.

Second, it must be established whether a grammatical sentence exists that is identical to

(12), except that drug druga has been replaced with a syntactic equivalent (i.e. a NP) that does

not convey SR. Consider for example, the grammatical sentence in (13), in which drug druga

has been replaced with the NP ix ‘them’. In Russian, this pronoun cannot be coreferential with

the subject. Crucially, this sentence is acceptable in a context in which Ivan and Marija do not

love each other, suggesting the sentence does not convey a SR relation between the entities.

(13) Context: You are in a zoo with your friend Kolja. Your friends Ivan and Marija do not like each other, but both independently recommended going to the zoo, particularly because they both love one kind of animal at the zoo. You don’t remember which animal though. You ask Kolja, and he points to the giraffes and says:

Ivan i Marija očen' ljubjat ix. Ivan.NOM and Marija.NOM really love.NPST.3PL them.ACC ‘Ivan and Marija really love them.’

Therefore, since this sentence is grammatical but does not convey SR, I suggest that drug druga, the expression that was replaced in this sentence, is a reciprocal expression.

Now, consider the adverb očen' ‘really’ in the sentence in (12). Intuitively, this adverb is not a reciprocal expression, but I show here this adverb fails the definition of a reciprocal expression. This adverb fulfills the first criterion, in that it is realized in a sentence compatible with SR. However, it fails on the second criterion. For example, consider the following sentence, which is identical to the sentence in (12) except that the adverb očen' ‘really’ is replaced with nepovole ‘unwillingly.’ Crucially, my consultants judged this sentence as acceptable given the context in (12):

(14) Ivan i Marija nepovole ljubjat drug druga. Ivan.NOM and Marija.NOM against.will love.NPST.3PL each other.ACC ‘Ivan and Marija love each other against their will.’

The sentence in (14) must in fact convey a SR relation between entities, as demonstrated by the fact that the same sentence is unacceptable in the context in (15):

(15) Context: Ivan loves Marija, but you know Marija does not love him back. You say:

#Ivan i Marija nepovole ljubjat drug druga. Ivan.NOM and Marija.NOM against.will love.NPST.3PL each other.ACC ‘Ivan and Marija love each other against their will.’

This context establishes a loving relation such that one individual loves the second individual, but

this love is not returned. The sentence is unacceptable in this context, suggesting that this

sentence must convey SR. Since this sentence is acceptable in the context in (12) but not in (15),

I propose that the adverb očen' ‘really’ is not a reciprocal expression.2

Note that based on this definition, the verb kiss in the sentence in (16) is also a reciprocal expression:

(16) Context: Your friends John and Mary just started dating. You know that they agreed to start dating after they kissed each other on the lips. You say:

John and Mary kissed.

This sentence expresses a SR relation between the two individuals. If we replace the verb with the verb eat, the SR relation is no longer entailed:

2 There is a problem with this test in that although it correctly excludes expressions from meeting the definition of a reciprocal expression, it is too inclusive because of the second criterion in (11). Certain expressions can make a sentence that would otherwise express a relation of SR unacceptable for reasons not related to the SR meaning. For example, if we replaced očen' with the adverb včera ‘yesterday’, the sentence would be grammatical (i.e. syntactically acceptable) but unacceptable for semantic reasons because the verb has a non-past form. This is a problem, because this second criterion requires the sentence without the expression E to be grammatical but to not entail a meaning of SR, and unacceptable sentences do not entail meanings. Since the test does work in most situations, however, I will assume it to work properly, while acknowledging this problem. 17

(17) Context: You are a babysitter and are watching two very young children, John and Mary, who have very strict schedules of eating and sleeping. You fed John at 3 pm before you put him down for a nap. You fed Mary at 5 pm after she woke up from a nap. When their parents come home, you say:

John and Mary ate.

The sentence does not entail a SR relation between the individuals denoted by John and Mary, as

it is acceptable in a context in which these individuals did not eat each other. Although a context

could be found in which this sentence is compatible with a SR relation, the sentence itself does not entail a SR relation.

Although the verb kiss is a reciprocal expression, I will also want to specifically discuss

reciprocal expressions that are verbs, and so I also propose the following definition:

(18) A verb V is a reciprocal verb iff

(i) there exists at least one sentence A which contains the verb V that conveys a Strongly Reciprocal relation

AND

(ii) there exists at least one grammatical sentence B that is the same as sentence A except that the verb V is removed or replaced with an expression of the same syntactic category (i.e. a verb), such that sentence B does not entail a Strongly Reciprocal relation

The definition of a reciprocal verb is similar to that of a reciprocal expression, except that it

requires that the morphosyntactic element in question is a verb, instead of any morphosyntactic

type of word. All reciprocal verbs are reciprocal expressions, though not all reciprocal

expressions are reciprocal verbs.

I use the definition in (18) to identify whether a verb is a reciprocal verb in the same manner as I identified reciprocal expressions. A naturally reciprocal verb is a reciprocal verb that has no overt reciprocal morphological marking. All naturally reciprocal verbs are reciprocal verbs, but not all reciprocal verbs are naturally reciprocal verbs. Examples of naturally reciprocal verbs in English are kiss, fight, and chat.

I now introduce my terms for the semantics of reciprocity.

(19) A relation is a reciprocal relation iff:

(i) the relation is conveyed by a sentence containing a reciprocal expression

AND

(ii) One or more individuals stand in the relation to individuals other than themselves

To illustrate how the definition of reciprocal relation applies to some sentences and not others, consider the following sentence that I hold expresses a reciprocal relation.

(20) Pjat' mal'čikov šli drug za drugom v komnatu. five.NOM boys.GEN walk.PST.PL each behind other.INST into room.ACC ‘Five boys followed each other into the room.’

I propose that this sentence expresses a reciprocal relation first because it realizes drug druga. As

I have already established, drug druga is a reciprocal expression. Therefore, since the sentence realizes a reciprocal expression, this provides half of the justification for calling the relation conveyed by this sentence reciprocal. As for the second criterion, the relation conveyed in the sentence in (20) can be depicted as in Figure 6:

Figure 6: Depiction of the sentence Pjat' mal'čikov šli drug za drugom v komnatu. ‘Five boys followed each other into the room.’

In this relation, each boy is involved in the following relation with at least one other individual.

Thus the second clause of the definition of a reciprocal relation is satisfied, namely that at least one individual must be involved in the relation with at least one other individual. Therefore, I assume that this sentence expresses a reciprocal relation between five boys.

What relation, however, does it express? It is not SR, as not every boy stands in the same relation to every other boy. Dalrymple et al. (1998b: 173) formalize this relation as in (21):

(21) ∀x,y ∈A (x≠ y→ for some sequence z0, . . . , zm ∈A (x = z0 ∧ (Rz0z1 ∨ Rz1z0)∧ … ∧ (Rzm-1zm∨Rzmzm-1) ∧ zm= y))

“Each member x of A should be related to every other member y via a chain of R- relations, where we ignore which way the pairs making up the chain are related via the relation R.”

Since a reciprocal expression contributes this relation to the sentence such that the relation holds between two or more entities, the relation in (21) is a reciprocal relation.

I propose that we need the second clause in the definition in (19) in order to restrict the scope of the domain. Cross-linguistically there are relations that are conveyed by reciprocal expressions that I do not want to characterize as reciprocal. For example, consider the following

Polish sentence:

(22) Kiedy Róza i Adam mieszkali razem, zawsze się dręczyli. when Roza.NOM and Adam.NOM. live.PST.PL together always SIĘ torment.PST.PL ‘When Roza and Adam lived together, they always tormented each other/themselves.’

In this sentence, the marker się can have two different functions - reflexive, which corresponds to

the first interpretation where the denotation of Róza tormented herself and that of Adam tormented himself, and a reciprocal function where the two individuals tormented each other. In the first interpretation no relation holds between the denotations of Róza and Adam. Thus, even

though there are still more than two participants, the relation between the two individuals is not a

reciprocal relation. Therefore, based on my definitions, się is a reciprocal expression, but it can

convey both reciprocal and non-reciprocal relations.

I will make reference occasionally to symmetric and asymmetric relations:

(23) A relation R is symmetric iff ∀x,y ∈A(Rxy→Ryx) (Partee et al. 1993:40)

(24) A relation R is asymmetric iff ∀x,y ∈A(x≠y (Rxy→¬Ryx)) (Partee et al. 1993:40)

The sentence in (25) entails that the denotations of John and Mary stood in a symmetric relation.

The sentence in (26) does not entail a symmetric relation because the sentence is true even if the

denotation of Mary did not kiss that of John. It does not entail an asymmetric relation either, 21

however, because this sentence is acceptable in a context in which the denotation of Mary kissed that of John back. It is, however, compatible with an asymmetric relation.

(25) John and Mary kissed each other.

(26) John kissed Mary.

These are the basic definitions to be used throughout this dissertation; I will define other terms in the chapters in which they become relevant.

1.4 Background on the syntax and semantics of drug druga and the -sja verbs

In this section, I provide a basic introduction to the two reciprocal expressions under discussion in this dissertation.

1.4.1 Drug druga

Discussions of drug druga in grammars and other scholarship is sparse if it is present at all. Several prominent grammars of Russian, such as Vinogradov (1947), Ivanov (1990)

Timberlake (2004) and Wade (2011) have no mention of drug druga, for example. The grammars that do mention drug druga typically only note that it is one expression among several that can be used to describe reciprocal (vzaimnyj) meaning (e.g. Kamynina 1999).

The reciprocal expression drug druga is an anaphoric pronoun, much like each other, in its syntactic distribution. It is an NP and as an anaphoric pronoun it refers back to another NP in the sentence.

(27) Ivan i Marija videli drug druga v parke. Ivan.NOM and Marija.NOM see.PST.PL each other.ACC in park.LOC ‘Ivan and Marija saw each other in the park.’

In this sentence, drug druga refers back to the coordinated NP Ivan i Marija ‘Ivan and Marija’

and indicates that the entities denoted by the subject stand in the seeing relation to each other.

This expression occurs in an argument position.

In the sentence in (27), drug druga is the direct object of the verb videli ‘saw’ and is

assigned accusative case. Drug druga can also be used in the locative, genitive, dative, and

instrumental cases. Note that the second word of the expression inflects for case, while the first

word never changes form; it is only the second word on which the case endings are realized.3

The following sentences are from the Russian National Corpus (which I describe in more detail in

Section 1.1 below):

(28) Vse prepodavateli znajut rezul'taty drug druga. all.NOM teachers.NOM know.NPST.3PL results.ACC each other.GEN ‘All the teachers know each other’s results.’

3 The term ‘word’ when describing the two parts of the expression drug druga is in fact somewhat problematic, as the two parts of drug druga pass some tests for wordhood, while failing others. Support for considering them as words is that each part has its own phonological stress - the two parts are not assigned a single accent as a unit. They fail several other tests of wordhood though. The first element fails in that it never inflects, contrary to what would be expected if it were a free word. Similarly, the two parts cannot be separated by anything other than a preposition. Therefore, while throughout this dissertation I will occasionally refer to each part of the expression as a ‘word’, I acknowledge that this term is problematic. 23

(29) Vy podavali drug drugu repliki iz-za kamery? you.NOM give.PST.PL each other.DAT copies.ACC from camera.GEN ‘Did you give each other copies from the camera?’

(30) Oni zanimalis' tol'ko drug drugom. they.NOM occupy.oneself.PST.PL only each other.INST ‘They were occupied themselves only with each other.’

Drug druga cannot be the subject of a sentence and thus does not take nominative case. It also

does not occur in the locative case without a preposition, because NPs in Russian take the

locative case only as the object of a preposition. Drug druga may be used as the object of a

preposition, in which case it can take any of Russian’s five prepositional cases. When drug druga is the object of a prepositional phrase, the preposition is typically realized between the two words.

For example, note that the preposition o ‘about’ is realized between the two words in the sentence in (31):

(31) Olja i Jurij pišut drug o druge roditeljam. Olja.NOM and Jurij.NOM write.NPST.3PL each about other.LOC parents.DAT ‘Olja and Jurij write about each other to their parents.’

In Russian, the preposition o assigns locative case to its object. The sentence in (31) realizes this preposition as the head of the phrase containing drug druga, and therefore drug druga is in the locative case. The majority of the prepositions in Russian combine with drug druga in this manner.

There are, however, a few prepositions which are realized before drug druga. For example, the grammatical sentence in (32)a realizes the preposition blagodarja ‘thanks to’ before

drug drugu. The sentence in (32)b is ungrammatical when the preposition is realized between the two words.

(32) a. Oni sušestvujut tol'ko blagodarja drug drugu. they.NOM exist.NPST.3PL only thanks.to each other.DAT ‘They exist only thanks to each other.’

b. *Oni sušestvujut tol'ko drug blagodarja drugu. they.NOM exist.NPST.3PL only each thanks.to other.DAT (Intended: ‘They exist only thanks to each other.’)

There are some prepositions that can occur in either location. For example, the preposition mimo

‘past’ occurs in between the two words seven times in the Russian National Corpus and nineteen

times before both words. Which prepositions fall into each of these three categories is a subject

of debate and there is no consensus on why some prepositions occur before drug druga and some

in between (Es'kova 2011, Apresjan 1995). This is a subject that does not concern me in this

dissertation; I assume that the two different places the preposition can be realized should be

captured in an analysis of the prepositions and not of drug druga itself, and that where the

preposition is realized does not affect the semantics of drug druga. In this dissertation I typically

use examples where the preposition is realized between the two words, as the most frequent

prepositions exhibit this pattern.

Drug druga is unusual among reciprocal expressions as it has a diminutive form that is

only used very colloquially: drug družka. For example, the following example is a riddle taken

from a Russian language textbook (Smyth and Crosbie 2002:448):

(33) Živut dva bratca rjadkom, a drug družka ne vidjat. live.NPST.3PL two.NOM brothers.GEN near but each other.ACC NEG see.NPST.3PL ‘Two brothers live near-by, but they don’t see each other.’4

This form is particularly interesting for a historical reason too. Both words in the phrase are

etymologically related to the adjective drugoj ‘other’. However, folk etymology tells us that the

two words in the expression drug druga come from the word drug ‘friend’. This diminutive is a

result of this folk etymology, as the word družka is the diminutive form of drug ‘friend’

(Nedjalkov 2007c, Knjazev 2007).

I will not discuss this diminutive form further in this dissertation; although I predict that

the only difference in the meaning between the two forms is the same as between any expression

and its diminutive, more research would be required to determine whether this is so.

1.4.2 Verbs suffixed with -sja

In this section I provide background on verbs in Russian suffixed with -sja. These verbs

have many different functions. In Russian, each verb suffixed with -sja is lexically specified as to

what functional meaning it has, i.e. how it relates the NPs in the sentence. I therefore discuss

these functions to properly distinguish the reciprocal -sja verbs from other verbs suffixed with

-sja.

4 The answer to this riddle is glaza ‘eyes’.

1.4.2.1 Morphosyntax of the -sja verbs

The affix -sja5 is realized on verbs as the rightmost suffix after aspect, tense and number inflection, as in (34)a. It has an allomorph, -s'6 which is realized when it immediately follows a

vowel, as in (34)b, except after active participles when it has its full form regardless of whether it

occurs after a vowel or consonant (34)c:

(34) a. celujut-sja kiss.NPST.3PL-SJA ‘to kiss each other’

b. celovali-s' kiss.PST.PL-SJA ‘[they] were kissing each other.’

c. zanimajuščaja-sja studentka study.PTCP.NOM-SJA female.student.NOM ‘the studying student/the student who is studying’

The verbs suffixed with -sja can be divided into four categories based on their morphosyntactic

properties (Zarickij 1961, Vinogradov 1947, Gerritsen 1990, Israeli 1997). The first group of

these verbs are derived from any transitive, imperfective verb, and they have a passive meaning.

For example, the sentence in (35) realizes a passive verb, vspaxivat'sja ‘to be plowed’. The sentence in (36) comes from the Russian National Corpus (see Section 1.1 for more information about the corpus), and realizes the transitive verb the passive verb was derived from.

5 When -sja is realized after the consonants t or t’, the combination behaves as if it is the affricate /ts/. In this case, the IPA of the ending of the verb is /…tsə/. After other consonants, the IPA is /-sjə/.

6 This allomorph has the IPA [sʲ] 27

(35) Pole vspaxivalo-s' kolxoznikami. (Vinogradov 1947: 640) field.NOM plow.IMPF.PST.SG-SJA kolxozniks.INST ‘The field was plowed by kolxozniks.’7

(36) Každuju vesnu muž sestry ezdil v derevnju k materi, every.ACC spring.ACC husband.NOM sister.GEN go.PST.SG to country.ACC to mother.DAT

vspaxival priusadebnyj učastok, zaceival ego kukuruzoj … plow.IMPF.PST.SG garden.ACC plot.ACC sow.PST.SG it.ACC corn.INST ‘Every spring my sister’s husband went to the country to mother’s house, he plowed the garden plot and sowed it with corn …’

The verb suffixed with -sja in (35) has a passive meaning, as demonstrated by the fact that the subject of the sentence, pole ‘field’ does not have the role of AGENT; instead it has the role of

PATIENT. The valency of the verb in (35) is also reduced by one, as the AGENT of the sentence does not have an argument position; rather it is denoted by the word kolxoznikami ‘kolxozniks,’ which is in the instrumental case. This group of verbs is productive, by which I mean that -sja can attach to any transitive, imperfective verb to realize a passive meaning.

The verbs suffixed with -sja in the second category are derived from both imperfective and perfective transitive verbs. For example, the (a) verbs in (37) - (41) are all transitive verbs.

Note that those in (37), (39), and (41) are all imperfective, while those in (38), (40), and (42) are perfective.

7 Kolxozniks were workers on collective farms in the Soviet Union. 28

(37) a. celovat' Mišu kiss.IMPF.INF Miša.ACC ‘to kiss Miša’

b. celovat'-sja kiss.IMPF.INF-SJA ‘to kiss each other’

(38) a. pocelovat' Mišu kiss.PFV.INF Miša.ACC ‘to kiss Miša’

b. pocelovat'-sja kiss.PFV.INF-SJA ‘to kiss each other’

(39) a. myt' Mišu wash.IMPF.INF Miša.ACC ‘to wash Miša’

b. myt'-sja wash.IMPF.INF-SJA ‘to wash oneself’

(40) a. pomyt' Mišu wash.PFV.INF Miša.ACC ‘to wash Miša’

b. pomyt'-sja wash.PFV.INF-SJA ‘to wash oneself’

(41) a. xotet' Mišu want.IMPF.INF Miša.ACC ‘to want Miša’

b. xotet'-sja want.IMPF.INF-SJA ‘to feel like doing’

(42) a. zaxotet' Mišu want.PFV.INF Miša.ACC ‘to want Miša’

b. zaxotet'-sja want.PFV.INF-SJA ‘to feel like doing’

The verbs in this category do not have a passive meaning. Rather, depending on the verb, they have a variety of meanings, as suggested by the translations I provide for them, related to the middle voice (see Kemmer 1993). I will return to discussing the semantics of verbs of this category below in Section 1.4.2.2. The reciprocal -sja verbs under discussion in this dissertation are members of this category.

The verbs in the third group are derived from imperfective or perfective intransitive verbs. For example, the verbs in (43)a and (44)a are both intransitive verbs, and the verbs in

(43)b and (44)b are derived from them.

(43) a. belet' to.make.white.INF ‘to make white’

b. belet'-sja to.make.white.INF-SJA ‘to whiten/to become white’

(44) a. dumat' think.INF ‘to think’

b. dumat'-sja think.INF-SJA ‘to think’

The meaning of the derived verbs differs from the meaning of their base in a variety of

ways. These verbs are not relevant for this dissertation so I won’t go into here the meanings of

the derived verbs as the meanings vary and are a subject of debate.

The last group of verbs suffixed with sja has been called reflexivum tantum, or deponent8 verbs, i.e. those verbs whose bases do not have a separate lexical entry (Kemmer 1993, Knjazev

2007, Nedjalkov 2007a). I include in this group the semi-deponent verbs, those that are morphologically derived from a verb, but whose meaning is not related to the base verb. For example, the verbs in (45)a and (46)a do not have a base with a separate lexical entry at all, as indicated by the fact that a similar verb without the -sja is not a real word, and in fact these bases do not occur without the -sja affix in any other environment. The verb in (47)a does have an

8 The term deponent more traditionally has referred to a type of morphological realization in which two values belonging to the same feature, such as voice or tense, receive formal expression such that the first value is expressed by what normally expresses the second value, and the second value is expressed by what normally expresses the first value. The literature on reciprocity, however, has used the term deponent consistently to refer to reciprocal verbs that do not have a transitive counterpart from which they are synchronically derived (Kemmer 1993, Nedjalkov 2007, Knjazev 2007). 31

extant base that has a lexical entry (47)b, but the meaning of the verb with the -sja suffix differs from the meaning of the verb without -sja:

(45) a. borot'-sja ‘to struggle with each other’

b. *borot'

(46) a. nadejat'-sja ‘to hope’

b.*nadejat'

(47) a. videt'-sja ‘to meet with’

b. videt' see.INF ‘to see’

Trying to determine the meaning of the affix -sja has caused Slavists many headaches

over the years. Some have posited a variety of meanings for -sja (e.g. Vinogradov 1947) while

others have attempted to find a single, vague meaning for all of them (e.g. Jakobson 1932). I

discuss the semantics of this affix and outline some of the problems that either approach has to

solve in the next section.

1.4.2.2 The semantics of the -sja verbs

As can be seen from the translations of (37) - (47) above, the suffix -sja appears to have a wide range of functions. It is generally agreed that its main function is valency reduction, but it reduces the valency of a sentence in a variety of ways (Zarickij 1961, Gerritsen 1990, Knjazev

2007).

First, as noted above, when a transitive verb with the suffix -sja is imperfective, it can have a passive meaning. For example, the following sentence, taken from the Russian National

Corpus, realizes the verbs prodavat'sja ‘to be sold’ and čitat'sja ‘to be read’:

(48) A v dal'nejšem želaem, štoby lekcii po anatomii čitali-s' And in future.LOC wish.NPST.1PL that lectures.NOM on anatomy.DAT read.PST.PL-SJA

tol'ko Andreem Vezal'cem, synom imperatorskogo aptekarja. only Andrej.INST Vezalec.INST son.INST imperial.GEN pharmacist.GEN ‘And in the future, we wish that lectures on anatomy will be read only by Andrej Vezalec, the son of the imperial pharmacist.’

In this sentence, the NP lekcii ‘lectures’ is in the nominative case and is the subject of the verb.

The denotation of lekcii has the semantic role of PATIENT. The AGENT of the sentence is denoted

by the name of a person, Andreem Vezal'cem and is in the instrumental case. Thus, this sentence

exemplifies the passive voice.

The meanings of the verbs with -sja in other syntactic contexts are not so straightforward.

As mentioned before, -sja is a middle marker in that it can convey a number of the typical

functions of middle markers such as reflexivity, reciprocity, anticausativity, and antipassivity, to

name a few. See Kemmer (1993), Manney (2000), Steinbach (2002), Allan (2003) and Moyse-

Faurier (2008) for an overview of typical middle marker functions cross-linguistically. Many of the so-called reflexive pronouns/clitics in the Indo-European languages that are historically derived from the Proto Indo-European (PIE) reflexive pronoun *sve- or *se are middle markers and can convey typical middle voice meanings (Zarickij 1961, Maslova 2000). In this family, - sja is somewhat of an exception because it has been reanalyzed as an affix, whereas it is more common for the reflex of the PIE pronoun to have the status either of a full pronoun or a clitic.

There are two ways that these middle markers pattern semantically. For many of the other Indo-European languages, including for example French, Polish, and German, verbs marked with the middle marker are polysemous between some or all of the functions associated with middle markers (Kemmer 1993, Maslova 2007, Siloni 2012). This means that sentences that realize the pronoun/clitic are ambiguous in a metalinguistic context, i.e. when uttered out of the blue. In such languages, the meaning of the pronoun/clitic will depend largely on the syntactic and pragmatic context. For example, consider the following two Polish sentences, both of which realize the clitic się:

(49) a. Kiedy Róza i Adam mieszkali razem, zawsze się dręczyli. when Roza.NOM and Adam.NOM. live.PST.PL together always SIĘ torment.PST.PL ‘When Roza and Adam lived together, they always tormented each other/themselves.’

b. Kiedy Róza była sama, zawsze się dręczyła. when Roza.NOM be.PST.SG alone always SIĘ torment.PST.SG ‘When Roza was alone, she always tormented herself.’

In (49)a, while it is possible that the denotations of Roza and Adam tormented themselves (i.e. the denotation of Roza tormented herself and that of Adam tormented himself), the most natural

interpretation of się is reciprocal, based on the context that the two individuals live together. A reciprocal reading is impossible in (49)b however, because the subject of the verb dręczyła

‘tormented’ (Róza) is singular.

The verbs suffixed with -sja in Russian belong to a different group, which also includes

expressions in Lithuanian (Geniušienė 2007), Hebrew (Siloni 2012) and Hungarian (Ràkosi

2008). Verbs marked with the particular middle marker in these languages are typically not

ambiguous. For example, the sentence in (50) is acceptable only in Context 1 in (51) and not in

Context 2.

(50) Ivan i Marija myli-s'. Ivan.NOM and Marija.NOM wash.PST.PL-SJA ‘Ivan and Marija washed themselves before the concert.’

(51) Context 1: Ivan and Marija were at the gym exercising and so are hot and sweaty. They have to get ready for a concert so they come home and take showers separately.

Context 2: Ivan and Marija are an older couple and because of their age can’t raise their arms to wash their own backs. You ask their daughter who helps them wash. She says:

Context 1 establishes a reflexive relation for each entity, so that each individual washes themselves. Context 2, on the other hand, establishes a reciprocal relation between the individuals. Since the sentence in (50) is only acceptable in Context 1, it suggests that this -sja verb, myt'sja ‘to wash oneself’ can only have the reflexive function and not the reciprocal function. In Russian, each verb suffixed with -sja is lexically specified as to what functional meaning it has, i.e. how it relates the NPs in the sentence. As another example, the sentence in

(52) was taken from the Russian National Corpus. Note the meaning of the verb serdit' ‘to anger’

or ‘to make angry’ and the fact that it is a causative verb. 35

(52) A popytki zarodit' v nej somnenie eё ne serdjat i daže and attempts.NOM birth.INF in her.LOC doubt.ACC her.ACC NEG anger.NPST.3PL and even

ne volnujut ... NEG worry.NPST.3PL ‘And the attempts to birth doubt in her don’t anger her and don’t even worry her …’

Now consider the following sentence, which is acceptable in Context 1 in (54), but not Context 2

below:

(53) Ivan i Marija serdili-s'. Ivan.NOM and Marija.NOM anger.PST.PL-SJA ‘Ivan and Marija were angry.’

(54) Context 1: It is Ivan and Marija’s anniversary and they made reservations at their favorite restaurant months in advance. When they show up at the restaurant, however, the restaurant can’t find their reservation, and they are made to leave. They are both furious at the restaurant, since it is the restaurant’s error.

Context 2: It is Ivan and Marija’s anniversary and they both think that the other made reservations at their favorite restaurant. When they show up at the restaurant, however, the restaurant can’t find their reservation, and they realize that neither of them had called to make a reservation in the first place. They are both furious at the other for the mistake. This makes you happy because you’ve been wanting to break them up for years. You gleefully tell someone how they feel about each other:

Context 1 establishes a situation in which the denotation of Ivan and that of Marija are each

upset, but not at each other. Since the sentence in (53) is acceptable in this context, it suggests

that the suffix -sja has a decausative function when it derives the verb serdit'sja ‘to become

angry’. Since this sentence is not acceptable in Context 2, which requires that the denotation of

Ivan is angry at that of Marija and vice versa, it suggests that -sja cannot have a reciprocal function when on this verb.

In sum then, while some middle markers cross-linguistically compose with verbs to derive metalinguistically ambiguous verbs or verb phrases, verbs suffixed with -sja in Russian are typically not ambiguous (but see below for an exception to this). There is a lot of scholarship on the different meanings the -sja verbs have and what the empirically and theoretically best way is to characterize the different meanings the verbs have (e.g. Jakobson 1932, Gerritsen 1990, Israeli

1997, Timberlake 2004, Knjazev 2007). I do not enter into this discussion, but simply point the reader in the direction of the previous literature.

1.4.2.3 Reciprocal -sja verbs

In Russian, there are about 40 -sja verbs that have a reciprocal function:

videt'sja, ‘to meet’ bit'sja, ‘to fight’ vstrečat'sja, ‘to meet’ rubit'sja, ‘to cut each other’ pixat'sja, ‘to have sex’ streljat'sja, ‘to shoot each other’ otličat'sja,’to differ from’ dogovaryvat'sja, ‘to agree with’ sudit'sja, ‘to be in litigation with’ soveščat'sja, ‘to confer with’ rezat'sja, ‘to play cards with’ uslovit'sja, ‘to agree’ soveščat'sja, ‘to confer with’ prepirat'sja, ‘to bicker’ ob''jasnjat'sja, ‘to discuss’ prerekat'sja, ‘to bicker’ smešivat'sja, ‘to mix’ (‘to be mixed’) smešyvat'sja, ‘to mix’ obvenčat’sja, ‘to get married’ proščat'sja, ‘to bid farewell’ ženit'sja, ‘to get married’ sorevnovat'sja, ‘to compete with’ mirit'sja, ‘to make up with’ družit'sja, ‘to befriend’ družit'sja, ‘to befriend each other’ bodat'sja, ‘to butt each other’ celovat'sja, ‘to kiss each other’ klevat'sja, ‘to peck each other’ obnimat'sja, ‘to hug each other’ kusat'sja, ‘to bite each other’ tolkat'sja, ‘to bump into each other’ ljagat'sja, ‘to kick each other’ borot'sja, ‘to fight, struggle with’ pinat'sja, ‘to kick each other’ rugat'sja, ‘to fight’ oblivat'sja, ‘to splash each other’ torgovat'sja, ‘to haggle’ carapat'sja, ‘to scratch each other’ znakomit'sja, ‘to get to know, to meet’ branit'sja, ‘to scold each other’ zdorovat'sja, ‘to greet’ teret'sja, ‘to rub up against each other’ drat'sja, ‘to fight’ bratat'sja, ‘to fraternize with each other’ sražat'sja, ‘to overwhelm each other’ ženixat'sja, ‘to court each other

I will call these verbs reciprocal -sja verbs and will often shorten this to the -sja verbs, as I do not again discuss any of the other functions of the verbs suffixed with -sja. The above list was primarily compiled from lists or discussions in Gerritsen 1990, Israeli 1997, and Knjazev 2007.

None of these works claim to include every single reciprocal -sja verb, and likewise I do not make such a claim. The reciprocal -sja is entirely unproductive. New reciprocal -sja verbs cannot be derived from new verbs in the Russian language. To express a reciprocal relation between

entities with any verb but one of the reciprocal -sja verbs, one must use a different reciprocal expression, typically drug druga.

One exception to the general rule that verbs suffixed with -sja in Russian are not

ambiguous is that if such a verb has a reciprocal function, it also has an absolutive function (Israel

1997, Timberlake 2004, Knjazev 2007).9 When used in the absolutive function, the sentence

typically indicates that the denotation of the subject of the verb performs the event expressed by

the verb habitually. For example, the verb in (55) and (56), kusat'sja ‘bite (each other)’, has both

an absolutive meaning and a reciprocal meaning.

(55) Context: You have two new dogs you are trying to train who frequently bite people. When your friend comes to visit, you say:

“Ostorožno, sobaki kusajut-sja!”

careful dogs.NOM bite.NPST.PL-SJA ‘Careful, the dogs bite!’

(56) Context: You and a friend bring your dog to a park to run around, but when you get there, you see that there are already a number of dogs there that are fighting and biting each other. Being afraid for your dog’s safety, you say:

“Ostorožno, sobaki kusajut-sja!”

careful dogs.NOM bite.NPST.PL-SJA ‘Careful, the dogs are biting each other!’

The interpretation of the verb in (55), the absolutive reading, is the default reading. The reciprocal meaning has to be very salient for the hearer of the sentence in order for the reciprocal meaning to be understood, but such a reading is indeed possible. I will ignore the absolutive

9 Timberlake uses the term habitual instead of absolutive. 39

interpretation because no reciprocal relations are conveyed by such sentences. The examples I use throughout this dissertation will unambiguously convey reciprocal relations.

One more aspect to the semantics of the reciprocal -sja verbs is important. Kemmer

(1993) notes that reciprocal verbs cross-linguistically typically fall into a few categories in terms of the lexical meaning of the verb. Kemmer (104-105) writes that reciprocal verbs often express

“antagonistic action” e.g. rugat'sja ‘to quarrel’, borot'sja ‘to wrestle’, “affectionate actions” e.g. celovat'sja ‘to kiss’ obnimat'sja ‘to hug’, “encountering and associated social actions” e.g. vstrečat'sja ‘to meet’, proščat'sja ‘to say good-bye’, “actions denoting unintentional physical contact” e.g. teret'sja ‘to rub up against each other’ tolkat'sja ‘to bump into each other’, and “acts

of exchanging” e.g. obmenivat'sja ‘to exchange’. All of the reciprocal -sja verbs listed in (51)

above fall into one of these categories. This therefore suggests that reciprocal -sja verbs in

Russian are comparable in some ways to reciprocal verbs in other languages, in that they have

similar lexical meanings. When exploring the meaning of the -sja verbs throughout this

dissertation, I will use a variety of verbs from the different lexical categories to assure as much as

possible that my claims are true of the entire class of reciprocal -sja verbs.

1.4.2.4 Other Reciprocal Verbs in Russian

In addition to the -sja verbs under discussion in this dissertation, there are three other

types of reciprocal verbs in Russian. I introduce them briefly here because they are in some ways

very similar to the -sja verbs. After introducing them, I provide justification for not considering

them further.

The first are derived from transitive verbs. The derived reciprocal verb realizes the prefix

pere- and two suffixes: the imperfectivizing suffix -(y)va and -sja (Israeli 1997, Knjazev 2007).

For example, the first verbs in (57) are transitive verbs, and the second verbs are the derived reciprocal verbs.

(57) a. govorit' ‘to speak’ pere-govaryvat'-sja ‘to converse with each other’

b. branit' ‘to scold’ pere-branivat'-sja ‘to argue with each other’

When such transitive verbs realize only the suffix -sja, such as govorit’sja or branit’sja and not

the prefix, they do not have a reciprocal meaning. Therefore it is both the prefix and the suffix

that make these verbs reciprocal.

The reciprocal verbs of the second type are formed from intransitive verbs. The derived

verbs are prefixed by either raz(o)- or s(o)- and they are suffixed by -sja (Knjazev 2007). The

majority of these verbs involve spatial relations. For example, the first verb in (58) is an

intransitive verb of motion, and the second and third verbs exemplify the two different prefixes

and the meaning resulting from the derivation.

(58) idti ‘to go by foot’ razo-jti-s' ‘to separate’ so-jti-s' ‘to come together’10

Finally, most, if not all verbs, can be prefixed by so-, which is etymologically related to the

English prefix co-. For example, the first verbs in (59) can undergo derivation to result in the

second verb (Nedjalkov 2007b, Knjazev 2007).

10 Note that when the verb idti ‘to go by foot’ is prefixed, the d in the base disappears. 41

(59) a. čitat' ‘to read’ so-čitat' ‘to read together’

b. sušestvovat' ‘to exist’ so-sušestvovat' ‘to co-exist’

Although these three other types of reciprocal verbs exist in Russian, I consider them different reciprocal verbs from the -sja verbs because they exhibit a difference in their semantics. Only the

-sja verbs, and none of the verbs in the three groups directly above, are ambiguous between an absolutive and a reciprocal meaning. The three other types of reciprocal verbs only have a reciprocal function. While I ignore the absolutive meaning in this dissertation, this difference indicates that the semantics of these three types of reciprocal verbs are fundamentally different from the ones I am interested in. While research might show that there really are no semantic differences between these three and the -sja verbs in their reciprocal meaning, for the sake of simplicity, I do not again consider them again. Future studies of reciprocal expressions in

Russian beyond drug druga and the -sja verbs should certainly explore their semantics further.

Lastly, Russian, like English, exhibits naturally reciprocal verbs. In a study that looked at twenty mostly unrelated languages, Evans et al. (2011a) found that each of these languages exhibited naturally reciprocal verbs. Naturally reciprocal verbs are those that have no overt reciprocal marking, but still convey a reciprocal relation between entities. In English, examples of these verbs are argue, chat, and meet. Russian also has such verbs, for example razgovaryvat'

‘to chat’ and obsuždat' ‘to discuss’. I will discuss this group on and off throughout the

dissertation as they are relevant for languages outside Russian. I do not analyze them thoroughly,

but will simply comment on them here and there.

1.1 Methodology

In this section, I outline the various types of methodology I used to explore the semantics

of the two reciprocal expressions in Russian. My methodology for data collection involved three

procedures: (i) mining data from the Russian National Corpus, (ii) asking consultants to judge

the acceptability of Russian sentences in particular contexts, and (iii) asking consultants to

describe in Russian a situation that was shown to them as a video recording.

The first procedure was a corpus study utilizing data from the Russian National Corpus

(RNC). The RNC is a 300 million-word corpus with texts from the 18th-21st centuries. It includes written works of fiction and nonfiction, including newspaper articles, textbooks, diaries, etc., as well as written transcriptions of oral speech. The main purpose of using the corpus was as an exploratory task to find data that have not previously come to light.

I searched the RNC for instances of drug druga and the -sja verbs. In the relevant chapters I will provide more detail about how I searched for what, but I provide a brief overview here. To explore the meaning and distribution of drug druga, I mined 500 sentences from the corpus containing this expression in different cases and sometimes as the object of a prepositional phrase. See Table 1 for the distribution of the forms I examined. I searched for the form drug druga more often than the others, as this form is syncretic between the accusative and genitive.

Furthermore, accusative is the most common case in which drug druga occurs.

Form English translation Morphological Case Number of sentences I coded drug druga each other Accusative/Genitive 200 drug v druge in each other Locative 50 drug drugu to each other Dative 50 drug ot druga from each other Genitive 50 drug drugom by means of each other Instrumental 50 drug k drugu toward each other Dative 50 drug s drugom with each other Instrumental 50

Table 1: Details on the corpus study of drug druga

I analyzed the first 200 or 50 sentences that were returned as results in a search for these forms.

To explore the semantics of the -sja verbs, I extracted 25 sentences from the corpus

containing four different reciprocal -sja verbs, resulting in 100 sentences. These four verbs are

examples of the general lexical categories of reciprocal verbs that Kemmer (1993) posits, as

discussed in Section 1.4.2.3 above, namely affectionate actions (celovat'sja ‘to kiss each other’),

verbs of social interaction and encountering (zdorovat'sja ‘to greet each other’) , antagonistic

action (sražat'sja ‘to struggle with each other’), and verbs of unwanted physical touch

(carapat'sja ‘to scratch’). The verb celovat'sja is also included because it contributes a temporal

restriction that the other verbs do not (see Chapter 3 for more details). Including it in the corpus

study then is vital for understanding whether patterns I find are relevant for all the -sja verbs or

just those that do not introduce such a restriction.

I coded each of the sentences mined from the RNC with the following information: number of participants, type of reciprocal relation according to Dalrymple et al. (1998b) (e.g.

Strong Reciprocity, Intermediate Reciprocity, etc., to be discussed in Chapter 2), whether the participants were animate or not, the source, whether the subeventualities were simultaneous or sequential, the grammatical form of the verb (e.g. third person plural, past tense), the aspect of the

verb, the polarity of the sentence, and any other details I found interesting about the sentence.

The corpus is constructed so that searchers may see the sentence the string is in, and one sentence

before and one sentence after that root sentence. This amount of context was often not enough for

me to determine whether the subeventualities are simultaneous or sequential, for example, or

which relation was being described. Because determining the context was sometimes impossible,

and because the corpus study provides no negative evidence, I also worked with native speaker

consultants.

I worked with four native speakers. Three of these consultants grew up in Russia and one

in Eastern Ukraine. They are all bilingual with English and are living in the United States. Their

ages ranged from 28-53 when I started working with them. At that time they had been in the

United States for 3-20 years. Working with consultants has allowed me to discover the range of

possible meanings that are compatible with these two expressions within certain parameters.

Each of the sentences I provide that I obtained through elicitation was judged by at least three of

my consultants, but not always the same three consultants. Although this situation is not ideal, it

was necessary for practical reasons.

I used two different procedures with these consultants. First, for some of the elicited

data, I asked them to judge the acceptability of Russian sentences that I created (and had vetted

by a native speaker) in particular contexts, as well as to translate English sentences into Russian.

The responses elicited from this task show me how widely each expression can be used and 45

provides me with clear instances of when these expressions cannot be used. Such information is important because any hypothesis about meaning I posit for these expressions should encompass all possible meanings as well as exclude meanings that are not compatible with the expressions.

On the basis of these consultants’ judgements, I indicate each sentence’s acceptability using the following symbols:

(60) No mark Acceptable

# Unacceptable for semantic reasons

* Ungrammatical, i.e. unacceptable for syntactic reasons

? There is variation among speakers as to the sentence’s acceptability

I occasionally also use the symbol % when describing other authors’ work when a particular interpretation for a sentence is not available.

Second, I used video recordings to determine what native speakers consider the most natural utterance to use to describe a given relation. The videos were made by Evans et al.

(2004), and the results of their studies were published in Evans et al. (2011a). These recordings were designed to explore reciprocal expressions cross-linguistically, but have never been used with Russian data. They have been successfully used for twenty languages world-wide, which suggests that the videos do indeed elicit useful data from consultants. There are 64 videos which all depict two or more people or objects. The people perform various actions, and the objects are placed in various positions. I showed three consultants each of these videos and asked them how they would describe the videos most naturally. I then sometimes asked a follow-up question such

as “Can you say this instead?” “Can you use this expression?”. I recorded these answers and any comments the consultants provided.

I consider such a procedure as important because (i) consultants often find it easier to naturally describe a situation if they can observe it visually, rather than having it described to them with words, and (ii) this is a way to judge the most natural way Russians describe a certain situation. Evans et al. (2011a) successfully used recordings to elicit reciprocal expressions for certain contexts, and this procedure has been useful to me as well. I have analyzed this data qualitatively and provide the data I found significant in each relevant chapter of this dissertation.

I use data from all of these sources throughout the dissertation, often in conjunction with each other. Every Russian sentence I provide was vetted by native speakers and all acceptability judgments are based on the judgments of three speakers.

1.2 Looking forward

In this chapter I have provided motivation for a study of two reciprocal expressions in

Russian and have presented the goals of this dissertation. I have also introduced the terminology

I use throughout the dissertation and provided background on both drug druga and the reciprocal

-sja verbs. I finally outlined the methodology to be used to explore these two expressions. I now

turn in Chapter 2 to presenting my assumptions about the formal framework and reciprocity to be

employed throughout this dissertation.

2 Formal foundations

This chapter provides an overview of the formal framework I use in this dissertation to characterize the meaning of drug druga and the -sja verbs and the linguistic concept of reciprocity more generally. In Section 2.1 I introduce the formal framework I adopt to capture the meaning of sentences that realize a reciprocal expression. I then outline the two main

approaches to accounting for the meaning of each other in the formal literature and will motivate my decision to follow an approach in line with Dalrymple et al. (1998b), Sabato and Winter

(2005) and Dotlačil and Nilsen (2008). In Section 2.3 I discuss the different relations each other

conveys and provide evidence to argue that it is compatible with four relations. I develop an

analysis of a fragment of English including each other to demonstrate how the formal framework

can be used in an analysis that captures the meaning of this expression and finally propose basic

definitions of drug druga and the -sja verbs, to be revised throughout this dissertation, in Sections

2.5 and 2.6.

2.1 Formal framework

I work in the framework of truth-conditional semantics, which assumes that to know the

meaning of a sentence is to know its truth conditions, i.e. what conditions must be met in order

for the sentence to be true. I follow the Montagovian tradition of truth-conditional semantics in

that natural expressions of an object language, such as Russian in this dissertation, are first

translated into a translation language, and then each translation receives a model-theoretic

interpretation. In model-theoretic semantics, the notion of truth is relative to the model one adopts.

The syntax and semantics of Russian, English and my translations are couched in a

Combinatoric Categorial Grammar framework (Steedman 1996, 2000). The lexicon consists of triples of the form in (61):

(61) phon : CAT : ⟦sem⟧M,g

In this lexical entry, phon is a phonological string, CAT the syntactic category of the lexical item,

sem is the translation of the expression, and ⟦sem⟧M,g the interpretation of the translation of the expression in a model M under the assignment function g.

I assume three basic syntactic categories for the object language: N for noun phrase, S for sentence, and S’ for sentence radical. A sentence radical is a tenseless structure that is the result of combining the VP with its argument, i.e. the subject. Each syntactic category matches up with basic expressions of the object language, as exemplified in (62):

(62) Category Basic Expression

N Ivan, Marija

S Ivan kisses Marija.

S’ Ivan kiss Marija.

The translation language also consists of syntactic categories and basic expressions. In this language, basic expressions are non-logical constants and variables. I assume the following syntactic categories of basic expressions of the translation language: e for entities, ev for eventualities and t for truth values. If a and b are any types, then 〈a, b〉 is a type, and nothing else is a type.

I follow Dowty et al. (1981) in assuming the following syntactic rules for how basic expressions compose to result in complex expressions in the translation language (91-92):

(63) Syntactic Rules:

a. For each type a, every variable and every non-logical constant of type a is a member of the set of meaningful expression of type a

b. For any types a and b, if α is of type 〈a, b〉 and β is of type a, then α(β) is of type b

c. If ϕ and ψ are of type t, then so are each of the following:

¬ ϕ [ϕ ˄ ψ] [ϕ ˅ ψ ] [ϕ → ψ] [ϕ ↔ ψ]

d. If ϕ is of type t and u is a variable (of any type), then ∀uϕ is of type t.

e. If ϕ is of type t and u is a variable (of any type), then ∃uϕ is of type t.

I assume a tight correspondence between the syntactic categories of the object language expressions and the syntactic categories of the translation language expressions. The translation of syntactic type N is e, that of syntactic type S is t, and that of the syntactic type S’ is .

The denotations of the basic expressions of the translation language are specified in a model M = ⟨U, F, E⟩, where U is an algebra over entities, F an interpretation function, and E an 50

algebra over eventualities. I will discuss the algebras in more detail below in Section 2.1.1. The denotations of constants are given by F and those of variables are given by the variable assignment function g:

(64) Semantics of the Basic Expressions:

a. If α is a non-logical constant, then ⟦α⟧ M,g=F(α).

b. If α is a variable, then ⟦α⟧ M,g=g(α).

I assume the following semantic formulation rules of the basic expressions, again following

Dowty et al. (1981: 92). These rules mirror those for the syntax.

(65) Interpretation Rules of the Semantics

a. If α is of type 〈a, b〉 and β is of type a, then ⟦α(β)⟧ M,g=⟦α ⟧M,g(⟦β⟧ M,g)

b. If ϕ and ψ are of type 〈ev, t〉, then ⟦¬ ϕ⟧ M,g=1 iff ⟦ ϕ⟧ M,g=0; otherwise, ⟦¬ ϕ⟧ M,g=0

c. If ϕ and ψ are of type 〈ev, t〉, then ⟦ϕ ˄ ψ⟧ M,g=1 iff ⟦ ϕ⟧ M,g=1 and ⟦ ψ ⟧ M,g=1

d. If ϕ and ψ are of type 〈ev, t〉, then ⟦ϕ ˅ ψ⟧ M,g=1 iff either ⟦ ϕ⟧ M,g=1 or ⟦ ψ ⟧ M,g=1

e. If ϕ and ψ are of type 〈ev, t〉, then ⟦ϕ → ψ⟧ M,g=1 iff either ⟦ ϕ⟧ M,g=0 or ⟦ ψ ⟧ M,g=1

f. If ϕ and ψ are of type 〈ev, t〉, then ⟦ϕ ↔ ψ⟧ M,g=1 iff either ⟦ ϕ⟧ M,g=1 and ⟦ ψ ⟧ M,g=1 or else ⟦ ϕ⟧ M,g=0 and ⟦ ψ ⟧ M,g=0

g. If ϕ is of type 〈t〉 and u is a variable of type a, then ⟦∀uϕ ⟧ M,g= 1 iff for all e of type a, ⟦ϕ ⟧ M,g[e/u]= 1

h. If ϕ is of type 〈t〉 and u is a variable of type a, then ⟦∃uϕ ⟧ M,g= 1 iff for some e of type a, ⟦ϕ ⟧ M,g[e/u]= 1

I assume the two basic combinatoric rules in (66), Forward Function Application (FA) and Backward Function Application (BA). These rules compositionally determine the syntactic category and meaning of a complex expression ab from the syntactic categories and meanings of the expressions a and b that make up the complex expression. The rules compose the two expressions a and b through concatenation to result in the complex expression ab or ba. The syntactic category of a is A/B, and that of b is B. The syntactic part of the rule of FA, for example, says that A/B and B combine to form A. The meaning of a is ⟦a´⟧M,g and the meaning of

b is ⟦b´⟧M,g. The meaning of the complex expression is the meaning of a applied to the meaning

of b.

(66) a. Forward Function Application (FA):

a: A/B: ⟦a´⟧M,g b: B : ⟦b´⟧M,g

ab : A : ⟦a´(b´)⟧M,g FA

b. Backward Function Application (BA):

b: B : ⟦b´⟧M,g a: A\B : [[a´⟧M,g

ba : A : ⟦a´(b´)⟧M g BA

Finally, I also assume that the semantics of expressions can be beta-reduced. Beta reduction is a rule that allows the following conversion:

(67) λx[ … x … ](α) = [… α … ]

The conversion is the result of replacing all free occurrences of the variable in the first formula

with the constant in the second formula. For my purposes, it simplifies the notation.

With these basic assumptions presented, I introduce the algebraic structures of Krifka

(1998) that will enrich the semantics and allow reference to eventualities in my analyses.

2.1.1 Krifka (1998)

Krifka (1998) develops algebraic structures that capture the semantics both of pluralities and of the relation between individuals and events. Such complex structures will be necessary in my analyses of sentences that realize reciprocal expressions, because these sentences always express a relation between two or more entities and typically express two or more subevents.11

For example, consider the sentence in (68):

(68) John and Mary hit each other.

This sentence expresses two hitting relations that hold between the two entities denoted by the

subject. It is necessary to assume a structure that allows an accurate specification of the

extension of the predicate. Furthermore, this sentence expresses two subevents; one of the

denotation of John hitting the denotation of Mary, and one of the denotation of Mary hitting that of John. Because this sentence gives rise to a plurality of eventualities, it will also be necessary to have a framework in which to evaluate eventualities and subeventualities.

Here I adopt the algebra U, an algebra that imposes a structure on individuals, and E, an algebra that imposes a structure on events. U is a part structure, such that U = 〈U’, ⊕, ≤, <, ⊗〉.

I unpack this structure in the following bulleted list.

11 It is actually debated whether every reciprocal sentence expresses two or more events; see Chapter 3, Section 3.4.2 for more discussion. It is not debated, however, that at least some reciprocal sentences do express two or more events. 53

• U’ is the set of entities in the universe, containing both simple and complex individuals.

• The symbol ⊕ in the part structure represents the sum operation, which is a function from U’ × U’ to U’. The set U’ is closed under the sum operation. The domain of this function, the Cartesian product U’ × U’, includes the ordered pairs of simple and complex individuals. The range includes the complex pairs made up from both individuals in the ordered pairs of the domain. For example, if the simple individuals in U’ are {a, b, c}, then the set U’ includes these simple individuals plus all possible complex pairings of these individuals: {a, b, c, a⊕b, a⊕c, b⊕c, a⊕b⊕c}.This function from U’ × U’ to U’ is:

o Idempotent (∀ x ∈ U’ [x ⊕ x = x]). This says that for all x that are elements in U’, the sum of x and x is x.

o Commutative (∀ x, y ∈ U’ [x ⊕ y = y ⊕ x]). This says that for all x and y that are elements of U’, the sum of x and y is the same of the sum of y and x.

o Associative (∀ x, y, z ∈ U’ [ x ⊕ (y ⊕ z) = (x ⊕ y) ⊕z]). This says that for all x, y, z that are elements of U’, the sum of x and the complex individual y and z is the same as the sum of the complex individual x and y and z.

To illustrate the definition of the ⊕ operator, if the simple individuals in U’ are {a, b, c}, and the input to the sum function is an ordered pair 〈a, a〉, then the output is a. If the ordered pair is 〈a, b〉, then the output, a⊕b is the same individual as b⊕a. And lastly, given the input 〈a⊕b, c〉, the output (a⊕b)⊕c is the same individual as a⊕(b⊕c). For what follows, I will always assume that the simple individuals in U’ are {a, b, c}.

I now continue defining the elements of the U algebra.

• ≤ represents the part relation, defined as ∀x, y ∈ U’ [ x ≤ y ↔ x ⊕ y = y]. This says that for all x and y that are elements of U’, x is a part of y if and only if the sum of x and y is y. In U’, a is a part of a⊕b. Also, a is a part of a.

• < represents the proper part relation, defined as: ∀ x, y ∈ U’ [ x < y ↔ x ≤ y ˄ x ≠ y]. This says that for all x and y that are elements of U’, x is a proper part of y iff x is a part of y and x and y are distinct. In U’ , a is a proper part of a⊕b, though a is not a proper part of a.

• ⊗ represents the overlap relation, defined as ∀x, y ∈ U’ [x ⊗ y ↔∃z ∈ U’ [ z ≤ x ˄ z ≤ y]]. This says that for all x and y that are elements in U’, x overlaps with y iff there is an entity z that is also part of U’ such that z is a part of x and z is a part of y. In U’, a⊕b and b⊕c overlap because there is an individual, b, is a part of both a⊕b and b⊕c.

• Remainder principle: ∀x, y ∈ U’ [ x < y→∃ ! z[ ¬[z ⊗ x] ˄ x ⊕ z = y]]. This principle requires that for any two individuals, x and y, if x is a proper part of y, then there exists exactly one remainder z which does not overlap with x and which can be summed with x to give y. This remainder z is the difference between x and y. Assuming U’, if a is a proper part of b, then there is a remainder c that does not overlap with a, and b is the sum of a and the remainder c.

Having defined the U algebra for individuals, I now define the E algebra that imposes structure on eventualities.

12 E is the eventuality structure: E = 〈E’, ⊕E, ≤E,

structure:

12 Krifka’s (1998) framework does not explicitly include states. Rather, his framework accounts for different types of events. I see no problem, however, for my purposes, of including states in his framework, which should be constrained by the same axioms as events. 55

• E’ is the set of eventualities in the model. Each eventuality is either a state (eS) or an

event (eE). E’ contains simple and complex events and simple and complex states.

• ⊕E is the sum operation, similar to the sum operation in U. It is a subset of the functions E’ × E’ to E’ such that all eventualities participating in the function are of the same type (event or state). The range of this operation includes the complex pairs made up from both entities in the ordered pairs of the domain. The sum operation for eventualities has the same properties (idempotent, commutative and associative) as in U. E’ is closed

under the sum operation. Assume that the simple events in E’ are e1E, e2E and e3E. The

sum operation outputs the complex events e1E⊕Ee2E, e2E⊕Ee3E, e1E⊕Ee2E⊕Ee3E, for example. The same applies to the states in E’.

• ≤E,

• τ represents a temporal trace function from E’ to T’ that maps an eventuality to its running time.

• ∞E represents a temporal adjacency relation that is defined relative to the temporal adjacency relation for times (see below).

• ≪E represents a two-place temporal precedence relation in E’. This relation is also defined relative to the temporal precedence relation for times (see below).

Neither of these relations, adjacency or precedence, can be defined without a definition of the same operations for the time structure. T is a time structure in E such that T=〈T’, ⊕T, ≤T,

∞T, ≪T 〉. T’ is the set of simple and complex times in the structure. ⊕T, ≤T,

• ∞T represents a temporal adjacency relation. It is a two-place relation in T’ such that ∀ t, t’ ∈ T’ [ t ∞T t’ → ¬ t ⊗T t’] and ∀t, t’, t” ∈ T’[t∞T t’ ˄ t’ ≤T t” → t ∞T t” ˅ t ⊗T t”]. This definition says that for all t and t’ that are elements of T’, if t is temporally adjacent to t’, then it is not the case that t and t’ overlap. Furthermore, for all t, t’ and t” that are elements of T’, if t is temporally adjacent to t’ and t’ is a part of t” , then t is temporally adjacent to t”, or t overlaps with t”.

• ≪T represents a two-place temporal precedence relation in T’. It has the following properties:

o ∀t, t’ ,t” ∈ T’[[¬t ≪T t] ˄ [ t ≪T t’→ ¬t’ ≪T t] ˄ [t ≪T t’ ˄ t’≪T t” → t ≪T t”]]. For all t, t’ and t” that are elements of T’, t does not temporally precede itself, if t temporally precedes t’ then it is not the case that t’ temporally precedes t, and if t temporally precedes t’ and t temporally precedes t”, then t also temporally precedes t”.

o ∀t, t’ ∈ T’[t ≪T t’ → ¬t ⊗T t’]. For all t and t’ that are elements of T’, if t temporally precedes t’ then t and t’ do not overlap.

o ∀t, t’ ,t” ∈ T’[ t, t’ ≤T t” ˄ ¬ t ⊗T t’ → t ≪T t’ ˅ t’≪T t]. For all t, t’ and t” that are elements of T’, if t and t’ are parts of t” and t and t’ do not overlap, then t temporally precedes t’ or t’ temporally precedes t.

o ∀t, t’ ∈ T’[t ≪T t’ →∃t” ∈ T’[t, t’ ≤T t”]. For all t and t’ that are elements of T’, if t temporally precedes t’, then there exists a t” that is an element of T such that t and t’ are parts of t”.

• CE is the set of temporally contiguous events and is a subset of E’ with the properties:

o ∀e, e’ ∈ E’ [τ(e ⊕E e’) = τ(e)⊕T τ (e’)]. The run time of a complex entity made of any two eventualities in E’ is the same as the sum of the run times of the individual events. Assuming the same E’, for any complex eventuality such as e1 ⊕E e2, τ maps this complex eventuality onto one run time, and this run time is the sum of the run times of e1 and e2.

o ∀e, e’ ∈ E’ [e ∞E e’ ↔ τ(e) ∞T τ (e’)]. Any two eventualities in E’ are temporally adjacent iff the run times of these eventualities are temporally adjacent. For example, e1 and e2 are adjacent iff their run times are temporally adjacent.

o ∀e, e’ ∈ E’ [e ≪E e’↔ τ(e) ≪T τ (e’)]. For any two eventualities e and e’ in E’ e temporally precedes e’ iff the run time of e temporally precedes the run time of e’. If the temporal traces of an event e1 temporally precedes a second event e2, then e1 temporally precedes e2.

2.1.2 Analysis of a fragment of Russian

I provide here a syntactic and semantic analysis of the simple sentence in (69) as an illustration of the framework adopted in this dissertation. I first translate a sentence that realizes a singular entity and event and then discuss how to analyze complex entities and eventualities.

(69) Marija tancevala. Maria.NOM dance.PST.SG ‘Maria danced.’

The lexical entries of the basic expressions to be used in my derivation are in (70):

(70) Basic Expressions:

M,g Marija: N: ⟦m⟧

M,g tancevat' ‘dance: N\S’: ⟦dance’⟧

M,g -l PAST: S/S’: ⟦λP [∃e(P(e) ˄τ(e) ≪ now)]⟧

Note that in the translation of the past tense expression -l, I assume the Davidsonian event argument is bound by an existential quantifier that is introduced by the tense expression.

I provide the derivation of this sentence in Figure 7. This derivation shows how each of the basic expressions in (70) compose to derive the sentence in (69).

The derivation of the sentence, given these lexical entries, proceeds as in Figure 7:

M,g M,g 1. Marija: N: ⟦m⟧ tancevat’ : N\S’: ⟦dance’⟧

2. Marija tancevat’: S’: ⟦dance’⟧M,g (⟦m⟧M,g) BA

3. -l :S/S’: ⟦λP[∃e(P(e) ˄τ(e) ≪ now)] ⟧M,g Marija tancevat’: S’: ⟦dance’(m’)⟧M,g BR

4. Marija tancevala: S: ⟦λP[∃e(P(e) ˄τ(e) ≪ now)] ⟧M,g (⟦dance’(m)⟧M,g) FA

5. Marija tancevala: S: ⟦∃e(dance’(m) (e) ˄ τ(e) ≪ now) ⟧M,g BR

Figure 7: Derivation of the sentence Marija tancevala.

The verb tancevat’ ‘dance’ has the syntactic category N\S’ and denotes a function from an individual to events to truth values. It combines via Backward Function Application with an expression of type N, here Marija, to derive a sentence radical. This sentence radical has the syntactic type S’ and denotes a function from an event to a truth value. The semantic translation

of this sentence radical in line 2 of the derivation is beta reduced to result in the translation in line

3. The past tense expression has a syntactic category of S/S’ and combines with the sentence

radical to result in a sentence of type S via Forward Function Application. The denotation of the

tensed sentence is derived by applying the denotation of the past tense expression to the

denotation of the sentence radical. Finally, the last line of the derivation is equivalent to that in

the second to last line and has simply been beta-reduced.13

I repeat the part of the last line of Figure 7 which is the semantic translation of the

sentence in (69):

13 Note that in lines 3 and 4 the past tense -l combines with the sentence radical Marija tanceva- ‘Maria dance’ using Forward Application. The past tense phonology, however, does not precede the rest of the sentence as the FA rule dictates. I acknowledge that this is a problem for the derivation, but since tense is not a focus of this dissertation, I will not develop a more complex analysis here that accounts for the phonology. 59

(71) ⟦∃e(dance’(m) (e) ˄ τ(e) ≪ now)⟧M,g

This translation says that the sentence in (69) is true iff there is some event such that it is an event of the individual Marija dancing and this event occurred in the past. This translation appropriately captures the meaning expressed by the sentence in (69).

Having walked through a derivation with a simple entity and event, I now discuss the analysis of a sentence with complex entities and events. This analysis will make use of the algebras proposed by Krifka (1998).

(72) Ivan i Marija tancevali. Ivan.NOM and Marija.NOM dance.PST.PL ‘Ivan and Marija danced.’

The basic expressions of this sentence are in (73):

(73) Basic Expressions:

M,g Ivan: N: ⟦i⟧

M,g Marija: N: ⟦m⟧

M,g i ‘and’: (N\N)/N : ⟦⊕>⟧

M,g tancevat’: ‘dance’: N\S’: ⟦dance’>⟧

M,g -l PAST: S/S’: ⟦λP [∃e(P(e) ˄τ(e) ≪ now)] ⟧

The major addition is the expression i ‘and’ which denotes the sum operation. It first combines with an expression of syntactic type N, then with a second N to result in N, as outlined in the derivation in Figure 8 below:

M,g M,g 1. i: (N\N)/N : ⟦⊕⟧ Marija: N: ⟦m⟧

2. Ivan: N: ⟦i⟧M,g i Marija: (N\N) : ⟦⊕m⟧M,g FA

M,g M,g 3. Ivan i Marija: N : ⟦i⊕m⟧ tancevat’ : N\S’: ⟦dance’⟧ BA

M,g M,g 4. -l :S/S’: ⟦λP[∃e(P(e) ˄τ(e) ≪ now)]⟧ Ivan i Marija tancevat’: S’: ⟦dance’ (i⊕m)⟧ BA

5. Ivan i Marija tancevali: S: ⟦λP[∃e(P(e) ˄τ(e) ≪ now)] ⟧M,g ⟦(dance’(i⊕m))⟧M,g FA

6. Ivan i Marija tancevali: S: ⟦∃e(dance’(i⊕m) (e) ˄ τ(e) ≪ now)⟧M,g BR

Figure 8: Derivation of the sentence Ivan i Marija tancevali ‘Ivan and Marija danced.’

I repeat the semantic portion of the last line of the derivation in Figure 8 in (74):

(74) ⟦∃e(dance’(i⊕m) (e) ˄ τ(e) ≪ now) ⟧M,g

This translation predicts the sentence in (72) to be true iff there is an event such that it is an event

of the complex individual made up of the denotation of Ivan and Marija dancing, and this event

occurred before now. Once again this translation captures the meaning of the sentence in (72).

Having now provided an overview of my formal framework, I introduce the two main

approaches for developing a compositional analysis of each other in the formal semantic

literature, and I outline the assumptions that I will make for my own analysis. 61

2.2 Formal accounts of reciprocity

Within the formal semantic literature, there are two main approaches for developing a

compositional analysis of each other. Some linguists (e.g. Dalrymple et al. 1998b, Sabato and

Winter 2005) hold that each other is a generalized polyadic quantifier, i.e. a quantifier that maps

a set to a relation. Others (e.g. Langendoen 1978, Heim et al. 1991, Beck 2001) propose that the

meaning of each other is compositionally built up from its components and additionally rely on

operators that are optionally inserted into sentences to derive the meaning of sentences with each

other. I outline both of these approaches here. For this dissertation, I adopt the polyadic

quantifier approach, and I will provide my reasons after having summarized both.

Dalrymple et al. (1998b) propose that each other is a generalized polyadic quantifier that

is a function such that it maps the set denoted by the antecedent of the reciprocal expression to the

binary relation expressed by the verb. For example, in the sentence in (75) below, the set denoted

by the subject, i.e. Tom, Dick and Harry, is the antecedent of the reciprocal expression. This set

is then mapped to a seeing relation. They propose that the proposition of the sentence in (75) can

be represented by (76):

(75) Tom, Dick and Harry saw each other.

(76) RECIP({Tom, Dick, Harry}, λxy.saw(x, y)) (Dalrymple et al. 1998: 183)

This says that the RECIP function takes two arguments, one of which is the set of individuals denoted by the subject, and the second the seeing relation. They argue that RECIP can have three

different meanings (186):

(77) a. each pair of individuals in A may be required to participate in the relation R directly (FUL);

b. each pair of individuals in A may be required to participate in the relation R either directly or indirectly (LIN);

c. each single individual in A may be required to participate in the relation R with another one (TOT)

In addition to these three possible meanings of RECIP, they hold that the relation R can vary; it

can either be a symmetric relation that holds between two entities, such that both Rab and Rba

are possible, or an asymmetric relation that requires only Rab. The predicate determines which

of these two types of relations each other scopes over. For example, the predicate give measles to

is only compatible with asymmetric relations, while the predicate see is compatible with both

symmetric and asymmetric relations. By applying one of the RECIP operators to these two

different types of relations, six truth-conditionally distinct relations are compositionally

composed. Dalrymple et al. 1998b found that each other is compatible with five of these six

theoretically possible relations.

The first relation they explore is Strong Reciprocity (SR):

(78) Strong Reciprocity (SR):

|A|≥2 and ∀x,y ∈A (x≠y → (Rxy∧ Ryx))

A relation is strongly reciprocal when the relation holds of two or more entities and for all x and y

that are part of a set A, if x and y are distinct then x stands in the R relation to y, and y stands in

the R relation to x. Dalrymple et al. provide the following example from their corpus study (168):

(79) House of Commons etiquette requires legislators to address only the speaker of the House and refer to each other indirectly.

They argue that this sentence is only true when a SR relation holds between the denotation of the

legislators; if one legislator was allowed to refer to other legislators directly, then this sentence is

false.

The second relation they posit is Intermediate Reciprocity (IR):

(80) Intermediate Reciprocity (IR):

|A|≥2 and ∀x,y ∈A (x≠ y→ for some sequence z0, . . . , zm ∈A (x = z0 ∧ Rz0z1 ∧

. . . ∧Rzm- 1zm∧zm=y))

This definition says that IR is defined for a set of two or more members and that for all x and y that are elements of the set A, if x and y are distinct, then each entity stands in the R relation to another entity in a sequence as both the first and second arguments. Dalrymple et al. paraphrase it succinctly: “Every member of A is related directly or indirectly to every other member via the

relation R” (170). For example, they provide the following sentence (170):

(81) The telephone poles were spaced five hundred feet from each other.

Once again they argue that this sentence is true only if an IR relation holds between the entities denoted by the telephone poles.

The next relation they investigate is One-way Weak Reciprocity (OWR):

(82) One-way Weak Reciprocity (OWR):

A|≥2 and ∀x∈A ∃y∈A (x≠ y∧ Rxy)

This definition says that OWR is defined for a set of two or more members and that for all x that are elements of A there exists at least one y in A such that x and y are distinct and x stands in the

R relation to y. In other words, each entity in A stands in the R relation as the first argument to at least one other entity. They provide the following example (171):

(83) “The captain!” said the pirates, staring at each other in surprise.

They argue that it is impossible for each pirate to stare at more than one pirate at a time, and so this sentence exemplifies OWR because in order for this sentence to be true, each pirate must stand in the staring relation as the first argument, but not necessarily as the second.

Next, they posit the relation of Intermediate Alternative Reciprocity (IAR):

(84) Intermediate Alternative Reciprocity (IAR):

∀ ∈ ∈ ∧ ∨ |A|≥2 and x,y A (x≠ y͢ → for some sequence z0, . . . , zm A (x = z0 (Rz0z1 Rz1z0) ∧ . . .∧ (Rzm-1zm∨Rzmzm-1)∧zm=y))

They write that this definition says “Each member x of A should be related to every other member y via a chain of R relations, where we ignore which way the pairs making up the chain are related via the relation R.” (173). They provide the following example (173):

(85) The third-grade students in Mrs. Smith’s class gave each other measles.

It is important to remember for the interpretation of this sentence that measles is a disease that

you can only have once; if you have had it already you cannot contract it again. Therefore this

sentence describes a chain of relations of measles-giving, in which it is impossible for the

receiver to give it back to the one who gave it.

The last relation they define is Inclusive Alternative Ordering (IAO):

(86) Inclusive Alternative Ordering (IAO):

|A| ≥ 2 and ∀x ∈ A ∃y ∈ A (x ≠ y ∧ (Rxy∨ Ryx))

This relation is defined for any set that has two or more members and for all members of this set s there is at least one member y such that x and y are distinct and x stands in the R relation to y or y stands in the R relation to x. Dalrymple et al. write that: “every member x of the set A

participates with some other member in the relation R as the first or as the second argument, but not necessarily in both roles” (175). The example they provide is in (87) (174):

(87) He and scores of inmates slept on foot-wide wooden planks stacked atop each other – like sardines in a can – in garage-sized holes in the ground.

This sentence describes a situation where there are piles of planks upon which the inmates sleep.

Each plank stands in the stacked-atop relation as the first or second argument to at least one other plank. This relation differs from IAR only in that it allows that not all the entities in the relation are directly or indirectly related to each other via a sequence.

These are the five relations that Dalrymple et al. (1998b) found evidence for in a corpus study. These relations in fact can be ordered in an implicational hierarchy as in Figure 9. The relations at the top are logically stronger than the ones below them, in the sense that if a sentence

is compatible with SR, it is also compatible with the other five relations. Similarly, if a sentence

is compatible with IR, it is also compatible with OWR, IAR and IAO.

Strong Reciprocity

Intermediate Reciprocity Strong Alternative Reciprocity

One-way Weak Reciprocity Intermediate Alternative Reciprocity

Inclusive Alternative Ordering

Figure 9: Dalrymple et al.’s (1998b) implicational hierarchy of reciprocal relations

The relation Strong Alternative Reciprocity is defined in (88):

(88) |A| ≥ 2 and (∀x,y ∈ A) (x≠y ∧ (Rxy ∨ Ryx))

This definition requires the set A that denotes the antecedent of each other to have at least two

entities x and y, and then for all entities in A that are non-identical, the R relation holds between them such that x stands in the R relation to y or y stands in the R relation to x. Dalrymple et al.

posit this relation as falling out from their implicational hierarchy, but they find no sentence in

their corpus study that conveys this relation. They therefore predict that each other does indeed

convey this relation and that they simply did not find such a sentence in their corpus study. They

did not perform any elicitation tasks to follow up on this prediction.

Which relation each other conveys in a given sentence will depend on the linguistic and

non-linguistic context. Dalrymple et al. argue that a sentence that realizes each other conveys the

strongest relation, based on the hierarchy in Figure 9, that is compatible with the context. For example, they consider the following sentence (191):

(89) As the preposterous horde crowded around, waiting for the likes of Evans and Mike Greenwell, five Boston pitchers sat alongside each other: Larry Andersen, Jeff Reardon, Jeff Gray, Dennis Lamp and Tom Bolton.

The relation of sitting alongside is symmetric, as it is impossible to sit alongside a person and not

have them sit alongside you. Therefore the relation that this sentence expresses cannot be OWR,

SAR, or IAR. This leaves the relations of SR, IR and IAO. Since people only have two sides, SR

cannot hold of the relation since there are five individuals. Two relations are left: IR and IAO.

IR is the relation that is actually interpreted and it is the stronger of the two.

Such observations lead them to propose the following principle:

(90) Strongest Meaning Hypothesis (SMH): A reciprocal sentence S can be used felicitously in a context c, which supplies non-linguistic information I relevant to the reciprocal’s interpretation, provided the set c has a member that entails every other one:

c = {p | p is consistent with I and p is an interpretation of S obtained by interpreting the reciprocal as one of the six quantifiers in Figure 9}

In that case, the use of S in c expresses the logically strongest proposition in c

The SMH allows them to capture which of the five relations is conveyed in any given sentence if the context is known. To illustrate how this works, consider the sentence in (91):

(91) The girls followed each other into the room.

We know, based on world knowledge, that the following relations must be asymmetric; the girls went into a room (and the sentence says nothing about any of them coming back out), and so it is impossible for any girl to be followed and to follow the same girl in such a situation. The only asymmetric relations are OWR, IAR and IAO. The relation cannot be OWR, because OWR requires that each entity participates in the relation as the first argument, and so it is incompatible with this sentence. Therefore, we are left with IAR and IAO. IAR requires that the girls went into the room in one group, while the IAO would allow there to be groups of girls such that the girls within each group followed each other into the room. We have no further information to whittle down the choices further. We are therefore left with two interpretations of the sentence in

(91) in the set c., which is the set of propositions that are consistent with the context, that are

interpretations of the sentence in (91) obtained by interpreting the reciprocal as one of the six

quantifiers in Figure 9. One of these interpretations comes about when each other is interpreted

as contributing an IAR relation and the other when each other is interpreted as contributing an

IAO relation. According to the SMH, S must express the logically strongest proposition in c .

Since IAR is the stronger of the two, the interpretation of the sentence in (91) is obtained by interpreting each other as IAR.

It is by assuming the SMH that Dalrymple et al. (1998b) account for the different meanings that each other can contribute to a sentence and how they vary depending on linguistic and non-linguistic factors. This article has been fundamental for the study of reciprocity. For one, it is the first to propose a formal analysis that accounts for all of these relations. Second, it is the first to propose an analysis that makes use of a polyadic quantifier to capture the meaning of each other. Several future analyses of each other (e.g. Sabato and Winter 2005 and Dotlačil and 70

Nilson 2008) adopt this type of account, though they adapt the exact mechanisms. Lastly,

Dalrymple et al. (1998b) is the first to consider how the reciprocal expression is interpreted in context. Other works, both those also assuming polyadic quantifiers (e.g. Filip and Carlson 2001,

Sabato and Winter 2005) and those assuming the anaphor approach (e.g. Beck 2001), make reference to the SMH in their proposals. I now turn to summarizing the second approach to characterizing the meaning of reciprocal expressions.

There are several versions of this account (Langendoen 1978, Heim et al. 1991,

Sternefeld 1998, Beck 2001). In general these accounts have two things in common. First, they compose the meaning of each other from the two words in the expression. Heim et al., for example, argue that each contributes a distributive operator to a sentence, whereas other contributes a non-identity condition on the relation. Second, they make use of optional operators, such as the distributive operator, that are available for sentences that don’t realize a reciprocal expression. I will summarize in more detail here the account of Beck (2001), as it is the most recent account, and it incorporates elements from previous analyses.

Beck assumes that a distributive or a cumulative operator is optionally inserted into sentences in English. The distributive operator distributes the extension of the predicate to atoms.14 This operator is the one that captures the first interpretation of the (non-reciprocal)

sentence in (92), where each atomic child is in the extension of the predicate. The second

interpretation is collective, so that only the sum of all of the children is in the extension of the

predicate. This interpretation won’t concern me more here, as reciprocal expressions are not

compatible with a collective reading.

14 Actually Beck (2001) makes use of the Cover mechanism of Schwarzschild (1996), but I will ignore this for the sake of simplicity for the moment. I will introduce the Cover mechanism in Section 2.3.1 below. 71

(92) The children weigh 50 pounds. (i) Each child weighs 50 pounds. (ii) The children together weigh 50 pounds.

Beck represents the distributive operator with the symbol *. In a sentence with a transitive verb

such as that in (93), the distributive operator can apply to both the subject and the object,

capturing the reading in (93)(i). In this interpretation, Sue read ‘Fried Green Tomatoes’, Amy

read ‘Fried Green Tomatoes’, Sue read the ‘L-Shaped Room,’ and Amy ready ‘The L-Shaped

Room’. However, there is another interpretation of the sentence, namely the one in (93)(ii).15

(93) Sue and Amy read ‘Fried Green Tomatoes’ and ‘The L-Shaped Room.’ (Beck 2001: 80)

(i) Both Sue and Amy read both ‘Fried Green Tomatoes’ and ‘The L-Shaped Room’.

(ii) One of the women read ‘Fried Green Tomatoes’ and the other read ‘The L-Shaped Room.’

This second interpretation is not doubly-distributive, because it is not the case that every entity

denoted by the subject stands in the reading relation to every entity denoted by the object. Rather

there is a one-to-one correspondence such that either Sue or Amy read ‘Fried Green Tomatoes’

and the woman that did not read ‘Fried Green Tomatoes’ read ‘The L-Shaped Room’. Beck,

following Sternefeld (1998), calls this interpretation cumulative. She represents this operator

with the symbol **. Unlike the distributive operator, the cumulative operator does not atomize

entities; rather it atomizes relations. The following definition which Beck adopts is from

Sternefeld (1998: 304):

15 There are also other possible readings involving the collective reading, but these are not relevant here. 72

(94) For any two-place relation R, let **R be the smallest relation such that a. R ⊆ **R and b. If ∈**R and ∈**R, then ∈ **R

This says that a cumulative relation is a superset of a relation R and if the pair is in the extension of the cumulative relation (i.e. the relation is a mapping from a to b) and the pair is in the extension of the cumulative relation (a mapping from c to d), then the pair of the complex individual a⊕c and b⊕d (a mapping from a⊕c to b⊕d) is also in the cumulative relation.

Following Langendoen (1978), Beck holds that the distributive and cumulative operators are inserted into the derivation because of the properties of plural predicates. She makes use of both in her analysis of each other, arguing that whatever mechanism gives rise to the distributive and cumulative interpretations in transitive sentences also affects sentences with each other in the same way. Furthermore, she argues that the different readings in Dalrymple et al. (1998b) can be composed by manipulating these operators.

In addition to these operators, Beck argues that a non-identity restriction (which she defines as non-overlap; see below) is necessary to capture the meaning of sentences that realize each other. More specifically, Beck proposes that each other denotes a set whose members are all the members of the set denoted by each other’s antecedent, minus the individual that stands in the relation R as the first argument. She therefore ensures that the relation is not specified to hold between an entity and itself. To account for this meaning, she suggests that the underlying form of each other is actually the other ones among them as in (95)b.

(95) a. John, Mary and Sue hugged each other.

b. John, Mary and Sue hugged the other ones among John, Mary and Sue.

She assigns a meaning to each other based on the meaning of the definite determiner the, the adjective other, and the pronoun them. The structure of the other ones among them she posits is

as in (96):

(96) [max[*[[other x1](among) Pro3]]]

The word the introduces the maximality operator from Link (1983) and Sharvy (1980) (defined in

(97) below). The variable x1 represents the word ones, and Pro3 represents the pronoun them.

Both x1 and Pro3 refer back to the antecedent of each other.

She defines the maximality operator as in (97). In this definition, ɩ is the meaning of the,

which maps the constant to the individual in the world. ≤ represents the part-of function (as

defined in Section 2.1.1).

(97) Maximality operator:

Let S be a set ordered by ≤. Then max(S) = ɩs[s ∈ S & ∀s’ ∈ S [s’ ≤ s]]

This says that max(S) denotes the maximal individual of the set S. The maximal individual s is an element of S, and all elements s’ in S are parts of the maximal individual s. For instance, if S is as in (98), then max(S) is as in (99).

(98) S = {Mary, Jane, Sue, Mary⊕Jan, Jane⊕Sue, Mary⊕Sue, Mary⊕Jane⊕Sue}

(99) max(S) = Mary⊕Jane⊕Sue

The meaning of each other in (100) demonstrates how the max operator composes with the other elements in Beck’s analysis. It scopes over the set that is denoted by the distributed interpretation of other x1 among Pro3.

(100) each other = max(*λz[¬ (z ∘ x1) ∧ z ≤ x3]])]

She proposes that the word other contributes a non-overlap condition, bolded in the definition in

(100). This non-overlap condition ensures that no individual is entailed to stand in the R relation

with itself. The max operator applies to a distributed entity z. This entity z is part of the entity x3

which is assigned its meaning by the reciprocal expression’s antecedent. This entity z does not

overlap with x1, one particular member of the set x. The max operator picks out the maximal entity of z. In other words, each other denotes the maximal entity that is a part of the set denoted

by the subject, such that this part of the set denoted by the subject does not include one particular

individual (that will be constrained by the subject later via the syntax-semantics interface). To

illustrate, consider the sentence in (101), its Logical Form in (102), and its translation in (103), in

which MSB stands for the set that includes the individuals Mary, Sue and Bill (Beck 2001: 91): 75

(101) Mary, Sue and Bill saw each other.

(102) [[Mary, Sue and Bill]3[*[1[[max[*[[other x1] (of) Pro3]][*[2[t1 saw t2]]]]]

(103) MSB ∈ *λx[max(*λz[¬ z∘ x & z ≤ MSB]) ∈*λy [x saw y]]

This translation says that the sentence in (101) is true iff each atomic individual x in MSB saw each individual y, such that y is each atomic individual in the maximal set of the distributed z.

The set z is a set that does not overlap with x1 (one of the distributed entities in x) and is part of the denotation of the subject. In other words, each entity in the denotation of the subject stands in the seeing relation to the maximal set denoted by the subject that does not include that same entity. The denotation of Mary stands in the seeing relation to the denotations of Sue and Bill but not herself, the denotation of Sue stands in the seeing relation to the denotations of Mary and Bill, and the denotation of Bill stands in the seeing relation to the denotations of Mary and Sue. This translation then accounts for the SR interpretation of each other.

Beck argues that weaker interpretations of this reciprocal expression can be derived by utilizing the cumulative operator introduced above. The derivation is not as straightforward as simply substituting the cumulative operator for the distributive operator, however, since cumulativity applies to relations and not individuals. I will not include her analysis here, but will just say that her analysis accounts for many of the weaker reciprocal relations (but see my discussion below about Intermediate Alternative Reciprocity).

Although Beck’s analysis accounts for much of the data (see below for data it does not account for), in this dissertation I adopt the type of analysis in Dalrymple et al. (1998b) instead, for three main reasons. First, as Beck admits, her analysis cannot account for the Intermediate 76

Alternative Reciprocity relation. She instead proposes that this relation is composed in a different manner because only a handful of verbs in English are compatible with such a relation and all of these express temporal or spatial relations, e.g. be inside of, follow, lay on top of. She therefore

suggests that the IAR relation “comes about by a lexical process different from ordinary

reciprocity and is limited to the list of relations mentioned above [i.e. the spatial and temporal

relations]” (130). She thus proposes the following lexical meaning for the phrase inside each

other:

(104) inside each other = λX. ⟨X, X⟩ ∈ [**λx λy. x inside y or y inside x]

Therefore, while Beck does provide an account of IAR, her analysis does not lend itself to a

single analysis of the reciprocal relations in English.

Second, Beck’s account, as well as the others in this tradition, is tied directly to the form

of the expression each other. Heim et al. (1991) assume that each contributes the distribution,

and other contributes the non-identity clause. Beck assumes that each other can be transformed

into the others among them and then builds the semantics up from the individual elements, i.e. the

definite determiner, the adjective other and the pronoun them. I see this as a problem because it

has little direct cross-linguistic application; it is not immediately clear how one would apply this

analysis to the -sja verbs, for instance, since there is nothing in this expression that translates to,

or in any way could be transformed to, the others among them. Moreover, her analysis depends

on there being both a definite article to contribute the maximality operator and an expression to

contribute the non-overlap condition. I will show in Chapter 4 that the -sja verbs contribute a

very similar meaning to a sentence as each other and so it seems appropriate, if possible, to

analyze them in a similar manner as each other. It would be stretching the analysis of the -sja 77

verbs to argue that they contribute a definite operator, as the -sja verbs in no other way I am

aware of behave like a definite article. Similarly, even though the reciprocal expression drug

druga is more similar to each other than the -sja verbs in that it consists of two words and has

other in its meanings, drug druga literally means ‘other other’. There is no language internal

reason to posit that either of these expressions contributes a definite article. Therefore neither of

the two words is a clear candidate for introducing the maximality operator. The analysis of

Dalrymple et al. (1998b) is not form-specific and therefore is also not necessarily language-

specific. The polyadic quantifier approach then better accounts for cross-linguistic similarities

between reciprocal expressions.

Finally, there is experimental evidence that the polyadic quantifier approach better

captures the meaning of each other than Beck’s analysis does. Dotlačil and Nilsen (2008) start

from the facts (i) that Beck argues that the meaning of each other is really the meaning of the

others among them, and hence includes the definite determiner with the adjective other and (ii) that she proposes that this reciprocal expression requires the insertion of the distributive or cumulative operator. They propose that if her account is right, then it predicts that the subject

NPs of sentences that realize each other should have the same possible interpretations as subject

NPs of sentences that realize the other as direct objects. They perform the following two experiments to determine whether interpretations of these two expressions are sensitive to the type of NP.

They first test whether or not plural “expressions of non-identity” such as the others actually require the insertion of the * operator cross-linguistically. Their first study, in which

1150 speakers of Dutch participated, surveyed which types of noun phrases (i.e. distributive, definite plurals, coordinated NPs) can serve as antecedents for the phrase de anderen ‘the others’ in Dutch, and which types of NPs can serve as the subject of distributive predicates such as lose a 78

tooth and non-distributive predicates such as find (they do not provide the Dutch predicates they used in the experiment). To do this, they provided their participants contexts and sentences.

They asked their participants to judge whether the sentence was ‘possible’ or ‘not possible’ in the sentence. The participants were given a scale with six options from possible to not possible and were asked to pick one of the options for each context-sentence pairing.

They found that (i) sentences with distributive predicates are more acceptable with quantified DPs than with plural or coordinated NPs and (ii) sentences that realize de anderen are more acceptable with quantified DPs as antecedents than definite plurals or coordinated NPs.

They assume this means that sentences that insert the * operator to derive a distributive meaning are acceptable with quantified DPs. Such sentences are rarely if ever acceptable with definite plurals or coordinated NPs. Second, since they find the same co-occurrence patterns with de anderen as with the distributive predicates, they propose that just like the distributive predicates,

de anderen requires insertion of the * operator. This result supports Beck’s analysis of the others

as she proposes that it requires the distributive operator. Although their study was performed on

Dutch and not English, they assume that these expressions have the same semantic contribution

across languages. Indeed as they point out, result (i) from above has been found to hold of plural

definites in English as well (Brooks and Braine 1996, Kaup et al. 2002).

Dotlačil and Nilsen’s second experiment compares the results of the first experiment to acceptability judgments about sentences that realize the English reciprocal expression each other.

For this experiment, they gave 40 English speaking participants a picture and sentence. The participants were asked to judge whether the sentence was true or false given the picture. The sentences realized different types of subjects as the antecedents of either the phrase each other or the other.

This experiment showed that each other was acceptable with any of the three types of subjects (quantified, plural or coordinated). This result suggests that each other does not require a distributive operator to be interpreted, and thus that Beck’s definition of each other does not capture the distribution and behavior of each other. They conclude by stating that “This leaves the polyadic quantifier approach as better suited to handle the properties of each other”(264).

Therefore, because of these three reasons, I choose to follow Dalrymple et al. (1998b) in analyzing the reciprocal expressions in Russian as generalized polyadic quantifiers. My analysis will differ from theirs, however, in that I propose that the meaning of each relation is contributed solely by the reciprocal expression and is not composed from the possible type of relation (i.e. symmetrical or asymmetrical) plus the RECIP operator. Rather, I will propose different RDIST operators that contribute the entire meaning of the relation, including whether the relation is symmetrical or asymmetrical and how each entity participates in the event (directly, indirectly or with just one other entity). I will outline this analysis in more detail in Section 2.4.

2.3 Each other conveys four relations

In this section I outline the relations that I myself will use throughout this dissertation, as the set of relations I propose for each other is somewhat different from the set proposed by

Dalrymple et al. (1998b). Additionally, there has been some debate in the literature as to which

relations are appropriate to characterize the meaning of each other, and I will introduce other voices when it is relevant. I take up this issue of the meaning of each other here because

throughout this dissertation I will compare these relations with the relations that drug druga and

the -sja verbs are compatible with to discover uniformity and variation between the three

expressions.

Dalrymple et al. (1998b) find five different relations each other can convey: Strong

Reciprocity, Intermediate Reciprocity, One-Way Weak Reciprocity, Intermediate Alternative

Reciprocity and Inclusive Alternative Ordering, and they predict that each other should be compatible with Strong Alternative Reciprocity, all of which are defined in Section 2.2 above.

They do not find, however, a sentence that satisfies SAR that doesn’t also satisfy IAO, and as far as I know no subsequent work on reciprocity has found such a sentence. I therefore do not include SAR in my list of reciprocal relations.

Out of the remaining five relations, I also discard Intermediate Alternative Reciprocity

(IAR). I propose that IAR is in fact too weak to capture the meaning of English each other. I repeat the definition of IAR in (105):

(105) Intermediate Alternative Reciprocity:

∀ ∈ ∈ ∧ ∨ |A|≥2 and x,y A (x ≠ y͢ → for some sequence z0, . . . , zm A (x = z0 (Rz0z1 Rz1z0) ∧ . . .∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm= y))

“Each member x of A should be related to every other member y via a chain of R

relations, where we ignore which way the pairs making up the chain are related via the

relation R.” (173).

Dalrymple et al. show that their definition of IAR captures the personal relations expressed by the sentence in (106):

(106) The third-grade students in Mrs. Smith’s class gave each other measles.

This sentence conveys IAR because it is not possible to have measles twice, and so all of the relations expressed by the sentence must be asymmetrical. One could depict such relations as in

Figure 10:

Figure 10: Depiction of the sentence The third-grade students in Mrs. Smith’s class gave each other measles.

While I agree that the definition in (105) captures the meaning of this sentence, I argue that the relation as defined is too weak in that it predicts sentences to be acceptable when they are not.

Consider, for instance, the following sentence:

(107) ?John and Mary followed each other into the room.

Dalrymple et al. predict this sentence to be acceptable because each entity stands in the following relation as the first or second argument via a chain. For most native speakers, the sentence in

(107) is unacceptable, and intuitively the reason is that there are only two entities. For some, the 82

sentence in (108) is better, while the sentence in (109) that realizes a definite plural is unquestionably acceptable.

(108) ?John, Mary and Frank followed each other into the room.

(109) Context: You teach at an all-girls school. You have 20 girls in your class. After recess all the girls return to your classroom in a single file. You say:

The girls followed each other into the room.

It seems then that an easy adjustment to Dalrymple et al.’s definition of IAR would be to simply

change the required cardinality of participants in the relation to 3 instead of 2. This new

definition of IAR would then predict that the sentence in (107) is unacceptable, as well as the

sentence in (110) when describing the relation between two books as depicted in Figure 11.

Figure 11: Depiction of two books

(110) #The books are lying on top of each other.

Consider, however, the picture in Figure 12 and sentence in (111) that cannot acceptably be used to describe this configuration of three books:

Figure 12: Depiction of three books

(111) #The books are lying on top of each other.

The configuration of books in Figure 12 satisfies the truth conditions of IAR as each book is in a direct or indirect relation to every other book as either the first or second argument via a chain and the relation holds between more than two entities. Thus, the sentence in (111) should be acceptable if IAR is an appropriate relation for each other when the cardinality of participants in the relation is required to be 3 or more. Since the sentence is not acceptable, I propose that the definition of this relation provided in Dalrymple et al. (1998) overgenerates.

I instead suggest the following relation, which I call Revised Intermediate Alternative

Reciprocity (RIAR):

(112) Revised Intermediate Alternative Reciprocity:

|A|≥2 and ∀x,y ∈A (x ≠ y →for some sequence w0, . . . wm ≤ x ˄ ∃e0, . . . , en ≤ e (y = w0

∧ (R(w0)(w1)(e0) ∨ R(w1)(w0)(e0)) ∧ . . .∧ (R(wm-1)(wm)(en) ∨ R(wm)(wm-1)(en))∧ wm = z) ∧

∃x2 ∃y2 ∃z2 ∈ A (x2≠ y2 ∧ x2≠ z2 ∧ Rx2y2 ∧ Rz2x2)]

The RIAR relation requires A to have at least two participants and then each participant is related directly or indirectly to every other participant either as the first or second argument in the

R relation and there is at least one participant which participates as the first and the second argument in the R relation. This definition accounts for why the sentence in (111) is unacceptable in the context: there is no one entity that is involved in the lying relation as both the first and second argument. It furthermore predicts that the sentence in (107) is unacceptable because no entity stands in the relation as the first and second argument.

Following Sabato and Winter (2008), I also hold that IAO is not a reciprocal relation, but to argue this point, it is necessary to introduce the Cover mechanism of Schwarzschild (1996). I do this in the next section.

2.3.1 The Cover mechanism of Schwarzschild (1996)

Consider the picture in Figure 13 and the sentence in (113) that is acceptable in this context:

1 2 4 5

3 6

Figure 13: A possible depiction of the sentence The boys kicked each other.

(113) The boys kicked each other.

Since the sentence in (113) is acceptable in the context, I assume that each other is compatible with the relation depicted. In terms of the relations already surveyed, the only relation that captures the configuration of kicking relations between the boys is Inclusive Alternative

Ordering, the definition of which is repeated in (114):

(114) Inclusive Alternative Ordering (IAO):

|A| ≥ 2 and ∀x ∈ A ∃y ∈ A (x ≠ y ∧ (Rxy∨ Ryx))

This relation is defined for any set that has two or more members and for all members of this set there is at least one member y such that x and y are distinct and x stands in the R relation to y or y stands in the R relation to x.

It seems straightforward to simply analyze this relation as IAO, as each individual stands in the kicking relation to at least one other entity as the first or second argument. Note however that a SR relation holds between boy 1, boy 2 and boy 3, and a SR relation holds between boy 4, boy 5 and boy 6. It is therefore also possible to analyze this relation as SR that holds within partitions. Under this analysis, the strongest relation that holds within a partition could satisfy the

SMH. I adopt the latter analysis for the following two reasons.

First, consider the two contexts in (115) and the sentence adapted from Sabato and

Winter (2008: 4) that is acceptable in both of them:

(115) Context 1: You are looking at a photo, in which there are two singers. Singer 1 is looking into Singer 2’s eyes and vice versa.

Context 2: You are looking at a photo, in which there are four singers. Singer 1 and Singer 2 are looking into each other’s eyes and Singer 3 and Singer 4 are looking into each other’s eyes.

The singers are looking into each other’s eyes in the photo.

This sentence is acceptable in the first context, which describes a SR relation between the two singers. The sentence is also acceptable in the second context, when there four individuals and these individuals are partitioned such that each singer is only looking into one other singer’s eyes.

Based on what we know about the world, the relation of looking into each other’s eyes must be defined over two individuals as it is impossible to look into more than one person’s eyes at once.

Consider, for example, the following context: 87

(116) Context: You are looking at a photo in which there are four singers. Singer one is looking into singer two’s eyes, but singer two isn’t looking into singer one’s eyes. Singer three is looking into singer four’s eyes, but singer four is not looking into singer three’s eyes.

#The singers are looking into each other’s eyes in the photo.

The context in (116) sets up an IAO relation between participants, but not a partitioned SR relation. Significantly, the sentence in (116), the same as in (115), is not acceptable in this context. This unacceptability suggests that IAO is in fact too weak a relation for this sentence to convey, and therefore the analysis of such sentences as partitioned SR is superior.

The second reason to analyze a sentence in a context that establishes either a IAO or a partitioned SR interpretation as conveying the latter relation comes from Schwarzschild (1996).

He provides evidence that this partitioning effect is not in fact introduced by each other, but

rather by the optional distributive operator.

Schwarzschild is interested in accounting for the correct reading of the following

sentence, when the men denotes Rodgers, Hammerstein and Hart.

(117) The men wrote musicals.

Based on world knowledge, this sentence is false on a distributive reading, as it is not the case

that each man wrote musicals. It is also false on a collective reading, as it is not the case that the

three men wrote musicals together. It is only true on the reading that Rodgers and Hammerstein

wrote musicals together, and Rodgers and Hart wrote musicals together. To account for such a

reading, Schwarzschild, following work by Higginbotham (1980) and Gillon (1987), postulates the notion of a Cover. In the following definition of a Cover, P represents a plurality.

(118) C is a cover of P if and only if:

(i) C is a set of subsets of P

(ii) Every member of P belongs to some set in C

(iii) Ø is not in C

In the example above with the writers of musicals, the plurality P is as in (119), and the

Cover C is as in (120):

(119) P = {Hammerstein, Hart, Rodgers}

(120) C = {{Hammerstein, Rodgers}, {Rodgers, Hart}}

In this case, C is a Cover because every member of P is in one of the subsets in C, and there is no

empty set in C. Schwarzschild then proposes the following interpretive principle for Covers:

(121) [S NPplural VP] is true in some context Q iff there is a cover C of the plurality P denoted by NP which is salient in Q and VP is true for every element in C.

Crucially, the context is part of the truth conditions of this principle. If we were not aware that the speaker of the sentence in (117) was referring to Hammerstein, Hart and Rodgers, or if we did not 89

know who they were or who collaborated with whom, we would not know what Cover the speaker intended; the intended Cover is not specified in the semantics anywhere. By having this information that is salient in the context, it is possible to infer that the denotation of the VP applies to the relevant Cover, and not either to individuals or to the three men collectively.

Schwarzschild analyzes Cov as a free variable over sets of sets that is built into the meaning of the distributive operator. He proposes the following semantic rule for the distributive operator, where D is the * operator, x represents a plurality and α represents a verbal predicate:

(122) x ∈ ||D(Cov)(α)|| iff ||Cov|| is a cover of x ∧ ∀y[y ∈ || Cov|| → y ∈ || α|| ]

This says that the denotation of a plurality is in the extension of D(Cov)(VP) iff the free variable

Cov is a Cover of the denotation of this plurality, and for all y such that y is a member of the Cov

set, then y is in the extension of the VP. Schwarzschild proposes that this distributive operator is

an optional operator that is inserted to result in a distributed reading of the plural NP. Now the

distributed operator applies to Covers of the denotations of NPs. Note that a Cover can include

singleton subsets, which provides for interpretations for when the VP does apply to each

individual as opposed to sums of people.

Since a Cov variable is optionally inserted in non-reciprocal sentences, Schwarzschild

proposes that sentences that realize each other can also insert the distributive operator that

includes the Cov free variable and thus constrain the reciprocal relation as well between

participants. The reciprocal relation therefore applies to the subpluralities in the Cover of the set

denoted by the subject NP instead of to the set itself.

Returning now to the denotation of the boys in the sentence and context in (113) and

Figure 13, in this case, the Cover of the denotation of the NP the boys has two subsets as

members, each with 3 boys in it, and the SR relation applies to the subpluralities so that the

kicking relations are between the three boys in the first set and then separately among the three

boys in the second set.

I therefore assume that the partitioned reading of the sentence in (113) falls out from the

optional distributive operator that includes the Cov variable that can be inserted into sentences.

Thus, since my aim in this dissertation is to characterize the meaning of reciprocal expressions, and partitioned readings do not come about because of reciprocal expressions, I will not consider partitioned readings any differently than non-partitioned readings. I will not consider such readings again, except in passing to point out that they exist for drug druga and the -sja verbs as well.

I now return to the relation of Inclusive Alternative Ordering (IAO) as proposed by

Dalrymple et al. (1998b). I repeat the sentence they provided as an example of this relation (174):

(123) He and scores of inmates slept on foot-wide wooden planks stacked atop each other – like sardines in a can – in garage-sized holes in the ground.

This sentence is true if the planks are arranged in the following configuration (and if there are scores more of these) and one prisoner is sleeping on top of each of the planks:

1 2

Figure 14: Depiction of the sentence He and scores of inmates slept on foot-wide wooden planks stacked atop each other – like sardines in a can – in garage-sized holes in the ground.

This picture satisfies the truth conditions of IAO, as each entity (i.e. each plank) stands in the R relation to at least one other plank as the first or the second argument. The planks on the top participate only as the first argument, the planks on the bottom participate only as the second argument, and the plank in the middle as both arguments. Note that this relation allows there to be groupings of planks, so that the planks under prisoner 1 are not related via a chain, as is required for RIAR, to the planks under prisoner 2.

Crucially, as Sabato and Winter (2008) point out, if there is just one grouping, then this relation has the same truth conditions as IAR. This observation leads them to propose that IAO in fact is simply the partitioned reading of IAR, and thus that IAO is not an independent reciprocal relation. I follow Sabato and Winter in this, and so I do not analyze IAO as a relation that each other conveys. I hold that it is in fact the partitioned reading of RIAR. I thus remove a relation

from the set of relations argued for by Dalrymple et al. (1998b) and am left with four relations

that each other conveys: Strong Reciprocity, Intermediate Reciprocity, One-Way Weak

Reciprocity and Revised Intermediate Alternative Reciprocity.

2.4 Analysis of each other

Now armed with the formal framework I assume in this dissertation, the assumptions I make about the nature of each other, and the relations I propose that each other can convey, I provide here a derivation of a simple sentence that realizes each other to illustrate how everything comes together. I will derive the meaning of the sentence in (124):

(124) Context: You see your friend John hug his girlfriend Mary, while Mary hugs him back.

John and Mary hugged each other.

The lexical entry of each basic expression is in (125):

(125) Basic expressions:

John: N: ⟦j’⟧M,g

Mary: N: ⟦m’⟧M,g

and: (N\N)/N : ⟦⊕⟧M,g

M,g hug: (N\S)/N: ⟦hug’>>⟧

M,g -ed ‘PAST’: S/S’: ⟦λP [∃e(P(e) ˄τ(e) ≪ now)] ⟧

M,g each other: ((N\S’)/N)\(N\S’): ⟦λP >>λxλev[RDIST(P(x)(x)(ev))] ⟧

The lexical entry of each other requires further comment. RDIST is a constant, and it is mapped

by the interpretation function to a polyadic quantifier. RDIST in fact represents one of the four

different reciprocal relations: RDIST_SR, RDIST_IR, RDIST_OWR, and RDIST_RIAR. In this

sentence, RDIST will be interpreted as contributing a SR relation to the sentence. I propose the following definition of RDIST when it conveys SR:

(126) ⟦RDIST_SR⟧= ⟦[λP>>λxλev[∀y∀z(y ≤ x ˄ z ≤ x ˄ y ≠ z → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ P(y)(z)(e’) ˄ P(z)(y)(e”)))]⟧M,g

RDIST_SR is a function from P to a function from an individual x to a function from an

eventuality to a truth value such that RDIST is true (for a given P, x, and ev) iff for all y and z

such that y and z are distinct parts of x then there exist two distinct events that are part of ev and the y stands in the P relation to z for e’ and z stands in the P relation to y for e”.

As I have shown, each other is also compatible with three other relations, and I define each other for each of these relations here:

(127) Intermediate Reciprocity:

⟦RDIST_IR⟧=⟦λP>>λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → for some sequence z0,

…, zm ≤ x (∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e”˄ (a = z0 ∧ P(z0)(z1)(e’) ∧ … ∧ M,g P(zm-1)(zm)(e”) ∧ zm=b))⟧

RDIST_IR is a function from P to a function from an individual x to a function from an eventuality to a truth value such that RDIST_IR is true (for a given P, x, and ev) iff for all y and z such that y and z are distinct parts of x then for some sequence z0, … zm that are part of x, there exist two distinct subevents, e’ and e” of ev, and each individual stands in the P relation to each other individual directly or indirectly via a chain of P relations.

(128) One-way Weak Reciprocity:

⟦RDIST_OWR⟧= ⟦[λP>>λxλev[∀a∃b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’(e’≤ ev ˄ P(a)(b)(e’))]⟧M,g

RDIST_OWR is a function from P to a function from an individual x to a function from an eventuality to a truth value such that RDIST_OWR is true (for a given P, x, and ev) iff for all y and z such that y and z are distinct parts of x then there exists an event e’ that is part of ev and a stands in the P relation to b for e’.

(129) Revised Intermediate Alternative Reciprocity:

⟦RDIST_RIAR⟧= ⟦λP>>λxλev[∀a∀b (a ≤ x ˄ b ≤ x ˄ a ≤ x ˄ b ≤ x ˄ c ≤ x ˄

a ≠ b → for some sequence w0, . . . wm ≤ x (∃e0, . . . , en ≤ ev (a = z0 ∧ kiss’(w0)(w1)(e0) ∨

kiss’(w1)(w0)(e0) ∧ . . .∧ (kiss’(wm-1)(wm)(en) ∨ kiss’ (wm)(wm-1)(en))∧ wm = b) ∧ ∃c ∃d ∃g (c ≤ x ˄ d ≤ x ˄ g ≤ x ˄ c≠ d ∧ c ≠ g ∧ ∃e’∃e”(e’ ≤ ev ∧ e’ ≤ ev ∧ e’≠ e”˄ kiss’(c)(d)(e’) ∧ kiss’(g)(c)(e”))]⟧M,g

RDIST_RIAR is a function from P to a function from an individual x to a function from an

eventuality to a truth value such that RDIST_RIAR is true (for a given P, x, and ev) iff for all y and z such that y and z are distinct parts of x then for some sequence w0, . . . wm that are part of x there exist events ∃e0, . . . , en that are part of ev and each entity stands in the P relation via a chain of direct or indirect relations as the first or second argument to every other entity and there exist three individuals c, d, g that are distinct parts of x and two events e’ and e” that are distinct parts of ev and c stands in the kiss relation to d for event e’ and g stands in the kiss relation to c for event e”.

Each other can convey all of these relations, and as mentioned above, I assume that which one is conveyed is determined by the Strongest Meaning Hypothesis. The SMH predicts 95

that the interpretation of the sentence in (124) is SR, because it is the logically strongest relation that is compatible with the linguistic and non-linguistic context (i.e. there is nothing that disallows the SR relation).

Therefore, I assume the SR relation for the derivation of the sentence in (124). In Figure

15 below I provide a full derivation for this sentence, by showing how each atomic element in the sentence composes with other elements to result in the interpretable sentence in (124).

hug: (N\S’)/N : ⟦hug´⟧M,g each other: ((N\S’)/N)\(N\S’) : ⟦λPλxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ P(a)(b)(e’) ˄ P(a)(b)(e”))]⟧M,g

hug each other: N\S’: ⟦λPλxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ P(a)(b)(e’) ˄ P(b)(a)(e”))]⟧M,g (⟦hug´⟧M,g ) BA

John and Mary: N : ⟦j⊕m⟧M,g hug each other: N\S’: ⟦λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]⟧M,g BR

John and Mary hug each other: S’: ⟦λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]⟧M,g (⟦j⊕m⟧M,g ) BA

M,g -ed: S/S’: λP ⟦∃e(P(e) ˄ τ(e) ≪ now)⟧ John and Mary hug each other: S’: ⟦λev[∀a∀b(a ≤ j⊕m ˄ b ≤ j⊕m ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]⟧M,g BR

M,g John and Mary hugged each other: S: λP ⟦∃e(P(e) ˄ τ(e) ≪ now)⟧ (⟦λev[∀a∀b(a ≤ j⊕m ˄ b ≤ j⊕m ˄ a ≠ b →

97 ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]⟧M,g) FA

John and Mary hugged each other: S: ⟦∃e(λev[∀a∀b(a ≤ j⊕m ˄ b ≤ j⊕m ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]) (e) ˄ τ(e) ≪ now)⟧ M,g BR

John and Mary hugged each other: S: ⟦∃e[∀a∀b(a ≤ j⊕m ˄ b ≤ j⊕m ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]) ˄ τ(e) ≪ now)⟧ M,g BR

Figure 15: Derivation of the sentence John and Mary hugged each other.

The final interpretation in a model M under the assignment function g of the translation of the sentence in Figure 15 I repeat in (130):

(130) ⟦∃e[∀a∀b(a ≤ j⊕m ˄ b ≤ j⊕m ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ e’≠ e” ˄ hug’(a)(b)(e’) ˄ hug’(b)(a)(e”))]) ˄ τ(e) ≪ now)⟧M,g

This translation says that the sentence in (124) is true iff there is some event e such that for all a that are part of the complex individual of John and Mary and for all b, that are distinct from a that are part of the same complex individual, then there exist subevents e’ and e” of the event e, and e’ is an event of b hugging a, and e” is an event of a hugging b, and these events occurred in the

past. This translation then correctly predicts the meaning of the sentence in (124).

I have thus provided a derivation for a sentence that realizes each other as a base line to compare the two reciprocal expressions under discussion in this dissertation, drug druga and the

-sja verbs. I now provide basic definitions of each of these expressions that will be revised throughout this dissertation.

2.5 Preliminary formal definition of drug druga

As a place to start, I hypothesize that drug druga has the same meaning as each other,

since they are both pronouns and appear to have the same meaning. Thus, I hypothesize the

following lexical entry for drug druga:

M,g (131) drug druga: ((N\S’)/N)\(N\S’): ⟦λP >>λxλev[RDIST_SR(P(x)( x)(ev))]⟧

I have shown in Chapter 1 that drug druga is compatible with SR, and so the constant RDIST can certainly convey a SR relation. I therefore leave this definition as it is for now. I will further discuss in Chapter 5 whether the lexical entry in (131) accurately captures the meaning of drug druga and will also consider whether drug druga can convey the same four relations as each other.

2.6 Preliminary formal definition of the -sja verbs

Up to this point, I have called the second reciprocal expression under discussion in this dissertation the (reciprocal) -sja verbs without much comment. It might at first glance seem appropriate to instead call the affix -sja a reciprocal expression that combines in the syntax with a transitive verb to result in a reciprocal verb. The reciprocal affix -sja after all seems to contribute the same meaning regardless of the verb is it realized on (though see Chapters 3 and 4), and this meaning appears to be the same as drug druga or each other. These facts support an analysis where the reciprocal affix -sja has its own lexical entry and composes in the syntax with a transitive verb to result in a reciprocal verb. I will argue, however, that the -sja verbs are composed in the lexicon, not the syntax.

The lexicon I assume is abstract and does not attempt to represent the mental lexicon. I maintain that the lexicon is parsimonious, in that it holds only those entries that are necessary, based on the unpredictability of their meaning. It holds pieces of words (i.e. morphemes), words, and phrases. A particular form has a lexical entry if it cannot be decomposed into smaller morphemes.

Assuming these properties of the lexicon, we can straightforwardly analyze many of the reciprocal verbs as being generated in the lexicon. This group of reciprocal verbs are those

whose bases do not have a distinct lexical entry (e.g. borot'sja ‘to struggle with’, *borot'), which I

call here deponent verbs after Kemmer (1993), or those whose meaning cannot be calculated

based on its constituent parts, also called semi-deponent (e.g. videt'sja ‘to meet’, videt' ‘to see’).

The proposed lexical entries for these verbs are in (132).

(132) borot'sja ‘to struggle with each other’: N\S: ⟦λxλev[RDIST(struggle.with’(x)(x)(ev))]⟧M,g

videt'sja ‘to meet (each other)’: N\S: ⟦λxλev[RDIST(meet’(x)(x)(ev))]⟧M,g

Note that the RDIST constant makes an appearance again in these entries. I have shown in

Chapter 1 that the -sja verbs are compatible with SR, and so once again, I assume to begin with

that the RDIST can convey the SR relation. I will explore whether it can also convey other

relations in Chapter 4. In the semantics of these lexical entries, RDIST is simply a component of

the meaning of the -sja verb. The -sja verb combines first with an individual, supplied by the

subject of the verb, and then with an event argument, supplied by the tense morpheme.

An analysis of the other -sja verbs, the verbs whose meanings are compositional, poses

more of a problem. They could be analyzed like the deponent and semi-deponent verbs above, in

that their meanings are composed in the lexicon. Alternatively, the affix -sja could have its own

lexical entry such as in (133):

M,g (133) -sja: ((N\S’)/N)\(N\S’): ⟦λP >>λxλev[RDIST(P(x)( x)(ev))]⟧

This lexical entry, except for the phonological component, is exactly the same as for each other and drug druga. Such an analysis would allow a unified analysis of these -sja verbs and drug

100

druga and would explain why they have similar meanings. It would not explain, however, why all of the -sja verbs (deponent, semi-deponent and those transparently derived from a transitive verb) in fact pattern together more than the semantically transparent -sja verbs pattern with drug druga. Throughout this dissertation, I will demonstrate that the -sja verbs pattern together in ways that exclude drug druga, e.g. they can all participate in the discontinuous construction (see

Chapter 4). Therefore, to allow for a unified analysis of the -sja verbs, I assume that they are all generated in the lexicon and come into the syntax with both the type of lexical relation specified

(e.g. a kissing relation, or a meeting relation) as well as the reciprocal relation (e.g. SR).

Assuming all the -sja verbs are generated in the lexicon, a synchronic explanation must be found for why they all have the same functional (as opposed to lexical) meaning. The uniformity in the syntax and semantics is surprising if we posit that they are all independent lexical entries. I propose a lexical redundancy rule that picks out these verbs and generates the functional meaning of RDIST. The output of this rule is an entry like that in (132) for borot'sja

‘to struggle with each other’. The exact nature of the input for this rule is unimportant to this dissertation, as I am concerned with the syntax, semantics and pragmatics of these expressions, none of which would have access to the input of a lexical rule. I therefore do not attempt an answer of this question.

2.7 Summary

To summarize, in this Chapter I have provided an overview of my formal framework, including my assumptions of the semantics and syntax of basic expressions and how such expressions compose to result in meaningful constructions. I have also summarized the two major approaches to reciprocal expressions and provided reasons for why I assume that each other is a polyadic quantifier. From there I presented the six relations I believe each other is 101

compatible with and motivated positing these relations. Finally I provided a semantic derivation for a sentence that realizes a reciprocal expression.

Armed with this information, I now move on to discussing, in Chapters 3-5, the meaning of drug druga and the reciprocal -sja verbs. Note that for the remainder of the dissertation when presenting the semantic interpretation of the translation of an expression, I will not include the meaning brackets (⟦⟧M,g) that indicate that the translation is the semantic interpretation of the expression in a model M under an interpretation function g. I drop these brackets solely for simplicity and readability. Each of my semantic translations therefore is meant to be the interpretation of the given expression in a model under an assignment function.

102

3 Reciprocal expressions and temporality

Events expressed by a sentence can occur sequentially or simultaneously, and a variety of expressions can constrain the temporality of these events. For example, in English, temporal adverbs such as at the same time or one after the other can restrict events to simultaneity or sequentiality, respectively. In this chapter, I explore whether drug druga and the -sja verbs contribute a restriction on the temporal relations of events. Previous studies have suggested that reciprocal expressions can contribute such a restriction. Wierzbicka (2009) and Siloni (2012), for example, have suggested that the -sja verbs restrict events to simultaneity. Consider the following

context, which sets up simultaneous kissing events between the two individuals:

(134) Context: You see your friend Marija kiss Ivan, and you can tell that he is kissing her back. You say:

Marija i Ivan celovali-s'. Marija.NOM and Ivan.NOM kiss.PST.PL-SJA ‘Marija and Ivan kissed each other.’

Since the sentence that contains a -sja verb, celovat'sja ‘to kiss,’ is acceptable, I hypothesize that

sentences that realize this verb can expresses simultaneous events. The same sentence is

unacceptable, however, in the context in (135), in which the events occur sequentially:

103

(135) Context: You see your friend Marija kiss Ivan, but she took him by surprise. Only when she backs away from him does he then lean toward her and kiss her. You say:

#Marija i Ivan celovali-s'. Marija.NOM and Ivan.NOM kiss.PST.PL-SJA (Intended: ‘Marija and Ivan kissed each other.’)

Examples such as these have lead scholars to argue that the -sja verbs restrict the events to simultaneity and that reciprocity and temporality therefore interact.

In this chapter I find that drug druga does not contribute a temporal semantic restriction on events. I will, however, argue that it provides a pragmatic restriction such that the events are implicated to occur simultaneously. On the other hand, I provide data suggesting that only some of the -sja verbs semantically restrict the events expressed by the sentence to simultaneity, while others contribute no such restriction.

3.1 Previous literature regarding the interaction temporality and reciprocity

In Lichtenberk’s (1985) cross-linguistic survey of reciprocal expressions, the first work to

even consider that a connection between temporality and reciprocity is possible, he writes that

“the contrast between sequentiality and simultaneity of the relations in reciprocal situations is of

no consequence to reciprocal constructions” (24).16 Kemmer (1993) argues contra Lichtenberk

that reciprocal expressions can restrict the temporal structure of the events. She provides

evidence from a variety of languages suggesting that some reciprocal expressions implicate that

the events occur simultaneously. For example, in English, she argues that naturally reciprocal

verbs, such as kiss, only express simultaneous events. She provides the minimal pair in (136) to

16 Though Lichtenberk does not mention how many languages he studied, he provides examples from twenty-four languages.

104

demonstrate this. The sentence in (136)a is acceptable when there is no temporal adverb in the sentence. However, when the phrase one after the other is realized as in (136)b, Kemmer judges

the sentence as ungrammatical (as she marks the sentence with a *). She puts this down to the

fact that the reciprocal verb kiss is not compatible with an interpretation in which the events

denoted by it occur sequentially. It must be the case that the denotation of John kissed that of

Mary at the same time as the denotation of Mary kissed that of John.

(136) a. John and Mary kissed. (Kemmer 1993: 113)

b. *John and Mary kissed, one after the other.

Kemmer points out that sentences that realize the reciprocal expression each other with a

transitive verb, on the other hand, can express sequential events.

(137) a. John and Mary kissed each other.

b. John and Mary kissed each other, one after the other. (Kemmer 1993: 113)

Crucially, in these sentences, the verb kiss is not interpreted as a reciprocal verb. Rather the

same form is used as a transitive verb as in (138).

(138) John kissed Mary.

Kemmer analyzes the sentences in (137) as realizing the transitive verb kiss and the reciprocal

expression each other and holds that it is only the reciprocal expression, not the verb, that 105

contributes the reciprocal relation. She argues that sentences containing each other are

compatible with a sequential reading, as evidenced by the acceptability of the sentence including

the phrase one after the other in (137)b.

Based on these data, Kemmer’s interim conclusion is that reciprocal verbs in English

restrict the events to simultaneity while each other does not. She ultimately argues, however, that the temporal relation of events is not at stake here. Rather, she attributes the difference in grammaticality of the sentences in (136)b and (137)b to relative distinguishability of events.

While Kemmer does not define this term, she writes that when the distinguishability of events is lower, “the actions of the two participants are hardly distinguishable as separate actions, and are viewed as a single event” (112). When it is higher, “the actions carried out by the participants are relatively more distinguishable” (112). She provides the following example:

(139) a. John and Mary kissed each other for fifteen minutes. (Kemmer 1993: 113)

b. John and Mary kissed for fifteen minutes.

She writes that “it is not hard, despite the implausible situation portrayed, to get the reading that there was only one, albeit long, kiss involved” in the sentence in (139)b. She claims that the kissing relation in (139)a, however, is far more likely to be interpreted as involving multiple kissing events, presumably based on her own intuitions. Significantly, she is no longer comparing simultaneity or sequentiality but instead is treating multiplicity of events. In both sentences in (139), the kissing events are simultaneous. The difference is that in (139)a each participant probably participates as the agent and the patient in more than one highly

106

distinguishable (simultaneous) kissing event, while in (139)b each participant has the role of both

agent and patient for one event.

As another piece of evidence that relative distinguishability of events and not temporality

is at issue here, she writes that languages commonly convey two similar types of relations with

the same root but with different reciprocal expressions. One expression is used to express

relations for which the events are more distinguishable, and the other expression is used to

express events that are less distinguishable. For example, Norwegian exhibits the verb slå ‘to

hit,’ which can combine with the reciprocal expression hverandre ‘each other’ to describe

reciprocal hitting events. However, to express the relation of fighting, which Kemmer

characterizes as “an undifferentiated activity” (114), Norwegian speakers use the same verb, slå

‘to hit,’ with the addition of a suffixal reciprocal expression, resulting in the verb slå-ss ‘to fight’.

Kemmer argues that “the change of conceptualization involves a necessary shift of meaning to a

more complex, less differentiated activity than the exchange of two single blows” (115), and thus

that the meanings of slå hverandre ‘to hit each other’ and slå-ss ‘to fight’ differ in terms of

distinguishability of events. According to this analysis, what might seem to be a difference in the temporality of events is in fact a consequence of this parameter of distinguishability of events.

Kemmer proposes that whether a reciprocal expression is associated with a low or high relative distinguishability of events is a semantic property. If a reciprocal expression is associated with a low relative distinguishability of events, such as the reciprocal verb kiss, then

the events are required to be non-distinguishable. If on the other hand, a reciprocal expression,

such as each other, has a high relative distinguishability of events, then it is compatible with both

distinguishable and non-distinguishable events, such that the events can occur multiple times and

sequentially. Since these are semantic properties, I assume Kemmer holds that they are entailed.

107

I propose however, that the two properties associated with distinguishability of events, i.e. (i) simultaneity vs. sequentiality of the individual events and (ii) singular events vs. multiple events should not be considered together, as they are different, independent semantic properties.

Contrary to Kemmer’s claims, for example, the reciprocal verb kiss is indeed compatible with

multiple (simultaneous) kissing events. Consider the two sentences in (140), both of which realize

the same adverbial phrase requiring that a simultaneous kissing event occurred multiple times.

The sentence in (140)a realizes each other, and so based on Kemmer’s claim, we expect it to be

acceptable. The second sentence, however, realizes the reciprocal verb kiss, and so its

acceptability is surprising.

(140) a. John and Mary kissed each other every day for five years.

b. John and Mary kissed every day for five years.

Since the sentence in (140)b is acceptable, I propose that reciprocal kiss can after all express

multiple events.

On the other hand, I agree with Kemmer that reciprocal verbs such as kiss in English

must express simultaneous events, based on evidence such as that in (141), repeated from (136).

(141) a. John and Mary kissed. (Kemmer 1993: 113)

b. *John and Mary kissed, one after the other.

108

I therefore assume that the sentence in (140)b necessarily involves multiple simultaneous kisses, whereas that in (140)a allows for multiple, non-simultaneous kisses. Therefore, I suggest that the ability to express multiple events is distinct from the ability to express sequential events and that the two properties are not in fact correlated, as Kemmer proposes. The simultaneity of events is semantically entailed for reciprocal verbs, while the singularity of events isn’t.

On a theoretical level, another problem with Kemmer’s analysis is that she posits that relative distinguishability of events is a scale ranging from less distinguishable to more distinguishable. She does not provide evidence, however, for any intermediate point on the scale; reciprocal expressions are either associated with a high distinguishability of events or low distinguishability of events. The lack of any intermediate step suggests that the parameter in fact is not scalar and that the empirical generalizations should more simply be captured via semantic or pragmatic temporal restrictions. Some expressions introduce temporal restrictions and some do not. For these reasons, I choose to analyze the temporal contributions of reciprocal expressions (if any exist) as semantic and pragmatic restrictions on the temporality of the events expressed by a sentence.

Wierzbicka (2009) also argues that reciprocal expressions can contribute temporal restrictions. Working in the theory of Natural Semantic Metalanguage (NSM), she suggests that each reciprocal expression has a prototypical meaning. The prototypical meaning of each expression is different; thus the prototype of each other differs from that of drug druga, which

differs from that of the -sja verbs, etc. In NSM, the meanings of words are paraphrased using a

set of semantic primes, which are words that cannot be semantically reduced, i.e. are indefinable

in a simpler way. Some of these semantic primes are: good, bad, I, because, like. Relevant for my discussion of reciprocity and temporality, the following are also semantic primes: time, after, the same, this. In analyzing some of the reciprocal expressions, she uses the two complex phrases 109

(and therefore complex concepts): at the same time and after this. For example, part of the prototypical meaning of a sentence that realizes a reciprocal verb in English such as that in (142)

includes the complex phrase at the same time: “one of them does something to the other one’s

body; at the same time, the other one does the same thing to this one’s body” (115).

(142) They kissed.

Wierzbicka therefore argues that the prototype for reciprocal verbs in English includes a

restriction on the temporal relation of events. On the other hand, the prototype of a sentence that

realizes each other such as the sentence in (143) includes the phrase after this: “one of these two

people does something to the other one; after this, this other one does the same thing to this one”

(116).

(143) They kissed each other.

Wierzbicka therefore proposes that reciprocal expressions can include a temporal restriction in

their prototypical meaning, even if such restrictions are not entailed.17

The most recent work to discuss the interaction of temporality and reciprocity cross-

linguistically is Evans et al. (2011a). Each chapter of this joint project treats reciprocal

expressions in a different language, using the same stimuli to elicit sentences from consultants.

One of the five parameters explored in each chapter is simultaneity/sequentiality of the events.

17 Note that Wierzbicka proposes that the prototype of each other restricts events to sequentiality. This proposal is unusual in that I have found no other studies that argue that reciprocal expressions ever restrict events to simultaneity. However, since she provides no evidence for her claim, I will ignore it in this dissertation. 110

First, in the paper on Mundari (Austro-Asiatic, India), Evans and Osada (2011) suggest that the main reciprocal expression contributes a temporal restriction. Use of the Mundari reciprocal expression is very limited; the reciprocal expression is compatible with only a small subset of the relations that each other is. No example sentences are provided, but Evans and

Osada write that the main reciprocal expression is used to express simultaneous and sequential events of strongly reciprocal relations and melee situations, but only simultaneous events of chaining relations. In the introduction of the book, a melee situation is defined as: “A and C could each act on D who alone acts on C, etc.” (Evans et al. 2011:3) and chaining relations as “a linear series of events” (Evans et al. 2011:3). The simultaneous chaining relation in the video stimuli used by the researchers involved three participants: A, B and C. A delouses B, and B delouses C at the same time. Their consultants did not use a sentence with a reciprocal expression to express any of the sequential chaining relations. Their study is therefore suggestive that this reciprocal expression temporally restricts relations between events.

In the same volume, Burenhult (2011) writes that in Jahai (Mon-Khmer, Malaysia and

Thailand) the main reciprocal expression is not preferred to describe a context in which strongly reciprocal relations hold between participants when the subevents are sequential rather than simultaneous. Burenhult describes a video in which there are two people: Person A first follows

B, and then the two participants switch, so that B follows A. Out of four consultants, not one used the reciprocal expression to describe this video. Similarly, only one of her participants used this expression to describe a relation in which there are two participants, and A looks at B then looks away, and then B looks at A and looks away. Both of these videos depict sequential events, and the majority of Burenhult’s consultants did not use the reciprocal expression to describe them. Significantly, her consultants described the same relations with simultaneous events using

111

the reciprocal expression. These results suggest that this expression in Jahai also introduces a restriction of simultaneity on the temporal interpretation of the events conveyed by the verb.

Although the results of these two studies are suggestive, the researchers did not determine whether the particular reciprocal expressions in Mundari and Jahai can ever be used to describe the sequential events, and they did not obtain negative evidence, i.e. what is not possible in the language. Therefore, while their results are suggestive of an interaction, they are not conclusive.

In sum, others have argued that reciprocal expression can contribute temporal restrictions, but the data they provide to argue for such claims is often insufficient to provide the full range of empirical generalizations needed to propose an analysis. In this chapter I explore the distribution of drug druga and the -sja verbs utilizing data obtained through elicitation and a

corpus study to determine the empirical generalizations regarding the interaction of temporality

and reciprocity in Russian. I will first define the terminology I use throughout this chapter in

Section 3.2. The rest of the chapter is then broken up into two sections, one on drug druga in

Section 3.3 and one on -sja in Section 3.4. In Section 3.5 I will provide general conclusions and

implications of the data and analyses of both expressions.

3.2 Definitions of terms

As I pointed out in Chapter 1, the terminology used in studies of reciprocity is rarely

clearly defined. The same holds true for the terminology of temporality in these studies. The

studies cited here all use the terms simultaneity or sequentiality with little or no definition of the

terms. For example, Kemmer (1993) does not explicitly define either of these terms, and Evans

et al. (2011a) and the works within only define simultaneity as when each “pairwise event”

happens at the same time (32), and they do not define sequentiality. Here, I take care in defining

112

my terms. Such precision allows others to better understand my arguments and allows for

comparison with other studies.

In Section 2.1.1 of Chapter 2, I outlined the formal framework I adopt based on Krifka

(1998) that includes reference to events. Here, using this framework, I define terms that are

relevant to a discussion of temporality.

A subevent is an event that is a proper part of another event. For example, eE1 is a subevent of the complex event eE1 ⊕E eE2 because it is a proper part of the complex event, as defined in Chapter 2.

A substate is a state that is in the domain of a proper part relation. For example, eS1 is a

substate of the complex state eS1 ⊕E eS2. I will not discuss states as different from events here,

though such an exploration may be interesting and important; it is simply outside of the scope of

this dissertation.

Eventualities are strictly simultaneous if they map onto the same run time:

(144) ∀e, e’ ∈ E’ [SSIMUL (e, e’) ↔ τ(e) = τ (e’)]

Throughout my study of the interaction of temporality and reciprocity, I have found no sentence

that entails that strict simultaneity holds between events. I have found simultaneity, as defined in

(145), to be entailed, and thus throughout this chapter I will refer to this relation between events instead of strict simultaneity. Whether events are ever required to be strictly simultaneous is unclear at this point in our explorations of reciprocity, except for making clear what simultaneity is not.

113

Eventualities are simultaneous iff they temporally overlap:

(145) ∀e, e’ ∈ E’ [SIMUL (e, e’) ↔ τ (e) ⊗E τ (e’)]

Eventualities are sequential iff they do not temporally overlap (and thus participate in the

temporal precedence relation):

(146) ∀e, e’ ∈ E’[SEQ(e, e’) ↔ ¬ [τ(e) ⊗E τ(e’)]]

Having defined my terms, I now turn to exploring the temporal interpretations of

sentences that realize drug druga and the -sja verbs.

3.3 Temporality and drug druga

In this section I explore the temporality of sentences that realize drug druga in detail. The only work that mentions that drug druga might affect the temporal configuration of event arguments is Wierzbicka (2009). When examining this expression, she notes that “drug druga implies simultaneity” (127), but provides no evidence for this claim.

3.3.1 Temporal entailments and drug druga

Consider the two contexts and sentence in (147), which I presented to my consultants.

Each of my three consultants judged the provided sentence as acceptable in both contexts.

114

(147) Context 1: You are an elementary school teacher and have two troublesome students: Fedja and Griša. During recess Fedja, unprovoked, hits Griša on the shoulder. Griša recoils, then punches him back. You saw everything and you explain to the principal:

Context 2: You are an elementary school teacher and have two troublesome students: Fedja and Griša. During recess Fedja and Griša start yelling at each other and then suddenly attack each other at the same time, going for each other’s shoulders. You saw everything and you explain to the principal:

Fedja i Griša bili drug drugu po plečam. Fedja.NOM and Griša.NOM hit.PST.PL each other.DAT on shoulders.DAT ‘Fedja and Griša hit each other on the shoulders.’

In Context 1 in (147), it is clear that the subevents, the two hitting events, do not temporally

overlap. In Context 2, on the other hand, we understand that the subevents do overlap: the hitting

events occurred at the same time. Since all three of my consultants judged the utterance as

acceptable in both contexts, I hypothesize that the reciprocal expression drug druga is compatible

with contexts in which the subevents expressed by the sentence are simultaneous or sequential.

In short, drug druga does not require that its sentence entail simultaneous or sequential events.

3.3.2 Temporal implicatures and drug druga

Although drug druga does not entail simultaneity or sequentiality, I propose that it

implicates that the events expressed by the sentence that realizes drug druga are simultaneous.

Specifically, I argue that drug druga gives rise to a Generalized Conversational Implicature.

Grice (1989) recognized a difference between Particularized Conversational Implicatures

(PCIs) and Generalized Conversational Implicatures (GCIs). Grice writes that PCIs leave “no

room for the idea that an implicature of this sort is normally carried by saying that p” (Grice

1989: 37). About GCIs on the other hand, he writes that they arise when “the use of a certain

115

form of words in an utterance would normally (in the absence of special circumstances) carry

such and such an implicature” (Grice 1989: 37). Levinson (2000:16) proposes that the difference

between Particularized and Generalized Implicatures is the following:

(148) The distinction between PCIs and GCIs:

An implicature i from utterance U is particularized iff U implicated i only by virtue of specific contextual assumptions that would not invariably or even normally obtain

An implicature i is generalized iff U implicates i unless there are unusual specific contextual assumptions that defeat it

Additionally, he writes (42) that:

GCIs are inferences that appear to go through in the absence of information to the contrary; but additional information to the contrary may be quite sufficient to cause them to evaporate. Thus the mode of inference appears to have two important properties: it is a default mode of reasoning, and it is defeasible.

Levinson does not conclusively define his term ‘default’, but rather directs the reader to

various theories of default logic. Informally, a default meaning is a natural presumption a

recipient will make upon hearing an utterance. The term defeasible, for my purposes, has

the same meaning as cancelable.

To illustrate the differences between the two types of implicatures, consider the

conversation in (149):

116

(149) A: I’m out of petrol.

B: There is a garage round the corner.

Implicature: The garage is open and sells petrol.

The implicature that arises in (149) is a PCI, because B’s utterance gives rise to the implicature only in that particular context of A’s statement. We could also imagine the following conversation, in which B” implicates something different with the same sentence:

(150) A”: Why are there so many mechanics in the restaurant?

B”: There is a garage round the corner.

Implicature: The mechanics work in the garage and since it is close by, they have come to

the restaurant to eat.

Since the same sentence can give rise to different implicatures, these implicatures are categorized as PCIs. On the other hand, consider the sentence and one of its implicatures in (151), and the questions in (152) which could provide a context for the sentence.

(151) Some students at the football game are wearing scarlet and gray.

Implicature: Not all students at the football game are wearing scarlet and gray.

117

(152) a. What were the students at the football game wearing?

b. Why can’t I find a gray t-shirt in the store?

c. What are OSU’s school colors?

The implicature in (151) arises in all three contexts, suggesting it is a GCI and not a PCI. This implicature is also tied to a particular form - in this sentence, the word some introduces the

(scalar) implicature. The sentence also gives rise to PCIs, but the PCIs differ depending on the

context.

To argue that drug druga gives rise to an implicature of simultaneity, I provide several

types of evidence. One of the defining characteristics of GCIs are that they are tied to a particular

piece of form as opposed to PCIs which are non-detachable, i.e. they arise no matter how the

proposition is expressed. Evidence from my corpus study suggests that the implicature of

simultaneity arises from the expression drug druga and not the utterance as a whole or another

expression of the utterance.

For the corpus study, I pulled 500 random sentences containing drug druga from the

Russian National Corpus (RNC) and tagged each sentence for temporality. Out of the 500

sentences, thirty-four sentences express sequential subevents (6.8%) and 279 (55.8%) express

simultaneous events. The remainder of the sentences I was unable to classify as simultaneous or

sequential, as the context given in the corpus was limited. These numbers suggest that although it

is possible to describe sequential eventualities with sentences containing drug druga, it is not

common.

However, the difference in the number of examples may not indicate anything about drug

druga, as I do not know whether sequential and simultaneous events have an equal probability in

118

general of being described in Russian in general. What is more telling, therefore, is that five of the thirty-four sentences expressing sequential events contain either the phrase po očeredi ‘in turn’ or the adverb poperemenno ‘alternately’. For example,

(153) Russkie i nemcy po očeredi vybivali drug druga iz Russians.NOM and Germans.NOM in turn drive.out.PST.PL each other.ACC from

odnogo naselёnnogo punkta. one.GEN occupied.GEN point.GEN ‘The Russians and Germans in turn drove each other out from the same occupied area.’

Crucially, these adverbs constrain the interpretation of the events so that the only interpretation of a sentence that realizes these adverbs is that the events occur sequentially. Although the five sentences containing a temporal adverb that requires sequentiality only make up 14.7% of the thirty-four sentences denoting sequential subevents, compare this percentage to that of sentences denoting simultaneous subevents in which there is a temporal adverb (odnovremenno

‘simultaneously’) forcing simultaneity: 0.36% (one time out of 279).18

Thus, adverbs requiring sequentiality co-occur with drug druga in sentences that entail

sequentiality more often than adverbs requiring simultaneity co-occur with drug druga in

sentences that entail simultaneity. This fact suggests that sequential adverbs are needed more

than simultaneous adverbs to cancel the proposed implicature of simultaneity. It could, however,

be that sequential adverbs are simply used more often in Russian throughout the language than

simultaneous adverbs. Crucially, this turns out not to be the case. In the RNC, po očeredi ‘in

18 Ideally we’d run a chi-square test on these numbers, but the low number of temporal adverbs in my sample causes problems for statistical inference. Though the raw counts are the best we can do, the difference in percentage between the two conditions is suggestive of significance. 119

turn’ occurs 2,837 times, poperemenno ‘alternately’ 937 times, and odnovremenno

‘simultaneously’ occurs 22,481 times. 19

The simultaneous adverb occurs six times more in the corpus than the two sequential

adverbs combined. This distribution is opposite that of utterances containing drug druga,

suggesting that sentences containing drug druga co-occur with sequential adverbs more than

other sentences. I propose that speakers use sequential adverbs because sentences that realize

drug druga give rise to an implicature of simultaneity. They use the adverb to cancel this

implicature.

These data from the corpus study suggest that it is indeed drug druga and not another

element of the sentence that contributes the implicature. Since the distribution of the temporal

adverbs with sentences containing drug druga is different from the distribution of all other

sentences, it suggests that it is the presence of drug druga that licenses the implicature. If it were

the verbs in the sentences, or any other element, that introduced the implicature, we would expect

the distribution of sentences containing drug druga to have the same co-occurrence pattern with

temporal adverbs as sentences without drug druga. Therefore these data suggest both that (i)

sentences realizing drug druga give rise to an implicature of simultaneity and (ii) that it is drug

druga which contributes the implicature.

Another piece of evidence that drug druga gives rise to an implicature of simultaneity is

that the implicature is calculable, i.e. that the person hearing the utterance is able to infer the

implicated meaning. I make use of Levinson’s I-heuristic as described below to calculate the

implicature.

As is well known, Grice claimed that implicatures arise due to the fact that speakers want

to cooperate in communication and in doing so follow or violate four different maxims, which he

19 As of March 27, 2013 120

called the maxims of quantity, quality, manner and relevance. Levinson (2000), instead of proposing conversational maxims like Grice, posits heuristics. He argues that a problem with

communication is that given an utterance, there is an infinite set of premises that might be

relevant for the calculation of an implicature. He suggests that instead of maxims, humans use

heuristics to narrow the possible meanings to the one intended. Although he proposes three

heuristics, I am only interested here in his I(nformativeness)-heuristic:

(154) I-heuristic: (Levinson 2000: 32) What is simply described is stereotypically exemplified.

Example:

The blue pyramid is on the red cube.

GCI: The pyramid is a stereotypical one, on a square, rather than, e.g., a hexagonal base.

He assumes that the NP the blue pyramid is simply described in terms of the shape of the

pyramid, because there are no additional words in the sentence that provide information about the

shape of the pyramid. Therefore, the phrase the blue pyramid is uttered to refer to a pyramid with

the most stereotypical shape, and this phrase is also understood to indicate a pyramid with the

most stereotypical shape. Furthermore, he assumes that the relation between the pyramid and

cube is simply described because it is expressed by the simple preposition on. He contrasts this

preposition with the almost synonymous preposition supported by. Levinson does not consider

supported by a simple preposition, and therefore a relation between two entities expressed by this

preposition would not be simply described. Since the relation on expresses is simply described,

the relation between the pyramid and cube is stereotypically exemplified. The implicatures arise,

for example, that the pyramid is squarely resting on the cube, and there is nothing in between the

pyramid and the cube. 121

Levinson describes the principle behind this heuristic in more detail (114-115):

(155) I principle:

Speaker’s (S) maxim: the maxim of Minimization. “Say as little as necessary”; that is, produce the minimal linguistic information sufficient to achieve your communicational ends (bearing Q-principle in mind).

Recipient’s (R) corollary: the Enrichment Rule. Amplify the informational content of the speaker’s utterance, by finding the most specific interpretation, up to what you judge to be the speaker’s m-intended point, unless the speaker has broken the maxim of Minimization by using a marked or prolix expression.

Specifically:

(a) Assume the richest temporal, causal and referential connections between described situations or events, consistent with what is taken for granted.

(b) Assume that stereotypical relations obtain between referents or events, unless this is inconsistent with (a)

(c) Avoid interpretations that multiply entities referred to (assume referential parsimony); specifically, prefer coreferential readings of reduced NPs (pronouns or zeros).

(d) Assume the existence of actuality of what a sentence is about if that is consistent with what is taken for granted.

In order to argue that it is this heuristic which licenses the implicature, I make the assumption that

an utterance such as that in (156) is “simply described,” or in other words, “minimally specified”

(Levinson 2000: 37) in terms of temporality.

(156) Marija i Ivan videli drug druga. Marija.NOM and Ivan.NOM see.PST.PL each other.ACC ‘Marija and Ivan saw each other.’

122

The sentence is minimally specified, at least in regard to temporality, because it does not realize an expression that explicitly contributes information as to how the temporality of the event arguments of the sentence should be interpreted. The sentence contrasts with that in (157) in which the adverbs odnovremenno ‘simultaneously’ or po očeredi ‘in turn’ are realized.

(157) Marija i Ivan videli drug druga odnovremenno/po očeredi. Marija.NOM and Ivan.NOM see.PST.PL each other.ACC simultaneously/in turn ‘Marija and Ivan saw each other at the same time/in turn.’

This sentence is not minimally specified regarding temporality because it does realize an

expression that explicitly constrains the temporality of the seeing events.

To calculate the implicature that the events occur simultaneously, it also must be

determined what is the most stereotypical interpretation of the temporal configuration of events.

Levinson defines stereotypes as “connotations associated with meanings, but not part of them,

which nevertheless play a role in interpretation” (115). While this definition tells us what sort of

thing a stereotype is, it does not tell us how to determine the character of a stereotype. He does,

when explaining this heuristic, describe the meaning a piece of form contributes as stereotypical

if it is what we assume, or expect, it to be. For example, he says that we might expect a spoon in

a cup to describe a metal spoon in a ceramic cup, since we expect a spoon to be metal and a cup

to be ceramic. As for the preposition on, he suggests that the most stereotypical meaning is that it

introduces a relation between two entities such that one is on top of the other, is supported by the

other and there is nothing intervening between the two entities. Crucially, the stereotypical

meaning is tied to the word (in a culture) itself and not to the language. There is nothing more

stereotypical in English about metal, for example, than plastic, and yet the word spoon still

implies that the object is made out of metal. 123

Given this information about what a stereotypical interpretation is, I hold that simultaneous events are more stereotypical for sentences that realize drug druga than sequential

events. Recall that in the corpus study I performed describe above, out of the 500 sentences,

thirty-four sentences express sequential subevents (6.8%) and 279 (55.8%) express simultaneous

events. Because sentences containing drug druga express simultaneous events more often than

sequential events, I propose that simultaneity is more stereotypical than sequentiality for

sentences that contain drug druga. Note that it does not matter whether simultaneous events are

more stereotypical than sequential events in general throughout Russian. Even though I have no

idea if sentences in Russian in general are more likely to express simultaneous events, the notion

of stereotypical meaning is tied to a form, such as spoon, or on, as discussed above. I therefore

assume as long as sentences containing drug druga are more likely to express simultaneity, then

simultaneity is in fact more stereotypical than sequentiality for drug druga. Based on this

finding, I propose that the most stereotypical interpretation of a sentence that realizes drug druga

is that the events occur simultaneously.

With these two assumptions, I can now calculate the implicature, providing evidence that

it indeed arises from a sentence that realizes drug druga. In my set of video stimuli, there is a video in which two people hit each other once at the same time. Although my participants used different verbs for hitting (bit', udarit', xlopnut'), they all agreed that they could use the structure

bit'/udarit'/xlopnut' drug druga ‘hit each other’ to describe the events in this video. I propose that given an utterance such as that in (158), an implicature arises that the hitting events occurred simultaneously. Support for this claim would be finding that this implicature is calculable.

124

(158) Oni bili drug druga. they.NOM hit.PST.PL each other.ACC ‘They hit each other.’

The calculation of this sentence is as follows:

(159) The speaker says: Oni bili drug druga. ‘They hit each other.’

The recipient assumes that the speaker is obeying the heuristic of “say as little as possible.”

The speaker doesn’t explicitly say whether the hitting events occurred at the same time or not.

The recipient assumes the most enriched temporal connections between described events, consistent with what is taken for granted.

The most enriched temporal connections between events would be that the events occurred either simultaneously or sequentially.

The recipient assumes the most stereotypical relations obtain between events, unless this is inconsistent with the most enriched interpretation. The most stereotypical relation between events is that the events occur simultaneously.

The only stereotypical interpretation that is consistent with the most enriched interpretation is that the events are simultaneous.

Thus, the recipient infers that the hitting events happened simultaneously.

The calculation in (159) shows that it is possible to follow the recipient’s line of reasoning to

infer that the events occurred simultaneously. It suggests that utterances which realize drug

druga conversationally implicate simultaneity unless an adverb or some other expression cancels

the implicature, or something in the context makes the implicature not arise in the first place.

In sum, I have argued that utterances containing drug druga do not semantically restrict

the temporality of the event structure but that they do give rise to a Generalized Conversational 125

Implicature that the events are simultaneous. Furthermore, I proposed that this implicature is

licensed by Levinson’s I-heuristic and have outlined how such an implicature is implicated and

inferred.

3.3.3 Formal analysis of drug druga

In Chapter 2, I outlined the formal framework I adopt and developed a fragment of Russian

in this framework in order to translate sentences containing drug druga. I proposed the following meaning for drug druga when it contributes a Strongly Reciprocal (SR) relation to a sentence:

(160) ⟦RDIST_SR⟧= λP>>λxλev[∀y∀z(y ≤ x ˄ z ≤ x ˄ y ≠ z → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ P(y)(z)(e’) ˄ P(z)(y)(e”)))]

RDIST_SR is a function from P to a function from an individual x to a function from an eventuality to a truth value such that RDIST is true (for a given P, x, and ev) iff for all y and z such that y and z are distinct parts of x then there exist two distinct events that are part of ev and the y stands in the P relation to z for e’ and z stands in the P relation to y for e”. This being the

case, I translate the sentence in (161) as in (162).

(161) Ivan i Marija videli drug druga. Ivan.NOM and Marija.NOM see.PST.PL each other.ACC ‘Ivan and Marija saw each other.’

(162) ∃e[∀y∀z(y ≤ i⊕m ˄ z ≤ i⊕m ˄ y ≠ z → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ e’≠ e” ˄ see’(y)(z)(e’) ˄ see’(z)(y)(e”))) ˄ τ(e) ≪ now]

126

This translation says that the sentence in (161) is true iff there is some event e such that for all y that are part of the complex individual of Ivan and Marija and for all z, that are distinct from y, that are part of the same complex individual, then there exist subevents e’ and e” of the event e,

and e’ is an event of y hugging z, and e” is an event of z hugging y, and these events occurred in

the past.

Note that the RDIST_SR relation generates subevents and that it does not semantically

restrict the temporality of these subevents in any way. I propose that the sentence does give rise

to the implicature in (163):

(163) Implicature: Ivan and Marija saw each other simultaneously.

Having provided an analysis of the temporal contribution of drug druga, I now turn my attention

to the -sja verbs.

3.4 Temporality and the -sja verbs

In the introduction to this chapter, I stated that Siloni (2012) and Wierzbicka (2009) both

claim that the -sja verbs restrict events to simultaneity.20

As I will show below, it is not the case that all reciprocal verbs in Russian require that the

subevents happen simultaneously. Both Wierzbicka (2009) and Siloni (2012) assume this

because they both consider only the verb celovat'sja ‘to kiss’ but intend their analyses to extend

to all reciprocal -sja verbs. As I will show, this verb does indeed entail that the subevents happen

simultaneously, but the majority of the forty or so reciprocal verbs in Russian do not contribute

20 Siloni in fact argues that these verbs express one symmetric event, but under my analysis this is equivalent to two simultaneous events. I discuss the differences between our analyses in Section 3.4.2. 127

such a temporal restriction on the events expressed by their sentence. After presenting my data, I will present the analysis in Siloni (2012), which builds on Carlson (1998), in detail and will discuss why I choose not to accept their single-event analysis for the -sja verbs.

3.4.1 Temporal entailments and the -sja verbs

As already introduced in Chapter 1, there are about forty reciprocal -sja verbs in Russian.

In this section, I ask whether the events introduced by these verbs are temporally restricted to simultaneity. In order to determine this, I first limit myself to verbs whose lexical meaning does not entail simultaneity. I therefore put aside the verbs in (164) whose lexical meaning does entail simultaneity.

(164) videt'sja, ‘to meet’ vstrečat'sja, ‘to meet’ otličat'sja,‘to differ from’ sudit'sja, ‘to be in litigation with’ rezat'sja, ‘to play cards with’ soveščat'sja, ‘to confer with’ ob''jasnjat'sja, ‘to discuss’ smešivat'sja, ‘to mix’ (‘to be mixed’) obvenčat'sja, ‘to get married’ ženit'sja, ‘to get married’ mirit'sja, ‘to make up with’ družit'sja, ‘to befriend each other’ pixat'sja, ‘to have sex’

Based on the lexical meaning of these verbs, it is impossible for the subevents expressed by sentences containing the verbs in (164) to happen sequentially. For example, if John and Mary are playing cards with each other (rezat'sja), then it is impossible that John is playing cards with

128

Mary at a time when Mary is not playing cards with him, and vice versa.21 These verbs differ

from the others in that it is impossible for these verbs to be judged acceptable in a context that

requires a sequential interpretation. Take, for example, the verb smešivat'sja, ‘to mix’ (‘to be

mixed’) and the grammatical sentence in (165):

(165) Zapaxi smešivajut-sja na kuxne. smells.NOM mix.NPST.3PL-SJA in kitchen.LOC ‘The smells are mixing in the kitchen.’

My consultants judged this sentence as acceptable given the context in (166):

(166) Context: Your sister is baking bread and borscht in the kitchen right now. The smell of bread is mixing with that of borscht, and the smell of borscht is mixing with that of the bread so that you can smell both.

21 In English, verbs of this type have an additional property. When a sentence containing a reciprocal verb, as in (i)a, is compared to a sentence containing the same verb but used transitively as in (i)b, the two sentences have the same truth conditions.

i. a. John and Mary met. b. John met Mary.

In both sentences in (i), it must be the case that John met Mary and Mary met John. Compare this pair, however, with that in (ii): ii. a. John and Mary kissed. b. John kissed Mary.

Here, the two sentences do not have the same truth-conditions. The utterance in ii is true if and only if John kissed Mary and Mary kissed John. The sentence in ii only entails that John kissed Mary. It could be the case that Mary kissed John back, but this is not necessarily the case. Thus, in English, we have a test for determining when a reciprocal verb would fall into this category. However, the picture is not so neat in Russian, as the meaning of the reciprocal verbs often cannot be calculated based on its root verb. For example, the reciprocal verb videt'sja ‘to meet’ is derived from the base verb videt' ‘to see’. The semantic connection between the two verbs is clear, but the reciprocal verb does not simply express the meaning of “reciprocal” seeing.

129

Creating a context in which the smell of bread mixed with that of borscht but that of borscht does not mix with that of the bread is not possible based on what we know about the world. I therefore hold that these verbs lexically restrict the events to simultaneity. I will discuss how they fit into my analysis below.

This leaves about thirty verbs which can, a priori, be interpreted as expressing

simultaneous or sequential relations:

(167) celovat'sja ‘to kiss each other’ obnimat'sja ‘to hug each other’ perepisyvat'sja,’ to correspond’ tolkat'sja, ‘to bump into each other’ borot'sja, ‘to fight, struggle with’ rugat'sja, ‘to fight’ torgovat'sja, ‘to haggle’ znakomit'sja, ‘to get to know, to meet’ zdorovat'sja, ‘to greet’ drat'sja, ‘to fight’ sražat'sja, ‘to overwhelm each other’ bit'sja, ‘to fight’ rubit'sja, ‘to cut each other’ streljat'sja, ‘to shoot each other’ dogovaryvat'sja, ‘to agree with’ soveščat'sja, ‘to confer with’ uslovit'sja, ‘to agree’ prepirat'sja, ‘to bicker’ prerekat'sja, ‘to bicker’ smešyvat'sja, ‘to mix’ proščat'sja, ‘to bid farewell’ sorevnovat'sja, ‘to compete with’ družit'sja, ‘to befriend’ bodat'sja, ‘to butt each other’ klevat'sja, ‘to peck each other’ kusat'sja, ‘to bite each other’ ljagat'sja, ‘to kick each other’ pinat'sja, ‘to kick each other’ oblivat'sja, ‘to splash each other’ carapat'sja, ‘to scratch each other’ branit'sja, ‘to scold each other’ teret'sja, ‘to rub up against each other’

130

Although Wierzbicka (2009) and Siloni (2012) claim that the reciprocal -sja verbs restrict

relations to simultaneity, I argue that only two of these verbs, namely celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’, introduce such a temporal restriction. Consider the context and unacceptable sentence in (168):

(168) Context: You see your friends Ivan and Marija in the park. Suddenly Ivan kisses Marija on the cheek. She is surprised, but when he’s done, she kisses him on the cheek in return. You say:

#Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed each other.’

In this example, the context establishes a relation of Strong Reciprocity between the denotations of Marija and Ivan. The sentence is not acceptable, I suggest, because the -sja verb requires the

events expressed by the verb to occur simultaneously. In fact, in order to convey the meaning

that Ivan and Marija kissed each other sequentially, one must use drug druga:

(169) Ivan i Marija pocelovali drug druga po očeredi. Ivan.NOM and Marija.NOM kiss.PST.PL each other.ACC in turn ‘Ivan and Marija kissed each other in turn.’

The sentence in (169) is acceptable in the context in (168). Thus, the evidence suggests that

celovat'sja ‘to kiss’ contributes a restriction of simultaneity to an utterance. Another piece of

evidence for this claim is that celovat'sja ‘to kiss’ can co-occur with the sequential adverb

po očeredi ‘in turn’, but significantly the only available interpretation for such a sentence is that pairs of people kissed simultaneously in turn: 131

(170) Context: There are several couples sitting in the park. Each couple kisses, simultaneously, in a row, i.e. person A and person B kiss, then once they have stopped, person C and person D kiss, etc.

Oni celovali-s' po očeredi. they.NOM kiss.PST.PL-SJA in turn ‘They kissed each other in turn.’

This sentence is not acceptable in a context in which two people kiss each other sequentially:

(171) Context: Marija and Ivan are kissing in the park. Marija first kisses Ivan, without reciprocation, and then once she’s stopped, Ivan kisses Marija.

#Oni celovali-s' po očeredi. they.NOM kiss.PST.PL-SJA in turn ‘They kissed each other in turn.’

Furthermore, my hypothesis that celovat'sja only expresses simultaneous subevents

predicts that a kissing event in which three participants (or any odd number of participants) are

involved will be unacceptable, or at least very marked, since three simultaneous kissing events

between three different participants is unusual based on what we know about the world. Indeed,

this prediction is born out, as exemplified by (172):

(172) Context: You see your three friends, Ivan, Marija and Larisa, kiss each other as they said good-bye. You say:

#Ivan, Marija i Larisa pocelovali-s'. Ivan.NOM Marija.NOM and Larisa.NOM kiss.PST.PL-SJA ‘Ivan, Marija and Larisa kissed.’

132

My consultants judged this sentence as unacceptable. I hypothesize, since the only difference between this sentence and the sentence in (168) is that there are three NPs, that this verb is not acceptable when three entities are in the extension of the predicate because a simultaneous kiss between three individuals is strange. This piece of evidence further supports my hypothesis that celovat'sja is not compatible with an interpretation in which the events are sequential. These

data, and others like it, suggest that celovat'sja ‘to kiss’ cannot be used in a context in which the

events between participants occur sequentially.

The same holds true for the verb obnimat'sja ‘to hug’, as demonstrated by the fact that my consultants judged the sentence in (173)a as unacceptable in a context in which two hugging events occur sequentially:

(173) Context: You see your friend Ivan hug Marija, while she does not hug him back. He then stops hugging her, and then she hugs him, while he does not reciprocate. You say:

a. #Ivan i Marija obnjali-s'. Ivan.NOM and Marija.NOM hug.PST.PL-SJA ‘Ivan and Marija hugged.’

b. Ivan i Marija obnjali drug druga po očeredi. Ivan.NOM and Marija.NOM hug.PST.PL each other.ACC in turn ‘Ivan and Marija hugged each other in turn.’

Again, one has to use a transitive verb obnjat' ‘to hug’ with drug druga to express the sequential

hugging events. Like with celovat'sja, ‘to kiss’, if obnjat'sja ‘to hug each other’ is realized with a sequential adverb, then there must be multiple pairs of simultaneous hugging events:

133

(174) Context: There are several couples sitting in the park. Each couple hugs, simultaneously, in a row, i.e. person A and person B hug, then once they have stopped, person C and person D hug, etc.

Oni obnimali-s' po očeredi. they.NOM hug.PST.PL-SJA in turn ‘They hugged each other in turn.’

This sentence is not acceptable in a context in which there are two sequential hugging events:

(175) Context: Marija and Ivan are hugging in the park. Marija first hugs Ivan, without reciprocation, and then once she has stopped, Ivan hugs Marija.

#Oni obnimali-s' po očeredi. they.NOM kiss.PST.PL-SJA in turn ‘They hugged each other in turn.’

Above I demonstrated that a sentence that realizes celovat'sja ‘to kiss each other’ is not

acceptable when the subject denotes three individuals, because based on what we know about the

world, three simultaneous kissing events between three individuals is (almost?) impossible. It is,

however, possible for three people to hug each other at the same time. We would therefore

expect a sentence that realizes obnjat'sja ‘to hug each other’ to be acceptable with a subject that denotes three people. Indeed, my consultants judged the following sentence in such a context as acceptable:

134

(176) Context: You see your three friends, Ivan, Marija and Larisa, hug each other in a group hug at the same time. You say:

Ivan, Marija i Larisa obnjali-s'. Ivan.NOM Marija.NOM and Larisa.NOM hug.PST.PL-SJA ‘Ivan, Marija and Larisa hugged.’

As I illustrated in examples (169), (170) and (172), the verb obnimat'sja ‘to hug’ shows similar

patterns as celovat'sja. Based on these data, I hypothesize that the verbs celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’ restrict the events expressed by the sentence to simultaneity.

Other -sja verbs are also compatible with a simultaneous interpretation of the events, as suggested by the context and acceptable utterance in (177).

(177) Context: When you walk into the room, you see and hear Ivan and Marija both yelling at each other at the same time. You later describe this by saying:

Kogda ja vošla v komnatu, Ivan i Marija rugali-s'. when I.NOM enter.PST.SG in room.ACC Ivan.NOM and Marija.NOM yell.at.PST.PL-SJA ‘When I entered the room, Ivan and Marija were yelling at each other.’

They can also, however, express sequential eventualities. For example, consider the context and

utterance in (178):

(178) Context: Marija and Ivan have been in a long distance relationship and recently decided to break up. Marjia writes Ivan a good-bye note. Later, Ivan writes such a note back to Marija.

Marija i Ivan poproščali-s'. Marija.NOM and Ivan.NOM say.goodbye.PST.PL-SJA ‘Marija and Ivan said good-bye to each other.’

135

Since the utterance in (178) was judged to be acceptable in this context, I hypothesize that this verb, poproščat'sja ‘to say good-bye’ can be used to express sequential subevents. The next sentence is also compatible with a sequential interpretation because it is acceptable in the provided context, but also because it realizes a temporal adverb:

(179) Context: You see a field with two cows in it. You watch as one of them butts the other one in the side. The victim frees itself from the first cow and butts it back. You tell a friend:

Korovy bodajut-sja po očeredi. cows.NOM butt.NPST.PL-SJA in turn ‘The cows are butting each other in turn.’

We know from the context that the cows are not butting each other at the same time - in fact it is

impossible for this to be the case since they are butting each other in the sides. Furthermore, the

adverb restricts the events to sequentiality. Therefore, I hypothesize that the verb bodat'sja ‘to

butt’ is compatible with an interpretation in which the subevents are sequential. For another

example, the following sentence was found on the Internet:22

(180) Pushkin i D’anthes streljali-s' s dvatcati šagov . . . Pushkin.NOM and D’anthes.NOM shot.PST.PL-SJA from twenty.GEN steps.GEN ‘Pushkin and D’anthes shot each other from a distance of 20 steps . . .’

(http://db.chgk.info/question/pentag03/15)

22 Accessed January 9, 2012. 136

It is well-known that Pushkin died in a duel with D’anthes, so we can assume that this sentence is describing their duel. In Russian culture, opponents duel with pistols. Each participant takes ten steps away from his opponent and turns around. The person accused shoots first. Only if the

accuser survives does he have a chance to shoot back. So, based on our knowledge about duels in

Russia in general, and on what we know about the life and death of Pushkin in particular, the author of this sentence must be describing sequential shooting events (provided of course s/he is familiar with the details). Since the author used the -sja verb streljat'sja ‘to shoot each other’ to describe these sequential events, I hypothesize that this verb too is compatible with an interpretation with sequential events.

The acceptability of the sentences in (178), (179), and (180) provides evidence that the

-sja verbs do not always contribute a temporal restriction of simultaneity to a sentence. Although

I do not provide the data here, I was able to establish a sequential interpretation for all these verbs. I argue that of the verbs in (167) above, none but the first two, celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’, restrict the events to simultaneity.

These data suggest that -sja verbs in Russian, in terms of temporality, are not a homogeneous category. One subset of these verbs, including those in (164) (the verbs that are lexically restricted to simultaneity) as well as celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’, restrict the events to simultaneity; otherwise the -sja verbs contribute no such restriction.

137

3.4.2 Previous analyses of reciprocal verbs and events

Until this point, I have been treating reciprocal utterances as expressing two or more

subevents. Thus, for the sentence in (181)a, I analyze the semantics as in (181)b, so that there are

two kissing events, one of John kissing Mary, and one of Mary kissing John. Both subevents are

bolded in the translation.

(181) a. John and Mary kissed.

b. ∃e(∀y∀z(y ≤ j⊕ m ˄ z ≤ j⊕ m ˄ y ≠ z → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄e’≠ e” ˄ kiss’ (y)(z)(e’) ˄ kiss’ (z)(y)(e”)) ˄ τ (e) ≪ now)

The verb kiss I analyze as a naturally reciprocal verb, and so it comes into the syntax from the lexicon packaged with both the type of relation specified (e.g. a kissing relation in (181)) as well as the restrictions on the configuration of the relation between participants (i.e. whether the relation is Strongly Reciprocal or Weakly Reciprocal, etc.) I maintain that a tense morpheme introduces an existential quantifier over the event argument. For example, the past tense morpheme has the meaning as in (182):

(182) -l PAST: S’\S: λP [∃e(P(e) ˄ τ(e) ≪ now]

The tense morpheme existentially binds only one event argument that is introduced by the

predicate P. Under my analysis, this event argument denotes an element in a semi-join lattice.

The events are organized as a semi-join lattice as described in Chapter 2, and so it consists of

parts, which I call subevents. These subevents are introduced by either the RDIST relation that is

138

contributed by a reciprocal expression such as each other or drug druga, or by the semantics of a reciprocal verb such as kiss or Russian celovat'sja ‘to kiss’. This type of analysis differs in some respects from other analyses of the nature of events. To illustrate the differences, I first provide some background on previous accounts.

The literature generally assumes that each verb or tense morpheme introduces one event argument that is existentially quantified over (Carlson 1998, Siloni 2012), the same as I do. This being the case, linguists have had to determine how a sentence such as that in (183) can have the two readings in (184)a and (184)b:23

(183) John and Mary lifted a piano.

(184) a. John lifted a piano, and Mary lifted a piano.

b. John and Mary lifted a piano together.

The interpretation in (184)a is the distributed reading where each atomic individual denoted by the subject and object NPs is in the extension of the predicate, leading to multiple events, in this example, of lifting. The interpretation in (184)b, on the other hand, has a collective reading when the denotation of the coordinated NP is taken to act as a group, leading to only one event of lifting.

It has been argued (e.g. Roberts 1987, Landman 1989, Schwarzschild 1996) that the collective reading is more basic in that it introduces no covert operators. The conjunct and introduces the sum operator, familiar from Krifka (1998) (see Chapter 2), and so the coordinated

23 Some speakers find only the collective reading acceptable, but this observation is beside the point here. 139

NP John and Mary has the meaning of the complex group j⊕m. For example, the semantic representation of the sentence in (183) under a collective reading is analyzed as in (185):

(185) ∃e(lift’(piano’)(j⊕m)(e) ˄ τ(e) ≪ now)

This translation says that there is some event such that it is an event of the complex individual made up of the denotations of John and Mary lifting a piano in the past. This translation ensures that it is the group made up of the individuals denoted by John and Mary that participate together in one event of piano-lifting.

The distributed reading occurs when there is a covert distributive operator inserted into the sentence that distributes both the denotation of the NP and that of the predicate, as introduced in Chapter 2. The distributed operator is defined in (186) following analyses in Link (1983),

Roberts (1987), Landman (1989) and Schwarzschild (1996):

(186) ⟦DIST⟧= λP> λxλev[∀a(a ≤ x ˄ ATOM(a) → ∃e’(e’≤ ev ˄ P(a)(e’)))]

This DIST operator ensures that for each entity a such that a is part of x and is atomic, there exists a subevent e’ of the event ev and e’ is an event of a having the property P. This operator optionally applies to the predicate, as in (187)a, which is beta reduced to the representation in

(187)b:

140

(187) a. lift a piano ⇒ λPλxλev[∀a(a ≤ x ˄ ATOM(a) → ∃e’(e’≤ ev ˄ P(a)(e’)))] (lift’(piano))

b. lift a piano ⇒ λxλev[∀a(a ≤ x ˄ ATOM(a) → ∃e’(e’≤ ev ˄ lift’(piano) (a)(e’)))]

This predicate then composes with the subject NP as in (188)a, and the beta reduced representation is in (188)b.

(188) a. John and Mary lift a piano ⇒ λxλev[∀a(a ≤ x ˄ ATOM(a) → ∃e’(e’≤ ev ˄ lift’(piano)(a)(e’)))](j⊕ m)

b. John and Mary lift a piano ⇒ λev[∀a(a ≤ (j⊕ m) ˄ ATOM(a) → ∃e’(e’≤ ev ˄ lift’(piano)(a)(e’)))]

Finally, the tense is added in (189)a, and beta reduced twice to result in the translation of the sentence in (189)c with a distributed interpretation.

(189) a. John and Mary lifted a piano ⇒ λP [∃e(P(e) ˄τ(e) ≪ now)](λev[∀a(a ≤ (j⊕ m) ˄ ATOM(a) → ∃e’(e’≤ ev ˄ lift’(piano)(a)(e’)))])

b. John and Mary lifted a piano ⇒ ∃e(λev[∀a(a ≤ (j⊕ m) ˄ ATOM(a) → ∃e’(e’≤ ev ˄ lift’(piano)(a)(e’)))](e) ˄τ(e) ≪ now)

c. John and Mary lifted a piano ⇒ ∃e(∀a(a ≤ (j⊕ m) ˄ ATOM(a) → ∃e’(e’≤ e ˄ lift’(piano) (a)(e’))) ˄ τ(e) ≪ now)

(189)c is true iff there exists some event e, such that for all a, if a is a part of the complex individual made up of the denotations of John and Mary and a is atomic, then is a subevent e’ that

141

is part of e and e is an event of a lifting a piano in the past. Thus, two separate subevents are generated; one of John lifting a piano and one of Mary lifting a piano. In sum, some sentences are assumed to have at least two interpretations, one which introduces the DIST operator, and one which does not (resulting in a collective reading).

Returning to reciprocity, Carlson (1998) argues that some predicates such as meet and kiss that realize a reciprocal verb, when used intransitively (i.e. reciprocally), are group-level predicates. By this he means that these verbs are only compatible with the collective interpretation and assign the thematic role of AGENT to their subject argument. For a sentence like that in (190), Carlson argues that the only reading available is the collective reading; the

DIST operator cannot be applied to the verb.

(190) John and Mary kissed.

One consequence of this analysis is that, since verbs express only one event and there is no operator in the sentence to distribute this event into subevents, this sentence conveys only one collective kissing event. Therefore, according to Carlson, reciprocal verbs in English express only one event and the symmetric meaning comes about because of the collective interpretation of the denotation of the subject and predicate. He argues that his approach accounts for the fact that reciprocal verbs in English only express seemingly simultaneous events as evidenced by the sentences in (191):

142

(191) a. #John and Mary kissed sequentially.

b. #John and Mary kissed one after the other.

Carlson argues that neither sentence is acceptable because the verb in each sentence expresses

only one event of mutual kissing. The phrases sequentially and one after the other are only

acceptable when they can scope over multiple events. Therefore, instead of analyzing this

phenomenon in terms of simultaneity and sequentiality, he argues that in fact these sentences

introduce only one event of meeting that is not distributed down to subevents.

Carlson contrasts the sentence in (190) with that in (192).

(192) John and Mary kissed each other.

This second sentence, he argues, expresses (at least) two events, one of the individual denoted by

John kissing that of Mary and one of the denotation of Mary kissing that of John. He provides

evidence like that in (193), which mirrors the sentences in (191) above:

(193) a. John and Mary kissed each other sequentially.

b. John and Mary kissed each other one after the other.

Since these sentences are acceptable, unlike those in (191), Carlson proposes that predicates

containing each other are not group-level predicates and thus can give rise to multiple events. He

143

follows Heim et al. (1991) in assuming that the reciprocal expression each other contributes a

distributive operator to the sentence.

My account differs from his, as I always analyze predicates such as kiss as introducing

two or more subevents. I account for the data in (191) by assuming that these verbs contribute

the RDIST relation to the sentence. One reason I choose a different analysis is that Carlson’s

analysis does not account for the fact that predicates like kiss are semantically transitive. He

argues that a group-level predicate “assigns a single thematic role to the group denoted by the

subject NP” (13) and that a single event “has at most one entity playing a given thematic role”

(40), which makes assigning both an AGENT and a PATIENT role to the group denoted by the subject NP impossible. This is problematic as kiss is semantically a transitive verb and requires both an agent and a patient in the semantics. One solution to this problem was proposed by Siloni

(2012), who argued for a “bundling” operation that bundles two thematic roles into one complex thematic role which is then assigned to the group.

Dimitriadis (2008) and Siloni (2012) also propose that predicates with reciprocal verbs express one event, while predicates containing each other express two or more events. They go

further, however, to argue that although reciprocal verbs express only one event, in order for the

thematic role rules to hold, these predicates also express underlying events that are not visible to

other semantic elements in the utterance. In Siloni’s (2012) analysis, a transitive verb undergoes

“reciprocalization” in the lexicon, resulting in a meaning which she symbolizes as VSYM. For example, in the English lexicon, there is a lexical entry KISS and one that is KISSSYM. She posits a meaning postulate which constrains VSYM by the verb’s non-symmetric (and transitive)

counterpart. The postulate states that VSYM entails two underlying events of the same type as the

V. For example, KISSSYM entails two underlying events of the same type as expressed by KISS.

The noun phrase then can be interpreted distributively, as are the events. The point here is that 144

although Dimitriadis and Siloni argue that predicates such as kiss introduce only one event, they

make use of underlying events to solve problems regarding thematic-role assignment and

symmetry.

As I have shown in Chapter 2 and in Section 3.3 above, I treat reciprocal expressions

such as each other or drug druga as introducing the RDIST relation, which takes a transitive verb and distributes over it to generate subevents. Naturally reciprocal verbs come packaged into the syntax/semantics with the distribution built into the lexical entry. I propose, contra Carlson

(1998), Dimitriadis (2008) and Siloni (2012), that naturally reciprocal verbs instead come packaged with the same RDIST meaning as each other or drug druga, except that the events are

restricted to simultaneity. I will also apply this analysis to the two verbs celovat'sja ‘to kiss each

other’ and obnimat'sja ‘to hug each other’.

My analysis can account for the same data presented above as the analysis in which only

one event is expressed by the reciprocal verb. However, my analysis allows for a more elegant

analysis of the -sja verbs that I introduce in Chapter 6. As a preview, there I will provide data

that in some contexts the -sja verbs are compatible with asymmetric relations between

participants. If we were to assume that the -sja verbs express only one event that is accessible to

the syntax and semantics, we would have to come up with a different analysis for the asymmetric

verbs as they would only express “half” of the event, or one of the subevents. Assuming an

analysis where each relation is associated with a distinct subevent will allow a more coherent

analysis of the -sja verbs than positing a single event in that chapter.

Additionally, I analyze sentences with reciprocal verbs as expressing two simultaneous

subevents because it allows for a more unified analysis across reciprocal expressions; under my

analysis, all reciprocal expressions introduce distribution of events. On Siloni’s account, the

145

distribution happens on two different levels and which interpretation a reciprocal expression has is dependent on its morphosyntactic structure. In this chapter I have provided evidence

demonstrating that the morphosyntax of the expression plays less of a role than she proposes for the Russian reciprocal expressions, suggesting a unified analysis is the better option. I therefore look at all reciprocal relations in terms of subevents because it allows for a more standard analysis between reciprocal expressions.

3.4.3 Formal analysis of -sja verbs

I have shown that a temporal restriction to simultaneity of events is lexically specified for the 13 (and possibly more) verbs listed in (164) above as well as for two -sja verbs in Russian,

celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’. The other -sja verbs are underspecified in terms of

the temporal relations between events. These are the empirical generalizations; I now turn to

capturing these findings in a formal analysis.

As outlined in Chapter 2, I analyze the -sja verbs as being generated in the lexicon, and so

they come to the syntax packaged with both the semantic relation, such as kissing or butting, but

also with the configuration of the relations specified. Thus, the lexical entry that I proposed in

Chapter 2 for the verb bodat'sja ‘to butt’ is as in (194):

(194) bodat'sja ‘to butt each other’: N\S’: λxλev[∀y∀z(y ≤ x ˄ z ≤ x ˄ y ≠ z → ∃e’∃e” (e’≤ ev ˄ e” ≤ ev ˄ e’≠ e” ˄butt’(y)(z)(e’) ˄ butt’(z)(y)(e”)))]

The syntax of this lexical entry says that bodat'sja combines with a noun phrase (of category N) to result in a sentence radical (a tenseless sentence). As for the semantics, this verb denotes a

146

function from an entity x to a function from an eventuality ev to a truth value such that bodat'sja applied to an entity x and an eventuality ev is true iff for all y and z that are parts of x, where y and z are distinct, there are events e’ and e” such that e´ and e” that are parts of ev, where e’ and e” are distinct, and e’ is an event of z butting y, and e” is an event of y butting z. The individual x is contributed by the subject of the sentence, and the eventuality ev is contributed by the tense morpheme. Note that there is no restriction on the temporality of the events. The tense morpheme will add a temporal trace that will map each subevent to its running time, but the lexical entry of the verb has no temporal restriction at all. This definition of the verb bodat'sja is

apt, as I provided evidence suggesting that this does not in fact introduce a temporal restriction on

its subevents of simultaneity or sequentiality.

The lexical entries of the other -sja verbs, on the other hand, will have to specify that the

subevents introduced by the verbs occur simultaneously. Thus, I propose the category and

specifically the semantics in (195):

(195) celovat'sja ‘to kiss each other’: N\S’: λxλev[∀y∀z(y ≤ x ˄ z ≤ x ˄ y ≠ z → ∃e’∃e” (e’≤ ev ˄ e” ≤ ev ˄ e’≠ e” ˄ kiss’(y)(z)(e’) ˄ kiss’(z)(y)(e”)) ˄ τ(e’) ⊗T τ (e’’))]

The syntax of this lexical entry is the same as for bodat'sja. As for the semantics, celovat'sja denotes an entity x such that for all distinct atomic y and z that are part of this entity, there are subevents e’ and e” of the event ev and e’ is an event of y butting z and e” is an event of z butting

y and the run times of the subevents e’ and e” overlap. The bolded portion is the clause that

constraints the temporality of the events to simultaneity.

Note that this analysis only requires, as I mentioned with defining my terms in Section

3.2, that the events expressed by the sentence temporally overlap and not that they map onto the 147

same run time. I propose that this is the best analysis, based on sentences such as that in (196),

which is acceptable in a context in which the hugging events overlap but do not map onto the same runtime.

(196) Context: You see your friend Marija hug Ivan, startling him. It takes a few seconds for him to start hugging her back. She then disengages while he still has his arms around her. He then lets her go a few seconds later. You say:

Marija i Ivan obnjali-s'. Marija.NOM and Ivan.NOM hug.PST.PL-SJA ‘Marija and Ivan hugged.’

In this context, one individual begins and ends a hugging event before the second individual.

Since this sentence is acceptable in the context, I suggest that temporal overlap as opposed to temporal identity is the best way to account for the temporal restriction introduced by these -sja verbs.

Another approach to accounting for the fact that some -sja verbs contribute the temporal restriction to simultaneity and some do not is conceivable. It is possible that the verbs that restrict their events to simultaneity, unlike the other reciprocal -sja verbs, are composed in the syntax, as opposed to being generated in the lexicon. In this approach, the transitive verbs such as celovat'

‘to kiss’ and obnimat' ‘to hug’ would compose with -sja in the syntax, and the verbs celovat'sja and obnimat'sja would not have their own lexical entry. Under this analysis, the suffix -sja would

have the following lexical entry:

(197) -sja: ((N\S’)/N)\(N\S’): λPλxλev[∀y∀z(y ≤ x ˄ z ≤ x ˄ y ≠ z →∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e” ˄ P(y)(z)(e’) ˄ P(z)(y)(e”) ˄ τ(e’) ⊗T τ (e’’)))]

148

This definition specifies that -sja composes first with a transitive verb, such as celovat' before then composing with an individual of type e and an event argument. Again, the bolded portion is the temporal restriction to simultaneity of events.

This analysis works well for celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’ as well as a few of the verbs whose lexical meaning requires simultaneous subevents, such as vstrečat'sja ‘to meet’ as these verbs are, as far as I can tell, fully compositional; there is no shift in the meaning of the type of relation itself when the -sja suffix is realized on the transitive verb. The only differences are that the verb with -sja is syntactically intransitive and stipulates that the entities denoted by the subject stand in the relevant relation to each other, while the verb without -sja is syntactically transitive and simply denotes a relation between the denotations of its subject and its object. Therefore, in terms of semantic compositionality, there is no reason to suppose that the transitive verb and the reciprocal verb are not related.

The main problem with this analysis is that, as I mentioned in Chapter 1, the reciprocal

-sja suffix is in no way productive. Under this analysis then there is a lexical entry -sja that co-occurs with only a few verbs. An explanation would have to be found for its unproductivity.

Similarly, it would have to be specified in the subcategorization information for the suffix that it composes only with certain verbs such as celovat' ‘to kiss’ and obnimat' ‘to hug’. Such an

analysis is possible but more complex than is desirable.

Therefore, there are reasons to analyze the verbs that restrict events to simultaneity as

being fully generated in the lexicon but also reasons to analyze them as composed in the syntax. I

do not know which option is correct, and so I don’t take a firm stance. Not taking a firm stance

has no detrimental consequences for the analyses in this dissertation. For the sake of ease,

however, throughout the rest of this dissertation, I will continue to refer to the -sja verbs as one

149

group (unless specifically noted otherwise) and treat them all as having been generated in the lexicon.

3.4.4 Discussion of the findings on the -sja verbs

My results that not all -sja verbs restrict the event structure of an utterance to simultaneity pose a problem for Siloni (2012). She argues that there are three types of reciprocal expressions; periphrastic expressions (i.e. anaphors such as each other), syntactic reciprocal verbs

(a verb and reciprocal expression combination formed in the syntax, e.g. French and Spanish se

V) and lexical reciprocal verbs (a verb and reciprocal expression combination formed in the lexicon - such as kiss and celovat'sja ‘to kiss’). Siloni finds that a number of syntactic and

semantic properties can be used to categorize reciprocal expressions into these three groups. One

such property to determine whether a reciprocal verb is syntactic or lexical is whether such a verb

expresses only one event, or whether it can express a multitude of events. As I noted in Section

3.4.2, Siloni claims that reciprocal verbs in Russian express only one event (which I call two

simultaneous events). I have shown that this claim is true only for some reciprocal verbs in

Russian. Since the other reciprocal verbs can express more than one event, as evidenced by the

fact that the subevents can happen sequentially, -sja seems to fall in between her two categories

of reciprocal verbs.

It is possible that her basic categories are essentially correct, but that the Russian

reciprocal expression is developing diachronically from one category (i.e. syntactic) to the other

(i.e. lexical). Such a development would make sense considering the Russian expression was at

one point an anaphor and thus composed only in the syntax. If this is the correct analysis, it

suggests that Siloni’s category of syntactic reciprocal verbs is a false category, in that

diachronically reciprocal expressions develop from anaphors to lexical reciprocal expressions and 150

that this development is really a series of changes that happen gradually through various intermediate steps. The syntactic reciprocal verbs then may not be as cohesive a group because reciprocal expressions develop certain properties of lexical reciprocal verbs in different orders.

Indeed, there is really no reason from a diachronic point of view to expect cohesiveness due to the piecemeal nature of lexicalization (Janda and Joseph 1988).

Alternatively, it could just be the case, without recourse to a diachronic explanation, that the properties that reciprocal expressions in each category exhibit are tendencies and not universals. It may be that if we looked at the distribution of other reciprocal expression cross- linguistically (or perhaps even if we look more closely at the data Siloni presents) we would find more exceptions to her classifications, which may not be related to different stages of diachronic development. Reciprocal expressions may follow patterns but need not strictly adhere to them.

On this view, reciprocal expressions are too varied on too many levels to definitively categorize them so that every semantic parameter correlates. I will return to this point in Chapter 6.

3.5 Conclusions

In this chapter I have explored the temporal properties of sentences containing drug

druga and the -sja verbs and have provided data suggesting that both expressions temporally restrict the event structure. Two -sja verbs, celovat'sja ‘to kiss’ and obnimat'sja ‘to hug’, restrict

the events to simultaneity, while the other -sja verbs contribute no such restriction. The pronoun

drug druga does not introduce an entailment of simultaneity, but rather, a Generalized

Conversational Implicature that the events occur simultaneously. This is the first study that

explores the temporal restrictions reciprocal expressions may contribute in a thorough and

systematic manner using both data from elicitation and a corpus study.

151

These findings have implications for an analysis of reciprocal relations. The formal literature that has been concerned with the relations each other conveys has not before considered whether any temporal restrictions should be considered. Since I have shown that some reciprocal expressions do indeed contribute such a restriction, I suggest that when discussing the different reciprocal relations available, in addition to the four relations I have argued each other to be compatible with, we also consider these relations with a restriction to simultaneity. However, since I also show that some reciprocal expressions are compatible with both a simultaneous and a sequential interpretation of the events, not all meanings should be constrained temporally.

Therefore, both the following relations should be included in this set:

(198) a. ∀x,y ∊A(x≠y→xRy)

b. ∀x,y ∊A∃e∃e’(x≠y→xRy ˄ τ(e) ⊗ τ (e’))

Note I have only shown that reciprocal expressions can restrict the event structure to simultaneity; neither of the expressions treated here restrict it to sequentially. While looking at two expressions does not provide a good sample of reciprocal expressions cross-linguistically, no other study that even glances at this issue suggests that a particular expression may restrict the event structure to sequentiality. This suggests two things. First, in our set of reciprocal relations, some relations should be restricted to having event arguments that are simultaneous only.

Second, this suggests something more fundamental about reciprocity. In particular, I hypothesize that if a reciprocal expression is compatible with sequential events, then it will also be compatible with simultaneous events. This hypothesis assumes an implicational hierarchy of temporality.

We already know that personal relations, such as those in Dalrymple et al. (1998b) form a scale

152

such that Strong Reciprocity is logically stronger than the other four meanings discussed. While simultaneity is not logically stronger than sequentiality in the same sense (as simultaneity does not logically entail sequentiality), the fact that reciprocal relations may restrict the events to simultaneity and not sequentiality suggests that simultaneity is higher up on an implicational hierarchy than sequentiality. A given reciprocal expression that is compatible with sequentiality is also compatible with simultaneity, but it is not necessarily the case that a reciprocal expression that is compatible with simultaneity will also be compatible with sequentiality.

If this hypothesis is correct, we should not expect to find a reciprocal expression that is compatible only with events that are sequential. That is, we should not find an expression in any

language that is compatible with a context in which A hugs B without reciprocation, then 10

minutes later B hugs A, and which is not compatible with a context in which A and B hug simultaneously. Finding such an expression would prove this hierarchy based on temporality to

be spurious.

Second, my findings raise the question of whether the way we have been defining

reciprocal expressions is a problem. Most studies, this one included, define a reciprocal

expression as one that is compatible with a Strongly Reciprocal relation between participants.

Strong Reciprocity is by design intended to reflect the most restrictive semantics, from which

other candidates for a reciprocal meaning deviate to varying degrees by loosening strong

reciprocity along the dimension of symmetry. Given my claim that temporality also forms a type

of scale, I ask whether our definition of Strong Reciprocity is specified enough. I suggest we

define an expression as reciprocal if and only if it is compatible with a strongly reciprocal relation

between participants for which the subevents are simultaneous, as it is the strongest claim about

the interaction that is compatible with the data. Under this definition, both drug druga and the

-sja verbs are still defined as reciprocal expressions. 153

In sum, I have shown that reciprocal expressions can contribute semantic and pragmatic temporal restrictions on sentences, suggesting we consider temporality when positing the set of reciprocal meanings. This finding poses questions about the nature of reciprocity and what factors other than personal relations should be taken into consideration when characterizing this linguistic concept.

154

4 The -sja verbs and the syntax-semantics interface

The reciprocal -sja verbs in Russian can occur in two different constructions. The simple reciprocal construction contains a reciprocal verb and has the basic form NP[NOM] V-sja, such as

in the sentence in (199).

(199) Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed.’

Russian also exhibits a so-called discontinuous reciprocal construction. In this

construction, in addition to the subject there is an obligatory comitative phrase. The antecedent

of the reciprocal expression is provided by two NPs instead of one, one of which is the object of

the preposition s ‘with’. The basic form of a sentence realizing the discontinuous construction is

NP[NOM] V-sja s NP[INST]. The denotation of both the subject and the comitative object participate in the relation conveyed by the verb. For example:

(200) Context: You are staying with your friends Ivan and Marija. Before they left for work, you saw them kiss each other. While describing their routine later, you tell a friend:

Ivan poceloval-sja s Mariej. Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed.’

155

In this sentence, the NP Ivan is the subject and the NP Marija is the complement of the comitative preposition s ‘with,’ and the kissing relation holds between these two individuals.

These constructions are called discontinuous because the set that is relevant for the reciprocal meaning is not realized by a single NP but by two. The realization of the relevant set is therefore

“discontinuous”. I will provide a more thorough description of the properties of both constructions in Section 4.1.

The verbs in the two different constructions in many ways seem identical; (i) the -sja verbs are identical in form, and (ii) if a -sja verb can participate in the simple construction, it can also participate in the discontinuous construction. For example, in both (201) and (202), the first sentence contains a -sja verb in the simple construction, and the second sentence contains a -sja verb in the discontinuous construction.

(201) Context: You look out your window to see your two cows, Zor'ka and Burenka, butting each other simultaneously. You say:

a. Zor'ka i Burenka bodajut-sja. Zor'ka.NOM and Burenka.NOM butt.NPST.3PL-SJA ‘Zor'ka and Burenka are butting each other.’

b. Zor'ka bodaet-sja s Burenkoj. Zork'a.NOM butt.3.NPST.3PL-SJA with Burenka.INST ‘Zor’ka and Burenka are butting each other.’

156

(202) Context: You are a teaching and you walk into your classroom one morning to see the boys hugging the girls and the girls hugging the boys. You say:

a. Mal'čiki i devočki obnimajut-sja. boys.NOM and girls.NOM hug.NPST.3PL-SJA ‘The boys and girls are hugging each other.’

b. Mal'čiki obnimajut-sja s devočkami. boys.NOM hug.NPST.3PL-SJA with girls.INST ‘The boys and girls are hugging each other.’

These sentences demonstrate that both bodat'sja ‘to butt’ and obnimat'sja ‘to hug’ can participate in both the simple and discontinuous construction, and that each verb has the same form in both syntactic contexts. These two examples also show that sentences realizing the discontinuous construction can sometimes be used in the same contexts as those containing the simple construction. This would seem to suggest that both constructions contribute the same meaning to a sentence. Consider, however, the context and two sentences in (203). Note that the sentence containing the simple construction in (203)a is not compatible with the context, while the sentence in (203)b containing the discontinuous construction is acceptable in this context.

(203) Context: You see your friend Ivan kiss his girlfriend Marija, but she was taken by surprise and didn’t kiss him back.

a. #Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA

b. Ivan poceloval-sja s Mariej Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed./Ivan kissed Marija.’

157

The context sets up an asymmetric kissing relation between the referents of Ivan and Marija - the denotation of the NP Ivan kisses the denotation of the NP Marija, but she does not kiss him back.

In a context that includes two participants that stand in this asymmetric relation, the simple construction is not acceptable while the sentence with the discontinuous construction is.

Where does the difference in the semantics come from? One possibility is that the -sja

verbs in the two constructions have different lexical entries, such that each contributes a different meaning to the sentence. This analysis seems reasonable, since the -sja verbs in the two constructions subcategorize for different types of arguments, which could correspond to differences in the semantics. A second possibility is that the verbs in the discontinuous construction are derived from those in the simple construction. Under this analysis, the meaning of the -sja verb in the discontinuous construction is derived from the meaning of the same verb in the simple construction. The difference in the semantics under such a possibility would come about as a result of the derivation and because of the differences in the syntax of the constructions. A third possibility is that the difference in meaning between the two constructions is due to the syntactic construction and that the differences in meaning fall out from the different constructions. A priori this analysis is preferable because it does not require two different lexical meanings for one piece of form and thus provides for a more parsimonious lexicon. I will ultimately argue for a hybrid of these possibilities. I will argue that some of the differences in meaning fall out from the different syntactic constructions, while some of the differences do not.

I conclude that different lexical entries are needed for the verbs in each construction. This hybrid approach accounts for both the empirical generalizations and theoretical assumptions I make about the constructions.

Before outlining my reasons for supporting this position, I first explore the syntax and semantics of both constructions in more depth. In Section 4.1, I provide an overview of the basic 158

properties of each construction. In Section 4.2, I summarize what the previous literature has said about the semantics of the discontinuous construction cross-linguistically, which provides a place for me to begin my investigations. Before delving into the semantics of each expression that make up these constructions, I provide commentary on the methods I used that are specifically relevant for my discussion of the -sja verbs in Section 4.3. In Section 4.4 I investigate the meaning of the simple construction, and in Section 4.5 I investigate the meaning of the discontinuous construction. I develop an analysis of both the simple and discontinuous construction in Section

4.6. I lastly discuss the theoretical implications of this chapter in Section 4.7.

4.1 Basic properties of the two constructions

4.1.1 Basic properties of the simple reciprocal construction

The simple reciprocal construction, as noted above, instantiates the string

NP[NOM] V-sja. The NP is of the syntactic type N, and the intransitive verb is of the syntactic type N\S’. The verb takes the subject as an argument to project a sentence radical.

The NP in a simple reciprocal construction is either coordinated, as in (204), plural as in

(205), or syntactically singular but semantically plural as in (206). The NP cannot have a

semantically singular denotation, as demonstrated by the unacceptability of the sentence in (207).

159

(204) Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed.’

(205) Mal'čiki obnimali-s'. boys.NOM hug.PST.PL-SJA ‘The boys hugged.’

(206) Gruppa obnimala-s'. group.NOM hug.PST.SG-SJA ‘The group hugged.’

(207) #Mal'čik obnimal-sja. boy.NOM hug.PST.SG-SJA (Intended: ‘The boy hugged.’)

Note that when the NP is coordinated by i ‘and’ as in (204), both NPs take the nominative case.

Although verbs whose subject are coordinated NPs like that in (204) typically have plural agreement, Russian also (rarely) exhibits nearest conjunct agreement, such that the verb agrees only with the nearest conjunct of the coordinated NP (Corbett 1983: 117):

(208) Emu zvonil i polkovnik Dobren’kij, i Borisov, he.DAT call.PST.SG.M and colonel.M.NOM Dobren’kij.M.NOM and Borisov.M.NOM

i kto-to ešče. and someone.NOM else ‘Colonel Dobren’kij and Borisov and someone else called him.’

In this sentence, the verb, zvonil ‘called,’ has singular agreement, even though its subject is a coordinated NP with three conjuncts. Based on a sentence like this, it is in fact impossible to know which conjunct the verb agrees with, as each one of the conjuncts is singular. Russian

160

verbs in the past tense also inflect for gender, and zvonil has masculine agreement, but since each one of the conjuncts in this sentence is masculine, we are left not knowing whether which NP the verb agrees with. The following sentence solves this problem, as the first NP is masculine and the second is feminine (Corbett 1983: 117):

(209) Teper' na nej byl sinij kostjum i novaja now on her.LOC be.PAST.SG.M blue.M.NOM suit.M.NOM and new.F.NOM

belaja bluzka. white.F.NOM blouse.F.NOM ‘Now she was wearing a blue suit and a new white blouse.’ (Literally:‘Now on her was a blue suit and a new white blouse.’)

In this sentence, the verb byl ‘was’ has singular and masculine agreement. Crucially, the only NP in the coordinated phrase that is masculine is the NP that is nearest the verb, sinij kostjum ‘blue suit.’ Examples such as these have led scholars to conclude that the verb in Russian can agree with the nearest conjunct. Nearest conjunct agreement is most likely to occur when the NP is realized after the verb, as in (208) and (209), and when one or more of the NPs are inanimate

(Corbett 1983). It can, however, occur, rarely, in other environments:

(210) Odna derevnja i odno selenije (Bošković 2010:11) one.F.NOM village.F.NOM and one.N.NOM settlement.N.NOM

bylo razrušeno. be.PST.SG.N destroyed.PTCP.SG.N ‘One village and one settlement were destroyed.’

161

In this sentence the coordinated NP occurs before the verb, and the verb still agrees with the nearest conjunct. Although I have found no examples of it, and my consultants have not accepted the sentences I have provided them, it is possible that if the subject of the simple construction is a coordinated NP, that the verb could be assigned agreement by the nearest conjunct. The fact that

Russian exhibits nearest conjunct agreement will become more relevant in my analysis of the discontinuous construction in Sections 4.6.2.1 and 4.6.2.2 below.

The simple construction also has a structure in which the subject is coordinated by s

‘with’ instead of the conjunct i ‘and’. In these cases the phrase, s Mariej ‘with Marija’ has the

status of a conjunct. The preposition s always assigns the instrumental case to its object.

(211) [Ivan s Mariej] pocelovali-s'. Ivan.NOM with Marija.INST kiss.PST.PL-SJA ‘Ivan and Marija kissed.’

The comitative coordination exemplified in (211) is syntactically very similar to coordination

with the conjunction i ‘and’. Subject NPs that are coordinated by i ‘and’ and those coordinated by s ‘with’ both typically require plural agreement on the verb, even if both NPs in the coordinated NP are singular.

The NPs in the coordinated construction form a constituent. Although strings of NP s NP

do not always form constituents in Russian, they do when in the simple construction, like in

(211). Evidence in support of the constituency of the bracketed words in (211) is that the

comitative phrase s Mariej ‘with Marija’ cannot be grammatically extracted from its position:

162

(212) *S Mariej Ivan pocelovali-s'. with Marija.INST Ivan.NOM kiss.PST.PL-SJA (Intended: ‘Ivan and Marija kissed.’)

(213) *Ivan pocelovali-s' s Mariej . Ivan.NOM kiss.PST.PL-SJA with Marija.INST (Intended: ‘Ivan and Marija kissed.’)

In (212), when the comitative phrase is fronted, the sentence is ungrammatical. The sentence in

(213) is also ungrammatical when the comitative phrase is realized non-adjacent to the host NP.

Furthermore, parentheticals cannot be inserted between the host NP and the comitative phrase:

(214) *Ivan, naprimer, s Mariej pocelovali-s'. Ivan.NOM for.example with Marija.INST kiss.PST.PL-SJA (Intended: ‘Ivan for example and Marija kissed.’)

The sentence is ungrammatical when the parenthetical naprimer ‘for example’ is realized between the NP and the comitative phrase. Based on these two tests, I assume that in the sentence in (211) that realizes the simple construction, the bracketed words form a constituent.

See Section 4.6.2.1 below for a more in depth comparison and discussion of the two types of coordination.

4.1.2 Basic properties of the discontinuous reciprocal construction

A sentence containing a discontinuous reciprocal construction realizes a subject NP, the

-sja verb, and an obligatory comitative phrase headed by the preposition s ‘with’. It might instantiate the string NP[NOM] -sja verb s NP[INST] as in (215), where the comitative phrase is

163

realized post-verbally, or NP[NOM] s NP[INST] -sja verb like that in (216) where the comitative

object is realized pre-verbally.

(215) Ivan poceloval-sja s Mariej. Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed.’

(216) Ivan s Mariej poceloval-sja. Ivan.NOM with Marija.INST kiss.PST.SG-SJA ‘Ivan and Marija kissed.’

I’m assuming that the verb first combines with the comitative phrase, then with the subject to project a sentence. I provide evidence for this order of composition in Section 4.6.2.2 below.

The two sentences in (215) and (216) are truth-conditionally identical despite the different word order. Russian word order is free in the sense that any of the major constituents can be grammatically realized in any position. SVO is the unmarked word order and deviations from this are typically analyzed as being for pragmatic reasons (Bailyn 2012).

Note that the sentence in (216) exemplifies the discontinuous construction and not the simple construction with a coordinated subject NP because the verb has singular, masculine agreement. This being the case, the subject of the verb must be Ivan. If the subject were Ivan s

Mariej, we would expect either plural agreement or singular, feminine agreement if the verb were

agreeing with the nearest conjunct. Therefore, I suggest that in this sentence, the NP Ivan is the

subject, and the comitative phrase does not form a constituent with the subject.

164

As evidence for its non-constituency, the comitative phrase s Mariej, unlike that in the sentence in (211), can be fronted as in (217), or a parenthetical can be inserted between the NP

Ivan and the comitative phrase, as in (218):

(217) S Mariej Ivan poceloval-sja. with Marija.INST Ivan.NOM kiss.PST.SG-SJA ‘Ivan and Marija kissed.’

(218) Ivan, naprimer, s Mariej poceloval-sja. Ivan.NOM for.example with Marija.INST kiss.PST.SG-SJA ‘Ivan, for example, and Marija kissed.’

The acceptability of these two sentences suggests that the comitative phrase in these two sentences does not form a constituent with the subject Ivan, and therefore these sentences do not exemplify the simple construction.

There are two immediate difficulties with working with the discontinuous construction when the comitative phrase is realized pre-verbally. First, although a sentence that realizes the discontinuous construction in which the comitative phrase is realized preverbally is grammatical, as the acceptability of (216) suggests, it is not frequent. No such examples were found in my corpus study and no consultant ever offered such a sentence as a way to describe a video or context.

Second, it is impossible to tell from the syntax of a sentence such as that in (219) whether the comitative phrase is a constituent of the subject or a comitative phrase in the discontinuous construction, since the verb is plural either way.

165

(219) Mal'čiki s devočkami celovali-s'. boys.NOM with girls.INST kiss.PST.PL-SJA ‘The boys and girls kissed.’

Therefore, for the remainder of this chapter, I focus on examples that unambiguously realize either the simple or the discontinuous construction. My findings will later allow me to use semantics to identify for sentences such as (219) whether it instantiates the simple or the discontinuous construction. In this way, this chapter makes a contribution in how to diagnose such sentences with semantics as opposed to syntax.

Both the subject and the object NPs in a sentence with the discontinuous construction can be either singular, as in (215) and (216), or plural as in (220) - (222):

(220) Ivan poceloval-sja s devočkami. Ivan.NOM kiss.SG.PST-SJA with girls.INST ‘Ivan and the girls kissed.’

(221) Devočki pocelovali-s' s Ivanom. girls.NOM kiss.PL.PST-SJA with Ivan.INST ‘The girls and Ivan kissed.’

(222) Mal'čiki pocelovali-s' s devočkami. boys.NOM kiss.PL.PST-SJA with girls.INST ‘The boys and the girls kissed.’

166

Therefore, for a sentence to realize the discontinuous construction, there must be an NP subject which can be syntactically and semantically singular or plural, a -sja verb24, and a comitative phrase, the object of which can be syntactically and semantically singular or plural.

4.2 Previous literature on the discontinuous construction

Many languages cross-linguistically have been shown to make use of the discontinuous

construction, including Russian, Hebrew, Greek, English, and Swahili. In each of these

languages, there is a construction which realizes a subject NP, a reciprocal verb, and a comitative

phrase. Cross-linguistically the discontinuous construction has been claimed to semantically

differ from the simple construction in two main ways: (i) it is only acceptable in contexts in

which the relations hold only between entities denoted by the subject and entities denoted by the

object argument, and not between entities in the same set (Dimitriadis 2004, 2008; Rákosi 2008;

Siloni 2012), and (ii) the denotation of the object is said to be either “less volitional” than that of

the subject (Rákosi 2008). By “volitional”, Rákosi, for example writes that volition is a semantic

feature that corresponds to one of the possible lexical entailments in that Proto-Agents may have

(à la Dowty 1991). An entity that is volitional has the desire to participate in the relation.

Dimitriadis analyzes this difference as a difference in “degree of agentivity,” which is compatible

with the claim about volition, as agents are typically volitional. In this section I discuss the

differences in what authors claim about various languages for the restrictions in (i) and (ii).

Dimitriadis explores discontinuous constructions in Greek. He claims that when the

discontinuous construction is realized, no relation between entities in the same set is entailed. He

provides, for example, the following example in Greek:

24 In Chapter 5, however, I argue that naturally reciprocal verbs can also participate in the discontinuous construction. I will discuss this addition there. 167

(223) O Yanis kje o Nikos tsakothikan me ti Maria. (Dimitriadis 2008: 9) the John and the Nick argued.RECP with the Maria ‘John and Nick argued with Maria.’

He writes that this example:

is either about an argument between John and Nick on the one part and Maria on the other, or possibly about two different arguments between Maria and each of the two men. In each case, the reciprocal relation must involve pairs consisting of one participant (possibly plural) from the syntactic subject, and one participant from the comitative argument. (9)

Thus, the sentence in (223) is acceptable in a context in which John and Nick argued as a team against Maria, or in which John argued against Maria, and Nick argued against Maria.

Dimitriadis thus claims that for the sentence in (223) to be true, it has to be the case that one participant in the arguing relation is drawn from the subject NP and the other is from the object

NP. He does not say whether this sentence is compatible with a context in which two members of

the same set stand in the R-relation to each other but just that such a relation is not asserted.

As for requirement (ii), that the individual denoted by the object is less volitional or

agent-like than that denoted by the subject, Dimitriadis provides the following example from

Greek:

(224) O Nikos filithike me to aghalma. (Dimitriadis 2008: 400) the Nick kiss.PST.RECP.SG with the statue ‘Nick engaged in a mutual kiss with the statue.’

168

Dimitriadis writes that “Many Greek speakers find [(224)] … to be acceptable.’ Therefore in

Greek, some speakers can interpret the sentence in (224) such that the individual denoted by the object is less agent-like.

Rákosi (2008) discusses the discontinuous construction in Hungarian. As for requirement

(i), he writes that “… reciprocity holds of the distinctly construed subject and oblique sets” (419).

He does not provide any discussion or examples of sentences that realize plural subjects or objects, and so based on his discussion it can’t be determined whether the members of the sets denoted by the subject and object in the discontinuous construction are constrained such that they don’t stand in the relation to each other.

He does, however, provide data that the denotation of the object of the discontinuous construction is less volitional than the subject. Rákosi provides the following example in

Hungarian:

(225) Én nem veszeked-t-em János-sal, ő veszeked-ett vel-em. I not quarrel-PST-1SG John-with he quarrel-PST with-1SG ‘I was not quarrelling with John, he was quarrelling with me.’ (Rákosi 2008: 423)

As he points out, this sentence would result in a contradiction if a reciprocal relation were required to hold between the subject and the comitative object. He takes such evidence to suggest that the entity denoted by the object can be less “volitional” than that denoted by the subject.

Therefore, like Dimitriadis, Rákosi does not explicitly state or show that reciprocity cannot hold between members of the same set. It is therefore unclear whether Dimitriadis and

Rákosi intend to argue that relations cannot hold between members of the same set, or whether they hold that such relations are simply not entailed. Either way, they do not provide the full set 169

of empirical generalizations needed to develop an analysis of this construction. They do both

make the claim that the object of the comitative phrase can be less “volitional” or “agent-like” in

both Greek and Hungarian.

Siloni (2012: 298) discussing Hebrew, on the other hand, writes that “. . . in the

discontinuous construction . . . reciprocity holds between the subject set and the oblique set

introduced by the preposition with, and not between the members of the subject set …”. She

provides the following example of the discontinuous construction in Hebrew:

(226) Ha-yeladim hitnašku im ha-yeladot. the-boys kissed.RECP with the-girls [‘The boys kissed with the girls./The boys and girls kissed.’]25

Describing this sentence, she writes “there were mutual kissing events between boys and girls;

but no kissing events within the set of boys or girls” (298). Siloni, therefore, does claim that for

Hebrew discontinuous constructions, no members of the same set can stand in a reciprocal

relation to each other.

Siloni provides the following example in Hebrew, which she says entails that the

denotation of Dan participated in the arguing relation as the first argument, but that additionally

the two interpretations in (i) and (ii) are both possible:

iii. 25 Siloni does not provide a translation for these sentences, so these translations are mine, based on her glossing and discussion. I also asked a speaker of Hebrew to verify my translations.

170

(227) Dan hitvake’ax im Dina (Siloni 2012: 299) Dan argued with Dina [‘Dan argued with Dina.’]

(i) Dina hitvakxa im Dan Dina argued with Dan [‘Dina argued with Dan.’]

(ii) Dan hitvake’ax im Dina, aval Dina lo hitvakxa ito. Dan argued with Dina but Dina NEG argued with.him [Dan argued with Dina, but Dina didn’t argue with him.’]

The interpretation in (i) requires that in addition to the denotation of Dan standing in the arguing relation as the first argument, the denotation of Dina does too (and hence they both participate as the second argument as well). The second interpretation only requires that the denotation of Dan

stand in the arguing relation as the first argument - the denotation of Dina is not required to.

Although Siloni presents such examples, she ultimately argues that the non-symmetric reading,

i.e. that in (ii), comes about because there are two different lexical entries for the verb hitvake’ax

‘argue’, one that requires that the entities denoted by both the subject and the object stand in both

the first and second argument of the relation, and a second lexical entry that has a transitive

meaning, such that only the entities denoted by the subject stand in the relation as the first

argument and only the entities denoted by the object stand in the relation as the second argument.

I will return to this point later after I have presented her analysis.

Siloni (2012) presents an analysis in which the lexical entry for the reciprocal verbs that

occur in the simple construction is as in (228)a, and that for the reciprocal verb in the

discontinuous construction is as in (228)b. She assumes that the lexical representation of

reciprocal verbs includes two entries, one that is realized in the simple construction and one that

is realized in the discontinuous construction. VSYM represents a lexically reciprocal verb, or a reciprocal verb that is generated in the lexicon like the -sja verbs, that imposes a relation of 171

symmetry on relevant pairs. The bracketed information represents the thematic role each verb imposes on its arguments. Lexically reciprocal verbs in the simple construction take one argument, and this argument’s thematic role in (228)a is a bundled Agent-Theme role.26 The reciprocal verbs in the discontinuous construction, with the lexical entry as in (228)b, take two arguments and impose the Agent-Theme role on the subject. Siloni writes that “the empty role

[Ø] is introduced into the structure by the preposition with (or parallel case morphology).” (310).

(228) a. VSYM [Ag-Th]

b. V’SYM[Ag-Th], [Ø-WITH] (Siloni 2012: 310)

She proposes that the translation of the sentence in (229), which realizes a discontinuous

construction in Hebrew, is as in (230):

(229) Dan hitnašek im Dina (Siloni 2012: 310) Dan kissed.REC with Dina ‘Dan and Dina kissed.’

(230) ∃e[kiss’SYM (e) & [Ag-Th](e, Dan) & [Ø-with] (e, Dina)] (Siloni 2012: 310)

This translation says the sentence in (229) is true when there exists an event such that this event is one of symmetrical kissing, and the individual Dan has the role of [Ag-Th] for this event and the individual Dina is assigned the thematically empty role by the preposition with. The empty role gets assigned via a meaning postulate that makes use of Landman’s (2000) ↓ operator. This

26 The bundling of the two roles happens through a meaning postulate that derives reciprocal verbs from their transitive counterparts (Siloni 2012). 172

operator maps atomic events onto the sum events, i.e. allows group events to specify the atomic events that make up the group event. This postulate then “defines V’SYM as entailing two underlying events of the same type (e1 and e2)” (310). The simplex roles that are bundled into the

[Ag-Th] thematic role are defined for these underlying events such that the subject is interpreted

as the agent of e1 and the theme of e2. The meaning postulate also requires “. . . the thematically

empty argument [Ø-with] [to be] interpreted as being in a symmetric relation with the subject

argument; more precisely, it is interpreted as the Agent role of the underlying event e2 and the

Theme role of e1” (311). The underlying representation then of (230) is as in (231) (Siloni 2012:

311):

(231) ∃e∃e1∃e2[↓(e) = (e1⊔e2) &[kiss(e1)&Ag(e1, Dan) & Th(e1, Dina) & kiss(e2) & Ag(e2, Dina) & Th(e2, Dan)]]

This translation says that the sentence in (229) is predicted to be true when the kissing event is a

sum of two underlying events. The first underlying event is one of kissing such that the

individual denoted by Dan is the AGENT and the individual denoted by Dina is the THEME and the second event is one where the denotation of Dina has the role of AGENT and that of Dan has the

role of THEME.

Her analysis partially captures the first empirical generalization discussed above, that

entities in the same set are not entailed to participate in the relevant relation with each other, as

the meaning postulate requires that the thematically empty argument be interpreted as being in a

symmetric relation only with entities in the subject. Although she does not provide an example,

her analysis predicts that if there are two entities denoted by a plural subject and two denoted by a

plural object, then this sentence would express eight underlying events such that each entity

173

denoted by the subject stands in the kissing relation to each entity denoted by the object for two distinct underlying events; for one as the AGENT and one as the THEME.

She proposes that the restriction that no entity stand in the relation to another entity in the same set is due to the meaning of the verb. The meaning of the verb in the discontinuous construction, V’SYM[Ag-Th], [Ø-WITH], contributes the requirement of symmetry, accessible underlying events, and the bundled thematic role and the meaning postulate is connected with the verb, not the construction itself. The verb picks up another argument, but it is the verb that assigns the underlying events and simplex thematic roles to the individuals.

Note, however, that Siloni’s analysis does not restrict entities of the same set from standing in the relation with each other; it only requires that each entity denoted by the subject is entailed to stand in a symmetric relation to each entity denoted by the object. Her analysis predicts that a sentence in which members of the set denoted by the subject stand in the relevant relation to each other is acceptable, and therefore, her analysis does not capture the empirical generalizations she proposes for Hebrew.

Her analysis purposefully does not account for the second empirical generalization, i.e. that the denotation of the object is less volitional or agentive than that of the object. She discusses this restriction in terms of volition, like Rákosi. She provides data suggesting that when the discontinuous construction realizes some verbs, such as hitvake’ax ‘argued’, the denotation of

the object can be (but doesn’t have to be) interpreted as having no volition.

(232) Dan hitvake’ax im Dina, aval Dina lo hitvakxa ito. (Siloni 2012: 299) Dan argued.RECP with Dina but Dina NEG argued with.him [‘Dan argued with Dina, but Dina didn’t argue with him.’]

174

When the discontinuous construction realizes other verbs, such as hitnašek ‘to kiss’, she argues

that the denotations of both the subject and object must be volitional (298):

(233) #Dan hitnašek im Dina, aval Dina lo hitnaška im Dan Dan kissed.RECP with Dina but Dina NEG kiss.RECP with Dan [Intended: ‘Dan kissed Dina, but Dina didn’t kiss Dan.’]

Siloni argues that some lexically reciprocal verbs, such as hitvake’ax ‘argued’ have two distinct lexical entries: one which expresses a symmetrical relation between entities and the other which expresses an asymmetrical relation. This second entry is semantically equivalent to a transitive verb. Other reciprocal verbs, such as hitnašek ‘to kiss’, have only one lexical entry which expresses a symmetrical relation between entities.

Siloni supports her claim based on the verb collide and its translations in various languages. It has been pointed out that this verb does not require that the denotation of its object stand in the collide relation as the first argument:

(234) The truck collided with the fence.

In the sentence in (234), it is only the truck that moves; the fence is only collided into. Siloni argues, however, that this sentence is compatible with an asymmetric reading because the verb collide has two lexical entries based on evidence from Hebrew. The verb hitnageš ‘collide’ in

Hebrew, when in the discontinuous construction, only has a symmetrical reading.

175

(235) Ha-mexonit hitnagša im ha-masa-it/*ha-gader. (Siloni 2012: 299) the-car collided with the-truck / the-fence [‘The car collided with the truck/the fence.’]

Since the sentence that realizes the NP ha-gader ‘the fence’ is unacceptable, she proposes that in

Hebrew the verb hitnageš ‘collide’ only has a symmetrical interpretation. In order to describe a situation in which the object is stationary and thus does not participate as the first argument of the relation, the preposition be ‘in’ is used in place of the preposition im ‘with’:

(236) Ha-mexonit hitnagša ba-gader (Siloni 2012: 299) the-car collided in.the-fence [‘The car collided with the fence.’]

Since the verb takes one preposition when it has a symmetric interpretation and another preposition when it does not have a symmetric interpretation (or at least is compatible with an asymmetric meaning - she does not discuss whether a symmetric meaning is possible), she argues that verbs in Hebrew that participate in the discontinuous construction have two distinct lexical entries, one with a symmetric meaning and one with an asymmetric meaning.

Siloni’s claim about two different lexical entries for these verbs requires more evidence.

It is not clear from her discussion how many verbs are like hitnašek ‘kiss’, that do not have an asymmetric interpretation, and how many verbs are like hitvakxa ‘argue’, that do have an asymmetric interpretation. Furthermore, although her claim seems plausible for the verb hitnageš

‘collide’ in Hebrew, she does not discuss whether other verbs in the language also take different prepositions depending on whether they have a symmetric or asymmetric meaning.

176

Assuming different lexical entries simplifies the task of deriving the discontinuous construction from the simple, as under this assumption only the symmetric reading of the discontinuous construction must be derived. The interpretation of sentences where the object has no volition can be accounted for with an entirely different analysis. She provides very little independent evidence for such an analysis, however, and indeed there is evidence against this

analysis. For example, the two discontinuous constructions are syntactically identical, the verbs in

them are identical in form, and they have similar semantic properties. Therefore more data is

required to support her claims.

As my summaries of Dimitraidis (2008), Rákosi (2008) and Siloni (2012) suggest, the

literature on the discontinuous has been concerned with exploring the empirical generalizations

and developing an analysis that accounts for the requirement that (i) entities denoted by the

subject must stand in the relation to entities denoted by the object, and (ii) the denotations of the

object are less volitional than those of the subject.

The works summarized above discuss how reciprocal relations hold between entities

when the relations are defined over two sets. The formal literature on reciprocity in English has

debated whether relations are ever entailed to be defined for two sets when a sentence realizes

each other, as is required for the discontinuous construction in other languages. Langendoen

(1978: 188), for example, provides the following depiction of a relation between individuals in two different sets.

177

Figure 16: Reproduction of Langendoen’s (1978: 188) depiction of reciprocal relations for two sets

He writes that the sentence in (237) can acceptably be used to describe the situation depicted in

Figure 16 :

(237) They released one another.

In this situation, no releasing relation holds between b and c, only between a on the one side and

b and c on the other. He proposes the following definition of a relation Weak Reciprocity for

Subsets (188):

(238) (∀x∈ A)(∃X1,X2, Y≠ ∅, Z ≠ ∅ ⊆ A)(x ∈ X1 ∧ x ∈ X2 ∧ x ∉ Y ∧ x ∉ Z ∧ X1RY ∧ ZRX2)

He calls this relation Weak Reciprocity for Subsets (WRS), because relations hold only between

members of different subsets. This relation specifies that “each member of the set A be a member

of a subset of A that bears the relation R to some nonnull subset of A not containing that member,

178

and that each member of A is a member of a subset of A that has the relation R borne to it by some nonnull subset of A not containing that member” (188).

Langendoen is in fact interested in the fact that the relation depicted in Figure 16 entails neither aRb nor bRa, but rather aRb⊕c, and he proposes the relation of WRS to capture this fact.

However, he also captures the fact that the two entities, b and c, that are members of the same set,

are not entailed to stand in the R-relation to each other. His definition does not stipulate that

members of the same set do not stand in the R relation to each other. It simply does not require

that members of the same set stand in the R relation to each other.

Langendoen eventually argues, however, that this partitioning of the individuals into two

subsets is not a result of the reciprocal operator. He notes that transitive sentences without a

reciprocal expression are typically also compatible with a context in which the relevant relation

holds between members of opposing sets. For example, in Context 1 in (239), the students are

partitioned into two subsets (the French students and the American students), and the books are

partitioned into two subsets (the books in French and the books in English). The sentence

containing a subject, transitive verb, and an object, is compatible with this (double) partitioning.

(239) Context: You are teaching a group of French and American students. You give them an assignment to pick any book they want to read. You notice that the French students choose books in French to read while the American students choose books in English. You say:

The students read books for an hour.

Since both the sentences in (237) and (239) are compatible with the reading where the denotation

of the subjects and objects are partitioned into non-overlapping subsets, it suggests that some

element common to both sentences optionally contributes the subsetted reading, and not the 179

reciprocal expression. Because of similarities like these in the possible range of interpretations for the reciprocal sentence in (237), Langendoen argues that a sentence which has an interpretation where the set of individuals is partitioned into two subsets is in fact derivable from its non- reciprocal counterpart. Because of this, Langendoen does not find a reason to include a semantic restriction against entities in the same subset participating in a relation with each other in his analysis of the meaning of each other.

Dalrymple et al. (1998b) also discuss sentences containing each other that can require that the relevant relation hold between individuals in different sets, and not between individuals in the same sets. Dalrymple et al. (1998b) provide the following example:

(240) The satellite, called Windsock, would be launched from under the wing of a B-52 bomber and fly to a “libration point” where the gravitational fields of the Earth, the Sun and the Moon cancel each other out. (180)

They argue that the canceling-out relation holds between the Earth and the Sun as one unit, the

Sun and the Moon as a second unit and the Moon and the Earth as a third unit, as depicted in

Figure 17:

180

Earth

Sun Moon

Figure 17: Reproduction of Dalrymple et al.’s (1998b: 181) depiction of the relations between the Earth, Sun and Moon.

They argue that which groupings are established from the entities mentioned in the example are determined by linguistic and contextual indicators that are not related to the reciprocal expressions themselves. Therefore, like Langendoen (1978), they decide that the reciprocal expression does not entail that the set denoted by the subject is partitioned into non-overlapping subsets. Rather, each other is simply compatible with this interpretation.

Since previous formal analyses of the relations reciprocal expressions are compatible with have focused solely on English each other, they have not taken into consideration expressions like the -sja verbs in the discontinuous construction where two non-overlapping sets, denoted by two non-coordinated NPs, need to be considered.

When discussing the discontinuous construction in Russian in Section 4.5 below, I will first explore the empirical generalizations structured around those already discussed in the literature which I have surveyed here. In this way, this chapter will make a contribution to typological studies. I will also provide a formal analysis of how to capture the generalizations

181

when the reciprocal expression under discussion is a verb and not a pronoun like each other, and

thus will contribute to the formal literature as well.

In Section 4.2 I have reviewed the literature on the discontinuous construction cross-

linguistically. I now move on to exploring the semantics of both the simple and discontinuous

constructions in Russian to discover the empirical generalizations an analysis must capture.

Before doing so, however, I have a few comments on the particulars of the methodology I used

for this chapter.

4.3 Methodological preliminaries about the corpus study

To gather data for this chapter, I used the three types of methodology outlined in Chapter

1: (i) elicitation with native speakers utilizing video stimuli, (ii) elicitation using contexts

constructed by myself, and (iii) a corpus study using the Russian National Corpus (RNC). I

obtained the majority of the data through elicitation, and this data collection progressed as I

outlined in Chapter 1. I provide information about the corpus study here that is relevant only to

the -sja verbs.

For the corpus study on simple reciprocal constructions, I examined 25 sentences for each

of the following verbs:

(241) celovat'sja ‘to kiss each other’ carapat'sja ‘to scratch each other’ zdorovat'sja ‘to greet each other’ sražat'sja ‘to struggle with each other’

182

Although 25 is a relatively small number of sentences, I decided not to look for more as I realized

that this methodology ultimately was not that useful; it is usually unclear what relation between

participants was being described given the limited contexts available. This problem highlights

the limitations a corpus study has for the study of semantics.

These four verbs are lexically representative of the various -sja verbs. I include

celovat'sja ‘to kiss’ because it is one of the verbs I discuss in Chapter 3 that contributes a

temporal restriction on the events denoted by its sentence. I chose carapat'sja ‘to scratch’ because it is one of the group of -sja verbs that denote physical actions usually performed by animals or at times by humans with enmity (others include bodat'sja ‘to butt’ or klevat'sja ‘to peck’). The verb zdorovat'sja ‘to greet’ was chosen because it exemplifies the verbs that are

social in nature such as rezat'sja ‘to play cards’ or perepisyvat'sja ‘to correspond with each

other’. The verb sražat'sja ‘to struggle with each other’ was chosen because it is an example of

the group of verbs related to physical or verbal struggle such as borot'sja ‘to struggle with each other’ and ssorit'sja ‘to argue with each other’.

In the corpus study for the simple construction, I only looked at plural forms of the verb, such as celujutsja ‘they are kissing’ and celovalis' ‘they kissed’. I excluded the singular forms because in the singular the -sja verbs are dominantly used in the discontinuous reciprocal construction. Singular -sja verbs are most commonly used in the discontinuous construction because, as mentioned above, a sentence with a -sja verb is grammatical only if the subject is semantically plural. It is possible for the subject to be syntactically singular with a plural denotation such as gruppa ‘group’ or sem'ja ‘family’, but these types of subjects are not frequent in the corpus. A search for a singular verb form typically returns sentences with semantically and syntactically singular subjects, which cannot occur in the simple construction. Therefore, searching for -sja verbs in the simple construction was only practically possible when looking 183

solely at plural forms of the verb. I do not know of any reason, however, that sentences with syntactically singular but semantically plural subjects should be compatible with different relations between participants than sentences with plural subjects.

For the corpus study on the discontinuous construction, I pulled sentences for each of the

same four verbs, but this time I searched for verbs with both plural and singular agreement and

simply manually sorted through the data to remove the sentences without a comitative phrase. I sorted until I had found 25 sentences for each of the four verbs.

4.4 The meaning of the simple reciprocal construction

In this section I discuss the meaning of the simple reciprocal construction by exploring

the relations it is compatible with. I will provide evidence suggesting that the simple reciprocal

construction is compatible with the same four relations as each other.

4.4.1 Strong Reciprocity (SR)

The data show that the simple reciprocal construction is compatible with Strong

Reciprocity, as defined in:

(242) Strong Reciprocity: ∀x,y ∈A (x≠y → (Rxy∧ Ryx))

Of course, the simple construction would have to be compatible with SR to be defined as

reciprocal in the first place, since compatibility with this relation is how I determine whether an

expression is reciprocal or not. Because I have provided data before that demonstrate this in

Chapter 2, I will not repeat them here.

184

The simple construction is also compatible with partitioned SR, but as I argued in

Chapter 2 for English, the partitioning effects are due to the Cover mechanism (Schwarzschild

1996) of the plural NP. Therefore, while I found many examples of partitioned SR such as that in

(243), I do not consider partitioned SR to be a different reciprocal relation than SR. The

utterance in (243) is compatible with partitioned SR because the participants are in groups and the

exchanging relation holds between all members of the group.

(243) Video Context: There are six people, who are divided into three groups, each consisting of two people. Each person exchanges something with the one other person in his/her group.

Oni obmenivajut-sja veščami. they.NOM exchange.NPST.3PL-SJA objects.INST ‘They exchanged objects.’

Partitioned readings are also possible for the remaining relations discussed in this chapter, but I will not mention it again as I hold that the effects are not introduced by reciprocal expressions.

4.4.2 Intermediate Reciprocity (IR)

Recall that Dalrymple et al. (1998b) define the IR relation as the following:

(244) Intermediate Reciprocity:

|A|≥2 and ∀x,y ∈A (x ≠ y→ for some sequence z0, . . . , zm ∈A (x = z0 ∧ Rz0z1 ∧

… ∧Rzm- 1zm∧zm=y))

185

This definition requires A to have at least two participants and for all participants x and y in A, if

x and y are distinct, then they stand in a direct or an indirect R relation to each other via a

sequence of R relations. Data from the video stimuli illustrates that the -sja verbs are compatible

with the IR relation:

(245) Video context: Six people stand in a line shoulder to shoulder. Each person gives an object to his/her neighbor(s) and receives an object from his/her neighbor(s). The exchanging relations between each pair happen sequentially (i.e. A gives an object to B, and then B gives an object to A) but within the entire group, giving relations are happening simultaneously.

Oni obmenjali-s' predmetami. they.NOM exchange.PST.PL-SJA objects.INST ‘They exchanged objects.’

In this context, each person stands in a direct or indirect giving relation with each other person, as

in the diagram in Figure 18:

1 2 3 4 5 6

Figure 18: Depiction of an Intermediate Reciprocal Relation

Participant 1, for example, stands in a direct relation to participant 2, but in an indirect relation to

participant 6. This example instantiates IR because each individual is in a direct or indirect

relation to every other individual. The fact that my consultants provided this utterance to describe

the context suggests that this sentence is acceptable in this context. It then provides support for

the hypothesis that -sja verbs are compatible with the IR relation. Similarly, I provided the

186

following context to my consultants, and they judged the sentence in (246) as acceptable given the context:

(246) Context: You are babysitting four boys and they don’t get along very well. Their names are Misha, Borja, Petja and Vanja. Misha and Borja are fighting, Borja and Petja are fighting, and Petja and Vanja are fighting. You later tell their parents:

Mal'čiki ssorili-s'. boys.NOM fight.PST.PL-SJA ‘The boys fought.’

In this context, each boy stands in a direct or indirect fighting relation with each other boy. The

configuration of fighting relations again is as in Figure 18. Since each boy stands in the direct or

indirect fighting relation, this sentence conveys the IR relation. This utterance then provides

further support for the claim that the simple construction is compatible with the IR relation.

4.4.3 One-way Weak Reciprocity (OWR)

The definition of One-way Weak Reciprocity in Dalrymple et al. (1998b) is as in (247):

(247) One-way Weak Reciprocity :

|A|≥2 and ∀x∈A ∃y∈A (x≠ y∧ Rxy)

The definition in (247) requires that A have two or more participants and then that for all x in A, there exists at least one y in A such that x and y are non-identical and x stands in the R relation to y.

187

I have found no sentences in the RNC that unambiguously express the weakest version of this relation nor did any of the video stimuli exemplify it. I constructed the following context to establish the weakest version, as in the context, each student participates as the first argument of

the relation, but only Miša and Olja participate as the second argument, and thus this context does

not establish any of the stronger relations as they all require either that each participant also

participates as the second argument, or that every entity stands in the relation to every other

entity.

(248) Context: You are a teacher and you see four of your students, Miša, Kolja, Olja and Tanja come into the room. Miša greets Olja, Kolja greets Miša, Olja greets Miša, and Tanja greets Miša. You’ve been worried about how those three are getting along, so you later tell another teacher:

Studenty pozdorovali-s' segodnja utrom. students.NOM greet.PST.PL-SJA today morning.INST ‘The students greeted each other this morning.’

Since each student participates as the first argument, this context describes a relation that satisfies the truth conditions for OWR. Similarly, the next context sets up a OWR bumping-into relation between the students:

188

(249) Context: You are an elementary school teacher and are watching four of your students, Miša, Griša, Olja and Tanja. play soccer. As you watch, Miša bumps into Griša, Olja bumps into Griša and Tanja, Griša bumps into Tanja and Tanja bumpts into Miša as they jostle for the ball. As an explanation for their state of disarray afterwards, you tell the principal:

M O

G T

Studenty tolkali-s' kogda oni igrali v futbol. students.NOM bump.into.PST.PL-SJA when they.NOM play.PST.PL at soccer.ACC ‘The students bumped into each other when they were playing soccer.’

My consultants also found this sentence acceptable given the context. I therefore propose that the

simple construction is compatible with OWR.

4.4.4 Revised Intermediate Alternative Reciprocity

Recall that I defined the Revised Intermediate Alternative Reciprocity (RIAR) relation as the

following:

(250) Revised Intermediate Alternative Reciprocity:

|A|≥2 and ∀x,y ∈A (x ≠ y → for some sequence w0, . . . wm ≤ x ˄ (x = w0 ∧ (R(w0)(w1) ∨

R(w1)(w0)) ∧ . . .∧ (R(wm-1)(wm) ∨ R(wm)(wm-1))∧ wm = y) ∧ ∃x2 ∃y2 ∃z2 ∈ A (x2≠ y2 ∧ x2≠

z2 ∧ Rx2y2 ∧ Rz2x2))

The RIAR relation requires A to have at least two participants, and then each atomic participant is related directly or indirectly to every other participant either as the first or second

189

argument in the R relation and there is at least one participant which participates as the first and the second argument.

Data suggest that the simple construction is also compatible with the RIAR relation. For example:

(251) Context: You are a teacher who is teaching a course through the internet. As part of the course, your students have to write to each other. Miša writes to Griša, Griša writes to Olja and Olja writes to Tanja. You tell your supervisor:

Studenty perepisyvajut-sja. students.NOM correspond.NPST.3PL-SJA ‘The students are corresponding.’

In this context, there are four students and each student is directly or indirectly related to each

other participant through the corresponding relation. Student 1 participates only as the first

argument (as no one is writing to student 1), student 4 participates only as the second argument

(as student 4 writes to no one) and students 3 and 4 both participate as the first and second

argument. Since there is at least one student (in fact there are two students - 3 and 4) that are both

the first and second participant, they both satisfy the other requirement of RIAR.

The following context is similar. It sets up an OWR kicking relation between cows.

190

(252) Context: You see four cows standing in a row in a field. As you watch, the first cow suddenly becomes agitated and kicks the cow next to it. The second cow kicks the third cow, and the third cow kicks the fourth cow. You say:

Korovy brykali-s' cows.NOM kick.PST.PL-SJA ‘The cows kicked each other.’

Again, in this context there is a sequence of kicking relations such that each cow stands in the

kicking relation directly or indirectly to every other cow, but not as the first and second argument.

Additionally, two cows, namely the second and third cows, both stand in the kicking relation as

the first and second argument. Since my consultants accepted these sentences in the provided

contexts, I hypothesize that the -sja verbs in the simple construction are compatible with RIAR.

4.4.5 Intermediate Alternative Reciprocity (IAR)

In Sections 4.4.1 - 4.4.3, I provided evidence suggesting that the simple construction is

compatible with the relations each other can express. I now investigate whether the simple

construction is compatible with Intermediate Alternative Reciprocity (IAR). As a reminder, this

is the relation posited by Dalrymple et al. (1998b) that I found each other not to be compatible

with.

As I demonstrated in Chapter 2, the relation Dalrymple et al. (1998b) posit for English,

IAR, is too weak to be expressed by each other. I repeat their definition here:

(253) Intermediate Alternative Reciprocity:

∀ ∈ ∈ ∧ ∨ ∧ |A|≥2 and x,y A (x≠ y͢ → for some sequence z0, . . . , zm A (x = z0 (Rz0z1 Rz1z0) … ∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm=y)) 191

IAR requires “each member x of A [to be] be related to every other member y via a chain of R- relations, where we ignore which way the pairs making up the chain are related via the relation

R.” (173).

I found no examples in the RNC that suggest that the simple construction is compatible with IAR. Through elicitation I found evidence suggesting that the simple construction is not compatible with IAR. For example, none of my three consultants accepted the provided utterance in the context in (254):

(254) Context: It’s your birthday and you have invited your friends Maša and Olja over to your apartment. Maša is already sitting on your couch when Olja arrives but Maša doesn’t see or hear Olja. Olja greets Maša, but Maša does not greet her back. You say:

#Maša and Olja pozdorovali-s'. Maša.NOM and Olja.NOM greet.PST.PL-SJA (Intended: ‘Maša and Olja greeted each other.’)

In this context, there are two people who stand in an asymmetric relation with each other. Since none of my consultants considered this utterance to be acceptable, I hypothesize that the -sja verbs in the simple construction in Russian are not compatible with IAR. The following context is similar to that in (254), except that there are now three entities denoted by the subject. The utterance is not acceptable in this context, suggesting that it is not the number of participants in the relation which makes the simple construction unable to convey the IAR relation.

192

(255) Context: It’s your birthday and you have invited your friends Maša, Olja and Tanja over to your apartment. Maša is sitting on your couch when Olja arrives, and Tanja arrives just after Olja. As Olja comes in she greets Maša, and as Tanja comes in she greets Olja.

You say:

#Maša, Olja i Tanja pozdorovali-s'. Maša.NOM Olja.NOM and Tanja.NOM greet.PST.PL-SJA (Intended: ‘Maša, Olja and Tanja greeted each other.’)

I therefore conclude that the simple construction in Russian, like each other, is not compatible

with IAR. I know of no other relations to explore for the -sja verbs. The corpus study turned up

no instances of relations other than the ones already explored for the -sja verbs and furthermore,

previous research into other reciprocal expressions does not turn up any other possible relations

either.

4.4.6 Summary

In sum, I have found that the -sja verbs in the simple construction are compatible with

four truth-conditionally distinct relations: Strong Reciprocity, Intermediate Reciprocity, One-way

Weak Reciprocity and Revised Intermediate Alternative Reciprocity. Like each other, the simple

construction is not compatible with Intermediate Alternative Reciprocity.

4.5 The meaning of the discontinuous reciprocal construction

As demonstrated in the introduction of this chapter, the simple reciprocal construction

and the discontinuous reciprocal construction are not always acceptable in the same contexts. In

this section, I explore the relations the discontinuous construction is and is not compatible with to

better compare the semantics of the two constructions. I first investigate, however, whether the

empirical generalizations suggested by previous literature are relevant for the discontinuous 193

construction in Russian as well to separate any restrictions the construction contributes from the relations themselves.

4.5.1 Semantics of the discontinuous construction in Russian

What first must be accounted for is that two different NPs are relevant and therefore the relations have to be defined on membership in these two sets. I find in Russian that the entities in

the two NPs are not pooled into one set; rather the relation draws the entities from the two sets, and it is required that the relation holds between a member of one set and a member of the other

set.

I propose that the entities are not pooled because relations are never entailed to hold between members of the same set. For example, the context in (256) sets up a relation such that each one of the entities denoted by the subject stand in the greeting relation to each one of the entities denoted by the object.

(256) There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each of the boys greets each girl, and each girl greets each of the boys.

Mal'čiki pozdravali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and the girls greeted each other.’

My consultants judged the sentence acceptable in the context, suggesting that relations between

entities of each set are not entailed.

Siloni (2012), in her discussion of Hebrew, suggests that the hypothesis that relations

between entities denoted by the same set are not entailed is too weak for Hebrew data. She 194

instead argues that the discontinuous construction restricts these relations from holding. At first, some of my data suggest that such a restriction holds for Russian too. Given the context in (257), in which each child stands in a strongly reciprocal greeting relation, regardless of gender, the data

support the hypothesis that entities are restricted from standing in the relation to other members of the same set. My consultants judged this sentence unacceptable given the context. The sentence in (257)b realizes the simple reciprocal construction and was judged acceptable.

(257) Context: There are six children in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each child greets each other child. You say:

a. #Mal'čiki pozdravali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and the girls greeted each other.’

b. Mal'čiki i devočki pozdravali-s'. boys.NOM and girls.NOM greet.PST.PL-SJA ‘The boys and the girls greeted each other.’

In order to express a relation that holds between all entities in the domain of discourse, the data suggest the simple construction must be used instead. Given that the sentence in (257)a is unacceptable, it would lead us to believe that it is not simply that the relations between members of the same set are not entailed, but that there is in fact a constraint against such relations. The following examples suggest the same conclusion using different verbs, ssorit'sja ‘to argue’ in

(258) and (259) and mirit'sja ‘to make peace’ in (260) and (261).

195

(258) Context: You are a teacher and are watching your students and suddenly the boys and girls start yelling at each other- each student yells at every other student. You say:

a. #Mal'čiki ssorjat-sja s devočkami. boys.NOM argue.NPST.PL-SJA with girls.INST (Intended: ‘The boys and girls are arguing.’)

b. Mal'čiki i devočki ssorjat-sja. boys.NOM and girls.NOM argue.NPST.PL-SJA ‘The boys and girls are arguing.’

(259) Context: You are a teacher and are watching your students and suddenly the boys and girls start yelling at each other. The boys are yelling at the girls and the girls are yelling at the boys. You say:

a. Mal'čiki ssorjat'-sja s devočkami. boys.NOM argue.NPST.PL-SJA with girls.INST ‘The boys and girls are arguing.’

b. Mal'čiki i devočki ssorjat'-sja. boys.NOM and girls.NOM argue.NPST.PL-SJA ‘The boys and girls are arguing.’

(260) Context: Your neighbors, Ivan and Marija have two children, Kolja and Tanja. You have recently seen them arguing and then saw that no one in the family was talking to anyone else. Today you wake up and you see them taking a walk together, happily chatting. You say:

a. #Ivan i Marija pomirili-s' s Koljej i Tanjej. Ivan.NOM and Marija.NOM make.peace.PST.PL-SJA with Kolja.INST and Tanja.INST ‘Ivan and Marija made peace with Kolja and Tanja.’

b. Ivan, Marija, Kolja i Tanja pomirili-s'. Ivan.NOM Marija.NOM Kolja.NOM and Tanja.NOM make.peace.PST.PL-SJA ‘Ivan, Marija, Kolja and Tanja made peace.’

196

(261) Context: Your neighbors, Ivan and Marija have two children, Kolja and Tanja. You know that the kids really want a dog, but the parents have said no. Because of this disagreement, the children have been mad at their parents and vice versa. The parents finally gave in and got a dog. You say:

a. Ivan i Marija pomirili-s' s Koljej i Tanjej. Ivan.NOM and Marija.NOM make.peace.PST.PL-SJA with Kolja.INST and Tanja.INST ‘Ivan and Marija made peace with Kolja and Tanja.’

b. Ivan, Marija, Kolja i Tanja pomirili-s'. Ivan.NOM Marija.NOM Kolja.NOM and Tanja.NOM make.peace.PST.PL-SJA ‘Ivan, Marija, Kolja and Tanja made peace.’

In all of these examples, when the context sets up a situation such that the relevant relation holds

between all entities denoted by both the subject and the object, the sentence that realizes the

discontinuous construction is unacceptable, whereas the sentence that realizes the simple

construction was accepted by my consultants. When the relation holds between only the

members denoted by the subject on the one hand and entities denoted by the object on the other,

both the simple and discontinuous constructions are acceptable.

This evidence suggests that the discontinuous construction is not acceptable with contexts

in which the entities denoted by the subject stand in the relevant relation and entities denoted by

object stand in the relevant relation to each other. Consider, however, the following examples:

(262) Context: There is a parent-teacher conference. Ivan and Marija are teachers. Elena and Aleksej are parents. Svetlana is both a teacher and a parent. During the evening, each of the teachers talks to each of the parents. Svetlana talks to both the teachers (Ivan and Marija) about her child as well as the parents (Elena and Aleksej) about their child.

a. Učitelja i roditeli soveščali-s'. teachers.NOM and parents.NOM converse.PST.PL-SJA ‘The teachers and parents conversed with each other.’

b. Učitelja soveščali-s' s roditeljami. teachers.NOM converse.PST.PL-SJA with parents.INST ‘The teachers and parents conversed with each other.’

197

My consultants all judged the sentence in (262) to be acceptable. This is surprising, as the context sets up a situation in which there are two sets, one of teachers and one of parents. One individual, Svetlana, is both a teacher and a parent and thus in both sets. If the restriction were to hold that members of the set denoted by the subject do not stand in the relation to each other, this sentence should be unacceptable to my consultants, now an individual in the set denoted by the subject stands in the conferring relation to two other members in the same set.

Similarly, consider the following context, which depicts a situation in which all the boys stand in the fighting relation to each other:

(263) Context: There are four boys on the playground, all of whom are wrestling around with each other. You say:

Mal'čiki borjut-sja s mal'čikami. boys.NOM fight.NPST.PL-SJA with boys.INST ‘The boys are fighting with each other.’

My consultants judged this sentence as acceptable, even though members of the set denoted by

the subject, i.e. each of the four boys, stands in the fighting relation to each other, and likewise

for the object.

Despite such examples when the set denoted by the object overlaps with that of the

subject, I have found that consultants do not accept sentences that realize the discontinuous

construction in which relations hold between members of the same set when there is no overlap

between the sets. Based on this evidence, I propose that the restriction against members of the

same set standing in a relation to each other is not in fact a semantic restriction. These findings

lead me in my analysis to only require that relations must hold between members of the set 198

denoted by the subject and members of the set denoted by the object; I do not restrict members of

the same set from participating in the relation with each other, despite the unacceptability of

many of the examples above. I instead propose that there is a pragmatic restriction against

entities in the same set standing in the relation to each other that accounts for the unacceptability

of the sentences in (257)a, (258)a, and (260)a. I hypothesize that this restriction is in fact an implicature. While the evidence above supports a claim that the implicature arises when the set

denoted by the subject does not overlap with the set denoted by the object, much more research is necessary to determine if this is in fact true more generally, or if there are other factors that license the implicature. I leave the pragmatic analysis for future research.

The second generalization discussed by the previous literature is that the members of the set denoted by the object are less ‘volitional’ than those denoted by the subject. Since volition is

tied to agentivity, I will assume this means that members of the set denoted by the object do not

have to stand in the relation as the first argument, as the first argument of relations tend to be assigned the role of AGENT. Accordingly, members of the set denoted by the subject do not have to stand in the relation as the second argument. I provide evidence here suggesting that such a claim can be made about the discontinuous construction in Russian as well. For example, the first context in (264) sets up a situation in which there is a symmetrical kissing relation between two individuals. The second context describes a situation in which only one person kisses a second person without reciprocation. The sentence that realizes the discontinuous construction is acceptable in both contexts:

199

(264) Context 1: You see your friend Ivan kiss his girlfriend Marija, and you can see that she kissed him back. You tell a mutual friend:

Context 2: You see your friend Ivan kiss his girlfriend Marija, but she was taken by surprise and didn’t kiss him back. You tell a mutual friend

Ivan poceloval-sja s Mariej Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed./Ivan kissed Marija.’

If the sentence is uttered in the second context, then it expresses one event of the denotation of

Ivan kissing that of Marija - a second event of the denotation of Marija kissing Ivan is not

expressed. In this case, the sentence in (264) has the same meaning as a transitive sentence. The sentence in (264) can be used in at least some of the same contexts as the sentence in (265), which contains a transitive verb (without -sja) that takes an accusative direct object, such as the second context in (264).

(265) Ivan poceloval Mariju. Ivan.NOM kiss.PST.SG Marija.ACC ‘Ivan kissed Marija.’

For an even clearer example that the denotation of the object of the comitative phrase does not

have to participate in the relation as the first argument, all consultants judged the following to be

acceptable:

(266) Ivan obnimaet-sja s zaborom. Ivan.NOM hug.NPST.SG-SJA with fence.INST ‘Ivan is hugging the fence.’

200

This sentence shows that it is acceptable for the denotation of the object of the comitative phrase to be inanimate. Since a fence is inanimate, it cannot be the case that it participated as the first argument of the hugging relation (unless the fence is personified, but all my consultants say this is not a necessary assumption about the nature of the fence). This example then provides evidence for the claim that the discontinuous construction is compatible with an asymmetric relation between participants of the two sets, in which the referent of the subject is involved in the relation as the first argument and the referent of the object is involved as the second argument.

Data from the RNC backs up the existence of this asymmetric interpretation of a sentence containing the discontinuous construction. Consider the sentences in (267) and (268):

(267) Tёtka dognala ego i sprašivaet: - Otčego vy so mnoj woman.NOM chase.PST.SG him.ACC and ask.NPST.3SG why you.NOM with me.INST

ne pozdorovali-s'? NEG greet.PST.PL-SJA ‘The woman chased him down and asked: - why didn’t you greet me?’

(268) Nakonec ja proščaju-s' s zdešneju žizniju i edu k finally I.NOM say.farewell.NPST.1SG-SJA with here.INST life.INST and go.NPST to

polku. regiment.DAT ‘I’m finally saying good-bye to my life here and I’m going to the regiment.’

In both of these examples, it is clear that the object of the comitative phrase does not

participate in the relation as the first argument. The greeting relation expressed by this sentence

involves the denotation of the subject vy ‘you’ as the first argument and the denotation of mnoj

‘me’ as the second argument. Crucially, there is no greeting relation with the individuals denoted

201

by the arguments reversed, i.e. so that the semantic value of mnoj ‘me’ acts as the first argument

and that of vy ‘you’ as the second.

In the sentence in (268), the speaker is saying farewell to an abstract concept, life, which

does not have the ability to bid the speaker adieu itself. Thus, the saying-farewell relation has as

the first argument the denotation of ja ‘I’ and as the second argument the denotation of zdešneju

žizniju ‘life here’. Again, the arguments are not reversed: zdešneju žizniju ‘life here’ cannot be the first argument. This suggests that it is possible for a -sja verb in a discontinuous construction to be compatible with one-way relations when the individuals denoted by the subject are involved in the relation only as the first argument and the individuals denoted by the object are involved in the relation only as the second argument.

These examples also show that the entities denoted by the subject do not have to participate as the second argument. Further evidence suggests that the individuals denoted by the object have to stand in the R relation as the second relation and those denoted by the subject as the first: the discontinuous construction is not compatible with a relation in which the denotation of the object participates only as the first argument and that of the subject only as the second argument. For example:

(269) Context : You see your friend Marija kiss her boyfriend Ivan, but he was taken by surprise and didn’t kiss her back.

#Ivan poceloval-sja s Mariej Ivan.NOM kiss.PST.SG-SJA with Marija.INST (Intended: ‘Marija kissed Ivan.’)

This context sets up a situation in which one individual, Marija, kisses a second individual, Ivan, without reciprocation. The sentence, in which the individual Marija is denoted by the object of 202

the preposition and Ivan by the subject is unacceptable to my consultants. I propose the unacceptability is because the individual that participates as the first argument of the relation is not denoted by the subject of the sentence and the individual that participates as the second argument is not denoted by the object of the preposition. This evidence suggests that in the discontinuous construction, the individuals denoted by the subject must participate as the first argument of the relation and those denoted by the object must participate as the second argument.

What is not clear from the examples above, since they involve NPs that denote singleton sets, is whether the sets are universally quantified over or existentially quantified over. My data suggests that both the subject and the object are universally quantified over. For example, the following context establishes a situation in which each student, denoted by the subject, participates in the haggling relation as the first argument. Each seller, denoted by the object, stands in the relation as the second argument.

(270) Context: You are a teacher of a study abroad program and you bring your ten American students to a small Russian market that only has five stalls. You know the students do not have a lot of money and so they each will buy one thing. Each student goes up to one stall such that there is a student at each stall.

Studenty torgujut-sja s prodavcami. students.NOM haggle.NPST.PL-SJA with sellers.INST ‘The students and sellers are haggling.’

The acceptability of this sentence is not surprising, since each student participates in the haggling relation as the first argument and each seller participates as the second relation. It is thus compatible with an analysis in which the entities are universally or existentially quantified over.

Consider, however, the following context, in which only one student haggles with each of the sellers:

203

(271) Context: You are a teacher of a study abroad program and you bring your ten American students to a small Russian market that only has five stalls. You know the students do not have a lot of money, and that only one student will buy something. This one student goes to each stall and haggles with each seller before buying something.

#Studenty torgujut-sja s prodavcami. students.NOM haggle.NPST.PL-SJA with sellers.INST ‘The students and sellers are haggling.’

If the denotation of the subject were existentially quantified over, this sentence would be acceptable in this context. The unacceptability of this sentence (the same sentence as in (270)) suggests that the denotation of the subject must be universally quantified over.

Similarly, the context in (272) establishes a relation such that each student participates as the first argument of the relation and only one seller as the second argument.

(272) Context: You are a teacher of a study abroad program and you bring your ten American students to a small Russian market that only has five stalls. As soon as the students walk into the market, they all gravitate to one stall and they all spend all their time at the one stall.

#Studenty torgujut-sja s prodavcami. students.NOM haggle.NPST.PL-SJA with sellers.INST ‘The students and sellers are haggling.’

Since this sentence is unacceptable, I suggest that the denotation of the object, just like that of the subject, is universally quantified over. In sum then, I propose that the discontinuous construction is restricted such that every individual denoted by the subject stands in the relevant relation as the first argument to at least one entity denoted by the object and each individual denoted by the object stands in the relation as the second argument to at least one entity denoted by the subject.

In this section, I have established the following generalizations:

204

(273) (i) Members denoted by the same set are not entailed to stand in the relation to each other

(ii) Members denoted by the same set are not restricted from standing in the relation to each other

(iii) The entities denoted by the subject are universally quantified over, such that each entity is entailed to stand in the relation as the first argument of the relation

(iv) The entities denoted by the object are universally quantified over, such that every entity must stand in the relevant relation as the second argument.

4.5.2 Consequences of semantic generalizations for the discontinuous construction

Because I established above that the relations for the discontinuous construction must be defined for two distinct sets, this requires a hypothesis as to how the relations for two sets are related to those for one set that have been explored so far in this dissertation. I propose we derive the relations for two sets such that the relations that were required to hold of each entity in one set now are required to hold between members of set A, that denoted by the subject, and members of set B, that denoted by the object. In the following definitions, A is always the set denoted by the subject, and B is always the set denoted by the object. Since no one has developed these relations in detail before, these derived relations are hypotheses to be tested throughout this chapter.

I repeat the definition of Strong Reciprocity in (274).

(274) Strong Reciprocity:

∀xy ∈A (x≠y → (Rxy ∧ Ryx))

205

SR requires that each individual in A stands in the R relation to each other individual. To derive the SR relation for two sets, I apply a mechanism such that each individual in each set is universally quantified over as both the first and second argument of the R relation, but that this relation only holds between members of different sets:

(275) Strong Reciprocity for two sets:

∀x ∈ A∀y ∈ B (x≠y → Rxy ˄ Ryx)

This definition says that for all x that are elements in A and all y that are elements in B, if x and y are distinct, then x stands in the R relation to y and y stands in the R relation to x. This relation differs from SR in that it stipulates that the entities, represented by x and y, are drawn from two different sets. One of the sets will be denoted by the subject of the sentence and the other will be denoted by the object of the comitative object. Note that the two sets A and B could be identical; nothing in the definition states that they are distinct. I do not want to require, after all, that A and

B are distinct sets, based on the evidence in (262) and (263) above that at times individuals can be members of both sets. The syntax/semantics interface will ensure that A and B are denoted by distinct expressions.

The definition of Intermediate Reciprocity is repeated in (276):

(276) Intermediate Reciprocity:

|A|≥2 and ∀x,y ∈A (x≠ y→ for some sequence z0, … , zm ∈A (x = z0 ∧ Rz0z1 . . . ∧Rzm- 1zm ∧ zm = y))

206

IR is defined for a set with two or more members. For all x and y that are elements in set A, if x and y are distinct, then for some sequence each x and y stands in the R relation as both the first and second argument to the entity before it in the sequence and the entity after it in the sequence.

To define this relation over two sets, it must hold that there is a sequence of entities that encompasses entities both in sets A and B such that each entity in set A stands in the R relation as the first and second argument to an entity in set B, and likewise for each entity in set B. These relations must hold such that there is an indirect relation between each entity in both sets. I suggest the following definition:

(277) Intermediate Reciprocity for two sets:

∀x ∈A ∀y∈B (x ≠ y→ for some sequence z0, …, zm (z0, z2, …∈A z1, z3, ... ∈B (x = z0 ∧

Rz0z1 ∧ … ∧ Rzm- 1zm ∧ zm = y)))

This definition says that for all x that are elements of A and all y that are elements of B, if x and y are distinct, then for some sequence such that each entity in the sequence is in either set A or set

B, each x in A stands in the R relation with an entity y in set B as the first and second argument.

The definition of One-way Weak Reciprocity is in (278):

(278) One-way Weak Reciprocity:

|A|≥2 and ∀x∈A ∃y∈A (x≠ y ∧ Rxy)

The definition requires that A have two or more members and then that for all x in A, there exists at least one y in A such that x and y are non-identical and x stands in the R relation to y. To derive

207

the relation for two sets, I propose that each entity in A must stand in the R relation as the first argument to at least one entity in B, and likewise for each entity in B:

(279) One-way Weak Reciprocity for two sets:

∀x∈A ∃y∈B ∀w∈B ∃z∈A (Rxy ∧ Rwz)

This definition says that all entities x in A stand in the R relation as the first argument to at least one entity y in B and each entity w in B stands in the R relation as the first argument to at least one entity z in A.

The next relation, Revised Intermediate Alternative Reciprocity is defined in (280):

(280) Revised Intermediate Alternative Reciprocity:

|A|≥2 and ∀xy ∈A (x ≠ y→ for some sequence w0, … wm ∈A (x = w0 ∧ (R(w0)(w1) ∨ R(w1)(w0)) ∧ … ∧ (R(wm-1)(wm) ∨ R(wm)(wm-1))∧ wm = y) ∧ ∃x2 ∃y2 ∃z2 ∈ A (Rx2y2 ∧ Rz2x2))

This relation is defined for a set A with two or more members and for all x and y that are elements of A, if x and y are distinct then for some sequence in A, each entity is directly or indirectly

related to each other entity and each entity stands in the R relation as the first or second argument

with another entity in the sequence, and there exists at least one entity in A that stands in the R

relation to another entity as the first argument and to a (possibly different) entity as the second

argument.

To derive RIAR for two sets, I propose that like for IR, each entity in both A and B is

related to each other entity via a direct or indirect relation such that each entity in A stands in the 208

R relation to an entity in B as the first or second argument and each entity in B stands in the R relation to an entity in A, and there is at least one entity in A such that it stands in the R relation

as both the first and second argument, and at least one entity in B such that it stands in the R

relation as both the first and second argument. I propose the following definition for RIAR for

two sets:

(281) Revised Intermediate Alternative Reciprocity for two sets:

∀x ∈A ∀y ∈B(x ≠ y → for some sequence w0, …wm where w0, w2, … ∈ A ∧ w1, w3, …∈ B (x = w0 ∧ (R(w0)(w1) ∨ R(w1)(w0)) ∧ … ∧ (R(wm-1)(wm) ∨ R(wm)(wm-1) ∧ wm = y) ∧ ∃x2 ∈ A ∃y2,z2∈ B (Rx2y2 ∧ Rz2x2) ∧ ∃x3 ∈ B ∃y3 ∃z3∈ A (Rx3y3 ∧ Rz3x3)))

This definition states that for some sequence, each entity in A must stand in the R relation as the first or second argument with at least one entity in B, and at least one entity x2 in A must stand in the R relation as the first argument with at least one entity y2 in B and as the second argument to at least one entity z2 in B, and at least one entity x3 in B must stand in the R relation as the first argument with at least one entity y3 in A and as the second argument with at least one entity x3 in

A. As an example, the relation between the letters in Figure 19 is one of RIAR for two sets, as each element is directly or indirectly related to each other element, such that each letter stands in the relation as the first or second argument to a letter in the opposite set. Furthermore, there is at least one letter in each set (B in set 1 and Y in set 2) that stands as the first argument of the relation to at least one entity in the opposite set and as the second argument of the relation to at least one entity in the opposite set.

209

A X

B Y

C Z

Set 1 Set 2

Figure 19: Depiction of Revised Intermediate Alternative Reciprocity for two sets

I have thus far derived the relations defined over two sets from the relations derived over one set that I found the simple construction to be compatible with. I will also investigate whether the discontinuous construction is also compatible with Intermediate Alternative Reciprocity for two sets. The simple construction was found not to be compatible with IAR, but there is no reason a priori that the discontinuous construction should be compatible with the same set of relations as the simple construction. The definition of Intermediate Alternative Reciprocity is in

(282):

(282) Intermediate Alternative Reciprocity:

|A|≥2 and ∀x,y ∈A (x≠ y→ for some sequence z0, …, zm ∈A (x = z0 ∧ (Rz0z1 ∨ Rz1z0) ∧

… ∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm=y))

IAR is defined for a set A with two or more members. It is true if and only if for all x and y that are elements in A, if x and y are distinct, then for some sequence of elements z0, … , zm in A, such

210

that x is the start of the sequence as either the first or second argument of R and y is the end of the sequence as either the first or second argument of R.

Deriving this for two sets, each entity in A must stand in the R relation via a chain of direct or indirect relations to each entity in B, and likewise for each entity in B. I propose the following definition:

(283) Intermediate Alternative Reciprocity for two sets:

∀x ∈ A ∀y ∈ B (x≠ y→ for some sequence z0, …, zm(z0, z2, … ∈A z1, z3, … ∈B (x = z0

∧(Rz0z1 ∨ Rz1z0) ∧ … ∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm=y)))

This definition says that for all x in A and all y in B, if x and y are distinct, then for some sequence of entities in A and B, all entities in A must stand in the R relation as the first or second argument via a chain to every entity in B.

When exploring these relations, it will also be necessary to consider the generalization I established that each member of the set denoted by the subject must stand in the relevant relation as the first argument and at least one member of the set denoted by the object must stand in the relevant relation as the second argument. Throughout the remainder of this chapter, I will call this restriction the Subject/Object Restriction. This restriction can be formalized as in (284):

(284) Subject/Object Restriction:

∀x∈ A ∃y ∈ B∀w∈ B ∃z ∈ A (Rxy ˄ Rzw)

Consider the relations for two sets I have so far posited:

211

(285) Strong Reciprocity for two sets:

∀x ∈ A∀y ∈ B (x≠y → Rxy ˄ Ryx)

(286) Intermediate Reciprocity for two sets:

∀x ∈A ∀y∈B (x ≠ y→ for some sequence z0, …, zm (z0, z2, …∈A z1, z3, ... ∈B (x = z0 ∧ Rz0z1 ∧ … ∧ Rzm- 1zm ∧ zm = y)))

(287) One-way Weak Reciprocity for two sets:

∀x∈A ∃y∈B ∀w∈B ∃z∈A (Rxy ∧ Rwz)

(288) Revised Intermediate Alternative Reciprocity for two sets:

∀x ∈ A ∀y ∈ B(x ≠ y → for some sequence w0, …wm where w0, w2, … ∈ A ∧ w1, w3, …∈ B (x = w0 ∧ (R(w0)(w1) ∨ R(w1)(w0)) ∧ … ∧ (R(wm-1)(wm) ∨ R(wm)(wm-1) ∧ wm = y) ∧ ∃x2 ∈ A ∃y2,z2∈ B (Rx2y2 ∧ Rz2x2) ∧ ∃x3 ∈ B ∃y3 ∃z3∈ A (Rx3y3 ∧ Rz3x3))))

(289) Intermediate Alternative Reciprocity for two sets:

∀x ∈ A ∀y ∈ B (x≠ y→ for some sequence z0, …, zm(z0, z2, … ∈ A z1, z3, … ∈ B (x = z0

∧ (Rz0z1 ∨ Rz1z0) ∧ … ∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm=y)))

Note that imposing the Subject/Object Restriction in (284) on discontinuous reciprocal sentences that satisfy SR and IR reciprocal relations does not change the truth conditions of these sentences, as the restriction is weaker in meaning than the relations themselves. The restriction does affect the truth conditions of the relations OWR, RIAR and IAR, however. Although OWR for two sets requires that each entity participates in the relation as the first argument, it does not universally quantify over the object such that each entity denoted by the object stands in the relation as the 212

second argument. RIAR and IAR involve a disjunct, such that the entities denoted by the subject must stand in the relation as the first or second argument and likewise for the set denoted by the object. These definitions therefore define a relation in which the entities in the set denoted by the subject are not entailed to stand in the relation as the first argument and those denoted by the object as the second argument. Adding the restriction therefore does change the truth-conditions of these relations. In what follows, I explore which relations the discontinuous construction is compatible with, but I take into consideration the restriction in (284) above. I will therefore only present contexts in which the relations between participants would not violate this restriction.

4.5.3 The discontinuous construction and two-set relations

I hypothesize that the discontinuous construction will be compatible with the derived version of the same relations that the simple construction was compatible with when including the

Transitive Restriction in (284) above. I explore this hypothesis throughout the remainder of

Section 4.5.

4.5.3.1 Strong Reciprocity for two sets

The first relation to be tested is Strong Reciprocity for two sets, defined in (290), repeated from (275):

(290) ∀x ∈ A∀y ∈ B (x≠y → Rxy ˄ Ryx)

213

This definition says that for all x in A and all y in B, if x and y are distinct, then x stands in the R

relation to y and y stands in the R relation to x. No relation between members of the same set is

entailed. Consider the context and utterance in (291):

(291) There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each of the boys greets each girl, and each girl greets each of the boys.

Mal'čiki pozdravali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and the girls greeted each other.’

In this context, each member of the set denoted by the subject (i.e. the boys) stands in a

symmetric greeting relation with every member of the set denoted by the comitative object (i.e.

the girls). The relation between the participants is strong reciprocity for subsets since each

participant stands in the greeting relation to every other participant in the other set. I therefore

conclude that the -sja verbs in the discontinuous construction are compatible with Strong

Reciprocity when two sets are involved.

4.5.3.2 Intermediate Reciprocity for two sets

The definition I provided for IR for two sets is repeated in (277):

(292) Intermediate Reciprocity for two sets:

∀x ∈A ∀y∈B (x ≠ y→ for some sequence z0, …, zm (z0, z2, …∈A z1, z3, ... ∈B (x = z0 ∧ Rz0z1 ∧ … ∧ Rzm- 1zm ∧ zm = y)))

214

B, each x in A stands in the R relation with an entity y in set B as the first and second argument via a chain of direct or indirect relations. The following context sets up an IR for two sets greeting relation between the boys and the girls.

(293) There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Miša says hello to Olja, Olja says hello to Kolja, Kolja says hello to Tanja, Tanja says hello to Mitja, Mitja says hello to Galja, and Galja says hello to Miša.

Miša Olja

Kolja Tanja

Mitja Galja

Mal'čiki pozdorovali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and girls greeted each other.’

Since my consultants judged this sentence as acceptable in the context, I propose that the

discontinuous construction in Russian is compatible with the IR relation for two sets.

4.5.3.3 One-Way Weak Reciprocity for two sets

I proposed the following definition for OWR for two sets:

(294) One-way Weak Reciprocity for two sets:

∀x∈A ∃y∈B ∀w∈B ∃z∈A (Rxy ∧ Rwz)

215

According to this definition, OWR for two sets holds when all x in A stand in the R relation to at least one y in B and all w in B stand in the R relation to at least one entity z in A.OWR is one of those relations whose truth conditions change when the Subject/Object Restriction repeated in

(295) is added.

(295) Subject/Object Restriction:

∀x∈ A ∃y ∈ B∀w∈ B ∃z ∈ A (Rxy ˄ Rzw)

With this restriction, a sentence will be compatible with OWR when each entity in set A participates in the relation as the first argument to at least one individual in set B, and each entity in set B stands in the relation as the first argument to at least one individual in set B.

Furthermore, each individual in B must stand in the relation as the second argument to at least one individual in set A.

The context in (296) establishes a OWR relation for two sets:

(296) Context: You are a teacher and you see four of your students, Miša, Kolja (both boys), Olja and Tanja (girls) come into the room. Miša greets Olja and Tanja, Kolja greets Olja, Olja greets Miša, and Tanja greets Miša. You’ve been worried about how those three are getting along, so you later tell another teacher:

Devočki pozdorovali-s' s mal'čikami segodnja utrom. girls.NOM greet.PST.PL-SJA with boys.inst today morning.INST ‘The girls and boys greeted each other this morning.’

In this context, each boy greets a girl, each girl greets a boy, and each girl is greeted by a boy.

This context exemplifies OWR for two sets because not every entity is greeted (no one greets

216

Kolja), and no relation holds between the girls, and no relation holds between the boys. Since my consultants judged this sentence as acceptable in this context, I propose that the discontinuous

construction is compatible with OWR for two sets.

4.5.3.4 Revised Intermediate Alternative Reciprocity

I repeat the definition of Revised Intermediate Alternative Reciprocity in (297):

(297) ∀x ∈A ∀y ∈B(x ≠ y → for some sequence w0, …wm where w0, w2, … ∈ A ∧ w1, w3, …∈ B (x = w0 ∧ (R(w0)(w1) ∨ R(w1)(w0)) ∧ … ∧ (R(wm-1)(wm) ∨ R(wm)(wm-1) ∧ wm = y) ∧ ∃x2 ∈ A ∃y2,z2∈ B (Rx2y2 ∧ Rz2x2) ∧ ∃x3 ∈ B ∃y3 ∃z3∈ A (Rx3y3 ∧ Rz3x3)))

Revised Intermediate Alternative Reciprocity is another one of the relations whose truth

conditions are weaker than those of the Subject/Object Restriction.

With this restriction, a sentence will be compatible with RIAR for two sets when there is a sequence such that each entity in A stands in the R relation as the first or second argument via a chain of direct or indirect relations with every entity in B, and likewise for every entity in B. It must also hold that there is at least one entity in A such that this entity stands in the R relation as the first argument to at least one entity in B and as the second argument to at least one entity in B and there exists at least one entity in B that stands in the R relation as the first argument to at least on entity in A and as the second argument to at least one entity in A. Lastly, the Subject/Object

Restriction requires each entity in A to stand as the first argument to at least one entity in B, and each entity in B to stand as the second argument to at least one entity in A.

The following context sets up a RIAR relation for two sets including the Subject/Object

Restriction with the visualization in Figure 20 below:

217

(298) There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Miša says hello to Olja, Kolja says hello to Tanja, Tanja says hello to Mitja, and Mitja says hello to Galja.

Mal'čiki pozdorovali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and girls greeted each other.’

Miša Olja

Kolja Tanja

Mitja Galja

Figure 20: Visualization of a Revised Intermediate Alternative Reciprocity for two sets relation

In this context, there is a greeting chain from Miša to Galja. Each boy stands in the R relation as

either the first or second argument to at least one girl, and likewise for each girl. There is at least

one boy that stands in the R relation as the first and second argument to a girl, and likewise for at

least one girl. Lastly, each boy stands in the R relation as the first argument to at least one girl,

thus satisfying the restriction. Since the sentence was judged as acceptable in the context, I

propose that the discontinuous construction is compatible with RIAR for two sets.

4.5.3.5 Intermediate Alternative Reciprocity

In Section 4.4.5, I provided evidence suggesting that the simple construction, like each other, is

not compatible with IAR. This being the case, the prediction is that the discontinuous

construction would also not be compatible with IAR. My data suggest, however, that the

discontinuous construction is compatible with IAR for two sets.

218

(299) Intermediate Alternative Reciprocity for two sets:

∀x ∈ A ∀y ∈ B (x≠ y→ for some sequence z0, …, zm(z0, z2, … ∈ A z1, z3, … ∈ B (x = z0 ∧ (Rz0z1 ∨ Rz1z0) ∧ … ∧ (Rzm-1zm ∨ Rzmzm-1) ∧ zm=y)))

This definition says that for all x that are elements of A and all y elements that are elements of B, if x and y are distinct, then for some sequence of entities in A and B, all entities x in A must stand in the R relation via a chain of direct or indirect relations as the first or second argument to each other entity in B. Recall that the truth conditions of IAR, like RIAR, change when the

Subject/Object Restriction is added.

Consider the following context and sentence:

(300) Context: You see your friend Ivan kiss his girlfriend Marija, but she was taken by surprise and didn’t kiss him back.

Ivan poceloval-sja s Mariej Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed./Ivan kissed Marija.’

Similarly, the example in (301) is repeated from (266) above:

(301) Ivan obnimaet-sja s zaborom. Ivan.NOM hug.3.SG.NPST-SJA with fence.INST ‘Ivan is hugging the fence.’

Each of these sentences is compatible with a relation in which the entity denoted by the subject stands in the relevant relation as the first argument to at least one entity denoted by the object,

219

and each object stands in the relevant relation as the second argument, thus fulfilling the

definition of IAR for two sets with the added restriction.

Note that the sentences in (300) and (301) are not compatible with any of the other relations. Even RIAR is too strong to capture the truth conditions of these sentences as there is not an entity in either the set denoted by the subject or that denoted by the object that stands in the

given relation as both the first and the second relation.

Similarly, my data suggest that sentences that realize subjects and objects that denote pluralities are also compatible with IAR for two sets:

(302) There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Miša says hello to Olja, Kolja says hello to Tanja, and Mitja says hello to Galja.

Mal'čiki pozdorovali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and girls greeted each other.’

Miša Olja

Kolja Tanja

Mitja Galja

Figure 21: Visualization of an Intermediate Alternative Reciprocity for two sets relation

In this context, each boy greets one girl and each girl is greeted by one boy. Once again, the

relation depicted here is IAR since each entity participates in the greeting relation as the first or

second argument, but never as both. 220

I suggest therefore that the discontinuous construction is compatible with IAR for two sets, which is surprising considering that the simple construction is not compatible with this

relation, meaning that this one relation cannot be derived from the simple construction. It is

furthermore interesting because each other is also not compatible with IAR, and as I will show in

Chapter 5, neither is drug druga. I will discuss the implications of this in Section 4.7.

4.5.4 Summary

In this section, I have provided evidence suggesting that the simple construction is

compatible with the same six reciprocal relations as each other: Strong Reciprocity, Intermediate

Reciprocity, One-Way Weak Reciprocity and Revised Intermediate Alternative Reciprocity.

When investigating the meaning of the discontinuous construction I established two empirical generalizations that have implication for the meaning of this construction: (i) the relations must be defined over two sets and (ii) each entity denoted by the subject must stand in the relation as the first argument to at least one entity in the set denoted by the object and each entity denoted by the subject must stand in the relation as the second argument to at least one entity in the set denoted by the subject. Taking into consideration these two generalizations, I redefined the relations for two sets and provided data suggesting that the discontinuous construction is compatible with Strong Reciprocity, Intermediate Reciprocity, One-Way Weak Reciprocity,

Revised Intermediate Alternative Reciprocity, and Intermediate Alternative Reciprocity.

4.6 A semantic analysis of the -sja verbs

I began this chapter with a question: how does the difference in meaning between the

simple and discontinuous constructions arise? In Sections 4.4 and 4.5 I explored in detail the

221

meaning of each construction, and I now return to this question better equipped to attempt an answer since it is clearer what the semantics of each construction are and how they differ.

I propose that the difference arises partly from the syntactic construction and partly from the verb itself. I develop this analysis in this section.

4.6.1 An analysis of the simple construction

I first develop an analysis of the simple construction. To do so, I provide a lexical entry for each expression in the simple construction to build up to a compositional analysis of such sentences.

4.6.1.1 Semantics of the subject noun phrase in the simple construction

As I introduced in Section 4.1.1, the subject NP that combines with the simple reciprocal

can be plural, syntactically singular but semantically plural, coordinated with the conjunct i ‘and’,

or coordinated with the conjunct s ‘with’. I assume that the type of subject NP does not have any

influence on the semantics of the simple reciprocal construction. Although I assume this based

on the fact that I have found no difference, this assumption requires more discussion because it

has been argued in the literature that coordination with i differs from coordination with s. Here I

provide the arguments for such a claim and then argue, following Dalrymple at al. (1998b), that

any differences in interpretation are pragmatic, not semantic. Although I recognize that this

means that without a pragmatic analysis, we are missing a part of the picture, I will not attempt to

formulate a pragmatic analysis in this dissertation for this expression. Determining what

pragmatic factors are at work here should prove an interesting study for future research.

222

McNally (1993) proposes that the meaning of comitative coordination (with the form of

NP s ‘with’ NP) differs from the denotation of “ordinary” coordination (with the form of NP i

‘and’ NP) in that comitative coordination denotes a group, whereas ordinary coordination denotes

a sum. She provides the following example:

(303) a. Anna i Maša vyigrali 1000 rublej. (McNally 1993: 373) Anna.NOM and Maša.NOM win.PL.PST 1000 rubles.GEN i. ‘Anna and Maša won 1000 rubles each.’ (2000 total) ii. ‘Anna and Maša won 1000 together.’ (1000 total)

b. Anna s Mašej vyigrali 1000 rublej. Anna.NOM with Maša.INST win.PL.PST 1000 rubles.GEN i. ‘Anna and Maša won 1000 rubles together.’ (1000 total)

In (303)a, the sentence realizes ordinary coordination, and has two different interpretations, one where the denotation of the complex subject is interpreted as distributed and one where it is interpreted as collective. The sentence in (303)b realizes comitative coordination and has only one interpretation, namely that the denotation of the coordination is interpreted as a group acting together. McNally explains this fact by arguing that the conjunct i ‘and’ creates a complex NP by summing the individuals in the NPs it coordinates. The denotation of the ordinary coordination in

(303)a is a⊕m. This sentence has two interpretations because the optional distributive operator can be applied to the predicate, which entails that each individual in the sum is in the extension of the predicate. Alternatively, the sentence could be interpreted without the distributive operator, in which case the denotation of the coordinated NP is interpreted as a group.

McNally argues that comitative coordination with s ‘with’, in contrast, exclusively denotes an “impure atom” (Link 1983). An impure atom is the output of a membership function

223

that maps sums of individuals into groups. Link notates this function with angular brackets ⟨⟩ .

Under this analysis, the denotation of the comitative coordination in (303)b is ⟨a⊕m⟩ and is

analyzed as an atom (even though it is “impure”). Because the distributive operator only

distributes over atoms in a sum, it cannot distribute the two individuals that are part of the atom,

and therefore the distributed interpretation is not available.

Dalrymple et al. (1998a) argue, on the other hand, that i ‘and’ and s ‘with’ when used as

coordinated conjunctions have the same denotation. Under their analysis both types of

coordination denote sums and not group-formation. They analyze the differences in meaning

between the two sentences in (303) as arising due to pragmatics.

They start by demonstrating that the coordinated NPs do not have such different

interpretations as McNally claims. They argue that for some speakers, the sentence in (303)a

above that realizes the conjunct i ‘and’ only has the second (collective) interpretation and not the

first (distributed) interpretation. Therefore, the sentences in (303) both have the same

interpretation for these speakers. Furthermore, they present the following example, in which the

sentence realizes a comitative coordinated phrase and has both a collective and a distributive

reading:

(304) Vasja s Petej pomogali pensioneram. (Dalrymple et al., 1998b: 627) Vasja.NOM with Petja.INST help.PST.PL pensioners.DAT ‘Vasja with Petja helped the pensioners.’

i. group reading (more prominent) ii. distributive reading (less prominent but accessible reading)

Based on these examples, I follow Dalrymple et al.’s (1998a) claim that the two coordinated phrases have the same truth conditional meaning. Therefore, I assume that the discussion of the 224

relations that the simple construction is compatible with applies to sentences with both types of subject NPs.

4.6.1.2 Compositional analysis of the simple construction

Given a sentence like that in (305) containing a simple construction, I assume that the denotations of the proper names Ivan and Marija do not change and thus are of type e and simply denote the individual in the world. The conjunction i ‘and’ I assume has the same meaning as

English and when used as a NP conjunction27, and therefore the denotation of the coordinated NP

in (305) is the sum of the individuals denoted by the NPs Ivan and Marija.

(305) Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed each other.’

Left to be determined is the meaning of the verb pocelovat'-sja ‘to kiss.’ Recall that in

Chapter 2 I proposed that a -sja verb has the following semantics when it conveys a relation of

Strong Reciprocity:

(306) kiss_SR: pocelovat'sja ‘to kiss each other’: N\S’: λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e”˄ kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]

The relation that this lexical entry entails is SR, but any of the other relations that the simple construction is compatible with could be plugged into this lexical entry. Now armed with

27 This assumption is something of a simplification, because, as is well known, there is a conjunct a in Russian that shares the functions of the English conjuncts and and but. I ignore this complication as I do not believe this affects the interpretation of coordinated NPs in Russian in a way that affects my analysis. 225

definitions of all of the expressions of the sentence in (305) that realizes the simple construction,

the translation of the sentence in (305) is:

(307) ∃e[∀a∀b(a ≤ (i⊕ m) ˄ b ≤ (i⊕m) ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ τ(e)≪ now)]

This translations says that the sentence in (305) is predicted to be true if and only if for every entity a and b that are parts of the denotation of the subject, there are subevents e’ and e” that are part of an event e such that e’ is an event of a standing in the kiss relation to b, and e” is an event of b standing in the kiss relation to a, and the event e happened before now. This definition then predicts that this sentence will be true in one of the contexts in which my consultants judged it to be acceptable, namely in a context in which the relation between the participants is strongly reciprocal. I therefore assume that the definition of the verb pocelovat'sja ‘to kiss’ is as in (306).

Each of the other reciprocal relations can also be plugged into the lexical entry of any of the reciprocal -sja verbs. The lexical entries of these verbs are in (308) - (310):

(308) kiss_IR: pocelovat'sja‘to kiss each other’: N\S’: λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → for some sequence z0, … , zm ≤ x (∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ e’≠ e”˄ (a = z0 ∧ kiss’(z0)(z1)(e’) ∧ … ∧ R(zm- 1)(zm)(e”) ∧ zm=b))))]

(309) kiss_OWR: pocelovat'sja ‘to kiss each other’: N\S’: λxλev[∀a∃b(a ≤ x ˄ b ≤ x ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ kiss’(a)(b)(e’)))]

226

(310) kiss_RIAR: pocelovat'sja ‘to kiss each other’: N\S’: λxλev[∀a∀b(a ≤ x ˄ b ≤ x ˄ a ≠ b → for some sequence w0, … wm ≤ x (∃e0, … , en ≤ ev (a = z0 ∧ kiss’(w0)(w1)(e0) ∨ kiss’(w1)(w0)(e0) ∧ … ∧ (kiss’(wm-1)(wm)(en) ∨ kiss’ (wm)(wm-1)(en))∧ wm = b) ∧ ∃c ∃d ∃g (c ≤ x ˄ d ≤ x ˄ g ≤ x ˄ c≠ d ∧ c ≠ g ∧ ∃e’∃e”(e’ ≤ ev ∧ e’ ≤ ev ∧ e’≠ e”˄ kiss’(c)(d)(e’) ∧ kiss’(g)(c)(e”)))))]

These are the four lexical entries I propose for each reciprocal -sja verb in Russian that are realized in the simple reciprocal construction. I now turn to an analysis of the discontinuous construction.

4.6.2 An analysis of the discontinuous construction

In this Section I provide an analysis of the discontinuous construction that captures the empirical generalizations established in Section 4.5.1 and that accounts for the compatibility of this construction with the five different reciprocal relations.

I first discuss the preposition s to determine what its syntax and semantics are. I then

argue that the fact that entities in the set denoted by the subject are entailed to stand in the relevant relation to those denoted by the object I propose will fall out naturally from the syntax- semantics interface. I will suggest two ways to account for the second generalization and in the end will tentatively support an option that requires the discontinuous construction to realize a

verb with a different lexical entry than that in the simple construction.

I then provide an analysis to capture the first generalization and will ignore the second

generalization temporarily. I return to the issue of the Subject/Object Restriction in Section

4.6.2.4.

227

4.6.2.1 Basic properties of comitative phrases in Russian

The literature typically recognizes three functions of the comitative phrase in Russian

(Dalrymple et al. 1998a, Dyła and Feldman 2003, Vassilieva and Larson 2005). Comitative phrases function as conjuncts (as in the simple construction), verbal adjuncts and nominal adjuncts, and sentences truth-conditionally differ depending on which type of phrase is realized.

Vassilieva and Larson (2005) provide the following example:

(311) Mal'čiki s devočkami tancevali. (Vassilieva and Larson 2005; 103) boys.NOM with girls.INST dance.PST.PL

a. ‘The boys and the girls danced.’ Comitative conjunct b. ‘The boys danced with the girls.’ Verbal adjunct c. ‘The boys who had/were with girls danced.’ Nominal adjunct (cf. ‘The boys with the blond hair danced.’)

They note that comitative adjuncts can also attach to the right of the VP, as in (312), in which

case, they only have one interpretation as a verbal adjunct:

(312) Mal'čiki tancevali s devočkami. boys.NOM dance.PST.PL with girls.INST ‘The boys danced with the girls.’

Vassilieva and Larson provide a number of diagnostics for determining the syntactic type of comitative phrase. First they demonstrate that coordinating comitative phrases differ in agreement patterns from either the nominal or verbal adjuncts. Vassilieva and Larson claim that in Russian, when the conjunction i ‘and’ coordinates two NPs, regardless of whether the NPs are singular or

228

plural, the coordinated NP requires plural agreement on the verb, as demonstrated by the two sentences in (313).

(313) a. Mal'čik i koška ušli domoj. (Vassilieva and Larson 2005: 104) boy.SG.NOM and cat.SG.NOM go.PL.PST home ‘The boy and cat went home.’

b. *Mal'čik i koška ušёl domoj. (Vassilieva and Larson 2005: 104) boy.SG.NOM and cat.SG.NOM go.SG.PST home (Intended: ‘The boy and cat went home.’)

Vassilieva and Larson argue that when the comitative phrase is a conjunct, it also requires plural

agreement on the verb, as in (314):

(314) Mal'čik s koškoj ušli domoj. (Vassilieva and Larson 2005: 104) boy.SG.NOM with cat.SG.INST go.PL.PST home ‘The boy and cat went home.’

When the comitative phrase is a nominal (315)a or verbal adjunct (315)b, however, the verb will have singular agreement if the subject is singular:

(315) a. [Mal'čik s koškoj]NP ušёl domoj. (Vassilieva and Larson 2005: 104) boy.SG.NOM with cat.SG.INST go.SG.PST home ‘The boy with the cat went home.’ (‘but the boy with the dog stayed.’)

b. Mal'čik [s koškoj ušёl domoj]VP. (Vassilieva and Larson 2005: 104) boy.SG.NOM with cat.SG.INST go.SG.PST home ‘The boy and cat went home.’

229

As I discussed above in Section 4.1.1, however, in Russian verbs can also agree with the

nearest conjunct. Because of this, the diagnostic that Vassilieva and Larson provide must be

tweaked to account for such agreement patterns. When both NPs are singular, the diagnostic

works in the following manner: if the verbal agreement is plural, then the two NPs are a

coordinated NP that is the subject of the verb. On the other hand, if the verb has singular

agreement, it does not necessarily indicate that the NPs make up a coordinated phrase. In this

case, we can only tell whether it is a coordinated phrase if the two NPs are of different genders

and the verb is the past tense. Since Russian verbs inflect for gender in the past tense, we can

determine from the gender agreement whether the verb is agreeing with the nearest conjunct or

with another conjunct. If it agrees with the nearest conjunct, the comitative phrase is not a

conjunct in a coordinated phrase. If, on the other hand, it agrees with the first NP in the string,

we can assume that the comitative phrase is not a conjunct.

I furthermore assume that when the comitative phrase has the status of a conjunct, then

the denotation of the host NP and that of the object both participate in the event expressed by the verb since the denotation of the object is in fact part of the subject. On the other hand, when the comitative phrase has the status of an adjunct, only the denotation of the subject participates in the event expressed by the verb - not the denotation of the object. The following example provides evidence that the difference in agreement patterns correlates with a difference in the semantics as well. The context sets up a situation in which the boy owns a cat, but the cat is not with him as he leaves. Crucially, a sentence that realizes a comitative conjunct should be unacceptable in this context, as the denotation of the object of the conjunct is required to participate in the event expressed by the sentence.

230

(316) Context: You bring your young child to the playground so often that you recognize most of the other children, but you don’t necessarily know their names. You know that one of the boys owns a cat. The cat is not with him today. Another parent comes up to you to ask you which child just left the playground. You say:

a. #Mal'čik s koškoj ušli domoj. boy.SG.NOM with cat.SG.INST go.PL.PST home ‘The boy and cat went home.’

b. Mal'čik s koškoj ušёl domoj. boy.SG.NOM with cat.SG.INST go.SG.PST.MASC home ‘The boy with the cat went home.’

Since the verb in (316)a has plural agreement, we expect the sentence in (316)a to express either

two events (one of the denotation of the boy going home, and one of the denotation of the cat

going home), or one event of a plural entity going home. Both interpretations should be

unacceptable given the context and indeed my consultants confirmed this unacceptability. Since

my consultants judged the sentence in (316)a as unacceptable, I propose that the comitative

phrase in this sentence is a comitative conjunct. The sentence in (316)b we expect to denote a

single event - that of the denotation of the subject, mal'čik ‘the boy’, going home, since (i) there is

singular agreement on the verb and so the agreement of the verb cannot be agreeing with both

NPs or (ii) there is singular, masculine agreement on the verb, and so this cannot be an example

of nearest conjunct agreement. Because of this agreement pattern, the denotation of the object,

the cat, should not participate in this event. My consultants judged this sentence acceptable given

the context, suggesting that the comitative phrase in (316)b is indeed an adjunct. The only

syntactic difference between the two sentences is that the verb in (316)a has plural agreement

while that in (316)b has singular agreement. This example therefore provides semantic evidence

to support the diagnostic concerning agreement patterns and comitative phrases.

231

As a second diagnostic to distinguish between comitative conjuncts and adjuncts, Russian does not allow discontinuous conjuncts, since they form a constituent. For example, the sentence in (317) realizes the conjunction i ‘and’ such that it and the second NP are realized post verbally, and my consultants judged it as unacceptable. I hypothesize it is unacceptable because the conjunct is not realized adjacent to the subject.

(317) *Ivan ušli domoj i koška. Ivan.NOM go.PL.PST home and cat.NOM (Intended: ‘Ivan and the cat went home.’)

Note that the sentence in (318) features s and not i as the head of the conjunct. This sentence that realizes the comitative conjunct (as forced by the plural agreement on the verb) is also unacceptable:

(318) *Ivan ušli domoj s koškoj. Ivan.NOM go.PL.PST home with cat.INST (Intended: ‘Ivan and the cat went home.’)

The comitative phrase is realized discontinuous from the host NP, just as the conjunct is in (317).

It is not acceptable and therefore provides a second diagnostic for determining whether a

comitative phrase is a conjunct or adjunct.

Based on these diagnostics, I will now establish that the comitative phrase in the

discontinuous construction is not a conjunct. The sentence in (319)a realizes the discontinuous

construction with a singular noun, Ivan, as the subject of the sentence and the singular noun

232

Mariej as the object of the comitative phrase. It is clear that the comitative phrase is an adjunct

immediately, since it is realized post-verbally.

(319) Ivan poceloval-sja s Mariej. Ivan.NOM kiss.SG.PST-SJA with Marija.INST ‘Ivan and Marija kissed.’

Similarly, consider the two sentences in (320) in which the comitative phrase is realized continuous with the subject. In these sentences it is not possible to use the adjacency diagnostic to determine whether the comitative phrase is a conjunct - we must depend instead solely on agreement. The verb in the sentence in (320)a realizes a verb with singular masculine agreement.

This agreement suggests that the denotation of the object of the comitative phrase, Marija, does not participate in the event expressed by the verb as the AGENT, because (i) the verb does not

have plural agreement and (ii) the verb does not have singular feminine agreement, which would

have to be the case for nearest conjunct agreement. The comitative phrase is therefore not a

conjunct.

(320) a. Ivan s Mariej poceloval-sja. Ivan.NOM with Marija.INST kiss.SG.PST-SJA ‘Ivan and Marija kissed.’

b. Ivan s Mariej pocelovali-s'. Ivan.NOM with Marija.INST kiss.SG.PST-SJA ‘Ivan and Marija kissed.’

The verb in the sentence in (320)b, however, has plural agreement, suggesting that both the denotation of Ivan and that of Marija participate in the event denoted by the verb and thus that the

233

comitative phrase is a conjunct. Note then that the sentence in (320)a exemplifies the discontinuous construction, and that in (320)b the simple construction.

Although I have provided evidence suggesting that the comitative phrase in the discontinuous construction is not a conjunct, I propose that it is not an adjunct either. In the end I will argue that it is instead a complement to the verb and thus a type of comitative phrase not discussed in the literature. I will first provide characteristics of nominal and verbal adjuncts to show how the comitative phrase in the discontinuous construction differs from those already analyzed, and in the next section I will use test the comitative phrases for complementhood.

As a diagnostic to distinguish nominal and verbal adjuncts, Vassilieva and Larson argue that verbal adjuncts can be realized to the right of the verb, while nominal adjuncts cannot be without having their interpretation changed. In the example in (321)a, the comitative phrase is realized to the left of the verb and it has one of two interpretations, one of which arises when the adjunct is interpreted as a NP and one when as a VP. The sentence in (321)b, on the other hand, realizes the comitative phrase to the right of the verb, and this sentence only has one interpretation, namely that of a verbal adjunct. The % symbol indicates that the interpretation is not available.

(321) a. Mal'čik s kotёnkom idёt domoj. (Vassilieva and Larson 2005: 106) boy.NOM with kitten.INST go.3.SG.NPST home i. ‘The boy with the kitten is going home.’ (NP) ii. ‘The boy is going home with the kitten.’ (VP)

b. Mal'čik idёt s kotёnkom domoj. (Vassilieva and Larson 2005: 104) boy.NOM go.3.SG.NPST with kitten.INST home i. %‘The boy with the kitten is going home.’ (NP) ii. ‘The boy is going home with the kitten.’ (VP)

234

When the comitative phrase is realized post verbally, the interpretation of the sentence in which the comitative phrase is a nominal adjunct is not available. Vassilieva and Larson therefore hold that nominal adjuncts cannot be realized disjoint from the phrase they modify.

One of the defining characteristics of the discontinuous construction, as already discussed, is that the comitative phrase can be, and often is, realized post-verbally. If we assume that nominal adjuncts must be realized right adjacent to the subject, then the comitative phrase of the discontinuous construction is not a nominal adjunct.

Although the comitative phrase patterns in the same way as the verbal adjuncts discussed by Vassilieva and Larson, I present data in the next section suggesting they are not actually adjuncts and are instead complements.

4.6.2.2 The comitative phrase is a complement

The diagnostic that is most widely used to test whether a phrase is an adjunct or complement is “obligatoriness” (Dowty 2000, Needham and Toivonen 2011). Complements are obligatory while adjuncts are optional. It has been pointed out that this is not a perfect test since it has been shown that some complements are optional and some adjuncts are obligatory

(Jackendoff 1990, Dowty 2000, Goldberg and Ackerman 2001, Culicover 2009). This test is therefore questionable and furthermore is problematic for my data. It is not possible to determine whether the comitative phrase is obligatory or not, due to the existence of the simple construction.

For example, the sentence in (322)a exemplifies the discontinuous construction as it realizes a subject NP, a -sja verb and a comitative phrase. The sentence in (322)b is formally the

same as in (322)a, except that the comitative phrase is not realized. Based on the obligatoriness

test, if the comitative phrase is a complement, we would expect the sentence in (322)b to be

235

ungrammatical. In this case, however, the sentence without the comitative phrase is acceptable because it realizes the simple reciprocal construction.

(322) a. Mal'čiki celovali-s' s devočkami. boys.NOM kiss.PST.PL-SJA with girls.INST ‘The boys and girls kissed.’

b. Mal'čiki celovali-s'. boys.NOM kiss.PST.PL-SJA ‘The boys kissed.’

The next logical step in determining obligatoriness is to test whether the comitative phrase is obligatory when the subject NP is semantically singular, as the simple construction cannot realize such a subject. For example, the well-formed sentence in (323)a contains a -sja verb and the comitative phrase s Mariej ‘with Marija’ in the discontinuous construction.

(323) a. Ivan poceloval-sja s Mariej. Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed.’

b. *Ivan poceloval-sja. Ivan.NOM kiss.PST.SG-SJA (Intended: ‘Ivan kissed.’)

The sentence in (323)b is formally exactly the same as (323)a, except that no comitative phrase is realized. Since this second sentence is ungrammatical, it seems to suggest that the discontinuous construction must contain the comitative phrase to be grammatical, and therefore the comitative phrase is obligatory. The problem with this example is, however, that it is impossible to know whether the sentence in (323)b is unacceptable because it realizes a discontinuous construction

236

that is missing the comitative phrase, or because it realizes the simple reciprocal construction with a semantically singular subject. Another diagnostic is therefore needed to determine the status of the comitative phrase in the discontinuous construction.

A second test to determine whether prepositional phrases are adjuncts or complements has been called the fixed preposition test (Needham and Toivonen 2011). This diagnostic assumes

that when a verb under a particular interpretation takes a prepositional phrase as a complement,

the verb requires a specific preposition. For example, since the verb trust in (324) is grammatical

with the preposition in as in (324)a, but not with the preposition on as in (324)b, Neeham and

Toivonen hypothesize that the prepositional phrase in (324)a is a complement to the verb trust.

(324) a. Kim trusted in her own abilities.

b. *Kim trusted on her own abilities. (Needham and Toivonen 2011:4)

Such a sentence can be compared to those in (325) that realize the verb rest. Since this

verb is compatible with prepositional phrases that include different prepositions (in, beside, on), it

can be hypothesized that none of the prepositional phrases in (325) are complements.

(325) a. Louise rested in the forest.

b. Louise rested beside the big tree.

c. Louise rested on the lawn. (Needham and Toivonen 2011:4)

Crucially, this test does not diagnose verbs, but rather verbs under particular expressions. For

example, even though the prepositional phrases in (325) are not complements, the verb rest does 237

take a prepositional phrase as a complement as in (326), when the subject of the sentence is the

NP argument:

(326) The argument rested on a single piece of evidence.

This distinction will be crucial to my argumentation about the comitative phrase of the

discontinuous construction.

I propose that the comitative phrase in the discontinuous construction is a complement

and not an adjunct, based on the fixed prepositional test. Consider, for example, the following

sentences:

(327) a. Ivan poceloval-sja s Mariej. Ivan.NOM kiss.PST.SG-SJA with Marija.INST ‘Ivan and Marija kissed.’

b. *Ivan poceloval-sja bez Marii. Ivan.NOM kiss.PST.SG-SJA without Marija.GEN (Intended: ‘Ivan kissed without Marija.’)

(328) a. Ivan proščaet-sja s Mariej. Ivan.NOM say.good.bye.NPST.SG-SJA with Marija.INST ‘Ivan and Marija say good-bye.’

b. *Ivan proščaet-sja bez Marii. Ivan.NOM say.good.bye.NPST.SG-SJA without Marija.GEN (Intended: ‘Ivan says good-bye without Marija.’)

238

The sentences in (327)a and (328)a are acceptable, but those in (327)b and (328)b are not, when the preposition s is replaced with the preposition bez. The unacceptability of these (b) sentences suggests that s cannot be changed out for a different preposition, and furthermore that the comitative phrase in this sentence is a complement.

The same problem about the diagnostic applies here as for the obligatoriness test when the subject is singular. It is not possible to tell whether the sentences in (327)b or (328)b are unacceptable because they realizes the discontinuous construction with a different preposition, or because they realize the simple construction and are thus unacceptable because the denotation of the subject is semantically singular.

We must then consider plural nouns, and consider the different interpretations of the simple and discontinuous construction. Sentences such as (329) are possible, in which a -sja verb can co-occur with a prepositional phrase headed by bez:

(329) My obščali-s' bez Ivana. we.NOM chat.PST.PL-SJA without Ivan.GEN ‘We chatted without Ivan.’

I propose that the sentence in (329) exemplifies the simple construction, because the only interpretation available for this sentence is that the entities denoted by the subject stand in the chatting relation to each other. Such sentences are not acceptable when the entities denoted by the subject do not stand in the relevant relation to each other. For example, consider the sentences in

(330):

239

(330) Context: You and your friends Marija and Ivan are always together and share everything. You usually spend time talking in the afternoon. Today, however, Ivan had to go to a doctor’s appointment. You and Marija therefore had a quick chat without him.

My poobščali-s' do užina bez Ivana. we.NOM chat.PST.PL-SJA before dinner.GEN without Ivan.GEN ‘We chatted before without Ivan.’

This context establishes a situation in which each individual denoted by the subject stands in the chatting relation to each other individual denoted by the subject. The sentence is acceptable in this context, as would be the case if the interpretation were that of either the simple or the discontinuous construction. The following context, however, sets up a situation in which there is no relation between the entities denoted by the subject. Since the discontinuous construction does not entail such relations, if such a sentence can instantiate the discontinuous construction, the sentence in (331) should be acceptable .

(331) Context: You and your friends Marija and Ivan are always together and share everything. You’re even in the same French class together. You have an assignment due tomorrow though, in which you have to go and have a quick conversation with a native French speaker. You and Marija don’t get to talk then today, but you go out separately and have a conversation with a French speaker. Ivan has to go to a doctor’s appointment and doesn’t do the assignment.

#My poobščali-s' do užina bez Ivana. we.NOM chat.PST.PL-SJA before dinner.GEN without Ivan.GEN ‘We chatted before dinner without Ivan.’

Since my consultants found this sentence unacceptable given the context, I propose that the only

interpretation such sentences have is the same that the simple construction has. I propose then

that the comitative phrase of the discontinuous construction cannot acceptably be replaced with

another prepositional phrase, and therefore the comitative phrase is a complement. I propose then 240

that the comitative phrase is an argument to the -sja verb, and it is in fact this predicate, the -sja

verb plus the comitative phrase, that makes up a discontinuous construction.

4.6.2.3 Accounting for the two sets reading

In Section 4.6.1.1 above, I supported Dalrymple et al.’s (1998b) analysis of the

preposition s ‘with’ arguing that s denotes sum-formation in sentences like that in (332), where

the subject is coordinated with the preposition s.

(332) Ivan s Mariej pocelovali-s'. Ivan.NOM with Marija.INST kiss.PST.PL-SJA ‘Ivan and Marija kissed.’

The s in the discontinuous construction, like that in (333), on the other hand cannot denote a sum-

formation because the comitative phrase (bolded) is not a nominal modifier.

(333) Ivan pocelovali-s' s Mariej Ivan.NOM kiss.PST.PL-SJA with Marija.INST ‘Ivan and Marija kissed.’

Because the comitative phrase is not a nominal modifier, as discussed in Section 4.6.2.1 above, its syntax does not match that of the coordinating s discussed by Dalrymple et al. (1998b).

Therefore, another meaning of s must be found.

I propose that the preposition s of the discontinuous construction is one of the so-called semantically-vacuous prepositions. Such prepositions are functional in nature, rather than lexical.

For example, in English, the preposition of has been argued to be semantically vacuous in 241

constructions such as all of the children or the son of Mary (Chomsky and Lasnik 1995,

Tremblay 1996, Heim and Kratzer 1998). Heim and Kratzer (1998), for example, suggest the

following definition of this vacuous of:

(334) ⟦of⟧= λx ∊ De.x

(“⟦of⟧ is that function which maps every individual in De to itself . . .” (62))

Similarly, I propose the meaning of s in the discontinuous construction is as in:

(335) s ‘with’: N/N: λxe[x]

This definition says that s is a function that maps every individual of type e (which will be

supplied by the object of the comitative phrase) onto itself, just as of does. Therefore, the syntax

and semantics of the comitative phrase s Ivanom will be as in (336).

(336) s Ivanom ‘with Ivan’: N: i

Since the comitative phrase is a verbal complement, the meaning of the -sja verb will be

applied to the meaning of s Ivanom.

Assuming the simplest analysis, I first (and temporarily, as the Subject/Object Restriction

still must be accounted for) propose that the meaning of the verbs that are realized in the

discontinuous construction have the same semantics as those in the simple construction. It is only

in terms of their combinatorics that they differ (I will revise this assumption later). The semantic

242

representation in (337) is of a verb in the simple construction that conveys a SR relation, and that in (338) is that of a verb in the discontinuous construction.

(337) kiss_SR: pocelovat'sja: N\S’: λxλe[∀a∀b(a ≤ x ˄ b ≤ x → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ e’≠ e”˄ kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]

(338) kiss_SR: pocelovat'sja: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]

Note that the syntactic difference between the two is that the verb in (337) takes one N to project a sentence radical, while that in (338) takes two, i.e. that in (337) is intransitive and that in (338) is transitive. Similarly, the semantics mirror the syntax. The semantics of the two verbs differ as the function denoted in (337) is a function from individuals denoted by the subject NP to a

(characteristic function of a) set of events, whereas that in (338) is a function from individuals

(denoted by the object) to a function from individuals (denoted by the subject) to a (characteristic function of a) set of events. The same reciprocal polyadic quantifier applies to both of the sets

(which is why the number of quantifiers, variables and relations in (338) are double the number in

(337)).

Now armed with the meaning of all the components in a sentence realizing a discontinuous construction, I propose the derivation of the sentence in (339) in Figure 22, given the definitions provided for -sja verbs in the discontinuous construction and for the preposition s.

The relation I assume here is Strongly Reciprocal.

243

(339) Context: There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each of the boys kisses each girl, and each girl kisses each of the boys.

Mal'čiki celovali-s' s devočkami. boys.NOM kiss.PST.PL-SJA with girls.INST ‘The boys and girls kissed.’

244

s ‘with’: N/N: λx[x] devočki ‘girls’: N: d

s devočkami: N: λx[x](d) FA

celovat'sja: (N\S’)/N: λxλyλev[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e” (e’≤ ev ˄ e”≤ ev (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))] s devočkami: N: d BR

celovat'sja s devočkami: (N\S’): λxλyλev[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b→ ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’))](d) BR

mal'čiki ‘boys’: N: m celovat’sja s devočkami: (N\S’) :λyλev[∀a∀b (a ≤ d ˄ b ≤ y ˄a ≠ b → ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’)) ˄ kiss’(b)(a)(e”)))]] BR

245 mal'čiki celovat’sja s devočkami: S’:λyλev[∀a∀b (a ≤ d ˄ b ≤ y ˄ a ≠ b → ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))])](m) BA

mal'čiki celovat’sja s devočkami: S’: λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → -l ‘PAST’: S/S’: λP [∃e(P(e) ˄ τ(e) ≪ now)] ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]) BR

mal'čiki celovalis' s devočkami: S: λP [∃e(P(e) ˄ τ(e) ≪ now)](λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]) FA

mal'čiki celovalis' s devočkami: S: ∃e(λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e” (e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))] (e) ˄ τ(e) ≪ now) BR

mal'čiki celovalis' s devočkami: S: ∃e(∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b →∃e’∃e” (e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”))) ˄ τ(e) ≪ now) BR

Figure 22: Derivation of a sentence that realizes the discontinuous reciprocal construction

245

I provide the final translation of the sentence in (339), repeated from the last line of the derivation in Figure 22 in (340):

(340) ∃e(∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b) → ∃e’∃e” (e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”))) ˄ τ(e) ≪ now)

This translation says that this sentence is true if and only if there exists an event e such that all entities a that are a part of the sum denoted by devočki ‘girls’ and all entities b that are part of the sum denoted by mal'čiki ‘boys,’ and a and b are distinct, then there are two subevents that are each part of the event e and e’ is an event of b kissing a and e” is an event of a kissing b and the event e occurred before now. This translation then captures the truth conditions we want, as the sentence in (339) conveys a SR relation (as we know from the context).

Note that this translation only entails that relations hold between members of different sets; no relation is entailed between members of the same set. Neither, however, is there a restriction against such relations. This translation then captures the empirical generalization that entities in the same set can at times participate in a relation with other entities in the same set, but that such relations are not required.

Having provided a preliminary analysis of the discontinuous construction, I now turn to providing an analysis that captures the generalization that each entity in the set denoted by the subject must stand in the relevant relation as the first argument with at least one entity in the set denoted by the object.

246

4.6.2.4 Accounting for the Subject/Object Restriction

I propose that the Subject/Object Restriction, repeated in (341), is contributed by the verb.

(341) Subject/Object Restriction:

∀x∈ A ∃y ∈ B ∀w∈ B ∃z ∈ A (Rxy ˄ Rzw)

I hold that it is the verb that contributes the restriction, because the restriction must have access to

the lexical predicate of the verb. Recall that I have proposed that the -sja verbs come from the

lexicon with both the lexical predicate and the relations specified, as in (342). If the restriction is

not on the verb, then the Subject/Object Restriction cannot pick up the lexical information it

needs from the verb, i.e. that the relation specified by (342) is a kissing relation, and that the

restriction needs to pick up relations of this type. If the restriction were contributed by any other

element in the sentence, it would not be able to restrict the relation.

(342) pocelovat'sja: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b) → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]

Assuming then that it is the verb that somehow contributes the restriction, I add this restriction to the semantics of the lexical entry in (342). The restriction is bolded in (343):

247

(343) pocelovat'sja: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b) → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃d∃f∀h∃j∃f’(c, j ≤ y ˄ d,h ≤ x ˄ f, f’ ≤ e ˄ (kiss’(d)(c)(f) ˄ kiss’(j)(h)(f’))))]

The bolded portion of the semantics says that for all entities c which are part of the entity y,

which is contributed by the subject, there is an individual d that is part of the entity x, that

contributed by the object, and a subevent f that is part of the event e, and all c kiss at least one d.

Furthermore, all entities h that are part of x stand in the kiss’ relation as the second argument to at least one entity j that is part of the entity y for some event f’.

I see three possibilities for how this lexical entry comes about. The first possibility is that in the derivation process from the intransitive verb in the simple construction to the transitive verb in the discontinuous construction the verb undergoes a lexical rule that contributes the restriction. Under this analysis, the input to the lexical rule would be the lexical entry in (344) below, and the output that in (343) above.

(344) pocelovat'sja: N\S’: λxλe[∀a∀b(a ≤ x ˄ b ≤ x → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ e’≠ e”˄ kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)))]

The problem, however, is that as I’ve discussed before in Chapters 2 and 3¸ I assume that the

reciprocal -sja verbs in the simple construction have the reciprocal relation built into their lexical

entry. It is therefore impossible to construct a lexical rule that will derive the semantics in (343),

because this lexical rule will not have access to the predicate. Since the lexical rule does not have

access to the predicate, it is impossible for it to add the right conjunct to the output of the rule, i.e.

the lexical entry that is realized in the discontinuous construction.

248

Because of this problem, I must reject the analysis that the verb in the discontinuous construction is derived via a lexical rule from that in the simple construction. I therefore propose the second possibility, that the -sja verbs in the discontinuous construction have different lexical entries from those in the simple construction and furthermore are not derived from them. The lexical entries for the discontinuous -sja verbs under this analysis are syntactically and semantically transitive and have the Subject/Object Restriction built into the entry.

The lexical entries for the -sja verbs in the discontinuous construction would be as in

(345)-(349):

(345) pocelovat'sja[SR]: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃d∃f∀h∃j∃f’(c, j ≤ y ˄ d,h ≤ x ˄ f, f’ ≤ e → (kiss’(d)(c)(f) ˄ kiss’(j)(h)(f’)))]

(346) pocelovat'sja[IR]: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → for some sequence z0, z2. …, ≤ x ∧ z1,z3 … ≤ y ∧ zm ≤ x ∨ zm ≤ y(∃e’∃e”(e’≤ e ˄ e”≤ e ∧ (a = z0 ∧ kiss’(z0)(z1)(e’) ∧ … ∧ kiss(zm- 1)(zm )(e”) ∧ zm = b) ˄ ∀c∃d∃f∀h∃j∃f’(c, j ≤ y ˄ d,h ≤ x ˄ f, f’ ≤ e → (kiss’(d)(c)(f) ˄ kiss’(j)(h)(f’))))]

(347) pocelovat'sja[OWR]: (N\S’)/N: λxλyλe[∀a∃b∀c∃d (a ≤ x ˄ b ≤ y ˄ c ≤ y ˄ d ≤ x ˄ a ≠ b ˄ c ≠ d → ∃e’∃e” (e’≤ e ˄ e”≤ e (kiss’(a)(b)(e’) ˄ kiss’(c)(d)(e”)) ˄ ∀g∃h∃f∀i∃j∃f’(g, i ≤ y ˄ h,j ≤ x ˄ f, f’ ≤ e → (kiss’(g)(h)(f) ˄ kiss’(j)(i)(f’))))]

(348) pocelovat'sja[RIAR]: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → for some sequence z0, …, zm ≤ x ∧ z0, …, zm ≤ y ∃e’∃e” (e’≤ e ˄ e”≤ e ∧ (a = z0 ∧ (kiss’(z0)(z1)(e’) ∨ kiss’(z1)(z0)(e’))∧ … ∧ kiss(zm- 1)(zm )(e”) ∨ kiss (zm ) (zm- 1) (e”)∧ zm = b) ∧ ∃a2∃b2∃c2∃d2∃g∃h (a2 ≤ x ∧ b2 ≤ y ∧ c2 ≤ y ∧ d2 ≤ y ∧ g ≤ e ∧ h ≤ e (kiss’(a2)(b2)(g) ∧ kiss’(c2)(a2)(h)) ˄ ∀c∃d∃f∀h∃j∃f’(c, j ≤ y ˄ d,h ≤ x ˄ f, f’ ≤ e → (kiss’(d)(c)(f) ˄ kiss’(j)(h)(f’)))]

249

(349) pocelovat'sja[IAR]: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e” (e’≤ e ˄ e”≤ e ∧ for some sequence z0, . . . , zm ≤ x ∧ z0, . . . , zm ≤ y ∧ (a = z0 ∧ (kiss’(z0)(z1)(e’)∨ kiss’(z1)(z0)(e’))∧ … ∧ kiss(zm- 1)(zm )(e”) ∨ kiss (zm ) (zm- 1) (e”) ∧ zm = b) ˄ ∀c∃d∃f∀h∃j∃f’(c, j ≤ y ˄ d,h ≤ x ˄ f, f’ ≤ e → (kiss’(d)(c)(f) ˄ kiss’(j)(h)(f’)))]

These verb lexical entries would compose first with the comitative phrase, then with the subject

to result in the translation of the sentence that realizes the discontinuous construction.

This analysis accounts for the meaning of the discontinuous construction, including the

two empirical generalizations previously discussed, as well as the different relations explored in

Section 4.5.3. Additionally, it accounts for the fact that the discontinuous construction is

compatible with a relation that the simple construction is not. Recall that I provided evidence

suggesting that the simple construction is not compatible with Intermediate Alternative

Reciprocity, while the discontinuous construction was compatible with the version of this relation

for two sets. This finding poses no problem for this analysis, as the lexical entries for the

discontinuous construction are not derived from those from the simple. Any analysis that

assumes the transitive lexical entries are derived from the intransitive versions would have to

account for this additional relation.

The main problem with this analysis is that it immensely multiplies the lexical entries

posited for each verb. I’m already assuming that each reciprocal -sja has four different entries.

This analysis assumes four entries for the intransitive -sja verb and five for the transitive. It is

therefore an inelegant solution and problematic if we assume a parsimonious lexicon.

The only other option in accounting for the empirical generalizations I see is to revert

back to analyzing the reciprocal -sja verbs as made up of two morphemes that compose in the

syntax; the verb, e.g. celovat' ‘to kiss’ plus the suffix -sja. Under this analysis, each verb only

250

has one lexical entry, while the -sja suffix has nine different lexical entries, four of which can be realized in the simple construction and five of which can be realized in the discontinuous construction. The Subject/Object Restriction is contributed by the -sja lexical entries that are realized in the discontinuous construction. The lexical entry for the -sja that contributes the relation of SR in the discontinuous construction would be as in (350):

(350) -sja: ((N\S’)/N)\((N\S’)/N): λR>>λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y → ∃e’∃e” (e’≤ e ˄ e”≤ e (R(a)(b)(e’) ˄ R(b)(a)(e”)) ˄ ∀c∃d∃f ∀g∃h∃f’(c,h ≤ y ˄ d, g≤ x ˄ f , f’ ≤ e → (kiss’(c)(h)(f) ˄ (kiss’(d)(g)(f))))]

This lexical entry syntactically composes with a transitive verb to result in a transitive verb. It

imposes the SR relation on the entities denoted by the subject and object of the verb such that the

entities denoted by the subject stand in the R relation to each entity denoted by the object, and it

additionally requires that each entity denoted by the subject stand in the R relation as the first

argument to at least one entity denoted by the object.

The benefit of this analysis is that it generalizes across the verb and as a result cuts down

on the number of proposed lexical entries. Furthermore, it more directly captures the relatedness

of the reciprocal -sja verbs, since it is the same -sja expressions that compose with each of the

verbs. Additionally, it accounts for the fact that the lexical content of each -sja verb is identical

to that of the related transitive verb that does not realize -sja, e.g. celovat' ‘to kiss’ and celovat'sja

‘to kiss’.

The problem with this analysis is what has already been discussed in Chapter 1 and 3; not

all the -sja verbs are composed of an extant verb + sja; many of the verb stems do not exist

without -sja. The example I used before is the verb borot'sja ‘to struggle’, since there is no verb

251

*borot'. Similarly, even if there is an existing verb without the -sja, many of the -sja verbs

cannot be semantically composed from the meanings of the verbal stem plus -sja. For example, the verb rezat'sja means ‘to play cards’. When the verb is realized without the -sja, the verb rezat' means ‘to cut’. Therefore, if we assume that it is the -sja affix that contributes the reciprocal relation and the added restriction, then we still only account for a handful of the -sja verbs; the rest must still have lexical entries that include -sja.

Of these three options for how to capture the restriction, i.e. (i) via a lexical rule, (ii) as part of the lexical entry for each reciprocal -sja verb in the discontinuous construction and (iii) as part of the meaning of the affix -sja, option (ii) is the most promising, as it generates less serious problems for the analysis. I will tentatively assume this analysis, even though I recognize the problems with it.

Given this assumption, I now can provide a complete compositional analysis of a sentence that realizes the discontinuous construction. I again suggest an analysis for the sentence in (351):

(351) Context: There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each of the boys kisses each girl, and each girl kisses each of the boys.

Mal'čiki celovali-s' s devočkami. boys.NOM kiss.PST.PL-SJA with girls.INST ‘The boys and girls kissed.

The lexical entries for each of the expressions in the sentence in (351) are in (352):

252

(352) mal'čiki ‘boys’: N: m

celovat'sja[SR]: (N\S’)/N: λxλyλe[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃d∃f∀g∃h∃f’(c,h ≤ y ˄ d, g≤ x ˄ f , f’ ≤ e → (kiss’(c)(h)(f) ˄ (kiss’(d)(g)(f))))]

s ‘with’: N/N: λxe[x]

devočki ‘girls’: N: d

The compositional analysis of this sentence is in Figure 23:

253

s ‘with’: N/N: λxe[x] devočki ‘girls’: N: d

s devočkami: N: λx[x](d) FA

celovat'sja: (N\S’)/N: λxλyλev[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ y ˄ i, g≤ x ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))] s devočkami: N: d BR

celovat'sja s devočkami: N\S’: λxλyλev[∀a∀b (a ≤ x ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ y ˄ i, g≤ x ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))] (d) FA

celovat'sja s devočkami: N\S’: λyλev[∀a∀b (a ≤ d ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ mal'čiki ‘boys’: N: m kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ y ˄ i, g ≤ d ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))] BR 254 mal'čiki celovat'sja s devočkami: N\S’: λyλev[∀a∀b (a ≤ d ˄ b ≤ y ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ y ˄ i, g ≤ d ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))](m) FA

mal'čiki celovat'sja s devočkami: N\S’: λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ m ˄ -l ‘PAST’: S/S’: λP [∃e(P(e) ˄ τ(e) ≪ now)] i, g ≤ d ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))] BR

mal'čiki celovali-s' s devočkami: S: λP [∃e(P(e) ˄ τ(e) ≪ now)]( λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e”(e’≤ ev ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ m ˄ i, g ≤ d ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))]) BA

mal'čiki celovali-s' s devočkami: S: ∃e(λev[∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ ev ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ m ˄ i, g ≤ d ˄ f , f’ ≤ ev → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’))))] (e) ˄ τ(e) ≪ now) BR

mal'čiki celovali-s' s devočkami: S: ∃e(∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ m ˄ i, g ≤ d ˄ f , f’ ≤ e → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’)))) ˄ τ(e) ≪ now) BR

Figure 23: Final derivation of a sentence that realizes the discontinuous reciprocal construction

254

The last line of this derivation is the translation of the sentence in (351). I repeat it in (353):

(353) ∃e(∀a∀b (a ≤ d ˄ b ≤ m ˄ a ≠ b → ∃e’∃e”(e’≤ e ˄ e”≤ e ˄ (kiss’(a)(b)(e’) ˄ kiss’(b)(a)(e”)) ˄ ∀c∃i∃f∀g∃h∃f’(c,h ≤ m ˄ i, g ≤ d ˄ f , f’ ≤ e → (kiss’(c)(h)(f) ˄ (kiss’(i)(g)(f’)))) ˄ τ(e) ≪ now)

This translation says that the sentence in (351) is true iff there exists an event e such that for all a that are part of the set of girls and for all b that are part of the set of boys and a and b are distinct, then there exist two subevents e’ and e” and e’ is an event of a kissing b and e” is a subevent of b

kissing a and for all c that are parts of the set of boys there exists an i that is part of the set of girls and a subevent f’ that is a part of e and for all g that are part of the girls there is an h that is part of the boys and a subevent f” that is part of e, then f is a subevent of c kissing h and f’ is an event of

i kissing g. This translation captures the truth conditions we want, as it entails that each girl kisses

each boy, but no relations between the girls are entailed and no relations between the boys are

entailed. Furthermore, it requires that each boy kissed at least one girl and each girl was kissed

by at least one boy, which accounts for the Subject/Object Restriction.

This analysis therefore accounts for the various readings of sentences that realize the

discontinuous construction. The fact that no members of the same set are entailed to stand in the

relevant relation with each other is captured by the syntax-semantics interface and no additional

semantic restriction is necessary to account for these data. On the other hand, I have had to posit

that the -sja verbs in the discontinuous construction have different lexical entries from those in

the simple construction.

Based on this analysis, I can answer the question originally posed at the beginning of this

chapter. I argue that the difference in meaning, or the reason each construction is compatible

with a different set of relations, is due to the different interaction of the reciprocal relation 255

(contributed by the -sja verb) with the structure (simple or discontinuous, i.e. one NP versus two

NPs) in conjunction with different lexical entries for the verbs in the constructions.

4.6.2.5 Implications of the analysis of the discontinuous construction:

In Section 4.2 above, I discussed what others have claimed for reciprocal relations in

Greek, Hungarian and Hebrew. Recall that Dimitriadis for Greek and Rákosi for Hungarian, made claims only that the entities must stand in the relation to entities in the other set, but provided no data to show that relations cannot hold between entities denoted by the subject or entities denoted by the object. Siloni, on the other hand, explicitly claimed that such relations do not hold, but also did not provide evidence to support this claim. Given the similarities of the discontinuous construction across unrelated languages (e.g. Russian, Hebrew, Hungarian), it’s an open empirical question whether sentences like that in (354), repeated from (262), which realize

the discontinuous construction in a context where the sets denoted by the subject and object

overlap, are acceptable in the languages for which the strong claim has been made or not. Further

data will help elucidate whether this is true or not.

256

(354) Context: There is a parent-teacher conference. Ivan and Marija are teachers. Elena and Aleksej are parents. Svetlana is both a teacher and a parent. During the evening, each of the teachers talks to each of the parents. Svetlana talks to both the teachers (Ivan and Marija) about her child as well as the parents (Elena and Aleksej) about their child.

a. Učitelja i roditeli soveščali-s'. teachers.NOM and parents.NOM converse.PST.PL-SJA ‘The teachers and parents conversed with each other.’

b. Učitelja soveščali-s' s roditeljami. teachers.NOM converse.PST.PL-SJA with parents.INST ‘The teachers and parents conversed with each other.’

Furthermore, I have shown that the fact that relations are required to hold only between entities denoted by the subject on the one hand and entities denoted by the object on the other hand falls out from the syntax. In Siloni’s (2012) analysis outlined in Section 4.2 above, a meaning postulate associated with the verb in the discontinuous construction associates pairs of entities with a subevent and then associates each entity with a thematic role. In my analysis, the fact that relations are not entailed between members of the same set is a consequence of the syntactic structure and no further semantic operators or restrictions need be posited. My analysis then provides for an elegant solution to a long-discussed problem.

The Subject/Object Restriction is also novel in that it more precisely accounts for the empirical findings. Previous literature has simply noted that the object is less “volitional” or

“agentive” but has not provided data suggesting that the subject must stand in the given relation as the first argument and that it is universally quantified over. Furthermore, although it has been an implicit assumption, it has not been shown that the object must stand in the relation as the second argument and that the denotation of the object is existentially quantified over. These details allow a formal analysis that is founded on clear empirical generalizations. I propose a

257

unified analysis for the discontinuous construction when it contributes a symmetric and asymmetric relation between participants.

4.7 Theoretical conclusions

In this chapter I explored the relations the simple reciprocal construction and the

discontinuous reciprocal construction in Russian are compatible with. No other formal semantic

study has yet to investigate the relations reciprocal expressions other than each other are

compatible with and this chapter therefore widens the scope of cross-linguistic semantic

exploration of reciprocal expressions. Furthermore, I proposed an analysis to capture the

meanings of each construction.

4.7.1 Reciprocal relations

I provided evidence suggesting that one expression, the -sja verbs in the simple reciprocal

construction, is compatible with four different relations between participants, while the other

expression, the -sja verbs in the discontinuous construction, is compatible with five different

relations.

Such results are interesting on two fronts. First, the four relations that the simple

construction is compatible with are the same four that as I discussed in Chapter 2 that each other

is compatible with. The reason for the semantic similarity cannot be due to etymological reasons,

as these two reciprocal expressions are not etymologically related. The similarity then suggests

that there is something fundamental about these four relations. I continue discussing this notion

throughout the rest of this dissertation and will return to it after my analysis of drug druga.

258

Additionally, however, I have provided data suggesting that the discontinuous construction is compatible with a relation that each other is not; Intermediate Alternative

Reciprocity. This finding suggests that this relation is indeed a reciprocal relation, even though it

is too weak for English to express with each other. This finding also addresses the analysis of

Beck (2001) that argues that in English the IAR relation is not a true reciprocal relation; rather it is a property of particular lexical items that they can express this asymmetric reading. In particular she shows that only predicates of spatial and temporal relation are compatible with the

IAR relation. Although she does not discuss implications this might have for cross-linguistic analyses, a natural assumption would be that IAR is never a relation and that whenever it holds it is simply because of the verb’s meaning. I have provided evidence in this chapter that predicates other than those of spatial and temporal relation are compatible with IAR, such as pocelovat'sia

‘to kiss’, obnimat'sja ‘to hug’, pozdorovit'sja ‘to greet’, and proščat'sja ‘to say good-bye’.

Therefore, I add IAR to the set of reciprocal relations. Future studies of reciprocity should certainly include it when exploring the semantics of reciprocal expressions cross-linguistically.

This chapter too informs discussions of the discontinuous construction cross- linguistically because it has not been pointed out before, beyond noting that it is compatible with an asymmetric reading, that the discontinuous construction, like the simple, is compatible with different relations between participants. I have shown that the discontinuous construction is compatible with SR, IR, and OWR, RIAR and IAR and that it indeed is therefore compatible with a different set of relations than the simple construction. I would be interested to find out whether this is the case for other reciprocal expressions that participate in the discontinuous construction.

It is already fascinating that there are so many formal similarities in the construction in different, unrelated languages, and it should be determined whether these formal similarities correlated with semantic similarities.

259

4.7.2 Significance of syntactic construction

I have provided evidence in this chapter suggesting that the apparently different relations the simple and discontinuous constructions are compatible with are in fact the same relations, and that the relations the discontinuous construction is compatible with can be derived from the relations the simple construction is compatible with. This therefore suggests that reciprocal relations can interact with the structure in ways not previously discussed in the formal literature, as no one has yet examined sentences with reciprocal expressions that draw entities from two non-overlapping sets.

Extending this implication, I propose that any semantic exploration of reciprocal

expressions cross-linguistically should include at least a description of the different syntactic

constructions an expression can occur in. It has proven vital in understanding the semantics of

the -sja verbs to have this information, because these verbs seem to have different semantics depending on the construction they are realized in. It is only through examining the interaction of the structure and meaning that we can adequately analyze the meaning of each construction and the meaning of the reciprocal expression. I suggest that future work on reciprocal expressions look not only at the relations each expression is compatible with, but also carefully discuss the syntactic environments that they are realized in, whether or not any differences in meaning are noted based on the syntax. It is only in this way we can determine how much of a role the syntactic construction plays in the semantics of reciprocity.

None of the studies of each other, for example, discuss the syntactic elements surrounding the expression. Because of this we do not know if the semantics of the expression correlate in any consistent way with the syntactic construction. For example, consider the two

260

sentences in (355). In (355)a, each other is a direct object complement to the verb, while in

(355)b, it is the object of a prepositional phrase and has the status of an adjunct.

(355) a. The boys saw each other.

b. The boys sat down near each other.

The formal literature does not point out that each other can occur in these two different contexts, since the syntax is not thought to be relevant. It is possible we are missing a part of the picture because the semantics of these two constructions have not been studied separately.

Syntax therefore helps us make sense of the different semantics of the two different constructions. I therefore argue that when exploring the meaning of reciprocal expressions cross- linguistically, the syntax needs to be considered more carefully than it has been in the past.

261

5 Accompaniment and reciprocity: the meaning of drug druga

5.1 Introduction

The anaphoric pronoun drug druga is compatible with Strong Reciprocity (SR), as

demonstrated in (356) (and previously in Chapter 1 in Section 1.3.1). In this context, each

individual slaps each of the other individuals.

(356) Context: Your friends Marija and Tanja are verbally fighting. Suddenly it turns violent, and you see Marija hit Tanja, then Tanja slaps Marija back. You say:

Marija i Tanja xlopnuli drug druga po ščёkam! Marija.NOM and Tanja.NOM hit.PST.PL each other.ACC on cheeks.DAT ‘Marija and Tanja slapped each other!’

In this sentence, drug druga is in the accusative case. Recall from Chapter 1 that this pronoun can also take the dative case as in (357) or be used as the object of a preposition as in (358).

When drug druga is the object of a preposition, it is assigned case by this preposition.

262

(357) Context: Your friends Marija and Ivan are working on a project together in the library. As you watch, you see Marija pass Ivan a book for him to read and then you see Ivan pass Marija a book. You say:

Marija i Ivan peredali drug drugu knigi. Marija.NOM and Ivan.NOM give.PST.PL each other.DAT books.ACC ‘Marija and Ivan gave each other books.’

(358) Context: You see your young students Fedja and Olja playing together before class. You see Fedja running after Olja, then she turns around and starts chasing him. You say:

Fedja i Olja gonjalis' drug za drugom. Fedja.NOM and Olja.NOM chase.PST.PL each behind other.INST ‘Marija and Ivan chased after each other.’

In (357), the two individuals stand in a strongly reciprocal giving relation, while in (358), Fedja

and Olja stand in a Strongly Reciprocal chasing-after relation. These sentences illustrate that

drug druga is still compatible with SR when used in cases other than accusative and when the

reciprocal expression is the object of a preposition.

These observations make the unacceptability of the sentence in (359)a surprising, when

the relation between the two individuals is Strong Reciprocity and the reciprocal expression is

used as the object of the preposition s ‘with’:

263

(359) Context: Your friends Ivan and Marija, who are dating, just left for the movies together. You get a phone call from someone asking for them. You say:

a. #Ivan i Marija pošli v kino drug s drugom. Ivan.NOM and Marija.NOM go.PST.PL to cinema.ACC each with other.INST (Intended: ‘Ivan and Marija went to the movies with each other.’)

b. Ivan i Marija pošli v kino vmeste. Ivan.NOM and Marija.NOM go.PST.PL to cinema.ACC together ‘Ivan and Marija went to the movies together.’

A sentence that my consultants did find acceptable in the same context is that in (359)b, that

realizes the adverb vmeste ‘together’ instead of drug s drugom ‘with each other’.

The unacceptability of (359)a is especially puzzling considering some sentences that realize the phrase drug s drugom ‘with each other’ are acceptable. For example, consider the sentences in (360)-(362) taken from the RNC:

(360) Kstati, rabotniki Hermes drug s drugom obščajut-sja kak by-the-way workers.NOM Hermes.GEN each with other.INST converse.NPST.3PL-SJA as

budto u nix vnutri mikroskopy. if at they.GEN inside microscopes.NOM ‘By the way, the Hermes workers converse with each other, as if they have microscopes inside of themselves.’

(361) Poka oba soveta … soperničajut drug s drugom za glavnuju meanwhile both.NOM council.GEN compete.NPST.3PL each with other.INST for main.ACC

rol' v partii. role.ACC in party.LOC ‘Meanwhile, both councils … compete with each other for the main role in the party.’

264

(362) … čem plotnee naši specialisty budut rabotat' drug s that consistent.COM our.NOM specialists.NOM be.NPST.3PL work.INF each with

drugom, tem lučše eto otrazitsja na sostojanii našix ekonomik… other.INST that better it.NOM reflect.NPST.3SG on condition.LOC our.GEN economists.GEN ‘… the more consistently our specialists work with each other, the better it will be reflected on the condition of our economists …’

These sentences demonstrate that the phrase drug s drugom can be acceptably realized.

Moreover, from an English perspective, we would expect the sentence in (359)a to be

acceptable because given the same context, the English sentence in (363), which is a direct

translation of the Russian sentence in (359)a, is acceptable:

(363) John and Mary went to the movies with each other.

The expression drug druga thus appears to have an unexpected restriction on its usage from a

language internal as well as external point of view. This chapter is dedicated to determining why

drug druga is not acceptable in sentences such as that in (359)a. The unacceptability suggests

that drug druga and each other are not as similar as they have appeared so far in this dissertation.

One possible hypothesis for the difference is that drug druga actually has a very different

meaning and usage than each other and their only similarity is that they both can convey a

relation of SR. In this chapter I will show, however, that drug druga and each other are very

similar in the ways that formal semanticists have analyzed reciprocity in the past - i.e. in terms of

the personal relations each is compatible with.

An alternative hypothesis is that a parameter not before discussed in the formal literature

on reciprocity is relevant here, that of accompaniment. I will argue that this hypothesis best 265

explains the data. In particular, I will provide evidence suggesting that drug druga, unlike each other, cannot be realized in a prepositional phrase that expresses the relation of accompaniment and that this is the reason for the ungrammaticality of (359)a.

This chapter is structured as follows. In Section 5.2 I explore the relations between participants that drug druga can express. In Section 5.3 I introduce the semantic concept of accompaniment and the terms I will use to discuss it. I argue here that there is a restriction against realizing drug druga as the object of an accompaniment phrase. I argue in Section 5.4 that when the phrase drug s drugom ‘with each other’ is acceptably realized, it is as the complement to the discontinuous construction discussed in Chapter 4. In Section 5.5, I show that there are a few cases that my analysis cannot account for and I discuss a possible explanation for these apparent counterexamples. I conclude in Section 5.6.

5.2 Reciprocal relations and drug druga

In this section I investigate the meaning of drug druga to determine how similar it is to

the meaning of each other in terms of relations between participants. I will provide evidence suggesting that drug druga is compatible with the same set of relations as both each other and the

-sja verbs in the simple construction, suggesting that these three expressions are semantically comparable.

The data for this section was both extracted from the RNC and elicited from consultants.

For the corpus study, I pulled 200 sentences with the expression drug druga in the accusative case, 50 with it in the dative case, 50 for the instrumental case, and 50 each with the following prepositional phrases: drug v druge ‘in each other’, drug ot druga ‘from each other’, drug k drugu

‘to each other’ and drug s drugom ‘with each other’, to assure that any conclusions I make are not

266

just based on one case or syntactic structure. I chose these prepositional phrases because they include prepositions that take different cases (locative, genitive, dative and instrumental respectively).

5.2.1 Strong Reciprocity

In addition to being compatible with SR (see (356) and (357) above), drug druga is also compatible with Partitioned SR, as demonstrated by the following sentence found in the RNC:

(364) V preddverii Dnja vljublёnnyx na um nevol'no prixodit in threshold.LOC day.GEN lovers.GEN on mind.ACC involuntarily come.NPST.3SG

starinnaja legenda o tom, čto vsex ljudej jakoby pazdelili na old.NOM legend.NOM about that that all.ACC people.ACC allegedly divide.PST.PL in

polovniki, kotorye potom byli razbrosany po vsemu miru, i teper' halves.ACC which.NOM then be.PST.PL scattered around all.DAT world.DAT and now

eti polovinki stremjatsja najti drug druga. these.NOM halves.NOM strive.NPST.3PL find.INF each other.ACC

‘On the threshold of the Day of the lovers, to my mind involuntarily comes an old legend about the fact that each person allegedly was divided into halves, which then were scattered across the world, and now these halves are striving to find each other.’

This sentence expresses a relation in which the denotation of each ‘half’ is in a partition with the denotation of its other ‘half’, and SR holds within these partitions, as each half is trying to find its other half. Thus, drug druga is compatible with Partitioned SR. Again, however, as I noted in

Chapter 4 in connection with the -sja verbs, I do not consider this partitioning effect to be due to the reciprocal expression. Rather the partitioning effects are due to the Cover mechanism

(Schwarzschild 1996) of the plural NP. I therefore do not consider partitioning effects further in this chapter. 267

5.2.2 Intermediate Reciprocity (IR)

The following two examples demonstrate that drug druga is compatible with Intermediate

Reciprocity. The first example, taken from the RNC, describes four different groups of people,

and each group replaces each other group in shifts.

(365) V 1988 godu v peščere rabotalo 4 ekspedicii, posledovatel'no in 1988 year.LOC in cave.LOC work.PST.SG 4 expedition.GEN successively

menjavšie drug druga i prabotavšie po obščemu planu … replace.PTCP each other.ACC and work.PTCP on common.DAT plan.DAT ‘In 1988 there were 4 expeditions working in the cave that successively replaced each other and worked on a common plan …’

The key here is that if each group just replaced each other once, so that each group worked only

once, this sentence would not exemplify IR, because in that case not every group would be

directly or indirectly related to every other group as both the first and the second arguments.

However, since the groups work in shifts, each works more than once. The changing relation

could be depicted as in Figure 24.

1 2

4 3

Figure 24: Depiction of an Intermediate Reciprocal relation

268

Therefore, since every group participates as both the first and second relation, and each participant is directly or indirectly related to each other via a chain, this sentence is compatible with IR.

The following example, elicited with the video stimuli, is also compatible with IR:

(366) Video Context: Four people are sitting next to each other on a bench in a row.

Oni sidjat rjadom drug s drugom. they.NOM sit.NPST.3PL next.to each with other.INST ‘They are sitting next to each other.’

In this example, every participant is in a symmetrical relation with his/her neighbor, and each participant is directly or indirectly related to each other participant. Thus, this context and sentence provide another example of the compatibility of drug druga with IR.

5.2.3 One-Way Weak Reciprocity

My data also show that drug druga is compatible with OWR. Consider the following constructed context and the utterance I provided my consultants:

269

(367) Context: Over the summer break, you and your three friends spent time with each other. Everyone went to Fedja’s apartment, since it’s in the center, but Fedja only visited Nataša in return.

You Fedja

Nataša Saša

You say:

Letom, my xodili drug k drugu. summer.INST we.NOM go.PST.PL each to other.DAT ‘In the summer, we visited each other.’ (Literally: ‘In the summer, we went to each other.’)

This context sets up a OWR relation because each individual participates in the relation as the

first argument, but not necessarily as the second. Only Fedja and Nataša participate as the second

argument. Since my consultants judged this sentence as acceptable given the context, it provides

evidence that drug druga is compatible with OWR.

An example from the RNC also provides evidence that drug druga is compatible with

OWR. The following sentence expresses a disguised-as relation between the denotation of the

processes.

(368) Evolucija real'nyx system prestavljaet soboj složneju evolution.NOM real.GEN systems.GEN present.NPST.3SG self.INST complex.ACC

superpoziciju raznoobraznyx dinamičeskix processov, zavualirovannyx drug superposition.ACC different.GEN dynamic.GEN processes.GEN disguised.PTCP.GEN each

drugom … other.INST ‘The evolution of real systems presents a complex superposition of different dynamic processes, disguised as each other …’

270

I propose that this sentence expresses a OWR relation because each process is disguised and thus

its denotation stands in the disguised-as relation as the first argument. Furthermore, each process

is only disguised as one other process and thus stands in the disguised-as relation as the second

argument with one other process. The relation is therefore weaker than SR, and I hold that it

exemplifies OWR. Therefore, based on such evidence, I propose that drug druga is compatible with OWR.

5.2.4 Revised Intermediate Alternative Reciprocity

The pronoun drug druga is also compatible with revised IAR, as demonstrated by the following examples. In the example in (369), the seven musical groups replace each other on the stage sequentially.

(369) Gruppy, otkrytye “Ekzotikoj” (“Rjaba Mutant” “Bigudi”, Soda “Volga” groups.NOM opening.PTCP.NOM “Ekzotika” (“Rjaba Mutant” “Bigudi” Soda “Volga”

“Brat'ja po razumu”, F.R.U.I.T.S. “Burunduk kvartet”, Los Chikatilos), smenjali “Brat'ja po razumu” F.R.U.I.T.S. “Burunduk kvartet”, Los Chikatilos) replace.PST.PL

drug druga na scene, kak maršrutki vozle oživlёnnoj stancii each other.ACC on stage.LOC like maršrutkas.NOM near lively.GEN station.GEN

metro. metro.GEN ‘The groups, opened by “Ekzotika” (“Rjaba Mutant” “Bigudi”, Soda “Volga” “Brat'ja po razumu”, F.R.U.I.T.S. “Burunduk kvartet”, Los Chikatilos), replaced each other on the stage like maršrutkas28 near a lively metro station.’

Only one group can appear on stage at once, and each group plays and is replaced by another

group. The chain-like replacing relation holds between each of the groups, and this relation is

28 A maršrutka is a form of transportation that is a cross between a taxi and a bus. It is essentially a shared taxi that typically has a set destination. 271

strengthened by the clause ‘like maršrutkas near a lively metro station’, since maršrutkas line up near a metro station and replace each other as they find passengers.

Similarly, in the following utterance, the plants replace each other as they bloom throughout the year:

(370) S rannej vesny do pozdnej oseni, smenjaja drug druga, from early.GEN spring.GEN until late.GEN fall.GEN, replace.VADV each other.ACC

cvetut badan, medunica, del'finnum, tysjačelistnik, , bloom.NPST.3PL bergenia.NOM, lungwort.NOM delphinium.NOM achillea.NOM,

monarda, lapčatka, zveroboj i jaskolka. bee balm.NOM, cinquefoil.NOM, St. John’s wort.NOM and chickweed.NOM ‘From early spring until late fall, replacing each other, bergenia, lungwort, delphinium, achillea, bee balm, cinque foil, St. John’s wort and chickweed bloom.’

Bergenia replaces lungwort, which replaces the delphiniums, etc. Each plant is directly or

indirectly related to each other plant through a chain. Therefore, these two sentences from the

RNC demonstrate that drug druga is compatible with the RIAR relation.

5.2.5 Interim summary

In this section I provided data suggesting that drug druga is compatible with the same six

relations that each other and the -sja verbs in the simple reciprocal constructions are. These data

support an analysis of drug druga as semantically equivalent to each other, based on how

semanticists have analyzed the semantics of reciprocity in the past. I demonstrated in the

introduction, however, that drug druga cannot be used in all of the same contexts as each other,

suggesting the two expressions do differ in some way. I explore this difference in the following

section.

272

5.3 Drug druga and accompaniment

In the introduction, I provided the following unacceptable sentence and noted that its unacceptability is puzzling, considering (i) it expresses a SR reciprocal meaning between the entities denoted by the subject and (ii) that other sentences do acceptably realize the phrase drug s drugom.

(371) Context: Your friends Ivan and Marija, who are dating, just left for the movies together. You get a phone call from someone asking for them. You say:

#Ivan i Marija pošli v kino drug s drugom. Ivan.NOM and Marija.NOM go.PST.PL to cinema.ACC each with other.INST (Intended: ‘Ivan and Marija went to the movies with each other.’)

Here, I propose that this utterance is unacceptable because drug druga cannot be used in an

accompaniment phrase.

In this section, I will provide background on accompaniment, including introducing the

terminology to be used throughout the remainder of this chapter. I will present a diagnostic for

determining whether a phrase introduces a relation of accompaniment. In Section 5.3.3, I will

provide evidence that drug druga cannot be used to express an accompaniment relation between

entities and in Section 5.3.4, I will provide an analysis.

5.3.1 Background on Accompaniment

Accompaniment, as a semantic concept, has proven difficult to define and characterize.

The formal literature has assumed that in English both the adverb together and a comitative

phrase headed by with can give rise to an accompaniment reading. Formal semanticists agree that

273

in order for a sentence to express accompaniment, two or more subevents must be expressed which have different arguments, and that these subevents must temporally and spatially overlap

(Lasersohn 1995, Kratzer 2003). They typically also accept that there is more to accompaniment than temporal and spatial overlap, but this “more” is harder to pin down.

For example, Lasersohn (1995) treats social accompaniment as a “non-logical notion which is not amenable to a formal definition” (235) but writes that for the reading of social accompaniment to arise, the referents must be “party” to the subevents expressed by the predicate. According to his analysis, the sentence in (372) gives rise to the social accompaniment reading.

(372) John and Mary went to the movies together.

The sentence in (372) expresses two subevents, e’, that of the referent of John going to the

movies, and e”, that of the referent of Mary going to the movies. The sentence has a reading of

social accompaniment because the referent of John is party to the subevent e” (that of the referent

of Mary going to the movies), and the referent of Mary is party to the subevent e’ (that of the referent of John going to the movies). If the individuals denoted by the sentence are party to the subevents expressed by the sentence, the social accompaniment reading arises.

Faller (2007) takes a different approach. She provides a compositional analysis for sentences that express accompaniment in Cuzco Quechua and proposes that an accompaniment reading requires, in addition to temporal and spatial overlap, that “the actions by the individual agents have a common beneficiary”(76). For example, she argues that in the sentence in (373),

274

each subevent has a beneficiary; the event of one individual laughing benefits that of the second individual, and the event of the second individual laughing benefits that of the first individual.

(373) Asi-puna-ku-n-ku pay-kuna pura. (Faller 2007: 78) laugh-RECIP-REFL-3-PL (s)he-PL amongst ‘They are laughing together/with each other.’ (description of a video clip in which two people are laughing)

She proposes this analysis in part because it is the suffix -ku in Cuzco Quechua that gives rise to the accompaniment reading, and this suffix, in one of its meanings, marks “coreferentiality of the subject with a (non-overt) beneficiary” (76). She does not explicitly discuss how this analysis might be extended for other expressions, but it seems possible that it could be extended to data in languages such as English and Russian.

Moltmann (2004) proposes that the adverb together introduces the property of integrated whole. In her analysis, an expression that has the property of being an integrated whole maps a sum of subevents to a truth value. It specifies that “the sum event composed of the subevents that the group members contribute must be an integrated whole.” (310-311). A sum event may fulfill the property of being an integrated whole in several ways. For example, integrity can be fulfilled by the subevents being mapped onto the same runtime. In the sentence in (374), the subevents have the property of being an integrated whole because they “coincide in time” (311).

(374) John and Mary went to the movies together.

275

This sentence is acceptable only in a context in which the event of the denotation of John going to

the movies and the event of the denotations of Mary going to the movies have the same runtime.

A sum of events may also have integrity if it “constitute[s] a certain type of single event” (312).

For example, Moltmann provides the following sentence:

(375) John and Mary solved a problem together. (Moltmann 2004: 312)

She writes that the relevant subevents are those “of the activities of John and of Mary that

contribute to the solution of the problem, but that are not themselves solvings of the problem.

Nonetheless, that sum has integrity in that it constitutes a solving of the problem” (312).

Moltmann proposes that together takes three arguments. It maps a sum of individuals onto a measure function which is mapped onto the integrated whole property. For example, she translates the sentence in (376) as in (377):

(376) John and Mary work together.

(377) ∃e(work(e,j+m) ˄ together(j⊕m, fe, INT-WH))

The fe function is a measure function that maps individuals onto a subevent which is the subevent that has the individual in its extension. A measure function is a function w from a structure (D, ⊕), which is closed under the sum formation operation to a structure (R, +), where

R is a set of real numbers and + is the operation of addition, such that for entities a, b, c ∈ D, if c

= a⊕b, then w(c) = w(a)+w(b).

276

The adverb together maps the complex individual denoted by the subject onto the sum of the subevents of which each individual in the group is in the extension. For example:

(378) John and Mary work together.

(379) ∃e(work(e, j+m) ˄ together(j⊕m, fe, INT-WH))

In this example, the measure function fe applies to j⊕m and will return the sum of the subevent of the denotation of John working and the subevent of the denotation of Mary working. The property of INT-WH, or integrated whole will apply to this sum, requiring that the subevents that are part of the sum of subevents are an integrated whole, as discussed above.

Although Moltmann does not discuss accompaniment directly, since others hold that together contributes the same semantic content as with (Lasersohn 1995, Kratzer 2003), her analysis could be extended to account for the meaning of accompaniment with. Under this assumption, with would also require that events are interpreted as an integrated whole.

The typological literature has been less concerned with characterizing accompaniment and rather has investigated how accompaniment is expressed and how it is related to other domains. For example, Lakoff and Johnson (1980) put forth the following universal: “With few exceptions, the following principle holds in all languages of the world: The word or grammatical device that indicates ACCOMPANIMENT also indicates INSTRUMENTALITY” (135). For example, the word with in English can take an object that is interpreted as the instrument by which one accomplishes something, as in the sentence in (380). The same preposition can also introduce a relation of accompaniment to a sentence as in (381).

277

(380) I ate the tart with a fork.

(381) John went to the movies with Mary.

The French preposition avec ‘with’ also can convey instrumentality or accompaniment as

demonstrated by the following two examples:

(382) J’enfonce le clou avec un marteau. (Stolz 1997: 116) I make.enter the nail with a hammer ‘I drive in the nail with a hammer.’

(383) Je suis parti avec mon frère. (Stolz 1997: 116) I am gone with my brother ‘I left with my brother.’

Lakoff and Johnson hold that cross-linguistically, as in English and French, a single marker

typically expresses both instrumentality and accompaniment.

More recent works have shown, however, that this proposed universal is not in fact

universal. Stolz (1997), for example, argues that it is actually more common for languages to use different grammatical devices to express accompaniment and instrumentality than to use a common grammatical device. Russian, for example, uses the preposition s to express accompaniment as in (384), but the instrumental case with no preposition to express instrumentality, as in (385):

278

(384) Ivan pošёl v kino s Mariej. Ivan.NOM go.PST.SG to movies.ACC with Marija.INST ‘Ivan went to the movies with Marija.’

(385) Marija pišet pis'mo ručkoj. Marija.NOM write.NPST.3SG letter.ACC pen.INST ‘Marija is writing a letter with a pen.’

Furthermore, it has been shown that what has been characterized simply as accompaniment in the past can in fact be further partitioned into distinct meanings such as ornativity and materiality

(Stolz 2006). Stolz writes that a comitative marker exemplifies ornativity when its object is “an ephemeral bodily property of an animate participant” (42). A comitative marker exemplifies a material usage when “the USER employs certain substances or other to achieve a certain goal”

(42). As examples of these subtypes of accompaniment, he provides the following example from

Estonian. The suffix -ga is a comitative marker. Stolz states that -ga has an ornative meaning in

the example in (386), as the word that the affix attaches to is a bodily attribute of the head of the

NP. The same marker has a material meaning in (387), as it marks a substance that was used to

achieve a goal, namely that of creating ice cream.

(386) Habeme-ga mees (Stolz 2006: 49) beard-COM man ‘bearded man’

(387) Sõin moosi-ga jääti-st (Stolz 2006: 49) eat.PST.1SG jam-COM ice.cream-ELA ‘I ate ice cream made of jam.’

279

Stolz therefore finds many different functions for comitative markers cross-linguistically. The challenge these findings pose for my work is to in determining which functions of a comitative marker are an accompaniment function and which are related but distinct functions.

In the following section, I will provide a diagnostic for determining whether a comitative

phrase has an accompaniment interpretation. By providing such a diagnostic, I am singling out a

semantic property of accompaniment, and so I am providing a partial characterization of the

lexical meaning of an accompaniment marker. I will therefore focus on the meaning that my

diagnostic diagnoses, though I recognize that there may be other properties of accompaniment

that I am not considering.

5.3.2 Definitions and a diagnostic for accompaniment

In this chapter I am concerned with phrases of the sort bolded in the sentence (388),

headed by the preposition s ‘with’, though I will often compare these phrases with English

equivalents, headed by with.

(388) Ivan pošёl v kino s Mariej. Ivan.NOM go.PST.SG to movies.ACC with Marija.INST ‘Ivan went to the movies with Marija.’

I consider the prepositional phrase in (388) an accompaniment phrase. A phrase is an accompaniment phrase if and only if it introduces a relation of accompaniment to a sentence. I follow Stolz (2006) by assuming that an accompaniment relation holds between two or more entities, at least one of which has the role of ACCOMPANEE, and one of which has the semantic

role of COMPANION. The ACCOMPANEE is the individual denoted by the argument of the verb,

280

and the COMPANION is denoted by an argument of another element, possibly a preposition such as

English with or Russian s.

As already mentioned, however, the prepositions with and s can be realized when they do not express accompaniment and therefore they have more than one meaning. Because of this, a diagnostic must be found to determine whether the prepositions with or s in a given sentence expresses an accompaniment relation.

Based on the fact that, as discussed above, the accompaniment with has the same

meaning as together, following McKercher (2001), I assume that accompaniment phrases can

acceptably co-occur with the adverb English together29, as in (389):

(389) John went to the movies together with Mary.

Since the prepositional phrase in (389) is compatible with the adverb together, I propose that with

Mary in this sentence is an accompaniment phrase. The prepositional phrase in (390), however, is not an accompaniment phrase, as it is not compatible with the adverb together.

(390) John and Mary are the two finalists in a 5-yard dash. In the last race, they are the only two running.

John competed (#together) with Mary.

29 I have found that some native speakers do not accept sentences such as that in (390) above with the adverb together. For these speakers, however, substituting the adverb along does result in an acceptable sentence, as in (i):

(i) John went to the movies along with Mary.

Therefore, while throughout this dissertation I will diagnose sentences with together, some speakers will have to use the adverb along instead. For these speakers, presumably the adverb along has the meaning that Moltmann (2004) proposed for together. 281

This diagnostic works because I assume, following Moltmann (2004), that the adverb together has only one meaning. Because it only has one meaning, unlike with, it allows me to diagnose comitative phrases for accompaniment readings.

In Russian, this diagnostic utilizes the adverb vmeste ‘together’. For example, the sentence in (391) is acceptable when the adverb vmeste ‘together’ is realized. Because of this, the prepositional phrase s Mariej ‘with Marija’ is an accompaniment phrase.

(391) Ivan pošёl v kino vmeste s Mariej. Ivan.NOM go.PST.SG to movies.ACC together with Marija.INST ‘Ivan went to the movies together with Marija.’

The sentence in (392), on the other hand, does not acceptably realize the adverb vmeste ‘together’ and therefore, based on my diagnostic, the prepositional phrase in this sentence is not an accompaniment phrase.

(392) Ivan and Marija are the two finalist in a 5-yard dash. In the last race, they are the only two running.

Ivan soperničaet (#vmeste) s Mariej. Ivan.NOM compete.NPST.3SG (together) with Marija.INST ‘Ivan competes with Marija.’

Since this sentence is not acceptable with the adverb vmeste ‘together’, I propose that the

prepositional phrase in this sentence is not an accompaniment phrase.

282

In what follows, it will be crucial to keep in mind that this diagnostic tests predicates and

not verbs. I assume that certain predicates are compatible with an accompaniment reading and

other predicates are not compatible with an accompaniment reading. Therefore, I hold that the

predicate idti v kino ‘go to the movies’ can co-occur with an accompaniment phrase, and the predicate soperničat' ’to compete’ cannot co-occur with an accompaniment phrase.

Consider, for example, the following two sentences, which acceptably realize both the predicate idti v kino ‘go to the movies’ and the adverb vmeste ‘together’.

(393) Každyj den' Ivan idёt v kino (vmeste) s Mariej. every.ACC day.ACC Ivan.NOM go.NPST.3SG to movies.ACC (together) with Marija.INST ‘Ivan goes to the movies (together) with Marija every day.’

(394) Ivan i Marija pošli v kino (vmeste) s Borisom. Ivan.NOM and Marija.NOM go.PST.PL to movies.ACC (together) with Boris.INST ‘Ivan and Marija went to the movies (together) with Boris.’

Because vmeste ‘together’ can co-occur with this predicate in all these examples, I propose that idti v kino ‘go to the movies’ is acceptable with an accompaniment phrase. I have not found a sentence in which vmeste ‘together’ is unacceptable with this predicate with a comitative phrase, suggesting that it is indeed the predicate that is acceptable with an accompaniment phrase and not the sentence itself.

On the other hand, the following three sentences, each of which realizes the predicate soperničat’ ‘to compete’, are all unacceptable with vmeste ‘together’.

283

(395) Ivan, Marija and Boris are playing tennis together. Ivan and Marija are on one team and are playing against Boris.

Ivan i Marija soperničajut (#vmeste) s Borisom. Ivan.NOM and Marija.NOM compete.NPST.3PL (together) with Boris.INST ‘Ivan and Marija compete with Boris.

(396) Yesterday, Ivan and Marija were the two finalist in a 5-yard dash. In the last race, they were the only two running.

Marija soperničala (#vmeste) s Ivanom. Marija.NOM compete.PST.SG (together) with Ivan.NOM ‘Marija competed with Ivan.’

(397) Ivan and Marija sometimes race to see who is the fastest runner.

Ivan inogda soperničaet (#vmeste) s Mariej. Ivan.NOM sometimes play.NPST.3SG (together) with Marija.INST ‘Ivan sometimes competes with Marija.’

Indeed I have not found a sentence that realizes the predicate soperničat' ‘to compete’ and a comitative phrase in which vmeste is acceptable. For example, the context in (398) establishes a situation in which the subevents occur at the same time and place and thus overlap temporally and spatially, as the formal semanticists have agreed must be the case for a reading of accompaniment to arise. The sentence in (398) is still not acceptable with vmeste, however.

(398) Context: At 3pm yesterday, who did Ivan race with?

Ivan soperničal (#vmeste) s Mariej. Ivan.NOM compete.PST.SG (together) with Marija.INST ‘Ivan competed with Marija.’

284

In what follows, I will first test predicates to determine whether they are compatible with

vmeste plus a comitative phrase. If they are compatible, then I predict the predicates will not be

compatible with the phrase drug s drugom. On the other hand, if a predicate is compatible with

drug s drugom, I predict it will not be compatible with vmeste s NP.

Determining which predicates in a language an accompaniment phrase can subcategorize for and which it cannot would be an interesting task in and of itself. It would provide insight into the nature of accompaniment. While I make claims about a number of predicates in Russian, a broader exploration of predicates is outside the scope of this dissertation and is therefore necessarily left to future research.

Finally, armed with a definition of an accompaniment phrase, I define the term comitative

marker. The term comitative is often used cross-linguistically to refer to a morphosyntactic

element that, in at least one of its uses, contributes a relation of accompaniment to a sentence

(Stolz et al. 2006, Sirola-Belliard 2011, Trawinski 2012). It is often also pointed out that markers

of comitativity, such as English with, can express other meanings such as instrumentality (399),

possession (400), manner (401), etc.

(399) John ate the pizza with a fork and knife.

(400) The boy with the curly hair ran by the house.

(401) Mary ate the pizza with gusto.

I follow the majority of the literature by using the term comitative marker to refer to any

morphosyntactic element that introduces a relation of accompaniment to a sentence, even if it can 285

also have a different semantic function. Therefore, with is a comitative marker, since it

contributes a relation of accompaniment even though it can also convey possession or manner. If

a comitative marker is a preposition, then the preposition with its object is a comitative phrase.

An accompaniment phrase is always a comitative phrase, but comitative phrases are not always

accompaniment phrases. The term comitative phrase refers to a morphosyntactic element (that is

defined based on one of its meanings), while the term accompaniment phrase refers to both the

morphosyntax and semantics of a phrase.

5.3.3 Evidence that drug s drugom does not express accompaniment

In this section, I provide data to support the hypothesis that drug s drugom cannot be

realized as an accompaniment phrase. Evidence in support of this hypothesis would be finding

that (i) if a predicate realizes drug s drugom, then this predicate cannot co-occur with a

comitative phrase with an NP object that is compatible with vmeste and (ii) if a sentence

acceptably realizes both a comitative phrase with an NP object and vmeste, then drug druga

cannot be realized as the object of the comitative phrase.

First, consider the following three contexts and sentences that realize drug s drugom that

my consultants judged acceptable:

(402) Context: Your family is trying to decide where to go on vacation. Most of your family wants to travel to Helsinki, but Ivan and Marija want to go to Budapest. You say:

Ivan i Marija soglasili-s' drug s drugom. Ivan.NOM and Marija.NOM agree.PST.PL-SJA each with other.INST ‘Ivan and Marija agreed with each other.’

286

(403) Context: Ivan and Marija don’t like playing tennis together. When they show up to the tennis courts, however, they find that everyone else in their group is paired up, and so they must play against together. You say:

Ivan i Marija igrali v tennis drug s drugom. Ivan.NOM and Marija.NOM play.PST.PL at tennis.ACC each with other.INST ‘Ivan and Marija played tennis with each other.’

(404) Context: What is the prize for winning this contest?

Učastniki soperničajut drug s drugom za zolotuju medal'. participants.NOM compete.NPST.3PL each with other.INST for gold.ACC medal.ACC ‘The participants are competing with each other for the gold medal.’

Since each of these sentences that realize drug s drugom was judged as acceptable, if my hypothesis is to prove correct, the comitative phrases that realize drug druga in each are not

accompaniment phrases.

Support for this proposal comes from the following examples that realize the same verbs

as those in (402) - (404) but without drug druga. If the drug s drugom phrases above are

accompaniment phrases, the sentences below should be acceptable with the adverb vmeste. For

example, the sentence in (405) realizes the same verb soglasit'sja ‘to agree with’ as in (402) but with the NP Marija as the object of the comitative phrase. The sentence in (405) was judged

unacceptable by my consultants.

(405) Context: Your family is trying to decide where to go on vacation. Most of your family wants to travel to Helsinki, but Marija very vocally wants to go to Budapest. Regarding Ivan, you say:

#Ivan soglasil-sja vmeste s Mariej. Ivan.NOM agree.PST.SG-SJA together with Marija.INST (Intended: ‘Ivan agreed with Marija./Ivan and Marija agreed.’)

287

Based on my diagnostic, since the sentence in (405) which realizes the adverb vmeste is not

acceptable, the comitative phrase s Mariej ‘with Marija’ is not an accompaniment phrase. Since this sentence realizes the same predicate soglasit'sja ‘to agree with’ as the sentence in (402), I suggest that the reciprocal comitative phrase in (402) is not an accompaniment phrase.

Similarly, the next example realizes the same predicate igrat' v tennis ‘to play tennis’ as

in (403), except that the prepositional phrase realizes an NP instead of drug druga.

(406) Context: Ivan doesn’t like playing tennis with Marija. When he showed up to the tennis courts, however, he found that everyone but Marija is already paired up, and so he must play against Marija. You say:

#Ivan igral v tennis vmeste s Mariej. Ivan.NOM play.PST.SG at tennis.ACC together with Marija.INST (Intended: ‘Ivan and Marija played tennis with each other.’ )

Since the sentence that realizes the phrase vmeste s Mariej ‘together with Marija’ is not

acceptable, I propose that the comitative phrase s Mariej in (403) above is not an accompaniment

phrase. Lastly, the sentence in (407) in comparison with the sentence in (404), both of which

contain the predicate soperničat' ‘compete’, suggests the same.

(407) Context: What is the prize for winning this contest?

#Amerikanec soperničaet vmeste s ital'jancem za zolotuju medal'. American.NOM compete.NPST.3PL together with Italian.INST for gold.ACC medal.ACC (Intended: ‘The American is competing with the Italian for the gold medal.’)

Since my consultants judged the three sentences that realize the comitative phrase with vmeste in

(405), (406) and (407) as unacceptable, I propose that the predicates realized in these sentences 288

are not acceptable with accompaniment phrases. Furthermore, this finding suggests that the reciprocal comitative phrases in (402), (403) and (404) are not accompaniment phrases. In addition to these predicates I have tested other predicates such as svjazyvat' ‘to tie’, govorit' ‘to talk’ and vstrečat'sja ‘to meet’. Indeed, I have not yet found a predicate through elicitation that is compatible with a reciprocal comitative phrase but which is also compatible with the phrase vmeste s NP. I have therefore shown that when drug s drugom occurs with a particular predicate, that predicate does not co-occur with an accompaniment phrase.

I now turn to showing that when an accompaniment s is realized in a sentence, that it cannot take drug druga as an object. I provided my consultants with the following three sentences that realize the adverb vmeste along with a comitative phrase that realizes a noun, all of which they judged as acceptable:

(408) You and your friends Ivan and Marija went to the movies together. The movie was so popular that even though Marija arrived early, she could only save one seat. You and Ivan arrived a little late. Ivan took the seat near Marija, and you sat alone. You later say:

Ivan sidel vmeste s Mariej v kino. Ivan.NOM sit.PST.SG together with Marija.INST at movies.LOC ‘Ivan sat together with Marija at the movies’

(409) Context: Who did Ivan listen to music with?

Ivan slušal muzyku vmeste s Mariej. Ivan.NOM listen.PST.SG music.ACC together with Marija.INST ‘Ivan listened to music together with Marija.’

(410) Context: Who did Anja walk with?

Anja guljala vmeste s Tanej. Anja.NOM walk.PST.SG together with Tanja.INST ‘Anya walked together with Tanja.’ 289

Since these sentences were accepted when the comitative phrase co-occurred with vmeste, I propose that the comitative phrases are accompaniment phrases and thus that the predicates in each of these sentences are compatible with an accompaniment phrase. Crucially, my consultants did not find the following three sentences acceptable that realize the same predicates as in (408) -

(410), but where the object of the prepositional phrase is drug druga:

(411) You and your friends Ivan and Marija went to the movies together. You arrived there a little late, so many of the seats were already taken. You had to sit alone, but Ivan and Marija got seats next to each other. You later say:

#Ivan i Marija sideli drug s drugom v kino. Ivan.NOM and Marija.NOM sit.PST.PL each with other.INST in movies.LOC (Intended: ‘Ivan and Marija sat with each other at the movies.’)

(412) Context: Your friends Ivan and Marija, while on a study abroad trip, were invited to listen to music in Oleg’s room. They decided against it however and instead went into Ivan’s room to listen to music alone. You say:

#Ivan i Marija slušali muzyku drug s drugom. Ivan.NOM and Marija.NOM listen.PST.PL music.ACC each with other.INST (Intended: ‘Ivan and Marija listened to music with each other.’)

(413) Context: The women in your college class were invited by the men to go walking in the park. They decided not to walk with the men however. You say:

#Devuški guljali drug s drugom. girls.NOM walk.PST.PL each with other.INST (Intended: ‘The girls walked around with each other.’)

To convey these meanings, my consultants instead used the adverb vmeste instead of drug s

drugom:

290

(414) You and your friends Ivan and Marija went to the movies together. You arrived a little late, so many of the seats were already taken. You had to sit alone, but Ivan and Marija found seats next to each other. You later say:

Ivan i Marija sideli vmeste v kino. Ivan.NOM and Marija.NOM sit.PST.PL together in movies.LOC ‘Ivan and Marija sat together at the movies.’

(415) Context: Your friends Ivan and Marija, while on a study abroad trip, were invited to listen to music in Oleg’s room. They decided against it however and instead went into Ivan’s room to listen to music alone. You say:

Ivan i Marija slušali muzyku vmeste. Ivan.NOM and Marija.NOM listen.PST.PL music.ACC together ‘Ivan and Marija listened to music together.’

(416) Context: The women in your college class were invited by the men to go walking in the park. They decided not to walk with the men however. You say:

Devuški guljali vmeste. girls.NOM walk.PST.PL together ‘The girls walked around together.’

This pattern also held for other predicates, such as delat' ‘to do’, begat' ‘to run’, and užinat' ‘to

eat dinner’. Therefore, since none of my consultants found the utterances in (411) - (413)

acceptable, these data further suggest that drug s drugom is not compatible with an

accompaniment reading.

5.3.4 Analysis of accompaniment phrases

I have so far provided evidence that the phrase drug s drugom cannot be used as an

accompaniment phrase. In this section I provide a formal analysis for accompaniment phrases in

Russian. I also explore where the accompaniment reading comes from and discuss why drug s

drugom cannot express accompaniment. 291

I propose that it is the preposition s that introduces the relation of accompaniment to a sentence. Prepositions are traditionally analyzed as assigning semantic roles to their internal argument (their object) and an external argument (either their governor or the subject of the predicate), and I assume the same here. In Chapter 4, I suggested that the preposition s is

ambiguous; I argued that one of the functions of s is as a semantically vacuous preposition. The s

I analyze here has a different meaning than that explored in Chapter 4. Here I propose that the accompaniment s assigns the semantic role of COMPANION to its object and the role of

ACCOMPANEE to the denotation of the subject and that the lexical entry of this s is as in (417):30

(417) s ‘with’: (N\S’\(N\S’))/N: λyλPλxλev[∀a ∀b(a ≤ x ˄ b ≤ y → ∃e’∃e”( e’ ≤ ev ˄ e” ≤ ev ˄ P(a)(e’) ˄ P(b)(e”) ˄ COMPANION(b)(ev) ˄ ACCOMPANEE(a)(ev)))]

This preposition s combines first with a NP (category N), and then with an intransitive verb (or a transitive verb that has already composed with its object) and finally with another NP to result in a sentence radical. It guarantees that all entities a such that a is part of x, which is supplied by the subject of the verb, stand in the P relation to a subevent e’, and all entities b such that b is part of y, which is supplied by the object of the preposition, stand in the same P relation to a subevent e”,

and the b entities have the semantic role of COMPANION for event e and the a entities have the semantic role of ACCOMPANEE for the same event. For example, consider the sentence in (418):

(418) Who did Ivan go to the movies with last night?

Ivan pošёl v kino s Mariej. Ivan.NOM go.PST.SG to cinema.ACC with Marija.INST ‘Ivan went to the movies with Marija.’

30 I assume that semantic roles are sets of lexical entailments (Dowty 1989, 1991). 292

The translation of this sentence I propose is as in (419):

(419) ∃e(∀a∀b(a ≤ i ˄ b ≤ m → ∃e’∃e”(e’≤ e ˄ e” ≤ e ˄ go.to.movies’(a)(e’) ˄ go.to.movies’(b)(e”) ˄ COMPANION(b)(e) ˄ ACCOMPANEE(a)(e)) ˄ τ(e) ≪ now))

This translation says that the sentence in (418) is true iff there is an event e such that for all a and b such that a is a part of the entity denoted by Ivan and b is a part of the entity denoted by Marija, there are two subevents e’ and e” of e and a stands in the going-to-the-movies relation to the subevent e’ and b stands in the same relation to subevent e” and each entity b, i.e. the individual

denoted by Marija has the semantic role of COMPANION and each entity a, i.e. the individual

denoted by Ivan, has the semantic role of ACCOMPANEE, and the event e (and therefore the subevents) occurs before now.

I entertain two hypotheses for the unacceptability of drug druga with accompaniment s.

The first hypothesis is syntactic: according to this hypothesis, s cannot combine with drug druga

as an argument. The restriction would have to be written into the syntactic component of the

lexical entry of s in (417). The second hypothesis is semantic: according to this hypothesis, drug druga cannot be assigned the semantic role of COMPANION.

The syntactic analysis would be motivated if we found other syntactic restrictions on the

combination of accompaniment s, such as a restriction against co-occurring with another

anaphoric pronoun such as sebja ‘self’. I have found, however, that the phrase s soboj ‘with oneself’ can occur as an accompaniment phrase.

293

In the RNC, a search for the phrase s soboj ‘with oneself’ returns 19,431 results. Many

of these phrases are in fact accompaniment phrases, based on my vmeste ‘together’ diagnostic.

For example, consider the following sentence that realizes vmeste ‘together’ and s soboj ‘with oneself’:

(420) Pust' ne sotnju, ne desjatok, no odnogo-dvux čelovek oni objazatel'no Let NEG hundred.ACC NEG ten.ACC but one-two.ACC person.ACC they.NOM necessarily.ADV

unesut vmeste s soboj v mogilu. take.NPST.3PL together with self.INST to grave.ACC ‘Let them take not a hundred or tens, but rather one to two people together with themselves to the grave.’

Since the predicate unesti ‘to take’ is compatible with the adverb vmeste, it is also compatible with an accompaniment phrase. This predicate also co-occurs with s soboj ‘with oneself’, suggesting that s soboj can have an accompaniment interpretation. The following sentence also from the RNC provides a second example of s soboj and vmeste occurring with the same predicate. In this sentence, based on the previous context provided, eё refers to a cat.

(421) No esli ty xočeš' priučit' eё guljat' vmeste s soboj, But if you.NOM want.NPST.2SG train.INF her.ACC walk.INF together with self.INST

povodok vse že ponadobit'sja - ne prostoj, a special'nyj. leash.NOM all.NOM EMPH is.need.INF NEG simple.NOM but specialized.NOM ‘But if you want to train her to walk together with yourself, a leash is needed - not a simple one, but a specialized one.’

294

Such examples are a problem for the syntactic analysis, since the existence of these examples suggest that s can select for an anaphoric pronoun.

On the other hand, I also find a problem with the semantic explanation, that drug druga is not compatible with the role of COMPANION. As part of my analysis above, I argued that the

accompaniment s semantically selects an argument to assign the COMPANION role to. Just as with

any role assignment, this s requires that the argument it assigns the role to have certain properties,

e.g. that the entity has to be animate. To argue that drug druga is not compatible with the

COMPANION role, it must be assumed that s requires that its object have a property that drug

druga does not have, or, alternatively, that drug druga have a particular property that s cannot

select for, and thus there is a violation in selection restrictions.

If we were to consider data solely from Russian, this explanation would appear to be superior to the syntactic explanation, as I do not know of any other restrictions on accompaniment s in Russian. A major challenge to this proposal, however, is that the accompaniment with in

English is compatible with each other. This being the case, a property would have to be found that each other has that drug druga does not, or vice versa, such that each other can be assigned the role of COMPANION, while drug druga cannot. Such a difference is theoretically possible and we have such little data on the semantics of reciprocal expressions cross-linguistically that we cannot depend on empirical data to argue either way. I have found no independent evidence though that suggests that each other has different properties than drug druga. Therefore, there also is a serious challenge to the semantic explanation for why the phrase drug s drugom is not used as an accompaniment phrase.

In sum, although I have presented two possible explanations for the restriction against the phrase drug s drugom expressing accompaniment, there are problems with each explanation. I

295

again propose that more data on more expressions cross-linguistically will help elucidate the best explanation for the restriction.

5.3.5 Interim summary

In Section 5.3, I have provided evidence supporting my proposal that the phrase drug s drugom is not realized as an accompaniment phrase and furthermore that this is because the accompaniment preposition s that does not subcategorize for drug druga.

I have however shown that the phrase drug s drugom is syntactically possible, and so I turn my attention in the next section to discussing when this phrase is acceptable and what its semantics are.

5.4 Drug druga and discontinuous constructions

Consider the following sentence, repeated from (402) above, that acceptably realizes

drug s drugom:

(422) Context: Your family is trying to decide where to go on vacation. Most of your family wants to travel to Helsinki, but Ivan and Marija want to go to Budapest. You say:

Ivan i Marija soglasili-s' drug s drugom. Ivan.NOM and Marija.NOM agree.PST-SJA each with other.INST ‘Ivan and Marija agreed with each other.’

Recall that I argued that the comitative phrase in this sentence is not an accompaniment phrase because the following sentence, repeated from (405) above, is not acceptable:

296

(423) Context: Your family is trying to decide where to go on vacation. Most of your family wants to travel to Helsinki, but Marija very vocally wants to go to Budapest. Regarding Ivan, you say:

#Ivan soglasil-sja vmeste s Mariej. Ivan.NOM agree.PST-SJA along with Marija.INST (Intended: ‘Ivan agreed with Marija./Ivan and Marija agreed.’)

Now, consider the following sentence taken from the RNC that also realizes the comitative phrase drug s drugom:

(424) Kstati, rabotniki Hermes drug s drugom obščajut-sja kak by-the-way workers.NOM Hermes.GEN each with other.INST converse.NPST.3.PL-SJA as

budto u nix vnutri mikroskopy. if at they.GEN inside microscopes.NOM ‘By the way, the Hermes workers converse with each other, as if they have microscopes inside of themselves.’

The predicate in the sentence in (424) is obščat'sja ‘to converse’. Based on my diagnostic, drug s drugom is not an accompaniment phrase because the following sentence which realizes a comitative phrase with an NP instead of drug druga and the adverb vmeste was not acceptable to my consultants:

(425) #Kstati, rabotniki Hermes vmeste s studentami obščajut-sja by-the-way workers.NOM Hermes.GEN along with students.INST converse.NPST.3PL-SJA

kak budto u nix vnutri mikroskopy. as if at they.GEN inside microscopes.NOM (Intended: ‘By the way, the Hermes workers converse along with the students, as if they have microscopes inside of themselves.’)

297

These instances of drug s drugom fail the diagnostic, and therefore the s in these phrases does not

contribute a relation of accompaniment.

I propose in fact that both of these sentences realize a discontinuous reciprocal

construction as discussed and analyzed in Chapter 4. Note that both of the verbs in these two

sentences are reciprocal -sja verbs. Since they are reciprocal -sja verbs, they can participate in

the discontinuous construction (all reciprocal -sja verbs can). Recall that the discontinuous

construction, as discussed in Chapter 4, instantiates the string V-sja s NP. I now propose that the

string V-sja drug s drugom is also an example of the discontinuous construction. Without

evidence contradicting such an analysis, the natural conclusion is that since the syntax and

semantics are the same, that the two constructions are in fact the same construction.

For the syntactic analysis, recall that the comitative phrase that is part of the

discontinuous construction is a complement. Therefore, if drug s drugom in the above sentences

is part of a discontinuous construction, then it too should be a complement.

Unfortunately, as before, the obligatory/optional test for the phrase is not applicable here.

As previously discussed in Chapter 4, when the discontinuous construction is realized without its

comitative phrase, it has the same form as the simple reciprocal construction and so is

grammatical. For example, the acceptable sentence in (426) is the same as that in (424) except

that the phrase drug s drugom is not realized. If the comitative phrase is an argument, we expect

the sentence in (426) to be unacceptable. The sentence without the comitative phrase is

acceptable, but this sentence in fact exemplifies the simple construction now.

298

(426) Kstati, rabotniki Hermes obščajut-sja kak budto u nix by-the-way workers.NOM Hermes.GEN converse.NPST.3PL-SJA as if at they.GEN

vnutri mikroskopy. inside microscopes.NOM ‘By the way, the Hermes workers converse as if they have microscopes inside of themselves.’

However, the second test I used in Chapter 4 to determine whether the comitative phrase is an adjunct or complement works here too. Recall that I used the fixed preposition test, which assumes that verbs that take prepositional phrases as complements subcategorize for a particular preposition and may not take a different preposition. Therefore, we expect that if the sentences above with drug s drugom realize the discontinuous construction, then the comitative phrase cannot be replaced with a phrase headed by a different preposition. Indeed, sentences that realize a different preposition in place of s are unacceptable to my consultants. For example, the following sentence, which is the same as the sentence in (424), except that it realizes the preposition bez ‘without’ as the head of the prepositional phrase, was judged as unacceptable:

(427) #Kstati, rabotniki Hermes drug bez druga obščajut-sja by-the-way workers.NOM Hermes.GEN each without other.GEN converse.NPST.3PL-SJA

kak budto u nix vnutri mikroskopy. as if at they.GEN inside microscopes.NOM (Intended: ‘By the way, the Hermes workers converse alone (lit. without each other), as if they have microscopes inside of themselves.’)

Since this sentence is unacceptable, I propose that the reciprocal comitative phrase in (424) is a complement. Similarly, the following sentence is the same as that in (422) above, except that it also realizes the preposition bez instead of s:

299

(428) #Ivan i Marija soglasili-s' drug bez druga. Ivan.NOM and Marija.NOM agree.PST.PL-SJA each without other.GEN (Intended: ‘Ivan and Marija agreed without each other/the other.’ )

Again, the unacceptability of the sentence in (428) suggests that the phrase drug bez druga cannot replace drug s drugom, furthermore suggesting that the phrase drug s drugom in the above sentences is a complement. Thus, drug s drugom has the same syntactic status in these

constructions as the comitative phrase in the discontinuous constructions does.

I also argue that the semantics of the string V-sja s NP are the same as for V-sja drug s

drugom. Recall that in the discontinuous construction, the reciprocal relation holds between

entities denoted by the subject and entities denoted by the object. Also recall that drug druga

refers to the entities that are denoted by the subject. Therefore, if drug druga can be realized in a

discontinuous construction, the verb would draw entities for both sets from the set denoted by the

subject; one set from the subject and one set from drug druga, and thus from the subject as well.

This construction would then require that all the entities denoted by the subject stand in the

relevant relation with each entity denoted by the subject. Furthermore, recall that the sets denoted

by the subject and object are quantified over, such that each entity denoted by the subject stands

in the relevant relation to at least one entity denoted by the object. If the discontinuous

construction can realize drug s drugom as its comitative phrase, then this would mean that each

entity denoted by the subject of the reciprocal discontinuous construction stands in the relation as

the first argument to at least one other entity denoted by the subject. I provide the following

sentence as an example of the reciprocal discontinuous construction in which these restrictions on

the relations between participants hold:

300

(429) Context: The family next door wants to buy a pet. The parents want to get a dog, but the kids, Tanja, Ljuda and Miša want to get a cat. You know there was a split in the family, and you ask who agreed with whom.

Tanja, Ljuda i Miša soglasili-s' drug s drugom. Tanja.NOM Ljuda.NOM and Miša.NOM agree.PST.PL-SJA each with other.INST ‘Tanja, Ljuda and Miša agreed with each other.’

In the context, the three children are in agreement, and therefore each child stands in the agreeing

relation as the first and second argument to each other child. The fact that this sentence is

compatible with this context suggests that a construction of the type V-sja drug s drugom is

compatible with a situation in which each member of the set denoted by the subject of the -sja

verbs stands in the relevant relation to at least one other member of the set denoted by the subject.

Compare this context and sentence with those in (430):

(430) Context: The family next door wants to buy a pet. Tanja and Ljuda, the two young girls, want to get a dog while Miša and Kolja, the boys, want to get a cat. You know there was a split in the family, and you ask who agreed with whom.

#Tanja, Ljuda i Miša soglasili-s' drug s drugom. Tanja.NOM Ljuda.NOM and Miša.NOM agree.PST.PL-SJA each with other.INST ‘Tanja, Ljuda and Miša agreed with each other.’

My consultants judged the sentence as unacceptable given the context. I propose it is

unacceptable because there is an entity denoted by the subject, namely Miša, that does not stand

in the relation as the first argument to another entity denoted by the subject. Consider, however,

the sentence in (431), where each entity denoted by the subject does stand in the agreeing relation

as the first argument to at least one other entity denoted by the subject.

301

(431) Context: The family next door wants to buy a pet. Tanja and Ljuda, the two young girls, while Miša and Kolja, the boys, want to get a cat. You know there was a split in the family, and you ask who agreed with whom.

Devočki i mal'čiki soglasili-s' drug s drugom. girls.NOM and boys.NOM agree.PST.PL-SJA each with other.INST ‘The girls and boys agreed with each other.’

My consultants judged this sentence as acceptable, suggesting that the same restrictions are placed on the relations between participants in a sentence that realizes drug s drugom as in a sentence that realizes the discontinuous construction.

Based on both these syntactic and semantic criteria, I therefore propose that the s in drug s drugom in the sentences above with obščat'sja ‘to converse’ and soglasit'sja ‘to agree’ has the same meaning as when in the discontinuous construction when it selects an NP. I therefore propose that one meaning of s when it occurs in the phrase drug s drugom is that of the s in the

discontinuous construction, i.e. it is semantically null and only has a functional meaning.

5.4.1 Naturally reciprocal verbs and drug s drugom

In Chapter 4, I proposed that the preposition s that is realized in the discontinuous

construction is a semantically null preposition. I originally defined the discontinuous construction

as one with the structure of V-sja s NP. In Section 5.3 of this chapter, I expanded the definition

to also include strings of the type of V-sja drug s drugom. Here I suggest that a discontinuous

construction can also have the string V[naturally reciprocal] s NP or V[naturally reciprocal] drug s drugom.

Note that the verb in this construction no longer must be a reciprocal -sja verb.

Recall that in Chapter 1 I defined a verb as naturally reciprocal if it has no overt marking of reciprocity, and yet it conveys a reciprocal relation. To argue that naturally reciprocal verbs

302

can participate in the discontinuous construction, I show that the comitative phrase that occurs with the verb is a complement and that the construction has the same meaning as the discontinuous constructions with a -sja verb, as I did in Section 5.4 above, when arguing that drug s drugom can be realized in the discontinuous construction.

To test the syntactic status of the comitative phrase, I again use the fixed preposition test.

The unacceptable sentence in (433) is the same as that in (432), except that it realizes the

preposition bez ‘without’ rather than s.

(432) Poka oba soveta … soperničajut drug s drugom za glavnuju meanwhile both.NOM council.GEN compete.NPST.3PL each with other.INST for main.ACC

rol' v partii. role.ACC in party.LOC ‘Meanwhile, both councils … compete with each other for the main role in the party.’

(433) #Poka oba soveta … soperničajut drug bez druga za meanwhile both.NOM council.GEN compete.NPST.3PL each without other.GEN for

glavnuju rol' v partii. main.ACC role.ACC in party.LOC (Intended:‘Meanwhile, both council … compete alone (lit. without each other) for the main role in the party.’)

Since my consultants judged it unacceptable, I propose that the comitative phrase is a complement of the verb soperničat' ‘to compete’, providing support for my claim that the sentence in (432) realizes a discontinuous construction.

In terms of its semantics, we have already seen that the s in this sentence does not express accompaniment. I now provide data suggesting that the meaning of the construction with

303

soperničat' ‘to compete’ has the same meaning as a discontinuous construction. Recall the

sentence in (291) from Chapter 4, which realizes a discontinuous construction with the verb

pozdravit'sja ‘to greet each other’. The context establishes a greeting relation between each girl

and each boy. The sentence in (291) is acceptable in this context.

(434) Context: There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each of the boys greets each girl, and each girl greets each of the boys.

Mal'čiki pozdravali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and the girls greeted each other.’

The context in (435) is similar to that in (291), in that it establishes a relation between each girl

and each boy. The sentence in this example is also acceptable, suggesting that the construction

with soperničat’ ‘to compete’ can be used in a similar context as the discontinuous construction.

(435) Context: Today at your school the kids are competing in running contests. Each race consists of one girl and one boy running. Each girl races with each boy, and each boy races with each girl.

Mal'čiki soperničajut s devočkami. boys.NOM compete.NPST3PL with girls.INST ‘The boys are competing with the girls.’

The sentence in (435) is compatible with a context in which each member of the set denoted by

the subject stands in the relevant relation, i.e. competing, with each member of the set denoted by

the object, just like sentences with a discontinuous construction with a reciprocal -sja verb. I

suggest that the s in the sentence in (435) has the same meaning of the s in the sentence in (291),

304

and therefore that this s is semantically null, the meaning of the preposition in the discontinuous

construction.

Furthermore, the construction with soperničat' ‘to compete’ has the same restriction on what relations can hold between which entities, namely that relations can only hold between entities denoted by the subject and entities denoted by the object. For example, the sentence in

(436) which again realizes the verb pozdravit'sja ‘to greet each other,’ is unacceptable in the context in which each student greets each other student, so that greeting relations hold between entities in the same set.

(436) Context: There are six kids in the room. Three of them are boys - Miša, Kolja and Mitja - and three are girls - Olja, Tanja and Galja. Each student greets each other student. You say:

#Mal'čiki pozdravali-s' s devočkami. boys.NOM greet.PST.PL-SJA with girls.INST ‘The boys and the girls greeted each other.’

Similarly, the sentence in (437) is unacceptable when the competing relation holds between each

entity, including competing relations between girls and competing relations between boys.

(437) Context: Today at your school all of your students, both boys and girls, are competing in one race. You say:

#Mal'čiki soperničajut s devočkami. boys.NOM compete. NPST.3PL with girls.INST ‘The boys are competing with the girls.’

Based on the fact that the construction with soperničat’ ‘to compete’ seems to impose the same

restrictions on which entities the relevant relation holds between as the discontinuous 305

constructions as pozdravit'sja ‘to greet each other’, I propose that the discontinuous construction can realize either a -sja verb or a naturally reciprocal verb.

Therefore, the comitative phrase in the sentence in (361) above, repeated in below in

(438), is the comitative phrase that is realized in a discontinuous construction. Examples of other sentences that realize the discontinuous construction with a naturally reciprocal verb are in (439) and (440) below.

(438) Poka oba soveta … soperničajut drug s drugom za glavnuju meanwhile both.NOM council.GEN compete.NPST.3PL each with other.INST for main.ACC

rol' v partii. role.ACC in party.LOC ‘Meanwhile, both councils … compete with each other for the main role in the party.’

(439) Ogromnye ob''ёmy dannyx . . . xranjatsja bessistemno i huge.NOM amounts.NOM data.GEN save.NPST.3PL.SJA unsystematically and

fragmentarno, i format ix xranenija ne pozvoljajut ix fragmentarily and formats.NOM their storage.GEN NEG allow.NPST.3PL them.ACC

svjazyvat' drug s drugom. connect.INF each with other.INST ‘Huge amounts of data are stored unsystematically and fragmentarily and the formats of their storage don’t allow them to connect with each other.’

(440) Ljudi budut razgovarivat' drug s drugom iz samyx otdalёnnyx people.NOM will.NPST.3PL talk.INF each with other.INST from most.GEN far.GEN

stran … countries.GEN ‘People will talk with each other from the farthest countries…’

306

I argue therefore that each of these three sentences realizes the discontinuous construction with a naturally reciprocal verb, and thus that the same semantic analysis that applies to discontinuous

constructions with a reciprocal -sja verb applies to these constructions as well.

5.4.2 Analysis of the discontinuous construction

In the preceding sections, I demonstrated that drug druga can be realized as the object of the

comitative phrase in the discontinuous construction. I now develop an analysis of this

construction with drug druga. In this analysis, the syntax and semantics of drug druga remain

the same; rather it is the preposition whose syntax and semantics are adapted to combine with

drug druga. Recall that I defined the s that is realized in the discontinuous construction as in

(441):

(441) s ‘with’: N/N: λx [x]

This s takes an object of syntactic type N and semantic type e and returns an object of the same

type. Its status as a complement is captured because it only has one argument that is filled by an

N of type e. This lexical entry has the following counterpart that combines with drug druga:

(442) s ‘with’: (((N\S’)/N)\(N\S’))/(((N\S’)/N)\(N\S’)): λP>[P]

This s has an argument of the syntactic and semantic type that drug druga is and returns an entity

of the same type.

307

(443) drug s drugom: ((N\S’)/N)\(N\S’): λPλyλev[∀a∀b(a ≤ y ˄ b ≤ y ∃e’∃e”(e’≤ ev ˄ e” ≤ ev ˄ P(a)(b)(ev) ˄ P(b)(a)(ev))]

This category takes a transitive verb as its first argument. The -sja verb in the discontinuous construction is syntactically transitive. After drug s drugom has combined with the transitive verb, it combines with the denotation of the subject. The transitive verb draws entities first from the set denoted by the object of the preposition and then from the set denoted by the subject and requires that the entities in the first set stand in the relevant relation to entities in the second set.

If the transitive verb combines with drug druga, it requires that the entities denoted by drug druga stand in the relevant relation to entities denoted by the subject of the verb. The entities denoted by drug druga, based on its definition, will be those entities that are denoted by the subject of the verb. Therefore, a sentence with the structure NP V-sja drug s drugom will require that the entities denoted by the subject stand in the relevant relation to the entities denoted by the subject. Such a result seems appropriate to capture the meaning of these sentences.

5.5 Possible counterexamples

So far, I have argued (i) that the phrase drug s drugom does not occur as an accompaniment phrase and (ii) that the phrase drug s drugom is realized as the comitative phrase in the discontinuous construction. In my research I have found a number of counterexamples to these claims, however. For example, since I used vmeste as a diagnostic for an accompaniment phrase, my analysis predicts that sentences that realize both vmeste and the phrase drug s drugom as an accompaniment phrase should not be found. The following three sentences, however, taken from the RNC realize both of these expressions:

308

(444) … v slučajax suprugov, proživšix uže desjatiletija vmeste drug s in cases.LOC spouses.GEN live.PTCP.GEN already ten.years.ACC together each with

drugom … other.INST ‘… in cases of spouses living already tens of years together with each other …’

(445) Obvertyvat' katuški lentoj sleduet do kreplenija ix wrap.INF spools.ACC ribbon.INST must.NPST.3SG until bindings.GEN them.ACC

vmeste drug s drugom. together each with other.INST ‘It is necessary to wrap the spools with ribbon to bind them together with each other.’

(446) Vse dyšit vmeste drug s drugom. everything.NOM breathe.NPST.3SG together each with other.INST ‘Everything breathes together with each other.’

These are the only three sentences that realize the string vmeste drug s drugom, where vmeste is realized immediately preceding drug s drugom. All three of my consultants accepted the sentence in (444), and two of the three consultants accepted the sentences in (445) and (446). See my discussion directly below about the verb žit’ ‘to live’ that is relevant for the participle

proživšix ‘living’ in (444), as these two verb forms have the same root. Through elicitation, I failed to find any other sentences that realized the predicate obveryvat' ‘to wrap’ with vmeste or dyšat' ‘to breathe’ with drug s drugom. I therefore consider these two sentences as exceptions,

though more work is called for to determine whether this is so.

Furthermore, consider the following three sentences, also extracted from the RNC:

309

(447) … čem plotnee naši specialisty budut rabotat' drug s that consistent.COM our.NOM specialists.NOM be.NPST.3PL work.INF each with

drugom, tem lučše eto otrazitsja na sostojanii našix ekonomik . . . other.INST that better it.NOM reflect.NPST.3SG on condition.LOC our.GEN economists.GEN ‘… the more consistently our specialists work with each other, the better it will reflect on the condition of our economics.’

(448) Prežde vsego eti knjaz'ja družno živut drug s before everything.GEN these.NOM kings.NOM friendly.ADV live.NPST.3PL each with

drugom. other.INST ‘First of all, these kings lived peacefully with each other.’

(449) I, nasmotrevšis' i nasnimavšis' kino, spjat drug s drugom and watch.alot.PTCP and shoot.alot.PTCP cinema.ACC sleep.NPST.3PL each with other.INST

besporjadočno -- kak v kino. indiscriminately.ADV as in cinema.LOC ‘And, having watched and shot so many movies, they sleep with each other indiscriminately- just like in the movies.’

The predicates used in these sentences are rabotat' ‘to work’, žit’ ‘to live’ and spat' ‘to sleep’.

Since these predicates co-occur with drug s drugom in each of the sentences, based on my claim above, that drug s drugom is not realized as an accompaniment phrase, my claims predict that these predicates cannot co-occur with vmeste. Consider, however, the three sentences in (450) -

(452), which realize these same verbs, the adverb vmeste, and a comitative phrase:

(450) …Vašington rabotaet vmeste s Londonom… Washington.NOM work.NPST.3SG together with London.INST ‘ …Washington works together with London…’

310

(451) Mat'-odinočka Keti … živёt vmeste s maloletnej dočer'ju v mother-single.NOM Keti.NOM live.NPST.3SG together with young.INST daughter.INST in

ne bol'šoj kvartirke … NEG big.LOC apartment.LOC ‘A single mother, Keti … lives together with her young daughter in a small apartment …’

(452) Maša spit vmeste s mamoj v malen'koj komnate. Maša.NOM sleep.NPST.3SG together with mother.INST in small.LOC room.LOC ‘Maša sleeps together with her mother in a small room.’

There are many examples of each of these verbs co-occurring with vmeste in the RNC. By searching just singular forms of these verbs in the past and present and asking for searches where vmeste immediately follows the verb (due to search limitations in the RNC), there are 29 occurrences of spat' vmeste ‘sleep together’, 138 instances of rabotat' vmeste ‘work together’, and 210 instances of žit' vmeste ‘live together’. Therefore, based on my diagnostic, the reciprocal comitative phrase that can co-occur with these verbs is an accompaniment phrase.

Such a conclusion poses the question as to why there are exceptions and furthermore what it is about these three verbs (though there might be more) that participate in the exception.

While I cannot be sure there are no other verbs that are acceptable with drug s drugom, I have not come across them. While having only three such verbs makes any hypotheses premature, I do have some further comments.

First, as far as I can tell, the phrase spat' drug s drugom ‘sleep with each other’ can only be used to discuss the sexual relations of entities, not the location where they sleep. For example, none of my consultants accepted the following sentence given the context:

311

(453) Context: Your small children Ivan and Tanja are brother and sister and are both afraid of the dark, so they often sleep in the same bed together. You tell a friend:

#Ivan i Tanja spjat drug s drugom. Ivan.NOM and Tanja.NOM sleep.NPST.3PL each with other.INST (Intended: ‘Ivan and Tanja sleep with each other.’)

They all accepted the following sentence in the same context:

(454) Ivan i Tanja spjat vmeste. Ivan.NOM and Tanja.NOM sleep.NPST.3PL together ‘Ivan and Tanja sleep together.’

Given these data, I propose that the phrase spat' drug s drugom ‘sleep with each other’ has an idiomatic meaning, and therefore the verb spat' ‘to sleep’ is not semantically equivalent to the verb spat' when with drug s drugom. The idiomatic spat' then might indeed be more similar to a naturally reciprocal verb, thus explaining why it can co-occur with drug s drugom.

I have no such explanation for the other two verbs. I do have some further data, however, that question the acceptability of drug s drugom with these two predicates. Consider, for example, the following two sentences I provided my consultants. Only one of my three consultants accepted the sentence in (455), while none of them accepted the sentence in (456).

(455) ?Ivan i Marija rabotajut drug s drugom v banke. Ivan.NOM and Marija.NOM work.NPST.3PL each with other.INST in bank.LOC ‘Ivan and Marija work with each other at a bank.’

312

(456) #Ivan i Marija rabotajut drug s drugom nad programmoj. Ivan.NOM and Marija.NOM work.NPST.3PL each with other.INST on program.INST (Intended: ‘Ivan and Marija are working with each other on a program.’)

The sentences in (457) and (458) are the same as those in (455) and (456), except that they do not realize the phrase drug s drugom. Since the sentences in (457) and (458) are acceptable, I propose it is the presence of drug s drugom in (455) and (456) that causes the sentences to be unacceptable for my consultants.

(457) Ivan i Marija rabotajut v banke. Ivan.NOM and Marija.NOM work.NPST.3PL in bank.LOC ‘Ivan and Marija work in a bank.’

(458) Ivan i Marija rabotajut nad programmoj. Ivan.NOM and Marija.NOM work.NPST.3PL on program.INST ‘Ivan and Marija are working on a program.’

Once again, for the sentences in (455) and (456), my consultants preferred a version of the sentences that included vmeste ‘together’ instead of drug s drugom.

Therefore, while the use of drug s drugom is certainly sometimes possible with these verbs, its usage is still restricted more than that of with each other. More data is needed to determine why such examples occur in the RNC, but my consultants do not always accept such sentences. It is possible that there is some difference between the sentences in the RNC and those that I provided my consultants that is significant, but if so, this factor is still unknown.

313

It is possible that my diagnostic is not sharp enough and that the comitative phrases in rabotat' drug s drugom ‘work with each other’ and žit' drug s drugom ‘live with each other’ are not actually accompaniment phrases. This hypothesis assumes that the s in these phrases is semantically distinct from the s that cannot take drug druga as an object. One s introduces a relation of accompaniment to a sentence, and the other has a different semantic contribution. One possibility is that this s is a collective operator that composes the denotation of the subject with that of the object of the comitative, such that both individuals participate in the event together, as a team. More work is needed to determine whether such a hypothesis can be supported.

I hypothesize, however, that the sentences that realize both vmeste and drug s drugom are on the edge of acceptability, such that some speakers accept them, some do not, and the linguistic and extra-linguistic context can tip the acceptability in one direction or another. I briefly mentioned such a situation in connection with the IAR relation and each other. I have found the sentence in (459) to be unacceptable for most speakers, but I have had one person tell me that it is acceptable. The sentence in (459) some people accept and some do not. And most people accept the sentence in (460), though one (a different person than the one who accepted the sentence in

(459)) has told me that the sentence in (461) is unacceptable.

(459) #John and Mary followed each other into the room.

(460) ?The three girls followed each other into the room.

(461) The girls followed each other into the room.

314

Therefore, there’s a possibility that the Russian examples with the phrases rabotat' drug s drugom and žit' drug s drugom are similar in that some people accept them in certain contexts, some people always accept such sentences, and some people never accept such sentences. Whether more concrete conclusions can be made about these possible counterexamples I leave to future work.

I have therefore found two predicates that can co-occur with an accompaniment phrase in

Russian. Other than these two verbs, however, as far as I can tell my hypothesis holds, in that drug s drugom, except in these two instances, is not used as an accompaniment phrase.

5.6 Implications and conclusion

I posed the question at the beginning of the chapter of why the sentence in (359),

repeated in (462), was judged unacceptable.

(462) Context: Your friends Ivan and Marija, who are dating, just left for the movies together. You get a phone call from someone asking for them. You say:

#Ivan i Marija pošli v kino drug s drugom. Ivan.NOM and Marija.NOM go.PST to cinema.ACC each with other.INST (Intended: ‘Ivan and Marija went to the movies with each other.’)

I have argued that the unacceptability of this sentence is due to the fact that when the preposition s introduces a meaning of accompaniment to a sentence, then this preposition does not select drug druga as an object and thus that drug druga is incompatible with the relation of accompaniment.

I have furthermore provided evidence suggesting that the preposition s does select for drug druga when s is semantically null and is part of a discontinuous construction. Lastly I explored a few exceptions to my proposal.

315

These conclusions leave the question of why drug druga is (mostly) not compatible with the relation of accompaniment, whereas English each other is compatible with it. One possible

explanation is that because the adverb vmeste can be used to express accompaniment, having drug s drugom be able to play the same role is redundant. For example, the following sentence was found to be acceptable with the adverb vmeste:

(463) Ivan i Marija pošli v kino vmeste. Ivan.NOM and Marija.NOM go.PST.PL to movies.ACC together ‘Ivan and Marija went to the movies together.’

A similar sentence that realizes the phrase drug s drugom instead of vmeste is not acceptable, as demonstrated by (462) above. This analysis at first glance seems appropriate because inevitably when I presented sentences with drug s drugom to my consultants, if the sentences were found unacceptable, when asked for a better variant, consultants always provided me with a sentence that realized vmeste.

This argument falls apart however, because as far as I can tell, all the acceptable sentences that realize drug s drugom can also be paraphrased with the adverb vmeste in place of the prepositional phrase:

(464) Ivan i Marija igrali v tennis vmeste. Ivan.NOM and Marija.NOM play.PST at tennis.ACC together ‘Ivan and Marija played tennis together.’

Since vmeste can be used in place of drug s drugom in all instances, an argument that assumes that sentences with vmeste block sentences with drug s drugom is problematic. Furthermore, in

316

English, the adverb together can be used in place of with each other in sentences that express a

relation of accompaniment (465)a and in sentences that do not (465)b:

(465) a. John and Mary went to the movies with each other/together.

b. John and Mary played tennis with each other/together.

Therefore, I argue that it is not the case that the existence of vmeste restricts drug s drugom from

also expressing accompaniment.

Therefore, I am left concluding only that reciprocal expressions in different languages are

compatible with different meanings. In some languages they are compatible with an

accompaniment relation and in some they are not. It is possible that such a restriction occurs only

with prepositions that introduce a relation of accompaniment because accompaniment and

reciprocity are closely related semantic domains. On the other hand it is also possible that there

are many such restrictions on reciprocal expressions cross-linguistically that are not related to

accompaniment. It is only through further investigations of reciprocal expressions that an answer

to this question will be found.

5.6.1 Drug druga is compatible with the same relations as each other

I have provided evidence suggesting that drug druga is compatible with the same set of

relations as each other and the -sja verbs in the simple reciprocal construction. Since three of the

four expressions discussed in this dissertation (each other, drug druga and the -sja verbs in the

simple construction) are compatible with the same set, it begs the question of whether the set of

317

relations each is compatible with is somehow fundamental to the semantic concept of reciprocity.

I will take up this question in more detail in Chapter 6.

5.6.2 Drug druga can be used productively as the object of prepositional phrases

In this chapter I have shown that drug druga can be used in prepositional phrases which can be headed by any preposition in the Russian language, and I have provided a formal analysis for such sentences. No other compositional analysis of a reciprocal expression has been suggested for when it is the object of a prepositional phrase. In my analysis, I have argued that drug druga has different semantics when it is the object of a preposition and when it is a direct object of a verb.

This analysis is particularly important because it has been suggested that some reciprocal expressions cross-linguistically cannot be used as the object of a prepositional phrase. For example, Gast and Haas (2008) have shown that the German reciprocal expression sich cannot be

used in its reciprocal function as the object of a preposition. When realized in a prepositional

phrase, it can only have a reflexive interpretation:

(466) Sie sahen eine Schlange neben sich. they saw a snake beside SE ‘They saw a snake beside themselves/#each other.’

Gast and Haas propose that the word sich has two lexical entries. One of these is a pronoun and functions as a reflexive anaphor. The second is a verbal clitic and functions as a reciprocal middle marker. The point here is that not all reciprocal expressions can function as the object of a prepositional phrase. Thus, when describing a reciprocal expression it is necessary to determine whether it can occur in a prepositional phrase and if so, whether there are any restrictions on

318

which prepositions it can co-occur with. Why such a restriction might hold is still unknown. It is only through investigation of other reciprocal expressions to determine if such a restriction exists outside German that we will clarify whether the restriction actually is related to the reciprocal meaning or not. If there is an interaction between reciprocity and the meaning prepositions contribute, this will prove interesting, as reciprocity is fundamentally about relations between

participants. Finding a restriction on the lexical types of these relations would be a new and

fascinating discovery.

5.6.3 Comitativity and reciprocity

This chapter also contributes to the literature on comitativity, particularly in regards to its

interaction with reciprocity. It has been noted before that reciprocity and comitativity are similar

and interact in several ways (Faller 2007, Sirola-Belliard 2004). Conceptually, reciprocity and

comitativity are both pluractional domains, as phrases that realize reciprocal or comitative

expressions express more than one event. Furthermore, in both Russian and English (though not

in every language), they both exemplify “event external” pluractionality, as opposed to “event

internal” pluractionality (Wood 2007). Event internal pluractionality is expressed by the lexical

meaning of the verb. A canonical example is the difference between the non-pluractional verb

bite and the pluractional verb nibble. Both the reciprocal expressions and the comitative markers in this dissertation are external to the verb and are linked even more closely than other pluractional domains such as verbal pluractionality.

As it turns out, the theoretical similarity between reciprocity and comitativity is supported by linguistic evidence. Many languages have one marker to indicate both comitativity

(or accompaniment) and reciprocity, such as some Bantu (Maslova 2007) and Oceanic (Moyse-

Faurier 2008) languages. Furthermore, as noted above, many, if not all, languages which have a

319

discontinuous construction use a comitative marker to express the second (non-subject) argument

(Nedjalkov 2007a, Rákosi 2004, Siloni 2012). The data in this chapter add a third interaction

between comitativity and reciprocity to contribute to our understanding of this interaction. I, in

fact, show that there is an interaction via a restriction. It is possible that this restriction holds

because the two domains semantically overlap and therefore using them both becomes redundant.

If this is the case, then studying this restriction as well as searching for others cross-linguistically

may provide insight into the overlap, providing information in how these domains are similar and

how they are different. I suggest that more research into this interaction will prove fruitful.

320

6 Conclusions

In this dissertation, I have explored the morphosyntax, semantics and pragmatics of the

Russian reciprocal expressions to determine what they can tell us about both the structure of

Russian and the linguistic concept of reciprocity more broadly. The main empirical findings of

this dissertation are that the distribution and meaning of two reciprocal expressions in Russian,

drug druga and the -sja verbs, differ from one another and from the English expression each

other in subtle ways not before discussed in the formal literature. In this chapter, after

summarizing the conclusions of each chapter, I discuss what further implications my findings in

this dissertation have for both theoretical semantics and for the task of identifying a taxonomy of

reciprocal expressions.

6.1 Summary of main empirical and theoretical conclusions of the dissertation

In Chapter 1 I introduced the topic of reciprocity and provided motivation for studying reciprocal expressions other than each other. I introduced the terminology and methodology to be used throughout the dissertation and provided background information on drug druga and the

-sja verbs.

In Chapter 2 I introduced the formal framework assumed throughout the dissertation. I presented five relations that Dalrymple et al. (1998b) argue each other conveys: Strong

Reciprocity (SR), Intermediate Reciprocity (IR), One-Way Weak Reciprocity (OWR),

Intermediate Alternative Reciprocity (IAR) and Inclusive Alternative Ordering (IAO), and argued

(i) following Sabato and Winter (2008) that IAO should not be considered a relation distinct from

321

IAR and (ii) that IAR overgenerates in that it predicts sentences to be acceptable that are not. I proposed a relation, Revised Intermediate Alternative Reciprocity that each other conveys. I also

summarized the two main approaches to analyzing the meaning of each other. One that was

argued for by Dalrymple et al. (1998b), Sabato and Winter (2005) and Dotlačil and Nilsen (2008),

treats each other as a polyadic quantifier. The other account developed by Langendoen (1978),

Sternefeld (1998) and Beck (2001) composes the meaning of each other from its parts. This

approach furthermore assumes that the different relations conveyed by each other arise due to

operators that can optionally be inserted into reciprocal and non-reciprocal sentences such as the

distributive and cumulative operators. I argued for the polyadic quantifier account. Throughout the dissertation I have shown that this approach can account for the meaning of both drug druga and the -sja verbs, as well as each other, and I have extended it to capture subtle differences in meaning between the expressions.

In Chapter 3, I explored whether or not drug druga and the reciprocal -sja verbs impose a temporal restriction of simultaneity or sequentiality on the events expressed by the verb. There has been debate in the literature about whether any reciprocal expression contributes such a restriction, and I started my explorations there. I found in the end that only some of the -sja verbs, those listed in (467), restrict the events to simultaneity.

322

(467) celovat'sja ‘to kiss’ obnimat'sja ‘to hug’ videt'sja ‘to meet’ vstrečat'sja ‘to meet’ otličat'sja ‘to differ from’ sudit'sja ‘to be in litigation with’ rezat'sja ‘to play cards with’ soveščat'sja ‘to confer with’ ob''jasnjat'sja ‘to discuss’ smešivat'sja ‘to mix’ (‘to be mixed’) obvenčat'sja ‘to get married’ ženit'sja ‘to get married’ mirit'sja ‘to make up with’ družit'sja ‘to befriend each other’ pixat'sja ‘to have sex’

The other -sja verbs, listed in (468), do not.

(468) tolkat'sja, ‘to bump into each other’ borot'sja, ‘to fight, struggle with’ rugat'sja, ‘to fight’ torgovat'sja, ‘to haggle’ znakomit'sja, ‘to get to know, to meet’ zdorovat'sja, ‘to greet each other’ drat'sja, ‘to fight’ sražat'sja, ‘to overwhelm each other’ bit'sja, ‘to fight’ rubit'sja, ‘to cut each other’ streljat'sja, ‘to shoot each other’ dogovaryvat'sja, ‘to agree with’ soveščat'sja, ‘to confer with each other’ uslovit'sja, ‘to agree’ prepirat'sja, ‘to bicker’ prerekat'sja, ‘to bicker’ smešyvat'sja, ‘to mix’ proščat'sja, ‘to bid farewell’ sorevnovat'sja, ‘to compete with each other’ družit'sja, ‘to befriend each other’ bodat'sja, ‘to butt each other’ klevat'sja, ‘to peck each other’ kusat'sja, ‘to bite each other’ ljagat'sja, ‘to kick each other’ pinat'sja, ‘to kick each other’ oblivat'sja, ‘to splash each other’ carapat'sja, ‘to scratch each other’ branit'sja, ‘to scold each other’ teret'sja, ‘to rub up against each other

323

Furthermore, I found that although drug druga does not impose a semantic restriction, it does

impose a pragmatic restriction of simultaneity. The findings of this chapter suggest that

reciprocal expressions can contribute both semantic and pragmatic temporal restrictions to a

sentence. Such a claim has not been pointed out in the formal literature and has only been

marginally explored in the typological literature.

In Chapter 4 I explored the meaning of the reciprocal -sja verbs. I showed that these -sja verbs participate in two different syntactic constructions, simple reciprocal constructions and discontinuous reciprocal constructions. I explored the meaning of the two constructions and found that the semantics of the constructions differ in three major ways. First, the simple construction draws entities from one set such that the given relation is defined over these entities.

The discontinuous construction, on the other hand, draws entities from two sets denoted by two different NPs, and relations are only entailed to hold between members of different sets. Second, the entities denoted by the subject and the object of the sentence with the discontinuous construction are universally quantified over such that each member of the set denoted by the subject must stand in the relevant relation to at least one entity denoted by the object as the first argument, and each member of the set denoted by the object must stand in the relevant relation to at least one entity denoted by the subject as the second argument. Third, the simple construction is compatible with the same four relations as each other, SR, IR, OWR and RIAR, while the discontinuous construction is compatible with the same four as well as IAR.

I proposed that the differences in meaning between the two constructions could either arise from a difference in the meaning of the -sja verbs in each construction or from the syntactic structure. I argued that the first semantic difference (that having to do with how relations are entailed to hold between members of different sets) is due to the syntactic structure of the constructions. The second difference proved harder to capture; in the end I tentatively accounted 324

for the difference in meaning by positing different lexical entries for the -sja verbs in the simple versus discontinuous constructions. Assuming different lexical entries also accounts for the fact that the two constructions are compatible with a different set of reciprocal relations between entities as well. I provided a formal analysis for both the simple and the discontinuous reciprocal

constructions that accounts for the differences in meaning between the two constructions.

In Chapter 5 I investigated the meaning of drug druga. The first major empirical finding

is that drug druga is compatible with the same set of relations as each other and the -sja verbs in

the simple contruction: SR, IR, OWR, and RIAR. I discovered, however, that the distribution of

drug druga is more restricted than that of each other. Namely, I provided evidence suggesting

that drug druga cannot be realized as the object of the comitative preposition s when this

preposition introduces a relation of accompaniment to a sentence. I first provided a diagnostic for

determining when a comitative preposition introduces a relation of accompaniment to a sentence

and then provided evidence suggesting that the phrase drug s drugom does not introduce this

relation. Rather, when this phrase is realized, it is as the object of the comitative phrase in the

discontinuous construction as discussed in Chapter 4.

I suggested that the best way to capture the semantics of the accompaniment preposition s

is that it assigns the thematic role of COMPANION to its object and the role of ACCOMPANEE to the subject of the sentence. I developed a formal analysis for drug druga and suggested two ways of accounting for its restriction against co-occurring with the accompaniment preposition. One hypothesis is that the preposition s when it introduces a relation of accompaniment does not semantically select drug druga as its object. Such an analysis is problematic, however, because the accompaniment s can take the anaphoric pronoun sebja ‘self’ as its object, s soboj ‘with oneself’. The second hypothesis is that drug druga cannot be assigned the role of COMPANION.

325

While I found no language internal reason to reject this hypothesis, I pointed out that the fact that

English each other is compatible with the accompaniment preposition with challenges it.

6.2 Implications for theoretical semantics

The main contribution of this dissertation consists of the detailed description of two

reciprocal expressions in Russian; drug druga and the -sja verbs. The description has

implications for our understanding of the semantics of reciprocity cross-linguistically as well as

for the structure of Russian.

6.2.1 Reciprocal relations

One central concern of reciprocal studies in the formal semantic literature has been the

relations each other is compatible with. My data suggest that drug druga and the -sja verbs in the

simple construction are compatible with the same set: Strong Reciprocity (SR), Intermediate

Reciprocity (IR), One-way Weak Reciprocity (OWR) and Revised Intermediate Alternative

Reciprocity (RIAR). None of these expressions are etymologically related, so another reason

must be found for their compatibility with the same four relations.

Note that I do not claim that as soon as an expression is lexicalized as a reciprocal

expression that it will necessarily have all four meanings, as might be compatible with a

Universal Grammar framework. Unfortunately we have very little information on reciprocal

expressions and the relations they convey, and so even though all the expressions considered in

this dissertation are compatible with these four relations, it is possible that other expressions are

compatible with a set of fewer relations. It is likely that the first meaning an expression conveys

once it has drifted from a previous meaning is Strong Reciprocity and that through lexical drift

other meanings such as Intermediate Reciprocity and Revised Intermediate Alternative 326

Reciprocity become associated with a particular expression. More data will determine the validity of such a prediction.

Recall, however, that the -sja verbs in the discontinuous construction are compatible with five relations; those compatible with each other, drug druga, and the simple construction, plus one more: Intermediate Alternative Reciprocity (IAR). This finding was significant because the formal literature is not in agreement as to whether this relation should be considered a reciprocal relation or not. As I argued in Chapter 2, this relation is too weak to account for the meaning of each other, and others have argued that this relation should only be considered part of the meaning of particular lexical items (e.g. Beck 2001, Sabato and Winter 2005, Dotlačil and Nilsen

2008). Their evidence for this claim is that sentences that realize each other are only compatible with verbs and prepositions that express a temporal or spatial relation, such as lie on top of each other and follow each other. Therefore, finding an expression that is more generally compatible with IAR is theoretically important because it places IAR back into the list of reciprocal expressions. While all reciprocal expressions must convey the strongest reciprocal relation, SR

(as it is by this relation that we define such expressions), they differ in terms of how weak the relations they can convey are. This leads to the question of whether there are reciprocal expressions out there that are only compatible with SR, or with a set of relations smaller than four. Although my findings make no predictions on this front, Zeshan and Panda (2011) claim that a reciprocal expression in Indo-Pakistani is compatible with fewer relations (see Section

6.3.3 below for a more detailed discussion). Future studies should investigate this line of inquiry further.

Moreover, the similarity in meaning between each other, drug druga and the -sja verbs suggest that composing the meaning from the two apparent words in each other, each and other, is not a tenable solution to accounting for the meaning of this reciprocal expression, as suggested 327

by Heim et al. (1991) and Beck (2001). A compositional analysis does not account for the similarity in meaning between these and other reciprocal expressions, as for example, drug druga

certainly does not contain a word like each which could be analyzed as a distributive marker as

in Heim et al.’s (1991) analysis, or a definite article as in Beck’s (2001) analysis. I therefore

suggest that reciprocal expressions contribute similar non-decomposable meanings to a sentence.

6.2.2 Implications for other linguistic areas

My investigation of drug druga and the -sja verbs also provides new insights into

linguistic areas such as comitativity, accompaniment, temporality, pluractionality, plurality, and

coordination. So far, these areas have largely been studied without considering reciprocity, and

the findings in this dissertation have implications for all of these. For example, I argued in

Chapter 3 that reciprocal verbs, such as meet or kiss in English, or celovat'sja ‘to kiss’ and

zdorovat'sja ‘to greet’ in Russian, in fact express two or more events, in contradiction to analyses

such as those in Carlson (1998) and Siloni (2012). Such a claim contributes to the study of

plurality and pluractionality by suggesting a new analysis for eventualities based on temporality

instead of number of events.

Similarly, in Chapter 5 I argued for an interaction between reciprocity and comitativity.

My study informs the field of comitativity by demonstrating that a diagnostic that works for

English also diagnoses accompaniment phrases in Russian, suggesting that this test is cross-

linguistically valid. Furthermore, I suggest that in Russian accompaniment and reciprocity are

not compatible. Other comitative markers should be explored to determine whether such a

restriction holds cross-linguistically. Determining whether other reciprocal expressions have

such a restriction could determine whether the incompatibility is due to restrictions on the

comitative preposition or on the reciprocal expression. If it is a restriction on the comitative 328

preposition, then syntactic and semantic accounts of certain comitative markers should account for such a restriction. If it is on the reciprocal expression, then the properties that some comitative markers assign to their object that violate the selectional restrictions of certain reciprocal markers should be explored, thereby informing the study of comitativity. Therefore, since reciprocity interacts with many different semantic domains such as temporality, pluractionality and comitativity, my study informs these domains as well.

6.2.3 The structure of Russian and cross-Slavic implications

Both drug druga and the -sja verbs in their reciprocal function are only briefly considered in grammars and the linguistic literature. Since each chapter of this dissertation explores their distribution and meaning in depth, this dissertation contributes to linguistic inquiry concerning the structure of Russian. My findings have the potential to answer questions concerning other expressions in Russian as well, or to suggest future investigations. For example, there are many similarities between drug druga and the other anaphoric pronoun in Russian, sebja ‘oneself’.

They both occupy argument positions, and they both refer back to another entity denoted by an

NP in the sentence. Furthermore, they both compete with verbs suffixed with -sja; recall from

Chapter 1 that one function of some verbs suffixed with -sja is reflexive, as in the verb moet'sja

‘to wash oneself’ or vyrazit'sja ‘to express oneself’. Because of these similarities, my results will

be of use to those interested in studying anaphora in Russian. I discussed one difference between

the two anaphors in Chapter 5, and other properties of the reciprocal expressions I’ve explored

could be other possible points of comparison, e.g. temporal restrictions or syntactic constructions.

Moreover, my empirical findings especially about the -sja verbs should be of interest to

scholars interested in cross-Slavic comparisons. As I discussed in Chapter 1, the suffix -sja

329

comes from the Proto Indo-European root *sve or *se, and all of the Slavic languages make use of this root today for similar functions such as reciprocity, reflexivity, passivity and anti-causativity.

The similarity in the forms and functions of the reflex of this root creates an interesting case study to discover subtle differences between the expressions in the different languages.

Cross-Slavic studies that investigate related expressions in different languages to discover and catalogue semantic similarities and differences are common (see especially studies about aspect e.g. Flier 1985, Dickey 2000; 2005, Janda 2004, and Richardson 2007). My findings in this dissertation have the potential to inform similar studies on the verbs suffixed with -sja. Certain functions of these verbs have been studied in great detail, such as the reflexive meanings

(Gerritsen 1990, Israeli 1997), but little work has been done on the extent to which the semantics of the reciprocal function are consistent or vary across languages. Although my study is not cross-Slavic, it provides novel data for such studies. My study contributes data and an analysis to such studies that could be extended to other Slavic languages.

6.3 Implications for typological work

In this section I will discuss how my findings inform a taxonomy of reciprocal

expressions. A preliminary taxonomy of reciprocal expression should be developed to discover

the uniformity and variation in reciprocal expression and meaning. I do not attempt to sketch a

full taxonomy, as I include only four reciprocal expressions in the taxonomy, but these results

should motivate future studies to include a variety of subtle factors in their analyses of a more

thorough taxonomy.

I here suggest semantic and pragmatic factors that should be considered when proposing

a taxonomy of reciprocal expressions. My goal in this section is to determine parameters that are

330

relevant for the characterization of reciprocity and my subgoal is to explore which logically possible types within these parameters are attested. I have already suggested a number of parameters that are relevant for the characterization of reciprocity. Here are the factors that an adequate taxonomy should capture that I have explored in this dissertation:

(469) (i) The syntactic construction a reciprocal expression occurs in

(ii) Semantic and pragmatic temporal restrictions on events

(iii) Which reciprocal relations an expression is compatible with

(iv) Whether the reciprocal expression is compatible with an accompaniment relation

I now present what the previous literature has to say about these factors; as will become clear, there is in fact very little discussion about any of these factors. There is a significant amount of work on developing a taxonomy based on the morphosyntactic form of reciprocal expressions,

but only two works that I am aware of have attempted to classify reciprocal expressions by their

semantic properties.

6.3.1 Argument versus non-argument

One implication from my discussions in Chapter 4 about the -sja verbs is that whether or

not a reciprocal expression can participate in the discontinuous construction is directly tied to its

syntactic status for one simple reason; the comitative phrase in the discontinuous construction is

an argument of a transitive verb. Since one of the defining features of the discontinuous

construction is that it draws entities from two distinct sets that are denoted by two arguments of a

331

transitive verb, both of these entities must be expressed. Therefore, based on my analysis, if a reciprocal expression, such as each other, can have an argument status, then it cannot participate in the discontinuous construction. On the other hand, if a reciprocal expression cannot occupy an argument position, then it may participate in the discontinuous construction. Some non-argument reciprocal expressions clearly cannot participate in the discontinuous construction (e.g. reciprocal adverbs such as mutually). Such a conclusion suggests that a taxonomy of reciprocal expressions should include whether a reciprocal expression can play the role of an argument.

This syntactic parameter has implications for the semantics. Recall from Chapter 4 that one of the distinguishing characteristics of the discontinuous construction is that it draws entities from two sets and that these sets are not pooled into one larger set. Other reciprocal expressions, such as each other and drug druga, are never entailed to be defined over two sets. In this way, the syntax of a reciprocal expression can have implications for the interpretation of the reciprocal relation.

The majority of works that classify reciprocal expressions does so on the basis of the way a reciprocal meaning (i.e. a Strongly Reciprocal relation) is encoded. Such taxonomies take into consideration more or less detailed differences between expressions (e.g. Faltz 1985, Nedjalkov

2007a, Haspelmath 2007). I here present the taxonomy suggested by Evans (2008), as it slices possible morphosyntactic variation very thinly to highlight the many differences in encoding cross-linguistically. I present it only as an example, so I will not walk through the all the terminology Evans uses to describe the morphosyntactic variation.

Evans provides the following types of reciprocal strategies. Although he does not explicitly define his term strategy, I assume, based on his usage, that he uses the term

332

to describe any expression or construction that entails that the denotations of one or more

NPs stand in a symmetric relation to each other.

1. Single Clause a. NP marking strategy [Argument-marking strategy] i. Bipartite quantifier ii. Reciprocal nominal 1. Possessed 2. Unpossessed iii. Reciprocal pronoun 1. Free a. Person-marked b. Unmarked for person 2. Bound a. Clitic b. Affix iv. Reciprocal role marking on NP (not well attested) v. Double role marking (also not well attested) b. Verb-marking strategies [Predicate-marking strategy] i. Morphological modification of verb ii. Auxiliary to verb iii. Lexical strategy (e.g. exchange, swap, quarrel) c. Conjunct strategy d. Modifier strategy31

2. Multiple clauses [multiple propositions] a. Conventionalized biclausal construction b. Zigzag summative constructions c. Fused multiple predicates i. Verb compounding with mutual predicate ii. Verb compounding with repeated one-way predicate iii. Symmetric signing iv. Fused contrastive subject

Figure 25: Evans’ (2008: 45) typology of reciprocal expressions

Evans first divides reciprocal strategies into single clauses (i.e. Bob and Sally saw each

other in the park.) versus multiple clauses (i.e. Bob saw Sally in the park, and Sally saw Bob

31 Evans also calls this strategy an ‘Adverbial Strategy.’ 333

there too.) Each of these types is divided into more specific types. Here I focus on the single clause strategies, because they are relevant for the reciprocal expressions under discussion in this dissertation. I direct the reader to Evans’ paper for more information on multi-clausal strategies.

Evans finds four major types of single-clause reciprocal strategies: NP marking, verbal marking, a conjunct strategy, and a modifier strategy. A reciprocal expression exemplifies an NP strategy when it occurs in a sentence with a verbal predicate, and the reciprocal expression “will either occupy an overt NP position (bipartite NP, equivalent-token, free reciprocal pronoun), or a bound pronominal slot on the verb or an accompanying auxiliary” (46). For example, the reciprocal expression each other in the sentence in (470) is an example of an NP strategy, because the expression occupies the direct object argument of the verb.

(470) John and Mary kissed each other.

The expression each other is a bipartite NP (one subtype of the NP strategy), because there are, at least on the surface, two distinct elements that make up the expression - the word each and the word other. This strategy is common among the Indo-European languages. The Russian

expression drug druga is also a bipartite NP.

A verbal strategy is any type of reciprocal expression that “marks[s] the predicate directly

to derive the meaning ‘mutually Pred’ or ‘share in Pred-ing’ from a basic unidirectional

[asymmetric] meaning” (67). Evans writes that “the commonest method is to use a verbal affix of

the type that effects valency changes … it is also possible for verbs to be marked by other sorts of

derivational means, including affixes, reduplication, etc.” (67). Within the verbal strategies, there

are two types that are relevant for this dissertation: morphological modification and lexical

334

strategies. The morphological modification strategy encompasses those reciprocal expressions

that attach by affixation to the predicate. For example, the reciprocal expression in the following

sentence from Kayardild (Tangic, Australia), the suffix -thu, exemplifies morphological

modification.

(471) Bil-da miila-thu-th. (Evans 2008: 68) 3PL-NOM delouse-RECIP-ACTL ‘They delouse each other.’

The reciprocal expression is a morphological modification strategy because it is a verbal affix

which in this example attaches to the verbal predicate miila- ‘delouse’. The -sja verbs are

examples of morphological modification of a verb.

The lexical strategy is also a type of verb marking strategies. These are reciprocal

expressions that are verbs “whose meaning already encompasses mutual activity, and which

(unlike e.g. ‘kiss’) cannot be used unidirectionally” (72). Examples in English are verbs such as

exchange and quarrel. This type is in fact somewhat problematic based on Evans’ definition of

verbal strategies above because he first states that verbal strategies are derived from an

asymmetric predicate, but the verbs that make up the lexical strategy are innately symmetrical -

i.e. they are not derived from an asymmetric base.

This taxonomy is valuable because it lays out which strategies are attested to encode

reciprocal meaning cross-linguistically and demonstrates the variation in the encoding. It also

suggests that certain types are not attested. As Evans points out, no reciprocal expression was

found that attaches to the two NPs in a transitive sentence that then entails that the denotations of

the two NPs stand in a symmetric relation to each other. Since this type has not been found, this

335

suggests either that there is some constraint against this type of reciprocal expression or that the conditions that would lead to such a reciprocal expression are rare. Either way, the typology suggests that the conditions in which such a reciprocal expression could theoretically arise should be explored to determine why such an expression is not attested.

Beyond these points, however, I do not find these divisions meaningful beyond being a simple taxonomy of what is attested. My results suggest that the syntax of an expression is theoretically interesting only in the broadest sense; I have provided evidence that whether a reciprocal expression can play an argument role in a sentence has implications for its semantics.

Evans’ (2008) taxonomy accounts for this important factor by grouping NP strategies, i.e. those that can play an argument role, into one group. He does not make this important point however, that it is whether an expression can occupy an argument position that differentiates these strategies from others, and that it is this property that will have repercussions for the expressions’ distribution and semantics. He focuses on form, i.e. realization via an NP, but he should instead be focusing on role relations. My research does not motivate any further classification, and whether any of his other classifications have any implications beyond a simple taxonomy is unknown at this point.

This one difference, the ability to occupy an argument position, is the only result of the dissertation based on the syntax of the expression. The other parameters I found to be relevant in this dissertation are semantic and pragmatic.

6.3.2 Temporality

I found that some reciprocal expressions contribute temporal restrictions on the events expressed by a sentence. This parameter of temporality logically has three semantic types: (i) a

336

reciprocal expression could contribute no restriction, (ii) it could contribute a restriction to simultaneity, or (iii) it could contribute a restriction to sequentiality. I have shown that two of

these are attested, (i) and (ii). I outline here what the previous typological literature has claimed

about the temporality of reciprocal expressions.

Siloni, as discussed in Chapter 3, claims that we can classify reciprocal relations into

types based on whether they express only one event (my two simultaneous events) or multiple

events (simultaneous or sequential under my analysis). She claims that reciprocal verbs such as

kiss in English and celovat'sja ‘to kiss’ in Russian only express one event, while sentences that

realize expressions such as each other or drug druga express multiple events. She argues that

cross-linguistically, reciprocal expressions can be classified based on whether the verbs they

combine with express one or more events. Siloni also explores other semantic ways in which

reciprocal expressions pattern, considering the readings of embedded clauses that realize

reciprocal expressions and whether reciprocal expressions can co-occur with an NP direct object.

She argues that how particular reciprocal expressions pattern in terms of these factors indicates

their locus of composition. Her main claim is that whether a reciprocal expression is formed in

the lexicon or the syntax has repercussions for its semantics. She explores reciprocal expressions

in Bulgarian, Czech, English, French, Greek, Hebrew, Hungarian, Italian, Japanese, Russian,

Serbo-Croatian and Spanish to determine how they pattern semantically. I here summarize her

arguments, and then will argue that the Russian data do not pattern as neatly as she claims, posing

problems for her classification and possibly for the assumption that such a classification is

possible in the first place.

The first semantic parameter Siloni explores, based on Heim et al. (1991), is what she

calls the I-we reading (though Heim et al.(1991) use this term for a different phenomenon and

simply call the following a puzzle of scope). Following work by Higginbotham (1980), Heim et 337

al. (1991), and Carlson (1998), she notes that sentences containing some reciprocal expressions in

an embedded clause have two distinct readings, the ‘we-reading’ and the ‘I-reading.’ For

example, the sentence in (472) has two possible readings, the ‘we reading’ as in (472)(i) and the

‘I reading’ as in (472)(ii):

(472) John and Paul said that they saw each other. (Siloni 2012: 263) (i) John and Paul said that they (i.e. John and Paul) saw each other. (ii) John said that he saw Paul, and Paul said that he saw John.

According to the reading in (472)(i), each individual reported a symmetric seeing event, while according to the reading in (472)(ii), each individual reported an asymmetric seeing event. As

Siloni points out, this difference becomes more salient when we consider a sentence such as that in (473), where the reciprocal expression is not embedded and thus only has one possible interpretation. This sentence is unacceptable because it is typically impossible, knowing what we do about the way the world is, for two individuals to defeat each other in one match:

(473) #John and Paul defeated each other in the final. (Siloni 2012: 263)

The sentence in (474), however, with the embedded reciprocal, is acceptable because it has both the ‘we reading’ (474)(i) and the ‘I reading’ (474)(ii).

(474) John and Paul said that they defeated each other in the final. (Siloni 2012: 264) (i) %John and Paul said that they (i.e. John and Paul) defeated each other. (ii) John said that he defeated Paul and Paul said that he defeated John.

338

For this sentence, the ‘we reading’ is not available, because it is nonsensical, just like the interpretation of the sentence in (473). However, since the ‘I-reading’ is available, which is not nonsensical, the sentence in (474) is acceptable. Thus, both the ‘I’ and the ‘we’ readings are available for sentences containing each other.

Siloni shows, however, that cross-linguistically not all sentences that contain an

embedded reciprocal expression have both the ‘we’ and ‘I’ readings. Specifically, for some reciprocal expressions, the ‘I-reading’ is not available. The French sentence in (475), which realizes the non-embedded reciprocal expression se, is unacceptable, similar to (473). However, unlike for the sentences containing each other, when the reciprocal expression is realized in an

embedded clause, as in (476), the sentence is still unacceptable.

(475) #Pierre et Jean se sont vaincus à la finale. (Siloni 2012: 264) Pierre and Jean SE are defeated in the final ‘Pierre and Jean defeated each other in the final.’

(476) #Pierre et Jean ont dit qu’ils se sont vaincus à la finale. (Siloni 2012: 264) Pierre and Jean have said that they SE are defeated in the final

Since the sentence in (476) is not acceptable, Siloni argues that the ‘I reading’ is not available for the se reciprocal expression in French. Similarly, for the English sentence in (477), Siloni claims that only the ‘we-reading’(477)(i) is available, not the ‘I-reading’(477)(ii):

(477) Mary and John said they collided. (Siloni 2012: 265) (i) Mary and John said that they (i.e. Mary and John) collided. (ii) %Mary said she collided with John, and John said he collided with Mary.

339

For a speaker to utter the sentence in (477), it has to be the case that both John and Mary individually report that the colliding relation was symmetric. This sentence would not be felicitous if they both reported an asymmetric colliding event, e.g. that Mary collided with John,

and John did not collide with Mary. Therefore, the reciprocal expression in (477) does not give

rise to the ‘I reading’.

Siloni therefore notes that some reciprocal expressions are compatible with both the ‘we-’

and ‘I-readings’, and some are only compatible with the ‘we-reading’. This being the case, it

provides another way to classify reciprocal expressions.

As a third way reciprocal expressions differ, Siloni explores the difference in behavior of the reciprocal constructions when they occur with a count adverbial, such as five times. She specifically is interested in how many events a sentence can express. If a sentence contains, for example, the count adverbial five times, she finds that some reciprocal constructions are compatible with two readings, one reading in which there are five events, and one reading in which there are five events per individual denoted by the subject. Other sentences are only compatible with a reading in which there are five events. She provides the sentence in (478) and claims that it only conveys a situation in which there are a total of five events of mutual kissing.

(478) John and Mary kissed five times. (Siloni 2012: 266)

On the other hand, she argues that the sentence in (479) is compatible with both readings in

(479)(i) and (479)(ii):

340

(479) Le laryngologue et le dentiste se sont examinés cinq fois. (Siloni 2012: 267) the laryngologist and the dentist SE are examined five times ‘The laryngologist and the dentist examined each other five times.’

(i) There were five events of examination. (ii) There were ten events of examination: five by the laryngologist, and five by the dentist.

Based on such evidence, she proposes that some reciprocal expressions are compatible with two readings that differ in terms of how many events are expressed, and some reciprocal expressions are compatible with only one such reading.

Siloni (2012) also notes that reciprocal constructions can suppress either an accusative object, as in (480) and (481), or a dative object as in (482) and (483). She finds though that some reciprocal constructions, when suppressing a dative object, can also realize a direct object, as in the French sentence in (482). Other reciprocal constructions that suppress a dative object, such as that in the Hungarian sentence in (483), cannot realize an accusative object.

(480) Pierre et Marie se sont embrassés. (Siloni 2012: 290) Pierre and Marie SE are kissed ‘Pierre and Marie kissed.’

(481) Jànos és Mari csókol-óz-t-ak. (Siloni 2012: 290) Jànos and Mari kiss-REC-PST-3PL ‘Jànos and Mari kissed.’

(482) Pierre et Marie se sont chuchoté (des mots d’amour). (Siloni 2012: 290) Pierre and Marie SE are whispered (some words of love) ‘Pierre and Marie whispered (words of love) to each other.’

341

(483) Jànos és Mari (*hízelgő szavak-at) sugdol-ódz-t-ak (Siloni 2012: 290) Jànos and Mari (flattering words-ACC) whisper-REC-PST-3PL

She provides evidence that a reciprocal expression in Spanish also follows the pattern in (480) and (482), while an expression in Hebrew follows that of (481) and (483), but claims this difference holds for a number of languages. She has therefore identified another way in which reciprocal expressions can vary cross-linguistically.

Siloni has therefore explored a number of factors, morphosyntactic and semantic, that should be considered for a complete characterization of reciprocity. She discusses temporality, in that she argues that sentences with certain reciprocal verbs express only one event (i.e. two simultaneous events) while sentences with other reciprocal expressions can express multiple (i.e. sequential) events. We both agree then that reciprocal expressions restrict events in different ways and that a semantic analysis should capture such restrictions.

Siloni goes further, however, and suggests that these semantic factors correlate with where the reciprocal expression composes with the rest of the predicate. She argues that some reciprocal expressions are composed in the syntax. Two different groups of reciprocal expressions are composed in this way, (i) periphrastic reciprocal constructions “where reciprocity is expressed by means of a reciprocal anaphor (each other),” (261) and (ii) syntactic reciprocal verbs, which consist of a verb and a clitic, exemplified in the French sentence in (484)

“whose reciprocity is not periphrastic in the sense of [the periphrastic reciprocal constructions] but nonetheless [is] composed only the course of the syntactic derivation” (261-262). Lexically reciprocal verbs, on the other hand, “describe a reciprocal situation without the aid of a reciprocal anaphor (languages, unlike English, tend to mark such verbs morphologically)” (261) and are composed in the lexicon. The verbs in (485) and (486) exemplify lexically reciprocal verbs.

342

(484) Jean et Marie se sont embrassés. (Siloni 2012: 290) Jean and Marie SE are kissed ‘Jean and Marie kissed.’

(485) John and Mary kissed.

(486) Ivan i Marija pocelovali-s'. Ivan.NOM and Marija.NOM kiss.PST.PL-SJA ‘Ivan and Marija kissed.’

Siloni argues that the semantic factors listed above, (i) temporality, (ii) whether a reciprocal

expression can co-occur with a NP object and (iii) the we/I reading data, correlate with the

different types of reciprocal expressions. She proposes that the reciprocal expressions that are

composed in the lexicon, i.e. the lexically reciprocal verbs, are the only ones that require that the

verb express only one event, while both of the types of reciprocal expressions composed in the

syntax (periphrastic reciprocal constructions and syntactic reciprocal verbs) can range over

multiple events. Similarly, only lexically reciprocal verbs are unable to co-occur in a sentence

with a NP direct object. Both the syntactic reciprocal verbs and the periphrastic reciprocal

constructions can co-occur with a direct object.

Siloni uses the differences in the semantics of each expression to motivate an analysis in

which some reciprocal expressions compose with the verb in the syntax and some compose in the

lexicon. The argument that the locus of composition has implications for the meaning of a

reciprocal expression is plausible, given that we know that expressions composed in the lexicon

are more likely to exhibit semantic drift and are more likely to have an idiomatic meaning than

expressions formed in the syntax. My data, however, pose a challenge for her claim that we can

judge whether an expression is formed in the lexicon or the syntax based on its form. My results

343

suggest that the reciprocal verbs in Russian are not a homogeneous category semantically, as some restrict their events to simultaneity and some do not. Siloni, however, holds that all lexical reciprocal verbs restrict events to simultaneity. Siloni and I both agree that all the -sja verbs are composed in the lexicon, and so my finding suggests that where a given expression is composed is unrelated to whether or not it contributes a temporal restriction. Therefore, while her basic claim might be right, that the locus of formation is important, my data pose a serious problem for her broad assumptions about the form being correlated with the locus of formation and thus its semantics.

The data from the embedded sentence readings split the reciprocal expressions differently. Siloni presents evidence suggesting that it is only periphrastic reciprocal expressions that are compatible with both readings; neither the syntactic reciprocal verbs nor the lexical reciprocal verbs are compatible with the ‘I reading’. Such data then suggest that for this parameter it is not where the reciprocal expression composes with the verb that is important; rather, this semantic difference correlates, according to Siloni, to whether the reciprocal expression is a clitic or morphological marking. This claim therefore assumes that the morphosyntactic form of a given expression can correlate with its semantics. This assumption clashes with most findings about the form-meaning interface. The null hypothesis is that there is no correlation between form and meaning. There is little evidence cross-linguistically that form and meaning ever directly correlate such that one meaning, for example one particular value associated with tense or number, is expressed using the same syntactic means. There is in fact plenty of evidence to the contrary, that the same meaning cross-linguistically or even language internally is expressed by different types of expressions. For example, in English, the comparative is expressed either morphologically, e.g. bluer, sweeter, or periphrastically, i.e. more confident, more sentimental. Since Siloni claims that the form of a reciprocal expression

344

correlates to its meaning, unlike other types of expressions, an explanation must be found for why the form matters for the reciprocal expressions when it does not in other cases.

Furthermore, data from more typologically diverse languages are needed to make a claim

about syntactically reciprocal verbs. All of the syntactically reciprocal verbs that Siloni studies

are from Indo-European languages, and the reciprocal expression in each language is derived

from the same source. Since these reciprocal relations are etymologically related, to make the

claim that syntactic reciprocal verbs have certain properties based on their form alone requires

data from genetically unrelated languages, or at least from a reciprocal expression that is not

related to the se expression throughout these languages.

Siloni herself proposes a compositional analysis of the three different reciprocal

constructions and is not interested in the broader implications for a taxonomy. Nevertheless, her

article provides insights into how a taxonomy of reciprocal expressions might be identified and

proposes a number of semantic factors that have not been used before to classify reciprocal expressions. Testing these parameters against a wider range of reciprocal expressions would help determine whether these parameters truly directly correlate with each other and if so, why such a correlation might exist.

6.3.3 Temporality and relations between participants

Evans et al. (2011a) explore five different parameters they hypothesize might be relevant for the acceptability of a reciprocal expression: number of participants, configuration (“possible permutations of who acts upon whom” (31)), temporality (whether the subevents occur

simultaneously or sequentially), event type (whether the event was one of kissing, following,

delousing, etc.), and symmetry between participants (“are the participants both actors and

345

recipients of the target event” (32)). They investigate whether any of these parameters are

relevant for determining whether a reciprocal expression is acceptable in a context or not in 20

languages that are mostly genetically and areally unrelated.

Their experiments consisted of showing the videos that I myself have used in this

dissertation to consultants and asking them to describe what the videos were depicting.

Responses were then recorded and analyzed to determine which reciprocal expressions were used to describe which events and which types of each parameter. They found that some reciprocal

expressions do contribute temporal restrictions to simultaneity, as discussed in Chapter 3, and that

the reciprocal expressions were compatible with different relations between participants. For

example, Zeshan and Panda (2011), in the chapter on Indo-Pakistani Sign Language, write that

the reciprocal expression in Indo-Pakistani Sign Language is not used to describe “chaining

events,” which are defined as a configuration between events such that A acts on B who acts on C

who acts on D “resulting in a linear series of events” (Majid et al. 2011: 32). Such a

configuration corresponds to the relation I have labeled as Revised Intermediate Alternative

Reciprocity (see Chapter 2 Section 2.3 for my discussion of this relation). I have found that each

other, drug druga, and the -sja verbs in both the simple and discontinuous construction can all

convey a RIAR relation, and so Zeshan and Panda’s (2011) study provides evidence for a

reciprocal expression that is compatible with a smaller set of reciprocal expressions, as I

hypothesized above in Section 6.2.

One more finding that is relevant for my discussion here is that Evans et al. (2011a)

found that how the reciprocal expressions pattern in terms of the five dimensions they can express

does not correlate with the syntactic form of the expression, contra Siloni (2012). Specifically,

Majid et al. (2011) write that NP strategy reciprocal expressions are not more similar to each other than they are to verbal strategy reciprocal expressions and vice versa. This suggests that 346

there are semantic parameters that correlate with each other that do not correlate with the

morphosyntactic form of the expression. This finding provides even more motivation to

systematically explore semantic parameters in the future to find similarities between expressions

that are not evident based on the form.

Evans et al. (2011a) and the chapters therein are again useful for determining the possible

different types of a given parameter. They explore both the parameter of temporality and that of relations. They too only found reciprocal expressions that restrict events to simultaneity, not any that restrict events to sequentiality. My study adds to their contributions in suggesting that

reciprocal expressions can pragmatically, via an implicature, restrict events to simultaneity.

Similarly, this dissertation does not suggest further reciprocal relations beyond those in Evans et

al. I do, however, provide evidence that the syntactic construction a reciprocal expression is

realized in can influence the relations an expression is compatible with.

While their results are suggestive of differences across a wide variety of languages, to be

conclusive negative data needs to be collected along with the data about what is possible in the

language. For example, the data about the Indo-Pakistani Sign Language reciprocal expression

presented above does not provide conclusive claims. Whether such a situation could be described with the reciprocal expression, even if it was not the first choice of the consultants, was not determined. My explorations, on the other hand, provide information on what is possible and impossible, leading to more conclusive results. For example, I was able to show that some of the reciprocal -sja verbs cannot be used to describe sequential events. Such a finding would have been impossible to conclude without eliciting negative data.

Additionally, in terms of temporality, I have provided evidence that reciprocal expressions can contribute a pragmatic restriction that events occur simultaneously. No study, as

347

far as I know, has tried to classify expressions based on the pragmatics of the temporality of events. I have argued that reciprocal expressions can implicate that events are simultaneous, and I suggest that future studies consider the pragmatics of events to determine whether other reciprocal expressions give rise to other temporal implicatures.

6.3.4 Accompaniment

The last parameter I have found to be relevant for the acceptability of reciprocal expressions is whether an expression can be realized as the object of a preposition that contributes a relation of accompaniment to a sentence.

The only study I know of to suggest an interaction of accompaniment and reciprocity is

Sirola-Belliard (2011), who discusses Finnish data. Finnish has two ways of expressing comitativity (i) with a postposition kanssa and (ii) with an affix -ine, which Sirolo-Belliard calls the comitative case marking. In her corpus study (which consisted of 2,000 sentences containing

-ine, and 1,000 sentences that contained kanssa), 55% of the sentences that realized kanssa

“express[ed] reciprocal action” (144), while 0% of those that realized -ine did. She concludes that

“[e]xpression of reciprocal action is common with the postposition kanssa … but impossible with the comitative case” (144). Therefore, she finds that the reciprocal expression is not compatible with one of the markers of comitativity in Finnish.

Moreover, Sirola-Belliard claims that whereas the postposition marks its object as having the same level of participation in the event expressed by the verb as the subject, the comitative

“encodes an asymmetrical relationship between two participants: as in comitative structures universally, the companion is subordinate to the accompanee and hence participates in the situation described by the lexical verb only through its association with the accompanee, never

348

independently” (144). I take this to mean that the comitative case marking introduces a relation of accompaniment between its host NP and the subject of the predicate. If this is the case, then it suggests that the reciprocal expression in Finnish does not co-occur with a marker of accompaniment in Finnish. It therefore mirrors the restriction that drug druga cannot be realized

as the object of the preposition s when it introduces a relation of accompaniment to a sentence.

Therefore, it is possible that there is a restriction against a reciprocal expression co-

occurring with an accompaniment marker in a language other than Russian. Sirola-Belliard did

not systematically explore this restriction in Finnish as it was not the point of her article, but her

claim is suggestive that this restriction is not unique to Russian. Therefore, based on our

findings, I suggest that future work explore this restriction cross-linguistically to see if an

explanation for it can be found once we have data on many different expressions.

6.3.5 A preliminary taxonomy

Having outlined the factors I have found to be relevant for a taxonomy of reciprocal

expressions, I provide a preliminary taxonomy of the reciprocal expressions under discussion in

this dissertation in Table 2, to suggest a way for future researchers to study reciprocal

expressions. In this table, the column labeled ‘Argument’ refers to whether or not the reciprocal

expression can play an argument role in the sentence. The third column is for whether or not the

particular reciprocal expression entails that a reciprocal relation holds between entities of

different sets. The fourth column shows whether reciprocal expressions temporally restrict events

and if so, how the events are restricted. The ‘Accompaniment’ column shows whether a

reciprocal expression is compatible with an accompaniment relation. Finally, the last column

shows which reciprocal expressions are compatible with which relations between participants.

349

Reciprocal Argument Entails drawing Semantic Accompaniment Relations Expression entities from 2 Temporal sets Restriction each other Y No None Y SR, IR, OWR, RIAR drug druga Y No None N SR, IR, OWR, RIAR

Intransitive N No None N/A SR, IR, -sja verbs OWR, RIAR

Transitive N Yes None N/A SR, IR, -sja verbs OWR, RIAR, IAR celovat'sja/ N N/Y Simultaneity N/A SR, IR, OWR, RIAR obnimat'sja

Table 2: Summary of proposed parameters for a taxonomy of reciprocity

I have found no reason to assume that any of these parameters are logically dependent on one another. Therefore, since there are five parameters in this taxonomy and two or more values for each parameter, this leaves us with many possible types of possible reciprocal expressions.

Based on the information in this table, I present a few hypotheses about what values within these parameters will be attested, as well as the types of reciprocal expressions I expect to find and those I do not expect to find. Much more research must be done to determine the validity of the claims made in these hypotheses, but they hold based on the data in this dissertation and the information available about each other.

350

(i) Whether or not a reciprocal expression can take an argument position divides the reciprocal expressions; those that are arguments cannot entail that relations hold between entities of two sets and cannot convey a IAR relation between participants

(ii) Some reciprocal expressions restrict events to simultaneity; reciprocal expressions do not restrict events to sequentiality

(iii) Only nonarguments can (though they don’t have to) restrict events to simultaneity

(iv) Only reciprocal expressions that participate in the discontinuous construction (e.g. the transitive -sja verbs) are compatible with the IAR relation

Therefore, in addition to fleshing out this table with more reciprocal expressions and more parameters, I suggest that future studies test the above hypotheses with a much wider range of data. Only by doing such research will we obtain an adequate cross-linguistic characterization of reciprocity.

My results in this dissertation contribute to our growing picture of what morphosyntactically, semantically, and pragmatically possible in reciprocity in natural languages.

Many more detailed formal semantic studies of reciprocal expressions in a typological diverse set of languages are needed, however, to develop truly cross-linguistic theories of reciprocal semantics.

351

Bibliography Allan, Rutger. 2006. The middle voice in ancient Greek. A study in polysemy. Amsterdam: Gieben.

Apresjan, Jurij. 1995. Izbrannye trudy,t. II, Integral’noe opisanie jazyka i sistemnaja leksikografija. [Selected word tom II, Integrated description of the language and systematic lexicography.]. Moscow.

Babby, Leonard. 1975. Impersonal verbs and their lexical specification. Slavic and East European Journal 19(2): 182-187.

Bailyn, John. 2012. The syntax of Russian. Cambridge: Cambridge University Press.

Beck, Sigrid. 2001. Reciprocals are definites. Natural Language Semantics 9: 69–138.

Bošković, Željko. 2010. Conjunct-sensitive agreement: Serbo-Croatian vs Russian. In Formal Studies in Slavic Linguistics: Proceedings of formal description of Slavic languages 7.5. Ed. by Gerhild Zybatow, Philip Dudchuk, Serge Minor & Ekaterina Pshehotskaya. 31- 47. Frankfurt am Main: Peter Lang.

Brooks, Patricia & Martin Braine. 1996. What do children know about the universal quantifiers all and each? Cognition 60: 235-268.

Burenhult, Niclas. 2011. The coding of reciprocal events in Jahai. In Reciprocals and Semantic Typology. Ed. By Nicholas Evans, Alice Gaby, Stephen C. Levinson and Asifa Majid, 163-176. Amsterdam: John Benjamins Publishing Company.

Cable, Seth. 2012. Reflexives, reciprocals and contrast. Journal of Semantics ffs020.

Carlson, Greg. 1998. Thematic roles and the individuation of events. In Events and Grammar, ed. S. Rothstein, 35-51. Springer Science and Business Media.

Chomsky, Noam & Howard Lasnik. 1995. The theory of principles and parameters. In The minimalist program. Ed. by Noam Chomsky. 13-127. Cambridge, MA: MIT Press.

Corbett, Greville. 1983. Hierarchies, targets and controllers: Agreement patterns in Slavic. London: Croom Helm.

Culicover, Peter. 2009. Natural language syntax. Oxford: Oxford University Press.

Dalrymple, Mary, Irene Hayrapetian, & Tracy Holloway King. 1998a. The semantics of the Russian comitative construction. Natural Language and Linguistic Theory 16: 597-631. 352

Dalrymple, Mary, Makoto Kanazawa, Yookyung Kim, Sam Mchombo & Stanley Peters. 1998b. Reciprocal expressions and the concept of reciprocity. Linguistics and Philosophy 21: 159-210.

Dalrymple, Mary, Sam A. McHombo & Stanley Peters. 1994. Semantic similarities and syntactic contrasts between Chicheŵa and English reciprocals. Linguistic Inquiry 25(1): 145-163.

Dickey, Stephen. 2000. Parameters of Slavic aspect: A cognitive approach. Stanford: CSLI Publications.

Dickey, Stephen. 2005. S-/z- and the grammaticalization of Slavic aspect. Slovene Linguistic Studies 5: 3-55.

Dimitriadis, Alexis. 2004. Discontinuous reciprocals. Ms. Utrecht institute of Linguistics OTS. http://www.let.uu.nl/~alexis.dimitriadis/personal/papers/discon-long-ms04.pdf

Dimitriadis, Alexis. 2006. Distribution over symmetric events. In Proceedings of the 7th International Conference on Greek Linguistics. Ed. by G. Tsoulas. York: University of York.

Dimitriadis, Alexis. 2008. The event structure of irreducibly symmetric reciprocals. In Event Structures in Linguistic Form and Interpretation, Ed. By Johannes Dölling, Tatjana Heyde-Zybatow & Martin Schäfer, 327-354. Berlin: Mouton de Gruyter.

Dotlačil, Jakub & Øystein Nilsen. 2008. The others compared to each other – Consequences for the theory of reciprocity. In Proceedings of the 18th Semantics and Linguistic Theory conference, Ed. by Tova Friedman and Satoshi Ito.

Dowty, David. 1989. On the semantic content of the notion of ‘thematic role.’ In Properties, types, and meaning II. Ed. by Gennaro Chierchia, Barbara Partee & Raymond Turner. 69- 129. Dordrecht: Kluwer Academic Publishers.

Dowty, David. 1991. Thematic proto-roles and argument selection. Language, Vol. 67(3): 547- 619.

Dowty, David. 2000. The dual analysis of adjuncts/complements in categorial grammar. ZAS Papers in Linguistics 17: 53-78.

Dowty, David, Robert Wall & Stanley Peters. 1981. Introduction to Montague Semantics. Dordrecht: Kluwer Academic Publishers.

Dyła Stefan & Anna Feldman. 2008. On comitative constructions in Polish and Russian. In Proceedings of the fifth European Conference on Formal Description of Slavic Languages. Frankfurt am Main: Peter Lang.

353

Es’kova, Natalija. 2011. Izbrannye raboty po rustistike. Fonologija. Morfonologija. Morfologija. Orforgrafija. Lekskografija. [Selected works on Russian studies. Phonology. Morpho- phonology. Morphology. Orthography. Lexicography.] Moscow: Jazyki slavjanskix kul’tur.

Evans, Nicholas. 2008. Reciprocal constructions: Toward a structural typology. In Reciprocals and Reflexives: Theoretical and Typological Explorations. Ed. by Ekkehard König & Volker Gast, 33-103. Berlin: Mouton de Gruyter.

Evans, Nicholas, Alice Gaby, Stephen Levinson, & Asifa Majid, eds. 2011a. Reciprocals and Semantic Typology. John Benjamins Publishing Company, Amsterdam.

Evans, Nicholas, Stephen C. Levinson, N.J. Enfield, Alice Gaby & Asifa Majid. 2004. Reciprocal constructions and situation type. In: Field Manual Volume 9, ed. by Asifa Majid, 25-30. Nijmegen: Max Planck Institute for Psycholinguistics.

Evans, Nicholas, Asifa Majid & Stephen Levinson. 2011b. Introduction: Reciprocals and semantic typology. In Reciprocal and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen Levinson & Asifa Majid, 115-127. Amsterdam: John Benjamins Publishing Company.

Evans, Nicholas & Toshiki Osada. 2011. Mundari Reciprocals. In Reciprocals and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen Levinson, & Asifa Majid, 115- 127. Amsterdam: John Benjamins Publishing Company.

Faller, Martina. 2007. The ingredients of reciprocity in Cuzco Quechua. Journal of Semantics 24: 255–288

Faller, Martina. 2012. Pluractionality and accompaniment in Cuzco Quechua. In Verbal Plurality and Distributivity. Ed. by Patricia Cabredo Hoffherr & Brenda Laca. 55-85. Berlin: Walter de Gruyter.

Faltz, Leonard. 1985. Reflexivity: A study in Universal syntax. New York: Garland.

Fiengo, Robert & Howard Lasnik. 1973. The logical structure of reciprocal sentences in English. In Foundations of Language 9(4):447-468.

Filip, Hana & Gregory N. Carlson. 2001. Distributivity strengthens reciprocity. Collectivity weakens it. Linguistics and Philosophy 24: 417-466.

Flier, Michael. 1985. The scope of Slavic aspect. Language 64(1): 147-150.

Frajzinger 2000. Coding of the reciprocal function: two solutions. In Reciprocals: Forms and functions. Volume 2. Ed. by Zygmunt Frajzyngier & Traci Walker. 31-62. Amsterdam: John Benjamins Publishing Company.

354

Gast, Volker & Florian Haas. 2008. On reciprocal and reflexive uses of anaphors in German and other European languages. . In Reciprocals and Reflexives: Theoretical and Typological Explorations. Ed. by Ekkehard König & Volker Gast, 307-346. Berlin: Mouton de Gruyter.

Geniušienė, Emma. 2007. Reciprocal and reflexive constructions in Lithuanian (with references to Latvian). In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 633-672. Amsterdam: John Benjamins Publishing Company.

Gerritsen, Nelleke. 1990. Russian reflexive verbs: In search of unity in diversity, volume 15. Amsterdam: Rodopi.

Gillon, Brendan. 1987. The readings of plural noun phrases in English. Linguistics and Philosophy 10(2): 199-219.

Goldberg Adele & Farrell Ackerman. The pragmatics of obligatory adjuncts. Language 77(4): 798-814.

Greenberg, Joseph H. 1957. The nature and uses of linguistic typologies. International Journal of American Linguistics 23(2): 68-77.

Grice, Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press.

Haspelmath, Martin. 2007. Further remarks on reciprocal constructions. In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 2087-2115. Amsterdam: John Benjamins Publishing Company.

Heim, Irene & Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell Publishers.

Heim, Irene, Howard Lasnik & Robert May. 1991. Reciprocity and plurality. Linguistic Inquiry 22(1): 63-101.

Heine, Bernd & Hiroyuki Miyashita. 2008. The intersection between reflexives and reciprocals. In Reciprocals and Reflexives: Theoretical and Typological Explorations. Ed. by Ekkehard König & Volker Gast, 169-223. Berlin: Mouton de Gruyter.

Higginbotham, James. 1980. Reciprocal Interpretation. Journal of Linguistic Research 1: 97-117.

Israeli, Alina. Semantics and pragmatics of the “reflexive” verbs in Russian. Munchen: O. Sagner

Ivanov, Valerij. 1990. Istoričeskaja grammatika russkogo jazyka. [Historical grammar of Russian.] Moscow: Prosveščenie.

Jackendoff, Ray. 1990. Semantic structures. Cambridge, MA: MIT Press.

355

Jakobson, Roman. 1932. Shifters, Verbal Categories, and the Russian Verb. Russian and Slavic Grammar: Studies 1931-1981. 41-58. The Hague: Mouton.

Janda, Laura. 2004. A metaphor in search of a source domain: The categories of Slavic aspect. Cognitive Linguistics 15 (4): 471-527.

Joseph, Brian & Richard Janda. 1988. The how and why of diachronic morphologization and demorphologization. Theoretical morphology: approaches in modern linguistics. Ed. by Michael Hammond & Michael Noonan, 193-210. Academic Press.

Kamynina, Aleksandra. 1999. Sovremmenyj russkij jazyk: Morfologija. [Contemporary Russian. Morphology.] Moscow: Moskovskij Gosudarstvennyj Universitet.

Kaup, Barbara, Stephanie Kelter & Christopher Habel. 2002. Representing referents of plural expressions and resolving plural anaphors. Language and Cognitive Processes 17: 405- 450.

Kemmer, Susanne. 1993. The Middle Voice. Amsterdam: John Benjamins Publishing Company.

König, Ekkehard. 2011. Reciprocals and semantic typology: some concluding remarks. In Reciprocals and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen C. Levinson & Asifa Majid, 329-339. Amsterdam: John Benjamins Publishing Company.

König, Ekkehard & Shigehiro Kokutani. 2006. Towards a typology of reciprocal constructions: Focus on German and Japanese. Linguistics 44(2): 271–302.

Knjazev, Jurij. Reciprocal constructions in Russian. 2007. In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 609-632. Amsterdam: John Benjamins Publishing Company.

Kratzer, Angelika. 2003. The event argument and the semantics of verbs, chapter 4. Unpublished ms., available at http://semantics archive.net.

Krifka, Manfred. 1998. The origins of telicity. In Events and Grammar. Ed. by Susan Rothstein. 197-235. Dordrecht: Kluwer.

Kruspe, Nicole. 2011. Reciprocal constructions in Mah Meri. In Reciprocals and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen C. Levinson & Asifa Majid, 329- 339. Amsterdam: John Benjamins Publishing Company.

Lakoff, George & Michael Johnson. 1980. Metaphors we live by. Chicago: University of Chicago Press.

Landman, Fred. 1989. Groups, 1. Linguistics and Philosophy 12(5): 559-605.

Landman, Fred. 2000. Events and Plurality. Dordrect: Kluwer Academic Publishers.

Langendoen, Terence. 1978. The logic of reciprocity. Linguistic Inquiry 9(2): 177-197.

356

Lasersogn, Peter. 1995. Plurality, conjunction and events. Dordrecht: Kluwer Academic Publishers.

Levinson, Stephen. Presumptive meanings: The theory of generalized conversational implicature. Cambridge MA: MIT Press.

Lichtenberk, Frantisek. 1985. Multiple uses of reciprocal constructions. Australian Journal of Linguistics 5(1): 19-41.

Lichtenberk, Frantisek. 2000. Reciprocals without reflexives. In Reciprocals: Forms and functions. Volume 2. Ed. by Zygmunt Frajzyngier & Traci Walker. 31-62. Amsterdam: John Benjamins Publishing Company.

Link, Godehard. 1983. The logical analysis of plurals and mass terms a lattice-theoretical approach. In Meaning, Use, and the Interpretation of Language. Ed. by Rainer Bauerle, Christoph Schwarze, & Arnim von Stechow. 302-323. Berlin: Walter de Gruyter.

Majid, Asifa, Nicholas Evans, Alice Gaby, & Stephen C. Levinson. 2011. The semantics of reciprocal constructions across languages. In Reciprocals and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen C. Levinson & Asifa Majid, 29-60. Amsterdam: John Benjamins Publishing Company.

Manney, Linda. 2000. Middle voice in modern Greek. Meaning and function of an inflectional category. Amsterdam: John Benjamins.

Maslova, Elena. 2007. Reciprocal and polyadic (remarkable reciprocals in Bantu). In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 2087-2115. Amsterdam: John Benjamins Publishing Company.

McGregor, William. 2000. Reflexive and reciprocal constructions in Nyulnyulan languages. In Reciprocals: Forms and Functions. Ed. By Zygmunt Frajzyngier & Traci Curl, 85-122. Amsterdam: John Benjamins Publishing Company.

McKercher, David. 2001. The polysemy of with in first language acquisition. PhD dissertation, Stanford University.

McNally, Louise. 1993. Comitative coordination: A case study in group formation. Natural Language & Linguistic Theory 11(2): 347-379.

Mihaliček, Vedrana. Instruments and accompaniments in Serbo-Croatian. In Issues in Slavic Syntax and Semantics, Ed. by Anastasia Smirnova & Matthew Curtis, 116-130. Cambridge Scholars Publishing.

Moltmann, Friederike. 2004. The semantics of together. Natural Language Semantics 12: 289- 318.

357

Moyse-Faurie, Claire. 2008. Constructions expressing middle, reflexive and reciprocal situations in some Oceanic languages. In Reciprocals and Reflexives: Theoretical and Typological Explorations. Ed. by Ekkehard König & Volker Gast, 169-223. Berlin: Mouton de Gruyter.

Nedjalkov, Vladimir. 2007a. Overview of the research. Definitions of terms, framework, and related issues. In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 231-334. Amsterdam: John Benjamins Publishing Company.

Nedjalkov, Vladimir. 2007b. Polysemy of reciprocal markers. In Reciprocal Constructions. Ed. by Vladimir Nedjalkov, 3-114. Amsterdam: John Benjamins Publishing Company.

Needham, Stephanie & Ida Toivonen. 2011. Derived arguments. In Proceedings of the LFG11 Conference, Ed. by Miriam Butt & Tracy Holloway King. CSLI Publications.

Partee, Barbara, Alice ter Meulen & Robert Wall. Mathematical methods in linguistics. 1993. Dordrecht: Kluwer Academic Publishers.

Rákosi, György. 2003. Comitative arguments in Hungarian. In Uil-OTS Yearbook 2003, 47-57.

Rákosi, György. 2008. The inherently reflexive e and the inherently reciprocal predicate in Hungarian; Each to their own argument structure. In Reciprocals and reflexives: Theoretical and typological explorations. Ed. by Ekkehard König & Volker Gast.

Richardson Kylie. 2007. Case and aspect in Slavic. Oxford: Oxford University Press.

Roberts, Craige. 1987. Modal Subordination, Anaphora, and distributivity. PhD diss., University of Massachusetts at Amherst.

Sabato, Sivan & Yoad Winter. 2008. Against partitioned readings of reciprocals. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.172.9678

Sabato, Sivan & Yoad Winter. 2005. From semantic restrictions to reciprocal meanings. Proceedings of 15th Amsterdam Colloquium.

Schwarzschild, Roger. 1996. Pluralities. Dordrecht: Kluwer Academic Publishers.

Sharvy, Richard. 1980. A more general theory of definite descriptions. Philosophical Review 89 (4):607-624.

Siloni, Tal. 2012. Reciprocal verbs and symmetry. Natural Language & Linguistic Theory 30:261–320.

Sirola-Belliard, Maija. 2011. Is there a future for the Finnish comitative? Arguments against the putative synonymy of the comitative case -ine and the postposition kanssa. In Case, animacy and semantic roles. Ed. by Seppo Kittila, Katja Vasti & Jussi Ylikoski. 135-153. Amsterdam: John Benjamins Publishing Company.

358

Stassen, Leon. 2000. AND-languages and WITH-languages. Linguistic Typology 4(1): 1-54.

Steedman, Mark. 1996. Surface Structure and Interpretation. No. 30 in Linguistic Inquiry Monographs. Cambridge, MA MIT Press

Steedman, Mark. 2000. The Syntactic Process. Cambridge, MA MIT Press.

Sternefeld, Wolfgang. 1998. Reciprocity and cumulative predication. Natural Language Semantics 6: 303–337.

Stolz, Thomas. 1997. Some instruments are really good companions—some are not. Theoretical Linguistics 23(1-2):113-200.

Stolz, Thomas. 2006. On comitatives and related categories: A typological study with special focus on the languages of Europe. Berlin: Walter de Gruyter.

Smythe, Sarah & Elena Crosbie. 2002. A comprehensive course in Russian. Cambridge: Cambridge University Press.

Timberlake, Alan. 2004. A reference grammar of Russian. Cambridge: Cambridge University Press.

Trawinski, Beata. 2012. Polish comitative constructions: Empirical investigation and formal description. PhD diss., Universität Tübingen.

Tremblay, Mireille. 1996. Lexical and non-lexical prepositions in French. In Configurations. Ed. by Anna Maria Di Sciullo. 79-98. Somerville, MA: Cascadilla Press.

Vassilieva, Masha & Richard K. Larson. 2005. The semantics of the plural pronoun construction. Natural Language Semantics 13: 101-124.

Vinogradov, Vikor. 1947. Russkij jazyk. Grammatičeskoe učenie o slove. [Russian language. A grammatical study on the word.] Moscow-Leningrad.

Wade, Terrence. 2010. A Comprehensive Russian grammar. Oxford: Blackwell.

Wierzbicka, Anna. 2009. Reciprocity. Studies in Language 33(1): 103-17.

Winter, Yoad. 2001. Plural predication and the Strongest Meaning Hypothesis. Journal of Semantics 18: 333–365.

Wood, Esther. 2007 The semantic typology of pluractionality. Unpublished Ph.D. dissertation, UC Berkeley.

Zarickij, Nikolaj. 1961. Formy i funkcii vozvratnyx glagolov na materiale drevnerusskogo jazyka. [Forms and functions of reflexive verbs in Old Russian.] Kiev: Izdatel’stvo kievskogo universiteta.

359

Zeshan, Ulrike & Sibaji Panda. 2011. Reciprocal constructions in Indo-Pakistani Sign Language. In Reciprocals and Semantic Typology. Ed. by Nicholas Evans, Alice Gaby, Stephen C. Levinson & Asifa Majid, 91-114. Amsterdam: John Benjamins Publishing Company.

360