Old Russian Possessive Constructions: A Construction Grammar Approach 9783110255041, 9783110255034

Table of contents :
Acknowledgements
List of abbreviations
1 Introduction
1 Possession and its neighbours
2 The case study
2.1 Earlier approaches
2.2 The text samples
3 Why construction grammar?
4 Semantic maps
5 A path through the book
2 A map of the possessive semantic space
1 Where does the meaning come from?
2 Relational nouns and slot fillers
2.1 Deverbal nouns
2.2 Other relational nouns
3 Construction meaning: reference points and intrinsic relationships
4 The nodes
4.1 Identification (ID and ID(slot)): Reference points on instance level
4.2 TYPE: Reference points on type level
4.3 LABEL: Strongly conventionalised constructions
4.4 Elaboration of relational nouns: ELAB(slot) and ELAB(part)
4.5 Elaboration of non-relational nouns (ELAB)
5 Meaning to form or form to meaning
6 Semantic maps
6.1 What does a semantic map show us?
6.2 Drawing vs. generation
6.3 A correspondence analysis plot
7 The working map
3 The constructions in Old Russian
1 Denominal adjectives
1.1 ADJ1: “True possessives”
1.2 ADJ2: “Relative” adjectives
2 The genitive
3 The dative
4 Mixed constructions
5 OCS and Old Russian
6 Summary
4 ID: Reference points on instance level
1 OCS
1.1 ID
1.2 ID(slot)
2 11th–14th century Old Russian
2.1 ID
2.2 ID(slot)
3 Further developments in the history of Russian
3.1 The ADJ1 construction
3.2 The ADJ2 construction
3.3 The genitive constructions
3.4 The dative construction
3.5 Mixed constructions
3.6 A snapshot of the 18th century: When did the genitive start expanding?
5 The ELAB nodes: Intrinsic relationships
1 Reference points or not?
2 ELAB(slot)
2.1 OCS
2.2 11th–14th century Old Russian
2.3 Further developments in the history of Russian
3 ELAB(part)
3.1 ELAB(part) in OCS and 11th–14th century Old Russian
3.2 Further developments in the history of Russian
4 ELAB
4.1 ELAB in OCS and 11th–14th century Old Russian
4.2 Further developments in the history of Russian
5 Conclusions
6 Types and conventionalised units: TYPE and LABEL
1 TYPE
1.1 TYPE in OCS and 11th–14th century Old Russian
1.2 Borderline cases
1.3 Further developments in the history of Russian
2 LABEL
2.1 OCS
2.2 11th–14th century Old Russian
2.3 Further developments in the history of Russian
3 Conclusions
7 Synchrony and diachrony
1 Synchrony: Division of labour vs. complementary distribution
1.1 Complementary distribution?
1.2 Division of labour
1.3 Map and territory
2 Diachrony
2.1 OCS vs. Old Russian
2.2 Development trends
2.3 The demise of the dative construction
2.4 The withdrawal of the ADJ2 construction
2.5 The weakening of the ADJ1 construction
2.6 From two genitive constructions to one?
2.7 A brief note on origins and causes
2.8 Diachronic paths
8 Concluding remarks
Appendix: Corpus and method
1 Text selection principles
1.1 Fair representation of each period
1.2 Geography
1.3 Literary genres
1.4 Text editions and manuscripts
1.5 Late copies of early manuscripts
2 The Old Russian text samples
2.1 Text genres
2.2 Periodisation of Old Russian texts
3 The OCS text sample
4 Excerpation, data registration and citation
5 Text excerpts
5.1 Old Russian
5.2 OCS
Notes
References
Index

Citation preview

Old Russian Possessive Constructions

Trends in Linguistics Studies and Monographs 237

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume

Hans Henrich Hock

De Gruyter Mouton

Old Russian Possessive Constructions A Construction Grammar Approach

by

Hanne Martine Eckhoff

De Gruyter Mouton

ISBN 978-3-11-025503-4 e-ISBN 978-3-11-025504-1 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data Eckhoff, Hanne Martine. Old Russian possessive constructions : a construction grammar approach / by Hanne Martine Eckhoff. p. cm. ⫺ (Trends in linguistics. Studies and monographs; 237) Includes bibliographical references and index. ISBN 978-3-11-025503-4 (alk. paper) 1. Russian language ⫺ To 1300 ⫺ Noun. 2. Russian language ⫺ 1300⫺1700 ⫺ Noun. 3. Russian language ⫺ Syntax ⫺ History. 4. Russian language ⫺ To 1300 ⫺ Grammar, Generative. 5. Russian language ⫺ 1300⫺1700 ⫺ Grammar, Generative. I. Title. PG2741.E24 2011 491.71701⫺dc23 2011035436

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany. www.degruyter.com

Contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

List of abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

1 1 2 2.1 2.2 3 4 5

Introduction . . . . . . . . Possession and its neighbours The case study . . . . . . . . Earlier approaches . . . . . . The text samples . . . . . . Why construction grammar? Semantic maps . . . . . . . A path through the book . .

. . . . . . . .

1 1 1 2 5 6 8 9

2 1 2 2.1 2.2 3

A map of the possessive semantic space . . . . . . . . . . . Where does the meaning come from? . . . . . . . . . . . . . . Relational nouns and slot fillers . . . . . . . . . . . . . . . . . Deverbal nouns . . . . . . . . . . . . . . . . . . . . . . . . . Other relational nouns . . . . . . . . . . . . . . . . . . . . . . Construction meaning: reference points and intrinsic relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . The nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . Identification (ID and ID(slot)): Reference points on instance level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TYPE: Reference points on type level . . . . . . . . . . . . . LABEL: Strongly conventionalised constructions . . . . . . . Elaboration of relational nouns: ELAB(slot) and ELAB(part) . Elaboration of non-relational nouns (ELAB) . . . . . . . . . . Meaning to form or form to meaning . . . . . . . . . . . . . . Semantic maps . . . . . . . . . . . . . . . . . . . . . . . . . What does a semantic map show us? . . . . . . . . . . . . . . Drawing vs. generation . . . . . . . . . . . . . . . . . . . . . A correspondence analysis plot . . . . . . . . . . . . . . . . . The working map . . . . . . . . . . . . . . . . . . . . . . . .

11 11 12 14 15

4 4.1 4.2 4.3 4.4 4.5 5 6 6.1 6.2 6.3 7

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

17 19 20 21 22 23 25 26 26 26 27 28 30

vi Contents 3 1 1.1 1.2 2 3 4 5 6

The constructions in Old Russian . Denominal adjectives . . . . . . . . ADJ1: “True possessives” . . . . . . ADJ2: “Relative” adjectives . . . . The genitive . . . . . . . . . . . . . The dative . . . . . . . . . . . . . . Mixed constructions . . . . . . . . . OCS and Old Russian . . . . . . . . Summary . . . . . . . . . . . . . .

4 1 1.1 1.2 2 2.1 2.2 3 3.1 3.2 3.3 3.4 3.5 3.6

ID: Reference points on instance level . . . . OCS . . . . . . . . . . . . . . . . . . . . . . . ID . . . . . . . . . . . . . . . . . . . . . . . . ID(slot) . . . . . . . . . . . . . . . . . . . . . 11th–14th century Old Russian . . . . . . . . . ID . . . . . . . . . . . . . . . . . . . . . . . . ID(slot) . . . . . . . . . . . . . . . . . . . . . Further developments in the history of Russian . The ADJ1 construction . . . . . . . . . . . . . The ADJ2 construction . . . . . . . . . . . . . The genitive constructions . . . . . . . . . . . The dative construction . . . . . . . . . . . . . Mixed constructions . . . . . . . . . . . . . . . A snapshot of the 18th century: When did the expanding? . . . . . . . . . . . . . . . . . . .

5 1 2 2.1 2.2 2.3 3 3.1 3.2 4 4.1

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . genitive . . . . .

The ELAB nodes: Intrinsic relationships . . . . . . . Reference points or not? . . . . . . . . . . . . . . . . . ELAB(slot) . . . . . . . . . . . . . . . . . . . . . . . . OCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11th–14th century Old Russian . . . . . . . . . . . . . . Further developments in the history of Russian . . . . . . ELAB(part) . . . . . . . . . . . . . . . . . . . . . . . . ELAB(part) in OCS and 11th–14th century Old Russian . Further developments in the history of Russian . . . . . . ELAB . . . . . . . . . . . . . . . . . . . . . . . . . . . ELAB in OCS and 11th–14th century Old Russian . . .

. . . . . . . . .

. . . . . . . . .

35 35 36 39 41 46 49 49 51

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . start . . .

53 53 53 59 68 68 74 81 82 85 87 89 89

. . . . . . . . . . .

. . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

91 93 93 95 95 104 111 115 115 117 118 118

Contents

vii

4.2 5

Further developments in the history of Russian . . . . . . . . . 126 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

6 1 1.1 1.2 1.3 2 2.1 2.2 2.3 3

Types and conventionalised units: TYPE and LABEL TYPE . . . . . . . . . . . . . . . . . . . . . . . . . . . TYPE in OCS and 11th–14th century Old Russian . . . . Borderline cases . . . . . . . . . . . . . . . . . . . . . . Further developments in the history of Russian . . . . . . LABEL . . . . . . . . . . . . . . . . . . . . . . . . . . OCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11th–14th century Old Russian . . . . . . . . . . . . . . Further developments in the history of Russian . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

131 131 132 134 134 135 136 140 144 148

7 1 1.1 1.2 1.3 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8

Synchrony and diachrony . . . . . . . . . . . . . . . . . . . Synchrony: Division of labour vs. complementary distribution Complementary distribution? . . . . . . . . . . . . . . . . . . Division of labour . . . . . . . . . . . . . . . . . . . . . . . . Map and territory . . . . . . . . . . . . . . . . . . . . . . . . Diachrony . . . . . . . . . . . . . . . . . . . . . . . . . . . . OCS vs. Old Russian . . . . . . . . . . . . . . . . . . . . . . Development trends . . . . . . . . . . . . . . . . . . . . . . . The demise of the dative construction . . . . . . . . . . . . . The withdrawal of the ADJ2 construction . . . . . . . . . . . The weakening of the ADJ1 construction . . . . . . . . . . . . From two genitive constructions to one? . . . . . . . . . . . . A brief note on origins and causes . . . . . . . . . . . . . . . Diachronic paths . . . . . . . . . . . . . . . . . . . . . . . .

151 151 152 153 155 156 156 157 158 161 162 168 171 173

8

Concluding remarks . . . . . . . . . . . . . . . . . . . . . . 179

Appendix: Corpus and method . . . . . 1 Text selection principles . . . . . . 1.1 Fair representation of each period . 1.2 Geography . . . . . . . . . . . . . 1.3 Literary genres . . . . . . . . . . 1.4 Text editions and manuscripts . . . 1.5 Late copies of early manuscripts .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . . . . .

. . . . . . .

. . . . . . . . . .

. . . . . . .

. . . . . . .

181 181 181 182 182 182 183

viii Contents 2 2.1 2.2 3 4 5 5.1 5.2

The Old Russian text samples . . . . . . . Text genres . . . . . . . . . . . . . . . . Periodisation of Old Russian texts . . . . The OCS text sample . . . . . . . . . . . Excerpation, data registration and citation Text excerpts . . . . . . . . . . . . . . . Old Russian . . . . . . . . . . . . . . . . OCS . . . . . . . . . . . . . . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

185 185 186 187 188 190 190 195

Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

Acknowledgements

The subject matter of this book has been with me for a long time. I had my first look at the Old Russian possessive constructions when I started working on my master’s thesis in 1998, and they were the subject of my doctoral dissertation, of which this book is the heavily reworked offspring. I would like to thank a number of people who have assisted and encouraged me along the way. All errors and shortcomings are of course my own. I am deeply indebted to my supervisors, Irina Lys´en and Kristian Emil Kristoffersen, for their encouragement, generosity, diplomacy and patience during my doctorate. I would also like to thank the participants at the Seminar of Cognitive Grammar at the University of Oslo for providing such a good environment for reading and discussion within the framework of Cognitive Grammar and construction grammar. Especially I would thank my friend and fellow martyr to genitive studies Ellen Hellebostad Toft for constant and engaging discussions on theoretical issues, and for being such an excellent reader and poser of difficult questions. I want to thank my past and present colleagues at the project “Pragmatic Resources in Old Indo-European Languages” at the University of Oslo for providing me with such an extraordinarily stimulating and exciting working environment. This book has benefited in a myriad ways from what I have learned in my years as a postdoc on the project. My special thanks to Eirik Welo and Dag Haug for reading parts of the book and giving sound and thorough advice from a different theoretical perspective. Dag Haug was also the one who lead me to believe I could write this book in no time, which of course was not true, but very useful all the same. I am very grateful to Mouton’s anonymous reviewer for some excellent suggestions. You made this book so much better. Many thanks to Anne Eilertsen, queen of the semicolon, for proofreading after my own perfectionist heart. Quoth the raven, “Furthermore!” Finally, my warmest thanks to my family; my children for coming along at precisely the right times and making me take breaks, and my husband Sturla Berg-Olsen not only for his great patience and generosity, but also for being such a tough, critical and well-informed reader, colleague and discussion partner.

List of abbreviations

ADJ1 ADJ2 DAT ELAB ELAB(part) ELAB(slot) GEN GENMOD GENUNMOD ID ID(slot) IJ IN J N N’ OCS OES OV PTC REFL SK

adjective type 1, “true” possessive adjective adjective type 2, “relative” adjective dative elaboration (by way of an intrinsic relationship) elaboration (filling the argument slot of a part noun) elaboration (filling the object-like argument slot of a relational noun) genitive modified genitive-marked noun or genitive-marked nominalised adjective/participle bare genitive noun identification (by a reference point) identification (by a reference point filling an argument slot) adjective derived with the suffix -žjadjective derived with the suffix -inadjective derived with the suffix -jadjective derived with the suffix -žn adjective derived with the suffix -žn’Old Church Slavonic Old East Slavic adjective derived with the suffix -ovparticle reflexive adjective derived with the suffix -žsk-

Chapter 1 Introduction

1. Possession and its neighbours Possession is a problematic concept. It is used in different ways by different authors, and is simply left undefined in much descriptive work. Nonetheless, “possessive” is a useful general label to put on the constructions analysed in this book, in the sense that all of them have meanings that would be described by most as “possession”. Attempts to define what a “possessive” relationship amounts to range from describing it as a relationship between two nominals determined by context to formulations of the maximally typical features of a possessive constructions, such as the possessive gestalt proposed in Taylor (1996:339–340). It is typical of many languages to have multiple, apparently hugely overlapping possessive constructions. The distinction between the prenominal spossessive and the postnominal of construction in English is a much-discussed example. Such constructions are interesting in many ways. The language in question may have two or more constructions with apparently very schematic semantics, but nonetheless distinct distributions. What, then, drives the distributional differences? Many possessive constructions have additional functions that are not possessive by any reasonable definition, and the different constructions often have different sets of such additional functions. What functions are the “neighbours” of possession, and how are they connected? This book will try to make sense of the various possessive-like functions and their neighbours by assuming a possessive semantic space and describing this space by way of semantic maps.

2. The case study: Adnominal possession in early Slavic and towards modern Russian The present exploration of the possessive semantic space is highly empirically driven. This book provides a close case study of constructional polysemy and synonymy in a family of constructions that can all to some extent express relationships commonly accepted as possessive. The test case is the

2 Introduction encoding of adnominal possession in early Slavic, as attested in Old Church Slavonic (OCS) and Old East Slavic (more specifically, 11th–14th century Old Russian).1 In both languages, we observe complex interactions between at least five different constructions – genitive, dative and adjective constructions of various kinds. Another interesting aspect of this group of constructions is that the encoding of adnominal possession is one of the most striking syntactic differences between the earliest attestations of Old Russian and modern Russian. Whereas Old Russian (and OCS) has at least five possessive constructions in a complex pattern of competition and division of labour, the expression of adnominal possession in modern Russian is dominated by the adnominal genitive. In early Slavic, on the other hand, the use of the adnominal genitive to express possession is severely restricted. Thus, the early Slavic constructions also form the point of departure of a fairly well-attested change to the interrelationships between a group of constructions competing in the possessive semantic space. This book therefore also offers a diachronic study of the development of Russian possessive constructions from the earliest attestations and through the 17th century (with a small test sample from the 18th century as well). Finally, modern Russian possessive constructions have been important in the semantic literature on possession, especially in the many works of Barbara Partee. A better understanding of their history may help to assess proposals about their present-day semantics and syntax, e.g. the claim in Partee and Borschev (2001) that modern Russian prenominal possessors2 are modifierlike and adnominal genitives are argument-like.

2.1. Earlier approaches The origins and history of Slavic adnominal possessives have interested scholars since the earliest days of Slavistics3 and are sketched in most general accounts of East Slavic historical grammar and syntax.4 . Possessives are also a recurring issue in literature on Common Slavic grammar and syntax and the branching of Slavic,5 and on OCS grammar and syntax.6 There are also more general works on possession across the Slavic languages and in a typological perspective, which touch upon the earliest attestations (Comrie 1976, Corbett 1987, Ivanov 1989), and on Slavic possession in an Indo-European perspective (Wackernagel 1908, Uryson 1980, Ivanov 1989).

The case study

3

There are a number of specialised works, each focusing on a specific facet of the history of possessive constructions in Russian. Bratishenko (1998, 2003, 2005) deals with the synchronic interrelationship between adjective and genitive constructions in the earliest East Slavic texts. Marojevi´c (1983a, 1983b, 1989) concentrates on possession in a stricter sense in the history of Russian, with emphasis on the origins of the first attested situation. Makarova (1954), Richards (1976) and Widn¨as (1958) all present diachronic surveys of the development of possessives in Russian. Zverkovskaja (1986) deals with the formation of Russian derived adjectives diachronically, and Uryson (1980) examines adjective formation and use in a single Old Russian text. Pravdin (1957) examines the possessive dative. None of these works aim to arrive at a full account of the interactions and history of every construction competing in the possessive semantic space. Furthermore, most of them only take possession in a stricter (but unfortunately often undefined) sense into consideration, not the entire set of functions where the possessive constructions compete. In these two respects, this book will hopefully be a contribution to a better understanding of Slavic possessive constructions and their history as well as to the general understanding of adnominal possessives. A review of the existing literature shows that a number of issues are still unresolved. Several authors (Flier 1974, Huntley 1984, Bratishenko 1998, 2003, 2005) have made it clear that the choice of possessive construction in OCS and Old Russian depends greatly on the properties of the possessor itself, and have set up various hierarchies of referentiality, animacy and specificity to account for this fact. Human and specific possessors to a very great extent favour one of the available adjective constructions (often simply referred to as “possessive adjectives”, ADJ1 in this book; see chapter 3, section 1.1). It has also been argued that this construction is inherently definite (Vaillant 1958:600). However, very little attention has been given to the properties of the possessee, the head noun in the possessive construction, and to the nature of the relationship between possessor and possessee. Only in passing have authors mentioned e.g. that subjects of deverbal nouns are usually realised as adjectives, whereas objects of deverbal nouns tend to be realised as genitives (Bratishenko 1998:153–158; Comrie 1976 makes this a general observation for his comparative study of modern Slavic languages, which is followed up by Corbett 1987:330). A study of early Slavic possessives is therefore likely to benefit from the interest in relational nouns and the exact nature of the possessive relation predominating in the semantic literature on possession.

4 Introduction In the earliest attestations of Slavic, there is a rather clear trend towards complementary distribution between adjective constructions and genitive constructions in at least some of the functions that can be called possessive. In Borkovskij’s formulation (1968:165–166), in the overwhelming majority of cases, the possessive genitive is found when the possessor is modified in some way, or is a nominalised adjective or participle, whereas bare genitives are extremely infrequent. Elsewhere, denominal adjectives are used. This is the focus in most general accounts of Russian historical syntax, and much effort has been put into various formulations of the conditions of the distribution. All this attention to complementary distribution is problematic in several ways: Firstly, several authors point out that the complementary distribution “rule” is hardly clear-cut; it is violated both by complex adjective constructions and bare genitive-marked nouns, a fact that is particularly clearly stated in Bratishenko 1998. Secondly, the exploration of the formal conditions on the adjective/genitive distribution has taken attention away from the semantically conditioned interactions between all the early Slavic possessive constructions. The interrelationship between the two different adjective constructions (or construction groups) is not well understood, nor are the conditions on where the adnominal genitive is used freely, and where it is not. There is little agreement on the status of the possessive dative construction, and it tends to be treated separately from the other possessive constructions. It is clearly a Slavic phenomenon, as it is fairly frequent in OCS and normally translates Greek adnominal genitives. However, it is often considered a Balkan phenomenon, as it is more frequent in East Bulgarian texts (Veˇcerka 1963:222; Veˇcerka 1993:198; Xodova 1963:134). In work on Old Russian, some authors state that it was rare (Stecenko 1977:54,101), whereas others say it was fairly frequent (Lomtev 1956:438; Borkovskij 1968:197– 198), and it is the opinion of Borkovskij (1949:362) that the possessive dative must have been in little use in the spoken language since it is so rare in Old Russian charters (gramoty). Richards (1976) dismisses it as a South Slavicism. Whatever opinions the authors hold, none of them systematically compare the distribution of the dative construction to those of the other constructions in the possessive semantic space, though Pravdin (1957:106–107) holds that in Old Russian, it was never fully synonymous with the possessive7 genitive. Comparing the dative construction’s semantics and distribution with those of the other possessive constructions would clearly be useful. There are also several points of disagreement when it comes to the historical development of the Slavic possessive constructions. Firstly, authors

The case study

5

posit very different Common Slavic systems. Some authors use the severe restrictions on the possessive genitive in the earliest attestations as an argument in favour of positing a Common Slavic state where the denominal adjectives dominate almost completely, and where the genitive plays a very small part in expressing possession (Marojevi´c 1989, Uryson 1980), while other authors posit a Common Slavic state that is very similar to the one attested in OCS (Richards 1976). Secondly, comparisons of the OCS and Old Russian systems are at best impressionistic, and the conclusions differ substantially. For instance, Richards (1976) claims that OCS had a cleaner complementary distribution between genitive and adjective constructions, whereas Bratishenko (1998:91) claims the opposite. Most authors state that the dative construction was more frequent and more freely used in OCS than in Old Russian, but it is not clear that this is the case. Thirdly, authors do not agree on the dating of the various changes in the distribution of possessive constructions in the history of Russian. Therefore, this book will use one OCS text sample and four text samples from Old and Middle Russian in order to contribute to a clearer understanding of why, when and how the genitive came to take over many of the functions of the denominal adjectives and all of the functions of the possessive dative.8

2.2. The text samples As all work in historical linguistics must be, this is a corpus study. At the time when the data work for this book was done, there were no proper electronic corpora of early Slavic texts (a situation which is now rapidly improving). The examples have thus all been hand-excerpted, and the text samples read in full, something which may have lead to some errors and omissions. In order to be able to compare OCS and the earliest attested Old Russian, and to do a diachronic survey of the development in Russian, I compiled five text samples: One of OCS texts, one of 11th–14th century Old Russian texts, and three samples of later Russian; one from the 15th century, one from the 16th century and one from the 17th century. The samples are sized in accordance with the number of years they span, and selected so as to give a balanced representation by geography and literary genre. The dating of the texts is by year of composition, not by the date of the manuscript (which may be several centuries later), but wherever there was a choice, preference was given to the edition based on the earliest manuscript. From each text

6 Introduction every construction involving an adnominal genitive, an adnominal dative or a denominal adjective was excerpted, with the following restrictions: – Only constructions with an overt head were excerpted – Constructions headed by numerals or quantifier nouns (such as “multitude”) were excluded (since dependent nouns in such constructions are consistently genitive-marked) – Constructions with bare genitive- or dative-marked prounoun dependents were excluded – The selection of denominal adjectives was limited as described in chapter 3, section 1.1 and 1.2 – Clearly qualitative adjectives were excluded The excerpts were first registered in a FileMaker database, and the data were then exported into and further processed in the statistics software package R. For a fuller description of the selection criteria and method of excerption, and a full list of the text samples, see the Appendix.

3. Why construction grammar? This book belongs to the usage-based variety of the construction grammar tradition in the sense that it subscribes to the notions shared by work by Ronald Langacker (1987, 1991), Adele Goldberg (1995, 2006) and William Croft (2001). As Goldberg (2006:chapter 10) points out, all flavours of construction grammar, and also Langacker’s mostly compatible Cognitive Grammar, take the view that a uniform type of description is possible for all linguistic units: Anything from a dependent morphological element or lexical item to a complex and abstract syntactic pattern may be described as a symbolic unit or a construction, a pairing of form and meaning. The constructions under scrutiny will not be described by way of any particular formalism. This choice is not a matter of principle; rather, the form–meaning split is used to concentrate on the semantics of the constructions, whereas the form side has not been worked out in detail. One of the most important implications of this view for the present book is the notion that even a relatively abstract syntactic pattern, a complex and (partially) schematic construction, has a semantic side to it, which may well be more schematic than that of a lexical item, but which in principle is of the

Why construction grammar?

7

same kind. An important consequence is that complex and schematic constructions may be – and often are – polysemous, and that such constructions may be in relationships of partial synonymy with other constructions.The meanings associated with a single construction, and the meanings shared by a group of formally unrelated, but partially synonymous constructions, can be seen as clusters in semantic space, and each construction’s distribution may be plotted onto a semantic map of such a cluster of meanings (see section 4). Constructions are organised in inheritance networks, and range from highly schematic (such as the Subject–Predicate construction, which is a schematic pattern from which all constructions with a subject and a predicate inherit) to highly lexically specific in the case of idiomatic expressions, such as spill the beans. This construction has a subject slot (and thus inherits from the Subject–Predicate construction), and the verb may be inflected, but otherwise it is quite inflexible. Constructions are also connected by semantic extension links and links that generalise over parts of different constructions. This book subscribes to the usage-based variety of this model, where the storage and prominence of a construction is deemed to depend on its actual use and frequency. Schematicity relations, semantic extensions and the results of usage frequencies will have a place in the analysis of the diachronic development of the Old Russian possessive constructions. Such an analysis is akin to the diachronic work on syntactic productivity by BarDdal (2008). Finally, construction grammar allows a careful analysis of the respective contributions of a construction schema and of its component parts. Possessive constructions are expected to have quite schematic meanings. Particularly the head noun, but also the possessor nominal, is expected to contribute considerably to the meaning of each instance of the construction. It could be argued that the meaning ascribed to the possessive construction as a whole might as easily be ascribed to the possessive morpheme in question (be it a case ending or an adjective suffix). Case endings and derivational suffixes are inextricably linked to the nouns to which they are attached, and a noun–suffix combination would also count as a construction on the present approach. Furthermore, at least the case endings have meanings that are highly conditioned by the type of element heading the case-marked noun. The meanings of the genitive, in particular, differ considerably depending on whether the genitive-marked noun is headed by a noun, a verb or a preposition, and it is not obvious that it is possible to give a coherent synchronic analysis of the meanings of a single suffix (or noun-suffix combination) across all head types, cf. the attempts in Berg-Olsen (2005), Toft (2010). The requirement that the head be a noun, as

8 Introduction well as the decision to look at the meaning of the construction as a whole, thus both seem justified.

4. Semantic maps Semantic maps are widely used in linguistics. They are most widespread in typology (e.g. Haspelmath 1997, 2003), but have also been put to use in smaller-scale comparisons between languages (Clancy 2006), and likewise in single-language diachronic studies – in comparisons between stages of one and the same language (Luraghi 2003). For a quantitative application of BarDdal’s work on syntactic productivity (2008), also involving semantic maps and cluster modelling, see Fedriani (forthc.) The rationale for using semantic maps in this book is twofold: Firstly, it has long been recognised that the semantics of possession is intricate, and that languages typically have multiple adnominal possessive constructions with substantial overlaps (e.g. Partee and Borschev 2001). The early Slavic situation is quite an extreme case, as there can be argued to be five or more overlapping constructions expressing this type of content. Semantic maps (based on cross-linguistic observations) are useful tools for single-language comparisons in such cases of rampant synonymy. Secondly, this is also a diachronic study. The book traces the development of possessive constructions from OCS, the earliest attestation of Slavic, up to 17th–18th century Russian, which is close to the present-day variety. During this period of time, the system of expressing adnominal possession changes substantially, but the changes take the form of slow and gradual semantic shifts – semantic extensions and retractions that can best be visualised on a map. Both for the synchronic analysis of complex interactions between partly synonymous constructions and for the analysis of the diachronic shifts, it makes sense to view the possessive constructions as form–meaning pairings interacting in a semantic space which is kept constant over time. Even though the semantics of each individual construction changes, the expectation is that each construction will have a semantic centre of gravity and a connected set of functions around this centre, and that their semantics will either expand or retract along connected paths in the semantic space (cf. Croft 2001:92–96).

A path through the book

9

5. A path through the book Chapter 2 discusses the existing literature on possession and relational nouns, and establishes a set of functional nodes for a map of the possessive semantic space based on a cross-classification of noun types and constructional meaning, using the analytic tools of Langacker (2000) and Taylor (1996) as a point of departure. A semantic map is then generated by way of a correspondence analysis of the data. Chapter 3 establishes the five main possessive construction types in the 11th–14th century Old Russian text sample and plots each of them on the map of the possessive semantic space. Although all constructions overlap to a smaller or greater extent, each construction is found to have a clear semantic centre of gravity. Chapters 4–6 each zooms in on natural groups of nodes on the map of the possessive semantic space and provides close studies of the distributions of constructions at each node in the OCS and 11th–14th century Old Russian text samples, exploring which forms can express which of the posited meanings, and under which circumstances? Each chapter also gives a diachronic overview of the development at each node up to the 17th century. Chapter 7 gathers the synchronic and diachronic findings from the preceding three chapters and evaluates the semantic map as a tool. The findings are used to establish tentative implicational connections between the map nodes. Chapter 8 gives some concluding remarks.

Chapter 2 A map of the possessive semantic space

The main theoretical tool of this book will be a semantic map of the possessive semantic space. Adnominal possession has been included in semantic maps by previous authors, e.g. the map of non-spatial case functions in Malchukov and Narrog (2009), where possessives are connected to agents, sources and benefactives. Likewise, the map of typical dative functions found in Haspelmath (2003) clearly shows how possessive datives might be connected to the greater dative semantic space. However, such maps have to my knowledge not been used to study the finer distinctions between the various subtypes of possession and how possessive constructions may compete and interact within a more fine-grained space.9 The first question we need to address is what functional nodes we need for such a map. The semantic literature on adnominal possession has focused on several aspects of the phenomenon, in particular on the contribution of the head noun’s semantics to the interpretation of the construction as a whole, and also on the possibility of positing independent constructional meanings for possessive constructions. This chapter adresses both issues: First, there is a discussion of the importance of relational nouns to the understanding of adnominal possessives. Second, as a viable proposal for constructional meanings in the possessive semantic space, the Cognitive Grammar analyses of English possessive constructions proposed in Langacker (2000) and Taylor (1996) are discussed. I shall argue that the nodes of a map of the possessive semantic map should be based on a cross-classification of the properties of the head noun and the main constructional meanings suggested in the Cognitive Grammar analysis. In the final subsection of the chapter I present such a configured map based on a correspondence analysis of the data sorted into the seven categories resulting from the cross-classification.

1. Where does the meaning come from? Constructions that can express adnominal possession in a wide sense are expected to have quite schematic meanings. This makes their component parts important. In particular, the head nouns of the constructions are expected to

12 A map of the possessive semantic space contribute substantially to the meaning of each instance of each construction. Some nouns are relational and provide a full-blown argument structure which to a large extent determines the interpretation of the entire construction. Nonrelational nouns may also provide strongly preferred interpretations, but there are some nouns that do not fall unambiguously into either class. Consequently, the nodes on our map will be based on a cross-classification of constructional meanings and type of head noun. The head noun may dictate the reading of a possessive construction to a great extent, and the construction is also expected to carry a meaning of its own. It is important not to lose sight of the possessor nominal either, as previous research on Slavic possessives and possessives in general strongly suggests that properties of the possessor nominal may influence the choice of possessive construction. An effort will therefore be made to work out what types of possessor nominals go with what constructions, especially what their status is with respect to animacy, information status and referentiality. This perspective will be maintained throughout the analysis.

2. Relational nouns and slot fillers Valency is a term usually associated with verbs, not nouns. However, it is a well-known fact that many nouns exhibit valency properties,10 and that these properties are crucial to the understanding of possessive constructions, since relational nouns normally will provide a lexical interpretation of the possessive relationship. Also, relational nouns are well known to behave differently fron non-relational ones cross-linguistically. For instance, Barker (forthc.) uses the English of construction as a diagnostic for relational nouns. The valency properties are most obviously seen with deverbal nouns, since they clearly have “slots” that can be filled by elements that would have been the subjects and/or objects of the corresponding verb, such as in Karajan’s interpretation of Tchaikovsky. Much of the literature has focused on such nouns (e.g. Chomsky 1970, Grimshaw 1990). But there are clearly also several classes of nouns that are either underived or derived from other parts of speech than verbs, but which nonetheless necessarily evoke both an entity, a relation to another entity and that other entity (the relatum). Kinship terms are good examples. There can be no reference to a word such as mother without the concept of children being evoked; the word clearly has an argument slot. The word niece likewise cannot be conceptualised without a relationship

Relational nouns and slot fillers

13

to another person, namely an aunt or an uncle. When it occurs in a possessive construction, such as Harriet’s niece, Harriet must be interpreted to be the aunt in the relationship, unless there is a very strong context to indicate otherwise. In this book I will assume that the major division is between relational and non-relational nouns. This division is hardly clear-cut, in the sense that some nouns which would normally be considered non-relational can in some cases occur as relational nouns, e.g. book, which can alternately be vieved as a physical object (non-relational), an artefact (relational, with a slot for its creator, be it a printer or an author) or a source of information (relational, with a slot for its subject matter).11 Nouns that are not inherently parts of wholes may appear in constructions which add this feature to their meaning.12 Items of clothing have some of the same associations as body parts and may in some contexts be construed as such.13 Finally, even indisputably relational nouns can sometimes be found in constructions where their argument slots are not filled. Non-relational (or sortal) nouns, then, are nouns such as book, dog, car.14 Nouns of this type do not necessarily invoke any specific relation to another entity, and therefore do not provide a fixed interpretation of the possessor when they head a possessive construction. However, in line with the observations about book above, quite a lot of nouns look like potential deverbal result nouns in the sense that it is easy to invoke the agent that brought about the action which produced them. Similarly, as mentioned, some nouns appear to be construed as parts of wholes in some cases, but not in others. Words denoting time units, such as day, are good examples. In phrases such as the seventh day of the month, day is clearly construed as an inherent part of month, i.e. the part-whole relationship between them is evoked (aided by the inherent relationality of the ordinal number). In phrases such as a beautiful day, on the other hand, day is clearly viewed as a whole in itself, not a building-block in a larger unit. In the actual analysis this is, however, unproblematic, as I will classify the constructions based on what argument slots are actually filled, not by the static properties of the head noun. Relational nouns, on the other hand, have one or more argument slots, or as Stefanowitsch (2003:430) puts it, they (consistently) invoke semantic frames with roles in them. The following two sections discuss and list some important subtypes of relational nouns to be used in the analysis.

14 A map of the possessive semantic space 2.1. Deverbal nouns Deverbal nouns are nouns that are transparently derived from verbs and retain (some of) their argument slots. Not all deverbal nouns reify a process. In fact, several classes of deverbal nouns instead denote one of the participants in the verb’s argument structure, and retain other roles from that argument structure to a greater or lesser extent. Taylor (1996:242–243) provides a classification of deverbal noun types based on which facet of the verbal predication is singled out for profiling,15 listed below and graded from more nominal (a–e) to more verbal (f). a) Agent nouns profile the (proto-)agent or uppermost participant in the action chain, such as invader, narrator. (1)

tvorecž nebu i zemli creator heaven-DAT/GEN and earth-DAT/GEN “the creator of heaven and earth” AN 24/4, 15th century16

b) Patient nouns profile the (proto-)patient of the process (draftee, appointee). (2)

avraam ideˇz poloˇzi zˇertvou bgvi where put sacrifice God-DAT Abraham ‘where Abraham put his sacrifice to God’ XID 22r/6, 11th–14th century

c) Result nouns profile an entity that comes into existence as a consequence of the process. Taylor gives bruise, dent, photograph as examples. This type is close to the non-relational nouns, since it has a subject slot that is often not filled. (3)

ponoˇsenija ponosjaˇstixż insulting-GEN insult ‘the insults of those who insult’ SBG 38/21, 11th–14th century

d) Manner nouns profile the manner in which a process is carried out by the trajector, e.g. walk as in He has a peculiar walk. (4)

zˇitie i vlastž imˇejai crskuju life and power having tsar-SK ‘having the lifestyle and power of a Tsar’ RCAM 16/7, 17th century

Relational nouns and slot fillers

15

e) Ability nouns profile the ability of the trajector to perform the activity. Taylor’s English example is speech as in He lost his speech. f) Nouns that that do not single out a particular facet of the process, but instead reify the process itself. Such nouns have argument slots that may or may not be filled.17 (5)

gnˇevż Boˇzij na tebˇe wrath God-IJ on you ‘God’s wrath at you’ PVC 37/21, 15th century

(6)

povelˇe zvoniti po vsemu gradu na sobranie ljudemż he-ordered ring over all city on collecting people-DAT ‘he gave orders to ring all over town to collect people’ PVC 28/22, 15th century

This classification is just an illustration of types of deverbal nouns that can typically be found. The text sample did not afford a sufficient number of examples to go into detail on the differences between various subtypes of deverbal nouns. It is also likely that the subtle differences between some of the types (such as manner nouns and ability nouns) may stem from the constructions they occur in rather than from the nouns themselves. Still, it is crucial to realise that the verbal character of these nouns, as well as the number of argument slots available with each type, is a matter of degree.

2.2. Other relational nouns Many relational nouns are not plausibly derived from verbs, but rather underived or derived from other parts of speech. Some of these are very similar to deverbal nouns,18 some are consistently relational, and some border on non-relational. The following types of relational nouns were prominent in the data underlying the current book; most types are taken from Taylor (1996), whereas others are suggested by Stefanowitsch (2003) and Barker and Dowty (1993). The list does not aim to be exhaustive. a) Kinship terms and terms denoting other human relationships. Kinship terms always imply reference to the relative(s) by which they are defined, i.e. sister always implies one or more siblings, mother implies children, uncle implies nephews/nieces. Kinship terms are very frequent in the Old Russian and OCS text samples.

16 A map of the possessive semantic space There are quite a number of nouns denoting various human relationships other than kinship which have an argument slot for another human being. Obvious examples are friend, neighbour, but also words denoting rulers, leaders etc. have this kind of relata, such as in (7). (7)

starˇeiˇsina klirikomż elder priests-DAT ‘the head of the clergy’ SBG 58/19, 11th–14th century

Many such nouns are similar to agentive deverbal nouns in meaning. I have, however, chosen to consider a clear morphological relationship to the corresponding verb a necessary and sufficient criterion for classifying a noun as deverbal. Thus, e.g. naslˇedžnikż “heir” will be considered an agentive deverbal noun, since it is derived from a verb, although it is very similar to non-deverbal relational nouns denoting human relationships. b) Nouns denoting inherent parts of wholes. Barker and Dowty (1993) divide this group into body part nouns (arm, head), other part nouns (wheel, chapter, handle) and boundary nouns (border, surface, corner). All these nouns must necessarily invoke reference to the whole to which they belong. (8)

na konžce groba at end grave-GEN ‘at the end of the grave’ PJul 114/16, 17th century

c) Representation nouns and information nouns.19 These two closely related noun types are similar to deverbal result nouns. Representation nouns designate “artefacts which represent, in some medium, another entity” (Taylor 1996:259) (portrait, statue, biography, painting). Such nouns invoke both reference to the creator of the artefact and to the entity that the artefact represents, and thus, as deverbal nouns, they have two argument slots that each may be filled by a possessor nominal. (9)

ikona styja bca icon holy-GEN mother-of-God-GEN ‘an icon of the holy Mother of God” XID 13v/4–5, 11th–14th century

Information nouns, such as report and version, are quite similar, and may be regarded as a subcategory of representation nouns. d) Deadjectival nouns and other nouns denoting abstract properties or characteristics, such as strength, beauty, shape, colour. These nouns have a

Construction meaning: reference points and intrinsic relationships

17

slot for the individual characterised by the quality denoted by the original adjective (see also Stefanowitsch 2003:428–429 and Barker and Dowty 1993). (10)

džrzostž poganžskuju nizżlagaemż insolence pagan-SK we-will-bring-down ‘we will bring down the pagans’ insolence’ SBG 49/19–20, 11th– 14th century

e) Nouns denoting influence without being deverbal. This is a category of miscellaneous nouns particularly denoting power and control. They have two argument slots: one for the one who has power/control and one for the one who is under his power/control. These nouns are not derived from verbs, but are semantically related to deverbal nouns with similar meanings. The category is prominent in the Slavic data. (11)

preimż vžsju vlastž Rusžsky zemlˇe having-taken all power Rus’-SK.GEN land-GEN ‘having taken all power over the land of Rus’’ SBG 60/16

3. Construction meaning: reference points and intrinsic relationships The literature on adnominal possession is substantial, see e.g. the wealth of references in Barker (forthc.). In this book I will largely use the analytic tools from Taylor (1996) and Langacker (2000), but I will contextualise these analyses and point out similar insights from work within other frameworks. Partee and Borschev (2000) point out that the “possessive” relation can come from three places: the context, an inherently relational noun or an inherently relational adjective (such as favourite). The Langacker/Taylor approach to English possessive constructions captures a similar insight: The head noun of a possessive construction may or may not provide the exact nature of the relation between possessor and possessee. A central question in the literature on possession is whether the possessive construction itself also contributes a meaning to the whole. Langacker and Taylor answer this question by positing that the possessive constructions in English each carry a highly schematic meaning. This meaning is most clearly seen with nouns that are not inherently relational (cf. Stefanowitsch 2003:429). For the English prenominal possessive construction (John’s hat), Taylor and Langacker posit this schematic constructional meaning to be a refer-

18 A map of the possessive semantic space















Figure 1. Langacker’s reference point analysis of the prenominal possessive (Taylor (1996:136), after Langacker (1991:172))

ence point situation, where a known possessor serves as a reference point in the sense that it is used to access and identify (uniquely) a (group of) less available referents. As Taylor (1996:17) informally puts it, the speaker “invites the hearer to first conceptualize [. . . ] the one entity (the possessor), with the guarantee that this will facilitate identification of the target entity (the possessee)”.20 The reference point situation is illustrated in figure 1. The dashed lines indicate the conceptualiser’s path: He first accesses the possessor entity, which then serves as a reference point (RP) to identify and access a target entity (T), i.e. the possessee, within the RP’s dominion (D), e.g. the set of entities sufficiently associated with RP to be identifiable from RP.21 For the English of construction, on the other hand, Langacker (2000:73– 90) appeals to the notion of an “intrinsic relationship”: If an entity X cannot be conceptualised without reference to some other entity or entities, X is conceptually dependent, and there is an intrinsic relationship between X and the other entity or entities. The notion of intrinsic relationship is most obviously seen with relational nouns, which all have one or more argument slots,22 see section 2. Deverbal nouns have slots for the subject and for any object(s) found with the corre-

The nodes

19

sponding verb (the destruction of the city), deadjectival nouns have slots for the nominal head found with the corresponding adjective (Mary’s beauty), and nouns inherently denoting parts have slots for the wholes of which they are parts (a slice of cake). However, Langacker also sees this same schematic meaning in the other usages of the of construction. In appositional constructions, such as the city of London, the relationship between the two nominals is one of identity, and hence the one cannot be conceived of as independent of the other. Similarly, he argues that the remaining cases of the of construction all involve the relationship between a non-relational noun and a very salient quality of the referent of that noun (a man of property, Joan of Arc). By this analysis, then, the constructional meaning of the of construction does not add to the relation inherent in the head nouns. The construction is simply a vehicle for the inherent relation between a relational noun and the noun filling its argument slot. This stands in opposition to the reference point analysis of the prenominal possessive construction, which specifies that the possessor is there to help the hearer identify the possessee and thus adds to the inherent or contextually available semantic relationship between the two nouns. However, the constructional meaning becomes “visible” in cases where the of construction is not headed by a relational noun: Langacker posits a generalisation of the fact that most of the occurrences have an argumental relation, and an extension of the very schematic notion of conceptual dependence to include certain usages with non-relational nouns, as in the examples above.

4. The nodes In our map of the possessive semantic space, the essential bipartition between reference point situations and intrinsic relationships will serve as the basis for the nodes, since the distinction appears to reflect two important poles in the semantics of possessive-like constructions. They will be cross-classified with other constructional properties: Is the reference point on instance or type level? Does the possessor fill the argument slot of a relational noun? Is the entire construction a conventionalised name for something or someone?

20 A map of the possessive semantic space 4.1. Identification (ID and ID(slot)): Reference points on instance level In a reference point situation, a reference point, which is one particular (group of) referent(s), is used to access a target, which is one particular (group of) referent(s). Taylor (1996:210) argues that for the possessor to be able to function as a reference point, it must have two properties: topicality and cue validity. Taylor’s term “topicality” differs from the usual understanding of “aboutness topics” in the literature (e.g. Lambrecht 1994). Rather, Taylor’s topicality seems to be a matter of information status along the lines of Prince (1981): The reference point/possessor needs to be more easily accessible than the target, which according to Taylor (1996:210, 219) means that it must either be accessible through the discourse context, directly or indirectly, or “automatically more easily accessed than others, regardless of discourse context”. “Cue validity”, on the other hand, is defined as the possessor’s ability to give reliable cues for the identification of the target (Taylor 1996:238). This notion is meant to explain why some entities with a clear semantic relation to the target are excluded from functioning as reference points. For example, parts rarely function as reference points for the wholes to which they belong. Possessors must be informative. With relational nouns, the relatum is usually so informative (i.e. has such high cue validity) as to exclude other interpretations, unless there is a very strong context to indicate otherwise. John’s wife will hardly be open to the interpretation “the wife (of somebody else) that John is sitting next to”, or other interpretations available with non-relational nouns. With non-relational nouns the range of possible identifiers is much greater, although Taylor (1996:261–264) argues that legal owners are natural reference points for many non-relational nouns, since a possession relation is typically an exclusive relation between a thing and a person.23 The criteria used to identify instance-level reference points are therefore the following: 1. The possessor serves to single out a referent or group of referents; the reference of the entire construction is not to a kind or type 2. The possessor is normally accessible from the context or from world knowledge, although marginally the possessor can be non-specific or kindreferring 3. The entire construction does not serve to conventionally name or label a particular referent24

The nodes

21

Normally, the possessor is uniquely identifiable, either from context or from world knowledge. It seems reasonable to assume that discourse-new, nonspecific or generic possessors in some cases may serve as reference points,25 but all of these possessor types are rare or marginal. Discourse referents (in the sense of Karttunen 1976) may be introduced for the first time in a possessive construction, but rarely are, as in A young man’s car broke down in the morning rush. He struggled to start it again. Non-specific possessors are possible e.g. in hypotheticals, but quite rare: If a thief breaks into a strong man’s house, he will run into difficulties.26 Likewise, kind-referring possessors are marginally possible in generic statements: The lion’s tail is long and bushy.27 Since these are all non-typical possessors, we may expect them to occur in other constructions than the typical discourse-available, normally also singular and frequently human possessor. The non-specific and kind-referring possessors are clearly similar to possessors in TYPE constructions, but I will consider them to belong to the ID nodes as long as the construction as a whole does not denote a type. In a reference point situation, the possessor may or may not fill a slot in the argument structure of a relational noun. Since it seems quite clear that this may serve as a dividing line between constructions (e.g. the alienable:inalienable distinction, and Barker’s (forthc.) use of the of construction as a diagnostic for relational nouns), there will be two separate nodes for reference point situations; ID for situations where the possessor does not fill an argument slot,28 and ID(slot) for situations where it does.

4.2. TYPE: Reference points on type level Under the reference point analysis Taylor (1996:293) also subsumes cases where the possessor is not uniquely identifiable, but denotes a type or kind, and where it does not serve to identify a referent or group of referents, but a category instead (mostly a subcategory of the head noun’s category). In these cases, the resulting construction does not involve reference to two discourse referents, and the construction is not a device used to anchor the head noun in the previous discourse, it just creates a “complex noun” which is available for further modification. In this way Taylor is able to include so-called possessive compounds (woman’s magazine) in his reference point analysis of the English prenominal possessive, and he also uses the same analysis to describe the transition between possessive compounds and regular compounds (dog food).

22 A map of the possessive semantic space Whether or not the bond between regular reference point situations and these situations is tenable, the TYPE function is clearly very relevant to some of the Old Russian adjective constructions. (12)

ovoˇstnoe xraniliˇste vegetable-N storeroom ‘a vegetable storeroom’ PVC 6/20, 15th century

For English, there are good diagnostics (cf. Taylor 1996:288–291) to separate “real” possessives from possessive compounds, but most of these are based on the fixed word order of English, and are hence not applicable to the early Slavic data of this book. The present classification therefore has to rely on the semantics alone. Occurrences classified as belonging to the TYPE node must fulfill the following three criteria: 1. The possessor nominal cannot be more than type identifiable in the sense of Gundel, Hedberg and Zacharski (1993)29 and does not denote an individual 2. The contribution of the possessor nominal cannot be more than to make the construction as a whole type identifiable 3. The possessor nominal must assign the whole construction to a category that is smaller than (or in some cases entirely different from) that of the head noun

4.3. LABEL: Strongly conventionalised constructions In some cases, even though the possessor (and often also the possessee) clearly has specific reference, the construction as a whole is highly conventionalised and has a restricted interpretation. Taylor (1996:295–296) starts his discussion of these cases with the socalled “onomastic compounds”, such as Halley’s comet or Occam’s razor. In these examples the possessor is clearly an individual, but such constructions still differ from regular possessives in two main ways: 1. The possessor does not need to be cognitively accessible, which indicates that it is not entirely referential (you can talk about Occam’s razor without knowing anything about William of Ockham)

The nodes

23

2. The expressions are conventionalised to the extent that they are often listed in dictionaries (Parkinson’s disease) or serve as complex proper nouns At least in English, there is sometimes clear syntactic evidence that such constructions occur as complex nouns, not as NPs (the usual Hobson’s choice), and semantically it is clear that in some of the cases the possessor nominal contributes to type identifiability only Thus, this is not an entirely uniform group. Instead, its members can be seen to fall into three main types, depending on how they deviate from our definitions of TYPE and ID respectively. The first type has a uniquely identifiable possessor which picks out a subtype. Thus, it is very close to the TYPE constructions, since the construction as a whole serves to name a subtype of the head noun’s type, as in Hobson’s choice, or a different type entirely, as in Adam’s apple. The second type has a kind-referring or type identifiable possessor, but the entire construction names something uniquely identifiable. In the early Slavic material, we see this particularly in constructions naming countries, languages and religions, where the possessor nominal is the name of an ethnic or religious group, and the head noun is the word for “country”, “tongue” or “faith”. The third group has a uniquely identifiable possessor that picks out a uniquely identifiable referent. Thus it is very close to the ID nodes, but it differs from them in that the entire construction serves as a name for the referent in question, and in that the possessor nominal needs not be cognitively accessible. This is clearly seen in examples such as Halley’s comet and St. Basil’s Cathedral. Given the considerable differences between the three types, they can be expected to have somewhat different encodings.

4.4. Elaboration of relational nouns: ELAB(slot) and ELAB(part) Constructions which can express possession can typically also be used to encode the inherent relation between a relational noun and a nominal filling a slot in its argument structure. It is reasonable to claim that in some of these cases, the slot filler does not serve as a reference point – it does not serve to access and identify the head noun. Rather, the slot filler just fleshes out the head noun with more information, as in the top of the mountain, a portrait of

24 A map of the possessive semantic space a young woman, and the construction does no more than encode the relation between the two. This analysis is clearly problematic if a construction is not exclusively used to code reference-point situations. This is the case for most, if not all, constructions in our early Slavic data. How are we to decide whether a reference point is involved, or just an intrinsic relationship? For English, Taylor uses his notions of cue validity and topicality to posit a subtle difference between relata which serve as reference points and relata which do not: With many relational nouns, both the prenominal possessive and the of construction are available, with alternations such as Rome’s destruction/the destruction of Rome. According to Taylor (1996), the choice of construction is determined by whether the relatum of the relation noun is cognitively accessible (topical, in Taylor’s terms) or not. Being intrinsic to the relational noun, the relatum has very high cue validity, but it is topicality which determines whether the relatum is a reference point or not. Ideally, therefore, an analysis of our data should take into account the information status of the slot fillers in all potential ID(slot) and ELAB(slot) constructions. In the practical analysis, I have chosen a more schematic solution, assuming that subject-like elaborations of relational nouns will normally be highly accessible (subject arguments of deverbal nouns, arguments of deadjectival nouns). Conversely, object-like arguments (object arguments of deverbal nouns, subject matter arguments of representational nouns, arguments of nouns denoting parts of wholes) will typically be less accessible in the context. I have therefore counted the former type as instances of ID(slot), and the latter type as instances of ELAB(slot), unless the elaborating noun was highly accessible in context. This simplification is based on the assumption that there is a strong asymmetry between subjects and objects both with verbs and with deverbal nouns when it comes to information status, distance to antecedent and referential persistence, as attested for English verbs e.g. in Brown (1983). However, it is not obvious that a free-word-order case language such as Old Russian should display the same patterns as the strongly subject-requiring fixed-word-order language English, nor that deverbal nouns should necessarily behave in the same way, especially since Slavic deverbal nouns are known to be ambiguous as to voice properties. In chapter 5, a small study of the information status and referential properties of such constructions in OCS supports the assumption that this is a viable solution. The correspondence analysis plot in chapter 2 separates ID(slot) and ELAB(slot) clearly, using

The nodes

25

data classified by the established principle. The analysis in chapter 4 also confirms that subject-like arguments are consistently encoded like other reference points are, not like object-like arguments. An additional concern is that languages may treat ELAB(slot) constructions headed by nouns denoting inherent parts of wholes differently than those headed by other noun types. As we shall see, this is certainly the case for the early Slavic data used in this book, where the adnominal genitive is virtually the only possible choice (see chapter 5, section 2). It is therefore necessary to add a node to the map to account for such cases. This node is labelled ELAB(part).

4.5. Elaboration of non-relational nouns (ELAB) Some cases of possessive-like constructions, such as (13), do not fall into either of the categories above. (13)

cˇ elovek muˇzeska polu person male-GEN sex-GEN ‘a person of the male sex’ PF 214/10, 16th century

These constructions do not involve reference points, but the head noun is not relational, nor does the construction as a whole denote a type or function as a conventionalised name or label for a particular referent. There appears to be three main types of constructions of this kind: 1. The two elements are in semantic apposition; the head and the modifier have the same referent (that scoundrel of a man) 2. There is a salient relationship between the two elements, e.g. between an artefact and the material of which it is made (a ring of gold) 3. The modifier denotes a non-accidental quality of the head noun (a man of integrity, Joan of Arc) As noted in section 3, Langacker (2000) subsumes such usages under his intrinsic relationship notion, e.g. they are similar to the ELAB(slot) constructions in the terminology of this book. In his opinion, the English of construction denotes an intrinsic relationship in the sense of conceptual dependence; the head noun cannot be conceptualised without the dependent. This meaning generalises over the ELAB(slot) and the other ELAB situations. A relational

26 A map of the possessive semantic space noun cannot be conceptualised without its inherent relatum (e.g. picture without its subject matter), but Langacker argues that the same holds for the situations on our list. Concrete objects cannot be conceptualised without the material they are made of, and an entity’s essential qualities are necessary parts of the conceptualisation of that entity. Finally, in the case of semantic apposition, the relation is one of identity, and thus conceptual independence is of course out of the question. Langacker’s analysis largely captures the same set of constructions as Pustejovsky’s constitutive qualia role (Pustejovsky 1995, see also Jensen and Vikner 2004).

5. Meaning to form or form to meaning The previous literature on early Slavic possessive constructions has not aimed for a precise understanding of the interrelationships between the constructions in the possessive semantic space. Thus, it does not do to presuppose any stable relationship between any of the forms with any of the postulated meanings. In the data collection phase of the work, the methodology was therefore based on going from form to meaning. Each occurrence was sorted by form, and the appropriate meaning selected from the list based on the context. Given the nature of the data – written data transmitted with a large amount of noise (see the Appendix) – there are of course a number of uncertain cases. As we shall see, the uncertainties are quite systematic. However, the data are presented the other way around. In the exposition I shall go from meaning to form. The semantic map is presented node by node, not construction by construction. Hopefully, this gives a clearer structure and a better understanding of both form, meaning and the interactions between constructions. The diachronic exposition also follows this pattern: For each node the OCS data are presented first, then 11th–14th century Old Russian, and then the subsequent stages in the Old Russian material.

6. Semantic maps 6.1. What does a semantic map show us? Semantic maps are in wide use in typology (Haspelmath 1997, 2003), but they have also been employed for smaller-scale cross-linguistic work (Clancy

Semantic maps

27

2006) and diachronic investigations (Luraghi 2003), and they have proved an efficient way of visualizing contrasts in large data sets. The idea behind semantic maps is what W¨alchli (2010) calls “similarity semantics”; a strictly empirical approach to meaning. The assumption is that the more similar two meanings are, the more likely they are to be expressed by a single form in any language. Thus, we expect linguistic expressions from many languages to cluster around the same functions. By visualizing such relationships between languages, semantic maps can “model general trends in the semantic organization of categories” (W¨alchli 2010). Traditionally, semantic maps are used to display allegedly universal category configurations where each function on the map is connected by a line to every other function that is expressed by the same means in at least one of the languages under consideration (cf. Haspelmath 1997, 2003, Luraghi 2003). Within this approach, it should be possible to make a full map of universal semantic or conceptual space, provided that a sufficiently comprehensive language sample is taken. The expectation is that of Croft’s Semantic Map Connectivity Hypothesis: “any relevant language-specific and constructionspecific category should map onto a CONNECTED REGION in conceptual space” (Croft 2001). In its strong version, this approach amounts to assuming a universal conceptual space. W¨alchli (2010) argues against the idea that semantic maps depict a universal mental semantic space. If this were the case, W¨alchli claims, semantic maps should be both robust (so that different data sets yield highly similar maps) and comprehensive (so that all semantic categories in data set are wellrepresented). He indicates that even small changes in the samplings may lead to big changes in the maps, and that rare categories are hardly discernible from noise. W¨alchli suggests that a more modest understanding of semantic maps is necessary, since semantic maps are “a tool for identifying the fundamental tendencies in the data” (W¨alchli 2010). Furthermore, he points out that computationally generated maps cannot claim to depict universal configurations. Rather, they are statistical and show only the most important and pervasive tendencies in the data.

6.2. Drawing vs. generation Two types of semantic maps are used in the linguistic practice: manuallydrawn maps (Haspelmath 1997, 2003, Luraghi 2003) and maps generated

28 A map of the possessive semantic space computationally (W¨alchli 2010, Croft and Poole 2008). A computational approach clearly has several advantages. It efficiently handles much larger data sets than hand-drawn maps feasibly can. As W¨alchli (2010) demonstrates, it is possible to use parallel data to reduce preconditioning of the data dramatically, since the similarity of linguistic expressions can be calculated by the extent to which they occur in the same situations in the text.30 In cases where the generated maps are based on preanalysed material (as is necessary when aligned parallel data are unavailable), computationally generated maps still have two major advantages: 1. The input data need not be binary, but can be proportional instead (modeling not just whether or not a category is present in a language, but to what extent it is present). 2. Semantic distance is modeled as Euclidean distance, giving better visualisations than the hand-drawn map can provide. Given the nature of our data, a semantic preanalysis is absolutely necessary, since we only have parallel data for the OCS text material. A third advantage of choosing a computational approach to semantic maps is the nature of the proposed nodes of our maps: The analysis in terms of reference points and intrinsic relationships carries several strong expectations about semantic extension and about certain meanings being more central and others more peripheral. We certainly expect strong kinship and therefore similar coding between ID and ID(slot), and likewise between ELAB and ELAB(slot), and ELAB(part) is expected to be different from the ELAB(slot) occurrences. We may also expect ID and ID(slot) to be central in the map, since these are commonly considered to be cases of “prototypical” or “paradigmatic” possession (Taylor (1996:339)). Computationally generated maps will model such similarities – if they exist – as proximity.

6.3. A correspondence analysis plot In my analysis I have chosen an approach based on a generated map, where dissimilarity is represented as distance. In the chapters ahead, I will indicate the boundaries of each construction in these maps, as is commonly done in traditional, hand-drawn maps. The configuration of the nodes and the distance between them will be determined by a correspondence analysis based on three diachronic stages of

Semantic maps

29

the full data set: the OCS sample, the 11th–14th century Old Russian sample and the Middle Russian sample from the 17th century.31 The configuration of nodes represents semantic similarity and dissimilarity between the functions, and it relies on the extent to which each of the constructions are attested, not only on attestation vs. non-attestation (cf. Cysouw 2007). An important reason for the choice of a generated map is the fact that, as Narrog (2010:233–234) points out, a traditional hand-drawn map of a semantic space with much overlap runs the risk of becoming vacuous; a map where all nodes are connected to all nodes does not carry much meaning, whereas a map that models similarity as closeness does. Differentiation by frequency is therefore crucial when dealing with constructions with substantial semantic overlaps, such as the possessive constructions under analysis in this book. Construction OCS ADJ1 ADJ2 DATIVE GENMOD GENUNMOD OES ADJ1 ADJ2 DATIVE GENMOD GENUNMOD 17th century ADJ1 ADJ2 DATIVE GENMOD GENUNMOD

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB(part)

ELAB

0 46 0 0 1

15 88 1 2 0

171 27 10 59 10

183 26 25 126 21

12 34 31 43 27

0 3 7 22 36

5 103 12 34 14

1 218 0 0 0

26 265 0 54 1

148 57 11 122 15

247 84 38 267 27

17 82 69 83 28

0 0 2 32 28

0 80 41 64 41

3 206 0 0 0

35 50 0 9 0

45 11 1 45 10

94 27 0 92 10

10 21 14 57 14

0 0 0 10 35

1 40 6 44 17

Table 1. Data set for the correspondence analysis, construction by node (OES = Old East Slavic, e.g. 11th–14th century Old Russian)

The data set contains the raw frequencies of each construction type at each node in each sample, shown in table 1. The correspondence analysis calculates one distance matrix accounting for the differences between the rows in the data set, and one distance matrix accounting for the differences between the columns in the data set. The distance matrices are represented

30 A map of the possessive semantic space as faithfully as possible in two two-dimensional scatterplots, which are then superimposed.32 Correspondence analysis was chosen as a method primarily because of its visualisation possibilities: A correspondence analysis plot can visualise a node configuration, in the same way a traditional semantic map does, at the same time as visualising semantic (dis)similarity between construction types in the same plot. It also handles proportion data well. In his defense of the classical hand-drawn semantic map, van der Auwera (2008) points out that its main strength is its inherent diachronic dimension. Since the traditional hand-drawn map is implicational, e.g. contains hypotheses about which functions may only be similarly coded if an intermediate function is also coded in the same way, it also indicates paths of diachronic development. As this book is an analysis of a very small set of very closely related languages and stages of the same language, the results of this study cannot pretend to provide valid grounds for such generalisations, and this is a weighty reason for not choosing the hand-drawn approach in this book. However, van der Auwera (2008:45) suggests that a combined approach between generation and hand-drawing might be the best way to preserve the diachronic dimension in semantic maps. In chapter 7, I will suggest some hypothetical relationships for such a future map of the possessive semantic space based on the findings in this book.

7. The working map As seen in section 4, the basis of the seven nodes for the map of the possessive semantic space is the juxtaposition of two very schematic meanings: reference point situations vs. intrinsic relationships. This makes it possible to make certain predictions for the map: We expect the nodes based on the reference point analysis (the two ID nodes, but also TYPE and to some extent LABEL) to be closer to each other than to the other nodes, and vice versa for the nodes based on the intrinsic relationship analysis (ELAB(slot), ELAB and ELAB(part)). In order to substantiate the choice of adding the ELAB(part) node, I ran two correspondence analyses, one to visualise the six basic nodes (subsuming ELAB(part) under ELAB(slot)) (figure 2 on page 31), and one with ELAB(part) split from the rest of the ELAB(slot) occurrences (figure 3 on page 32). The two plots immediately reveal that ELAB(part) should be a separate node,

31

17cent.DATIVE

1.0

17cent.GENUNMOD OCS.GENUNMOD OES.DATIVE OES.GENUNMOD

ELABslot

OCS.DATIVE

ELAB

0.5

Factor 2 (28 %)

1.5

2.0

The working map

17cent.GENMOD

OCS.ADJ2

0.0

OCS.GENMOD OES.GENMOD OES.ADJ2

LABEL


0.5

IDslot. ID

17cent.ADJ2

TYPE

17cent.ADJ1 OES.ADJ1 OCS.ADJ1


1.0


0.5

0.0

0.5

1.0

1.5

2.0

Factor 1 (58.5 %)

Figure 2. A correspondence analysis plot of the six basic nodes in OCS, 11–14th century Old Russian (OES) and Middle Russian (17th century)

since the correspondence analysis sets it clearly off from all the other nodes, including ELAB and ELAB(slot). The plot in figure 3 on page 32 shows a configuration of our seven nodes that supports our predictions.33 In particular, we see that ID and ID(slot) are extremely close, suggesting that reference point situations on instance level are coded largely in the same way, regardless of the relationality of the head noun. ELAB(slot) and ELAB are very close as well, indicating that intrinsic relationships are also largely coded the same way regardless of the relationality of the head noun. The exception is clearly ELAB(part), but we see that it is located much closer to the two other ELAB nodes than to any other nodes. There are clearly two axes of importance to the choice of possessive construction in the early Slavic data set: Factor 1 is visualised with ID and TYPE as its two poles, with LABEL intermediate between them. This axis thus appears to deal with the distinction between instance and type level

32 A map of the possessive semantic space

0.5

OCS.ADJ1 OR.ADJ1 17cent.ADJ1

ID IDslot

LABEL

0.0

OR.ADJ2


0.5

OCS.GENMOD 17cent.GENMOD

OCS.ADJ2

ELABslotELAB

OCS.DATIVE OR.DATIVE


1.0

Factor 2 (27.1 %)

TYPE

17cent.ADJ2

OR.GENMOD

OR.GENUNMOD 17cent.DATIVE


1.5

OCS.GENUNMOD


2.0

17cent.GENUNMOD ELABpart


1.0


0.5

0.0

0.5

1.0

1.5

2.0

Factor 1 (52 %)

Figure 3. A correspondence analysis plot with ELAB(part) as a separate node in OCS, 11–14th century Old Russian (OES) and Middle Russian (17th century)

constructions, and accounts for all reference-point based nodes. Factor 2, on the other hand, has ID and ELAB(part) as its two poles, with the other ELAB nodes as intermediary, and thus apparently deals with the distinction between instance-level reference points and intrinsic relationships. We see that the ELAB(slot)/ELAB cluster is central in the plot, indicating that these two nodes are affected by both the instance/type axis and the reference point/elaboration axis. When we look at the configuration of the construction types by period, we see that the plot indicates several diachronic trends. In particular, the dative construction and one of the adjective constructions appear to go through considerable changes during the period under consideration in this book. We shall return to the correspondence analysis plot as a tool for analysing semantic change in chapter 7.

The working map

33

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 4. A basic map of the possessive semantic space

In the following semantic maps, this basic configuration of semantic nodes will be maintained. Boundaries will be added to delimit single-language constructions. The base map is seen in figure 4. In figure 5 on page 34, the nodes are connected by lines to indicate which of them can be coded in the same way in my data set.34 Since, in fact, there is a construction that covers all nodes except one (one of the adjective constructions), all nodes are connected to all nodes, except ELAB(part) and TYPE. The connecting lines are differentiated by the strength of the connections. A finely dotted line indicates a single-construction connection, a dashed line indicates a connection shared by three constructions, and a fully drawn line indicates a connection shared by all five possessive constructions.35 We see that the connecting lines add to the information conveyed by the distances. The ID nodes and the ELAB/ELAB slot cluster are connected by fully drawn lines, and are thus encoded by the same constructions, but clearly in different proportions. The two most extreme nodes, TYPE and ELAB(part), are not connected to each other at all. TYPE is the node most weakly connected to the other nodes, by only a single construction. ELAB(part) is better connected, with three-construction connections to the other ELAB nodes and to the two ID nodes, and a single-construction connection to the LABEL construction.

34 A map of the possessive semantic space

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 5. A map of the possessive semantic space with connecting lines based on the Slavic datasets

Similarly, LABEL has three-construction connections to the two ID nodes, ELAB(slot) and ELAB, but only a single-construction connection to TYPE and ELAB(part).

Chapter 3 The constructions in Old Russian

The possession-like semantics found in our seven map nodes can be expressed in three main ways in early Slavic: with a denominal adjective construction, with an adnominal genitive construction or with an adnominal dative construction. There are good arguments in favour of splitting the adjective constructions into two main groups, and I shall also argue in favour of splitting the adnominal genitive into two separate constructions. In addition, there are various kinds of mixed constructions, involving either multiple adjectives or adjectives and genitives at the same time. In this chapter, the construction types are presented using 11th–14th century Old Russian data for illustration. To state the rough semantics of these constructions I will plot their boundaries on the semantic map from chapter 2, and also indicate their semantic centre of gravity in terms of frequency.

1. Denominal adjectives Old Russian has a rich and quite productive system of denominal adjective formation, and denominal adjective constructions are used extensively in cases where other Indo-European case languages typically use adnominal genitive constructions or compounds. In fact, as seen in table 2, in the OCS material nearly 90 % of all denominal adjective constructions are translations of Greek adnominal genitives. 36

ADJ1 ADJ2 GENMOD GENUNMOD DATIVE

genitive 94.7 80.9 93.2 72.8 92.9

dative 0 0 0.3 0 3.1

adjective 1.3 6.2 0 0 0

other construction 3.8 12 4.8 27.2 4.1

no equivalent 0.3 0.9 1.7 0 0

Table 2. Greek correspondences to OCS possessive constructions, per cent

The denominal adjectives are traditionally separated into two groups: possessive adjectives (pritjaˇzatel’nye prilagatel’nye) and relative adjectives (ot-

36 The constructions in Old Russian nositel’nye prilagatel’nye). This is a division that makes sense; the possessive adjectives are, as we shall see, possessive in a quite narrow sense, whereas the relative adjectives have a wider range of functions. However, since we also find the “relative” adjectives in reference point situations, at our ID and ID(slot) nodes, I wish to avoid this terminology, and shall refer to the two groups simply as type 1 adjectives (ADJ1) and type 2 adjectives (ADJ2). 37 Previous accounts of Slavic possessive constructions have included different selections of adjectives/suffixes, and this is probably one of the reasons why some of them arrive at such different results. Some authors, such as Corbett (1987), include only the unequivocally possessive adjectives in -ovand -in-; others, such as Richards (1976), Uryson (1980) and Bratishenko (1998, 2005), include varying numbers of additional suffixes, while others again, such as Widn¨as (1958) and Comrie (1976), do not specify what suffixes they are talking about at all. To avoid these problems, in this book the two groups will primarily be defined by the suffixes involved (with a few exceptions stated below). I will only comment briefly on the history and etymology of the suffixes; for an in-depth account, see Zverkovskaja (1986).

1.1. ADJ1: “True possessives” In this book, type 1 adjectives will be taken to include adjectives derived from nouns denoting persons and personifications, directly or metonymically, with the suffixes -ov-, -in- and -j-. In addition, and perhaps somewhat controversially, this group is deemed to include adjectives derived from nouns denoting persons with the suffix -žn’-. Finally, the group includes the adjective boˇzij “God’s”, which is derived with the suffix -žj-, but which still clearly patterns with the ADJ1 constructions. In the examples, the adjectives are glossed with their simplified suffixes: OV, IN, J, N’ and IJ. (14)

zˇena ivanova wife Ivan-OV ‘Ivan’s wife’ Z 32/253

(15)

gradż volžzinż city Olga-IN ‘Olga’s city’ PVrL 60/3–4

Denominal adjectives

37

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 6. Old Russian ADJ1 on the map of the possessive semantic space

(16)

synż Solomonž son Solomon-J ˇ 362/23 ‘Solomon’s son’ ZD

(17)

ot ruki gospodnja from hand Lord-N’ ‘from the Lord’s hand’ PBR 346/34

(18)

posobiemž Boˇziemž help God-IJ ‘by God’s help’ SBG 44/6

Of the four suffixes involved, only -ov- and -j- are in competition, both primarily deriving adjectives from o-stem nouns. -j- is older than -ov-, but was productive both in OCS and Old Russian, though more so in OCS. In this period, -ov- is in the process of expanding at the expense of -j-. The suffix -in- is generally used to derive adjectives from a-stem nouns (not necessarily feminines). The suffix -žn’- is involved in the formation of two quite different groups of adjectives: 1) Adjectives derived from stems denoting persons, mostly kinship terms or characteristic-property terms. 2) Adjectives that occur exclu-

0

50

100

150

200

38 The constructions in Old Russian

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 7. 11th–14th century Old Russian ADJ1 distributed on functions, n = 449

sively in long form, derived from adverb and noun stems, denoting position in space or time. The latter group has been excluded from consideration in this book, since they only marginally have possession-related functions. The semantics of the Old Russian ADJ1 construction is briefly sketched in the map in figure 6 on page 37. As indicated by the dashed line, the boundaries of the construction comprise both ID nodes, ELAB(slot) – but not ELAB and ELAB(part) – and LABEL.38 As we can see from figure 7 on page 38, LABEL is not particularly frequent with ADJ1, and ELAB(slot) is also considerably more rare than the two ID nodes. This is modeled by the inner fully drawn boundary. In this and the following figures, the fully drawn boundary will comprise only functions that account for more than 15 % each of the occurrences, in order to have a way of stating the construction’s semantic centre of gravity. It is clear that the ADJ1 construction is a construction predominantly encoding reference point situations on instance level in the 11th–14th century Old Russian sample.

Denominal adjectives

39

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 8. Old Russian ADJ2 on the map of the possessive semantic space

1.2. ADJ2: “Relative” adjectives The ADJ2 group includes adjectives derived from nouns with the suffixes -žsk- (glossed as SK), -žn- (glossed as N), and, with the exception of the adjective boˇzij, -žj- (glossed as IJ). Adjectives with clearly qualitative meanings, such as angelžskż in the meaning “angelic, with angel-like behaviour”, are excluded from the material. (19)

razumż cˇ lvˇcskż understanding man-SK ‘the understanding of man / human understanding’ PVM 244/10

(20)

vvergżˇse i v propastž smrtnuju having-thrown him in abyss death-N ‘having thrown him into the abyss of death’ PVrL 175/3

(21)

za korovije moloko for cow-IJ milk ‘for cow’s milk’ RP 39/429–430

0

50

100

150

200

250

40 The constructions in Old Russian

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 9. 11th–14th century Old Russian ADJ2 distributed on functions, n = 848

The suffix -žsk- is very productive, and, according to Zverkovskaja (1986:50), combines with any type of noun stem in 11th–12th century Russian. The suffix -žn-, on the other hand, rarely forms adjectives from animate noun stems. The suffix -žj- is etymologically related to -j-, and can sometimes be hard to distinguish from it. 39 Looking at the distribution of ADJ2 in the 11th–14th century Old Russian data, we find that the boundaries of this construction comprise every node on the map except ELAB(part), as shown by the dashed line in figure 8 on page 39. The construction is attested in all these functions. On the other hand, the boundary delimiting nodes that account for more than 15 % of the ADJ2 occurrences shows us that this construction, too, has a fairly small centre of gravity: TYPE accounts for 25.7 % of the occurrences, and LABEL for 31.3 %. As the graph in figure 9 on page 40 shows, however, ADJ2 is much more evenly distributed across the less central functions than ADJ1 is, with ID(slot), ELAB(slot) and ELAB each coming close to the 10 % mark. Figure

The genitive

41

9 also shows us that the number of uncertain cases is quite high; these are mostly occurrences that could be either ID/ID(slot) or TYPE. In examples such as (22), it is not obvious whether bole˛ rskye˛ has reference to some particular Boyars that could be used to identify their wives, or whether it is a non-referring expression and part of a TYPE construction. (22)

i zˇeny bole˛ rskye˛ muˇzei svoixż i ospodarevż ostali and women Boyar-SK men their and masters lost ‘and the Boyars’ wives / Boyar women have lost their husbands and masters’ Z 32/257

2. The genitive Old Russian can clearly use the adnominal genitive for possessive semantics, but as all of the literature agrees, this use is severely restricted. A simple (but imprecise) generalisation is to say that the adnominal genitive is in complementary distribution with the possessive adjectives: You can only use the genitive if adjectives are unavailable.40 Roughly, this means that the genitive is used when the possessor consists of more than two words, as in example (23), since adjectives could not normally be formed from complex NPs (see, however, section 4). The same holds when the possessor is itself an adjective or a participle, as in example (24). This must be seen as a strictly syntactic condition on the use of the adnominal genitive. (23)

pavla apsla domż Paul-GEN apostle-GEN house ‘Paul the apostle’s house’ XID 25v/4

(24)

kaznitž nas Bż naxoˇzenžem poganyx Pagan-GEN punishes us God invasion ‘God is punishing us with the Pagans’ invasion’ SL 462/35–36

However, there are clearly also semantic conditions on the use of adnominal genitives. Most of the literature will leave it at remarking that the adnominal genitive can be used freely, e.g. also with bare nouns, when a partitive shade of meaning is involved. Again, this is a simplification, as shown in the detailed analysis of Old Russian bare adnominal genitives in Bratishenko (1998). Her analysis mainly takes into consideration the semantics of the possessor nominal, and she finds that unmodified adnominal genitives are much

0

50

100

150

200

250

42 The constructions in Old Russian

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 10. 11th–14th century Old Russian GENMOD distributed on functions, n = 641

less likely to be found in the upper than the lower part of what she refers to as an agent/possessor hierarchy, which basically amounts to a fine-grained animacy hierarchy: proper personal noun > supernatural being > common noun for mature person > common noun for immature person > infant > animal > inanimate (Bratishenko 1998:162). She also argues that nouns in the lower part of this hierarchy are more likely to occur as bare genitives when they are semantically definite, since this on her analysis is a meaning not carried by the group 2 adjectives (Bratishenko 1998:92–93). Thus, these observations tie in naturally with the conditions on formation of group 1 and group 2 adjectives. Bratishenko does not take into consideration the properties of the head noun in the possessive constructions. The 11th–14th century Old Russian data used in this book (see figure 10 on page 42 and figure 11 on page 43), however, shows clearly that bare adnominal genitives are much more likely to be used if the head noun is relational, and in particular if there is no reference

43

0

10

20

30

40

The genitive

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 11. 11th–14th century Old Russian GENUNMOD distributed on functions, n = 144

point situation involved. That is to say, the main bulk of bare adnominal genitive attestations are found at the ELAB(slot) node. This subsumes the general observation that the bare genitive is freely used when partitivity is involved, as in example (25), but expands it. This generalisation does not hold just for nouns denoting parts of wholes (or used as such), but also for all relational nouns with object-like slots, such as example (26). (25)

i skoropii mnoˇzstvo zmii. multitude snake-GEN and scorpion-GEN ‘a multitude of snakes and scorpions’ PVrL 39/23

(26)

ot vragż i slugż zloby by enemies and servants evil-GEN ‘by the enemies and the servants of evil’ PVrL 41/10

44 The constructions in Old Russian

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 12. Old Russian GENMOD on the map of the possessive semantic space

I shall therefore argue that we are dealing not with one adnominal genitive construction, but with two separate form–meaning pairings. The first construction is severely restricted on the formal side; it can only be used when adjective formation is not an option. On the semantic side, it has the same properties as the corresponding adjective construction (ADJ1 or ADJ2) would have. The other adnominal genitive construction has no such formal restrictions, but has a different semantics, centering on the three ELAB nodes rather than on the ID nodes. It is important to note that we have no direct evidence for the existence of two such constructions. What we do have is a considerable difference in distribution between constructions with bare adnominal genitive nouns (GENUNMOD) and adnominal constructions with either a complex genitive NP or a genitive-marked adjective/participle (GENMOD), which is clearly seen in figures 10 and 11. All the occurrences in the GENUNMOD group would by definition have to belong to the unrestricted genitive construction. The GENMOD group, on the other hand, would include occurrences of both constructions. Example (23) would be highly unlikely to belong to the unrestricted genitive constructions, since there is just a single bona fide occurrence of a genitive-marked

The genitive

45

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 13. Old Russian GENUNMOD on the map of the possessive semantic space

adnominal personal name in the 11th–14th century Old Russian text sample (example (27)), whereas example (28) is quite likely to be, since the complex genitive NP krove bratja moeja fills the object slot of a deverbal noun. This is rather a rare function with ADJ1 and ADJ2, and rather a typical one with the GENUNMOD occurrences. (27)

lice zˇe ego – aky lice Iosifa face PTC his like face Joseph-GEN ˇ ‘but his face was like Joseph’s face’ ZAN 160/13

(28)

bratja moeja da budetž otmestžnikż Bż krove God blood-GEN brothers-GEN my-GEN that will-be avenger ‘may God be the avenger of my brothers’ blood’ PVrL 141/13

Even if we draw our semantic maps on the basis of the GENMOD and the GENUNMOD groups without any attempt of filtering out the unrestricted genitive construction occurrences from the GENMOD group, we get quite different maps with different centres of gravity (figures 12 and 13). The boundaries of the GENMOD group comprise both ID nodes, all three ELAB nodes and LABEL, whereas LABEL has only a single attestation from the

46 The constructions in Old Russian GENUNMOD group, and is therefore not included within the GENUNMOD group’s boundaries. GENMOD’s centre of gravity is small, comprising only ID (19 %) and ID(slot) (41.7 %), but ELAB(slot) is also well represented, at 12.9 %. GENUNMOD’s centre of gravity, interestingly, excludes ID (10.4 % of the occurrences), and comprises ID(slot) (18.8 %), ELAB(slot) (19.4 %), ELAB (28.5 %) and ELAB(part) (19.4 %). That is, whereas the GENMOD occurrences most frequently encode reference point (ID) situations, the GENUNMOD occurrences most frequently encode intrinsic relationships (ELAB). In the following, the GENMOD and the GENUNMOD groups will therefore be treated separately.

3. The dative Adnominal datives are also found in the Old Russian texts, covering some of the nodes on our map of the possessive semantic space. Unlike what we found with the genitive (and the denominal adjectives, for that matter), there are no formal restrictions on the use of the adnominal dative; it can have complex and unmodified dative possessors alike. (29)

bjaxu bo serdca ix, aky serdca lvomż were PTC hearts their like hearts lions-DAT ˇ ‘for their hearts were like the hearts of lions’ ZAN 170/11

(30)

pridoˇsa poslanii ot Stopolka na pogublenže Glˇebu came sent by Svjatopolk for destruction Gleb-DAT ‘those who were sent by Svjatopolk in order to kill Gleb [for the destruction of Gleb] came’ PVrL 136/16–17

In Old Russian, the construction is typically [NP-NOM, V, NP-DAT, NPACC],41 where the accusative-marked NP is the object, and the dative-marked object is more like an affected dative object of the verb than a dative possessor. It is often difficult to decide whether we are dealing with a [NP, NPDAT] construction in its own right, or just a part of an external possession construction:42 (31)

pritrepa slavu dˇedu svoemu Vseslavu ruined glory grandfather-DAT his-DAT Vseslav-DAT ‘he ruined the glory of/for his grandfather Vseslav’ SPI 53/11–12

47

0

10

20

30

40

50

60

The dative

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 14. 11th–14th century Old Russian adnominal dative distributed on functions, n = 181

In such cases of doubt, the function of the dative occurrence was classified as “uncertain”, something which makes the number of dative constructions with uncertain semantics quite high (see figure 14). There are highly varying opinions on the status of the “possessive” dative in the Old Russian system. There is no doubt that it is a Slavic phenomenon; it was frequent in OCS, and was often used to translate Greek genitives – Greek did not have possessive datives (Mrazek 1963:247). However, it is often considered a Balkanicism, e.g. by Richards (1976:261–262), as it is particularly frequent in East Bulgarian texts (Veˇcerka (1963:222), Veˇcerka (1993:198), Xodova (1963:134)). The frequency and distribution of the construction is assessed very differently in works on Old Russian syntax: Borkovskij (1968:197–198) claims that it is fairly frequent, whereas Sprinˇcak (1960:120,140) and Stecenko (1977:54, 101) both claim that it is rare, and Ivanov (1990:380–81) does not even include it. The most balanced view is found in Pravdin’s influential article (1957); he concedes that the adnomi-

48 The constructions in Old Russian

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 15. Old Russian adnominal dative on the map of the possessive semantic space

nal dative was probably a part of the Old Russian system, but that it was never entirely synonymous with the adnominal genitive (similarly in Marojevi´c 1983b:57). When we look at the distribution of the adnominal dative construction in 11th–14th century Old Russian (figure 14 and 15), we find that it is rather similar to that of the GENUNMOD occurrences, in that both constructions clearly centre on intrinsic relationships, with a centre of gravity comprising ELAB, ELAB(slot) and ID(slot), but not ID. The central difference between the two constructions is that ELAB(part) is contained in the centre of gravity of the GENUNMOD occurrences, but only marginally occurs with the dative construction.43 The dative construction, then, has a distribution that avoids the extremes on the map; neither TYPE, LABEL nor ELAB(part) are included. The construction’s centre of gravity excludes ID (6.1 %), but includes ID(slot) (21 %), ELAB(slot) (38.1 %) and ELAB (22.7 %).44 In the great majority of cases, we find that the dative-marked NP fills the argument slot of a relational noun, typically the object slot of a deverbal noun, as in example (30), much more frequently than is the case in the GENUNMOD group.

OCS and Old Russian

49

4. Mixed constructions As mentioned in section 2, the claim that adnominal genitives and denominal adjectives are in complementary distribution is an oversimplification. Bratishenko (1998) points out that a considerable share of the violations of this “rule” are cases where denominal adjectives are used even though the possessor nominal is modified in some way. We find examples where the denominal adjective is combined with a genitive, as in example (32), or even a dative modifier, as in example (33). Furthermore, there are double adjective constructions, as seen in example (34). The semantic centre of gravity of these constructions appears be the two ID nodes (8 out of 14 adjective + adjective constructions and all 15 adjective + genitive constructions in my 11th–14th century Old Russian text sample). (32)

zoloto kne˛ gini mojeje olenino gold princess-GEN my-GEN Olena-IN ‘my princess Olena’s gold’ DIK 91/24

(33)

v otni emu manastyri in father-N’ he-DAT monastery ‘in his father’s monastery’ (Suzdal’skaja letopis’ 1154, l 114, quoted from Bratishenko 1998:95)

(34)

vode˛ s soboju Volodimera. Jurževiˇca. brata Vsevoloˇza. bringing with self Volodimer Jur’eviˇc brother Vsevolod-J i Mstislavlja and Mstislav-J ‘bringing with them Volodimer Jur’eviˇc, the brother of Vsevolod and Mstislav’ SL 461/23–24

5. OCS and Old Russian Although this book primarily deals with possessive constructions in the history of Russian, the investigations of each node in the following three chapters will all contain a survey of the situation in OCS and a contrastive investigation of the situation in 11th–14th century Old Russian. The emphasis will be on the points where the two languages differ. There are a number of reasons for comparing OCS and Old Russian possessive constructions, perhaps particularly at the ID nodes. Firstly, OCS is

50 The constructions in Old Russian the earliest attested Slavic language. Therefore, the hypothesis that the OCS possessive constructions may reveal an earlier stage of the typologically and genetically quite exotic system in Old Russian seems tempting. In reality, however, scholars disagree on this issue. Richards (1976) assumes OCS possessive constructions to have been closer to an alleged Common Slavic state of complementary distribution. Bratishenko (1998:91), on the other hand, comes to the opposite conclusion, stating that “exceptions” from the complementary distribution rule are less frequent in Old Russian than in OCS. Neither Richards nor Bratishenko, however, base these claims on independent research on an OCS corpus. This is quite symptomatic of the comparisons of OCS and Old Russian possessive constructions in the literature: The comparisons are mostly quite impressionistic, and most scholars actually just assume that the situation is more or less identical in the two languages. In order to be able to assess the similarities and differences between the two, and, if possible, come to any conclusions about archaisms and innovations in the possessive constructions, a quantitative approach to both languages is certainly needed. Secondly, there is no doubt that the language of the OCS texts and of the subsequent non-canonical Church Slavic religious texts has had a vast influence, particularly on the earliest, but also on the somewhat later Old Russian. This also pertains directly to the issues raised in this book, as several scholars (e.g. Borkovskij 1968:197–198) have cited (Old) Church Slavic influence as a factor influencing the possessive semantic space. The most frequent hypothesis is that the possessive dative construction could be a Balkan innovation, and that it might be a pure syntactic loan from (O)CS. To evaluate to what extent this is true, and whether other phenomena related to possessive constructions may be due to OCS influence, an OCS corpus study is necessary. Finally, there is the fact that almost the entire OCS canon consists of translations from Greek. Greek has a typical Indo-European situation with a genitive construction virtually alone in the possessive semantic space, as seen in table 2. The OCS situation cannot be assessed properly with regard to its proximity to Common Slavic and its influence on Old Russian without taking into account the possible effects of translation on the language. The claims of various authors that the expansion of the genitive in Slavic was due to Greek influence (Uryson 1980, who claims that the genitive was not used possessively in Slavic before the Greek influence; Uspenskij 1987:301–306) can only be evaluated by looking at the actual translations of Greek genitives. The Greek original text is also a valuable clue to the precise meanings of

Summary

51

the OCS constructions. This considerably reduces the number of constructions with uncertain semantics, compared to the figures from the Old Russian corpus studies. The literature leads us to expect that the differences between possessive constructions in 11th–14th century Old Russian and OCS will be subtle, and this is indeed what we find.

6. Summary In this chapter I have argued that there are five main construction types that must be taken into consideration in a study of adnominal possession in Old Russian (and in early Slavic in general). There are two main groups of adjective constructions, where ADJ1 is more strictly possessive than ADJ2. Arguably, there are also two genitive constructions, one that is formally restricted to be in complementary distribution with denominal adjectives, and one that is freely used, but has functions that are less typically possessive. We have only indirect evidence for this bipartition. The distributions of these five main constructions, as shown in graphs and visualised in semantic maps, overlap considerably. However, none of them are wholly synonymous. Even the ADJ1 construction and the GENMOD group are not a perfect match, since the GENMOD group comprises more nodes on the map, and also interacts with the ADJ2 construction. Furthermore, we find that it is easy to pick out a semantic centre of gravity for each construction, in this analysis determined to include nodes accounting for more than 15 % of the occurrences of the construction in question. These centres of gravity are visualised as smaller areas with fully drawn boundaries in the maps. For the ADJ1 construction, the two ID nodes form the centre of gravity. In the ADJ2 construction, even though it covers every node on the map, the centre of gravity clearly is TYPE and LABEL. The two genitive groups also differ clearly from each other, GENUNMOD centering on ID(slot) and the three ELAB nodes, and GENMOD centering on ID and ID(slot), just like the ADJ1 construction. The dative construction has ID(slot), ELAB(slot) and ELAB as its centre of gravity, and appears to have a strong preference for relational head nouns. The GENMOD occurrences and the dative construction have clear similarities, but they differ in that ELAB(part) is marginal to the dative construction. Instead of a complementary distribution account, then, the possessive con-

52 The constructions in Old Russian structions in 11th–14th century Old Russian are better viewed as five separate polysemous, but partly overlapping constructions, each with a clear semantic centre of gravity, competing and interacting in the possessive semantic space. Due to the diachronic relationship between OCS and Old Russian, the unclear statements of the relationship between the two systems and the undoubted influence of OCS on Old Russian, the following three chapters will examine the nodes one by one, compare the OCS and Old Russian distributions at each, and give a survey of the further development in the history of Russian for every node.

Chapter 4 ID: Reference points on instance level

The occurrences that are classified as ID or ID(slot) all have a possessor that serves as a reference point on instance level to identify the head noun. The criteria used to identify the ID constructions, listed on p. 21, are repeated here: 1. The possessor serves to single out a referent or group of referents; the reference of the entire construction is not to a kind or type 2. The possessor is normally accessible from the context or from world knowledge, although marginally the possessor can be non-specific or kindreferring 3. The entire construction does not serve to conventionally name or label a particular referent Thus, the two ID nodes on our map are the ones covering what one might call the core possessive meaning, that is, meanings that e.g. include many of the properties in the possession gestalt defined by Taylor (1996:340). As we saw in chapter 3, all of our five main constructions could have the ID and ID(slot) meanings, although some of them only marginally. We also found that it may make a difference whether the possessor fills an argument slot, and such cases were consequently labelled ID(slot), as opposed to plain ID. As we shall see, there are indications to suggest that unmodified genitives and datives are both more acceptable at the ID(slot) node than at the ID node. It therefore makes sense to keep this separation throughout the presentation in this chapter. Since the two functions are so closely related, I will nevertheless present them together, stage by stage.

1. OCS 1.1. ID In the OCS sample, we find that the two predominant ways of expressing ID semantics – reference point situations where the possessor does not fill an argument slot of the head noun’s – are ADJ1 and GENMOD (fig. 16). The ID

0

10

20

30

40

50

54 ID: Reference points on instance level

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 16. OCS constructions at the ID node, per cent, n = 289

node is the node where we find the closest thing to a proper complementary distribution between genitives and adjectives. 59.2 % (171 out of 289 occurrences) are ADJ1 constructions. The adjectives are in all but a very few cases (such as example (37)) formed from nouns denoting specific persons,45 often proper nouns. (35)

vż domu iˇekovli in house Jacob-J epi ton oikon Iakˆob ‘in the house of Jacob’ Mar. Luke 1:33

(36)

vż dvorż arxiereovż in courtyard high-priest-OV eis tˆen aulˆen tou arkhiereˆos ‘into the court of the high priest’ Zogr. John 18:15

OCS

(37)

55

blgnż gż bż izlvż blessed Lord God Israel-OV Eulogˆetos kurios ho theos tou Israˆel ‘Blessed be the Lord, Israel’s God’ Mar. Luke 1:68

The second group of denominal adjective constructions, ADJ2, is much less frequent at this node of the map, constituting 9.3 % (27 out of 289) of the occurrences. When we look at the nouns from which the adjectives are formed, we find that they pattern very differently from those found in the ADJ1 constructions. They may denote persons, but in those cases they have plural reference, as in (38), and in 14 out of 16 cases they are kind-referring; in (39) the reference is to the entire Jewdom. Apart from these examples, the possessors are either inanimate and concrete as in (41) or abstract as in (40), and in most cases they have kind reference (40) or specific reference accessible from world knowledge. Example (41) is the only one that can reasonably be said to have a specific possessor accessible from context. (38)

kżnigy proskye˛ books prophet-SK hai graphai tˆon prophˆetˆon ‘the prophets’ scriptures’ Savv. Matthew 26:56

(39)

bˇe zˇe blizż pasxa ijudeiska was PTC near Passover Jew-SK eˆ n de eggus to paskha tˆon Ioudaiˆon ‘Now the Passover of the Jews was near’ Zogr. John 11:55

(40)

imatż svˇeta zˇivotżnago he-will-have light life-N hexei to phˆos tˆes zˆoeˆ s ‘he will have the light of life’ Zogr. John 8:12

(41)

i se opona crkvżnaja razdra se˛ and behold curtain church-N tore REFL kai idou to katapetasma tou naou eskhisthˆe ‘and behold, the curtain of the temple was torn’ Savv. Matthew 27:51

These two adjective constructions are complemented by constructions from the GENMOD group, which constitutes 20.4 % (59 out of 289) of the ID occurrences. The possessor nouns in the GENMOD group match both the ones

56 ID: Reference points on instance level in the ADJ1 occurrences, such as in (42), and the ones in the ADJ2 occurrences, such as in (43). (42)

otroka svoego vż domu davda in house David-GEN servant-GEN his-GEN en oikˆoi Dauid paidos autou ‘in the house of his servant David’ Mar. Luke 1:69

(43)

svoego otż blgago sżkroviˇsta srdca from good treasure heart-GEN his-GEN ek tou agathou thˆesaurou tˆes kardias ‘out of the good treasure of his heart’ Mar. Luke 6:45

It is therefore reasonable to claim that if the restricted genitive construction is in complementary distribution at the ID node, it has to be with both the ADJ1 and the ADJ2 construction. Constructions with bare genitive nouns (GENUNMOD) are very rare at the ID node, there are only ten such occurrences in the OCS sample. Two of these cases are occurrences of the partially lexically specific construction [vż NP-GEN mˇesto] ‘in the place of / instead of ’ (44). As we shall see in the Old Russian data, this construction consistently occurs with the genitive, whether the genitive-marked NP is complex or not. (44)

vż ryby mˇesto zmijo˛ podastż emu in fish-GEN place serpent will-give him anti ikhthuos ophin autˆoi epidˆosei ‘he will give him a serpent instead of fish’ Mar. Luke 11:11

Four of the remaining cases have possessor nouns denoting persons (45) or a group of persons (46), all in the plural. Three occurrences have an abstract possessor noun (47), and the final occurrence has a concrete inanimate noun. None of the possessor nouns were accessible from discourse, as a typical reference point would be; they were either uniques which were accessible from world knowledge, non-specific or had kind reference. (45)

iˇze sżnˇedajo˛ tż domy vżdovicž who devour houses widows-GEN hoi katesthiousin tas oikias tˆon khˆerˆon ‘who devour widows’ houses’ Savv. Luke 20:47

OCS

(46)

ne idˇete na po˛ tž je˛ zkż on road Gentiles-GEN not walk eis hodon ethnˆon mˆe apelthˆete ‘do not walk onto the road of the Gentiles’ Ass. Matthew 10:5

(47)

po obyˇcaju prazdžnika by custom feast-GEN kata to ethos tˆes heortˆes ‘according to the custom of the feast’ Mar. Luke 2:42

57

The types of possessor nouns found in the GENUNMOD group at the ID node thus appear to be rather similar to the types we found with the ADJ2 construction: Again, human possessors are only found in the plural; the rest of the possessor nouns are inanimate and often abstract, and none of the possessors were found to be both specific and accessible in discourse. A small group of adnominal dative constructions (only 10 out of the 289 ID occurrences) can arguably be classified as belonging to the ID node. All of these correspond to adnominal genitive constructions in the Greek. It is not entirely clear that all of these really belong at the ID node, since it is easy to see a benefactive shade of meaning with many of them, even though that shade does not seem to be present in the Greek original. This holds e.g. for the six cases where the head noun of the construction is svˇetż or svˇetilžnikż (48) (48)

svˇetilžnikż tˇelu estż oko lamp body-DAT is eye ho lukhnos tou sˆomatos estin ho ophthalmos ‘the lamp of the body is the eye’ Ass. Matthew 6:22

In example (49) we have a fairly clear example of an ID relationship without any benefactive shade. (49)

vˇecˇ nˇei zˇizni vˇenžcž eternal-DAT life-DAT crown ton stephanon tˇes aiˆoniou zˆoeˆ s ‘the crown of eternal life’ Supr. 11:154/14–15

The possessor nouns include both animates and inanimates, concrete and abstract nouns. There are few examples, but again we notice a preponderance of marginal and untypical possessors: Only two are both specific and acces-

58 ID: Reference points on instance level sible in discourse (50), whereas the rest have either kind (48), non-specific or unique/encyclopaedically known reference (49). (50)

cˇesarž [. . . ] svoi vˇerˇe crżkżvż prositż emperor his-DAT faith-DAT church begs ho basileus [. . . ] tas tˆes oikeias pisteˆos [. . . ] ekklˆesias epizˆetei ‘The Emperor [. . . ] asks for the churches of his own faith’ Supr. 16:200/16

At the ID node, then, the dative construction appears to be quite marginal in OCS. Finally, there are 11 cases of mixed constructions, equally split between adjective + adjective (51) and adjective + genitive constructions (52). Both types could be either coordinating, as in (51), or subordinating, as in (52) and (53). (51)

otż arievy zˇe i makedonˇe strany from Arios-OV PTC and Makedonios-OV countries tou de Areiou merous kai Makedoniou ‘from Arios’ and Makedonios’ countries’ Supr. 16:201/15–16

(52)

na dvor arxiereovż. naricaemaago kaiafa on courtyard arch-priest-OV so-called-GEN Caiaphas-GEN eis tˆen aulˆen tou arkhiereˆos tou legomenou Kaiapha ‘in the courtyard of the high priest called Caiaphas’ Ass. Matthew 26:3

Ten of these occurrences have specific human possessors and involve at least one type 1 adjective. The single exception has an animal possessor and a type 2 adjective (53). (53)

otż uzdy konžnye˛ cˇesare˛ from bridle horse-N king-J ek tou khalinou tou hippou tou basileˆos ‘from the bridle of the king’s horse’ Supr. 16:193/9–10 46

To summarise, at the ID node in OCS we see a rather clean division of labour between adjective constructions and a restricted genitive construction. Given the scarcity of bare genitives, we must expect most of the occurrences in the GENMOD group to belong to this restricted genitive construction. However, this is not a situation of perfect complementary distribution, as we find exam-

OCS

59

ples both of complex adjective constructions and bare genitives. Strikingly, the complex adjective constructions are found in situations with prototypical possessors, i.e. specific human beings. The bare genitive possessors, on the other hand, are all less typical possessors: inanimates, often abstracts, and humans only in the plural. Furthermore, none of the examples have specific possessors – in fact, they pattern with the ADJ2 constructions at the ID node in that respect, see e.g. example (38) on page 55. The dative constructions with ID semantics are very few and to some extent dubious, but the few examples found in the sample show a similar tendency as the bare genitive constructions: they do not have prototypical possessors, but rather tend to have non-specific and kind-referring ones. These findings support and expand on Bratishenko’s general observations in Old Russian (1998), in particular concerning the use of the agent/possessor hierarchy: Possessors both high and low on the hierarchy may break the tendency towards complementary distribution. ADJ2, GENUNMOD and the dative construction, rare as they are, appear to belong to the lower end of the hierarchy.

1.2. ID(slot) Looking at the distribution of OCS constructions at the ID(slot) node in figure 17, we immediately observe that the distribution is different from the one at the ID node. As table 3 confirms, the difference between the two distributions is significant (p-value = 0.0007075). If we check the constructions against the totals one by one, however, we see that it is only the GENMOD group that is significantly different at the two nodes (p-value = 0.01363).47 ID ID(slot)

ADJ1 171 183

ADJ2 27 26

GENMOD 59 126

GENUNMOD 10 21

DATIVE 10 25

sum 277 381

Table 3. Distribution of OCS constructions, ID and ID(slot), mixed, constructions excluded; p-value 0.0007075 (Fisher’s Exact Test, two-sided)

The difference between the ID constructions and the ID(slot) constructions is that in the ID(slot) constructions, the head noun is relational, and the possessor nominal fills (one of) its argument slot(s). A number of different relational noun types are attested as head nouns, but the most frequent ones are dever-

0

10

20

30

40

60 ID: Reference points on instance level

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 17. OCS constructions at the ID(slot) node, per cent, n = 390

bal (103), kinship terms (89), body parts (69), realm nouns (30), deadjectival nouns (29) and nouns denoting other human relationships than kinship (16). As we can see in table 4, the noun types are not evenly distributed across the construction types. Most strikingly, the dative constructions have a strong preponderance of body part heads, whereas these are much rarer with the ADJ1 construction, which has a large proportion of kinship term heads instead. The GENUNMOD group is the only one which appears to have an affinity for head nouns denoting inherent parts of wholes – not surprising, given the agreement in the literature that partitive shades of meaning makes bare genitives acceptable. 46.9 % (183) of the constructions at the ID(slot) node are ADJ1 constructions. Again, the possessors are consistently human, singular and specific referents.

OCS

ADJ1 ADJ2 DATIVE GENMOD GENUNMOD

body part 9.8 23.1 44.0 25.4 4.8

deadj. 2.7 26.9 4.0 9.5 19.0

deverbal 23.5 30.8 24.0 32.5 4.8

kin/hum.rel. 41.5 7.7 20.0 11.9 23.8

part 0 3.8 4.0 1.6 23.8

realm 14.8 0 0 1.6 4.8

61

other 7.7 7.7 4.0 17.5 19.0

Table 4. Types of relational nouns in OCS ID(slot) constructions, per cent by construction

(54)

cˇ e˛ da avraamle˛ children Abraham-J tekna tou Abraam ‘Abraham’s children’ Zogr. John 8:39”

(55)

eˇ ko uslyˇsa elisavetż cˇelovanie mariino when heard Elisabeth greeting Mary-IN hˆos eˆ kousen ton aspasmon tˆes Marias ‘when Elisabeth heard Mary’s greeting’ Mar. Luke 1:41

Some of the ADJ1 occurrences classified as ID(slot) are highly conventionalised, and might have been classified as LABEL occurrences instead. This is for instance the case with all the 27 occurrences headed by a noun denoting a realm, which are all occurrences of the collocation “the kingdom of God” (56). These occurrences were nonetheless retained at the ID(slot) node. (56)

otż toli csrstvie bˇzie blagovˇestvuutż se˛ REFL from then kingdom God-IJ preaches apo tote hˆe basileia tou theou euaggelizetai ‘since then, the good news of the kingdom of God is being preached’ Mar. Luke 16:16

Likewise, a number of the ADJ1 occurrences headed by kinship terms appear to be conventionalised patronymics, on the form [name-ADJ1, synż]. There are five cases that appear to be rather clearly conventionalised (57) and six less certain cases. Because the share of uncertain cases I have kept them at the ID(slot) node, but classifying them at the LABEL node would also have been an option, see chapter 6. (57)

uzrˇe ina dżva bratra. iakova sna zebedeova i ioana saw other two brothers James son Zebedee-OV and John

62 ID: Reference points on instance level eiden allous duo adelphous Iakˆobon ton tou Zebedaiou kai Iˆoannˆen ‘he saw two other brothers, James the son of Zebedee and John’ Ass. Matt. 4:21 6.7 % (26) of the constructions at the ID(slot) node are instances of ADJ2. The head nouns here are primarily deverbal, deadjectival or denote body parts, and the possessor nominals follow the same pattern as the ADJ2 constructions at the ID node: The adjectives are either formed from noun stems denoting persons, but with plural reference (58), or from inanimate noun stems, either concrete (59) or abstract (60). (58)

vż ro˛ cˇe cˇ lscˇe into hands man-SK eis cheiras anthrˆopˆon ‘into the hands of men’ Savv. Mark 9:31

(59)

onż zˇe vżstavż zaprˇeti vˇetru. i vlżne(n)iju vodżnumu he PTC having-risen rebuked wind and raging water-N ho de diegertheis epetimˆesen tˆoi anemˆoi kai tˆoi kludˆoni tou hudatos ‘but having risen he rebuked the wind and the water’s raging’ Mar. Luke 8:24

(60)

otż b(ga)atžstviˇe i slastžmi zˇiteiskyimi from riches and sweetness life-SK hupo merimnˆon kai ploutou kai hˆedonˆon tou biou ‘by the (cares and) riches and pleasures of life’ Mar. Luke 8:14

Again we find that the possessor frequently has non-specific or kind reference; this is the case in 16 of the 26 occurrences. We see this in examples (58) and (60). 10 occurrences have specific possessors, but only four of these are old or inferrable in the discourse, as we see in (59), while the remaining six had unique possessors, accessible from world knowledge. In (61), for instance, the reference is not to prophetkind, but to a restricted, generally known group of prophets. (61)

da sżbo˛ do˛ tż se˛ kžnigy prrˇcžskyje˛ that will-fulfill REFL books prophet-SK hina plˆerˆothˆosin hai graphai tˆon prophˆetˆon ‘so the scriptures of the prophets might be fulfilled’ Ass. Matt. 26:56

OCS

63

The GENMOD group is the second largest at the ID(slot) node, constituting 32.3 % (126) of the occurrences. The head nouns in this group are primarily deverbal (62), body part nouns (63) or kinship terms. Looking at the possessor nouns, we again find that they match the entire range of possessors possible with both ADJ1 and ADJ2 constructions, from human, singular, specific possessors in (62) and (63) to inanimate possessors in (64) and (65). (62)

cˇ lˇca priˇsestvie sna coming son-GEN man-SK.GEN hˆe parousia tou huiou tou anthrˆopou ‘the coming of the son of man’ Ass. Matthew 24:27

(63)

otż krżve avelˇe pravedżnaago from blood Abel-GEN righteous-GEN apo tou haimatos Habel tou dikaiou ‘from the blood of Abel the righteous’ Ass. Matthew 23:35

(64)

za zˇivotż vžsego mira for life all-GEN world-GEN huper tˆes tou kosmou zˆoeˆ s ‘for the life of all the world’ Zogr. John 6:51

(65)

obrˇete bogoborżnyˇe eresi potopż duˇsžnyi found god-defeating-GEN heresy-GEN inundation soul-N heuren tˆes theomakhou haireseˆos ton kataklusmon tˆon psukhˆon ‘he found the God-defeating heresy’s inundation of the souls’ Supr 16:191/10–11

As was the case at the ID node, the GENUNMOD group is small, comprising 5.4 % (21) of the total occurrences, which is not significantly higher than the GENUNMOD share at the ID node (p-value = 0.3521, Fisher’s exact test, two-sided). Apart from the fact that these constructions are the only ones that appear to have an affinity with head nouns denoting inherent parts of wholes (66), the distribution of head noun types is unremarkable. (66)

pri koreni drˇeva by root tree-GEN pros tˆen rhizan tˆon dendrˆon ‘by the root of the tree’ Ass. Matthew 3:10

Although the bare genitive occurrences are few, there appears to be a dis-

64 ID: Reference points on instance level cernible difference in the distribution of possessor nouns between the ID and the ID(slot) occurrences. When we exclude three occurrences of lexically specific constructions requiring the genitive regardless of modification, as in (44) and (67), we see in table 5 that there are more human singulars, as in examples (68) and (69), and inanimate concretes (70) in the ID(slot) group. The difference between the two distributions is statistically significant (p-value = 0.02144, Fisher’s exact test, two-sided). All the occurrences are common nouns. human singular human plural animal concrete abstract

ID 0 4 0 1 3

IDslot 7 1 1 6 5

Table 5. Animacy of possessor nouns in GENUNMOD occurrences, ID vs. ID(slot) in OCS

(67)

vż slˇedż isa ido˛ narodi mnozi in track Jesus-GEN went crowds big eˆ kolouthˆesan autˆoi okhloi polloi ‘after Jesus went great crowds’ Ass. Matt. 4.25

(68)

abie vżzżpi ocž otroˇce˛ te immediately cried-out father child-GEN euthus kraxas ho patˆer tou paidiou ‘immediately the father of the child cried out’ Savv. Mark 9:24

(69)

na vžso˛ silo˛ vraga on all power enemy-GEN epi pasan tˆen dunamin tou ekhthrou ‘on all the power of the enemy’ Mar. Luke 10:19

(70)

eˇ ko zˇe pribliˇzi se˛ kż vratomż grada when PTC approached REFL to gates city-GEN hˆos de eˆ ggisen tˆei pulˆei tˆes poleˆos ‘as he drew near to the city’s gate’ Mar. Luke 7:12

Recall that among the GENUNMOD occurrences at the ID node there were none that had discourse-accessible possessor nouns with specific refer-

OCS

65

ence, as we would expect from an optimal reference point. At the ID(slot) node, however, we find that, excluding example (67), 9 out of 20 occurrences are both specific and accessible in discourse, as exemplified in (68) and (70). A further four are uniques and accessible through world knowledge (69), whereas the remaining seven occurrences are non-specific or kindreferring. We find that there is a statistically significant difference in frequency between the ID and ID(slot) nodes when we compare the shares of discourse-accessible possessor nouns and the remaining types (p-value = 0.0292, Fisher’s exact test, two-sided). A caveat is clearly in place here, since the number of examples is very low, and there may be arguments against collapsing the world-knowledgeaccessible uniques with the non-specific and kind-referring possessor nouns. Nevertheless, in conjunction with the data on possessor animacy in the two groups, this may serve to indicate that bare genitives are used more freely as reference points when they fill a slot in the head noun’s argument structure, than in ID constructions when they do not. The ID(slot) GENUNMOD occurrences are better reference points than the ID occurrences, since they are more likely to be human and less likely to be abstract, and since there is at least some evidence that they tend to be more available in discourse. As we recall, the share of GENMOD is significantly larger at ID(slot) than at ID, even though the share of GENUNMOD is not. In my opinion, this, combined with the data on the distribution of bare genitive possessor nouns at the ID and ID(slot) nodes, may be taken as a piece of evidence that the complementary distribution between adjective constructions and a restricted genitive construction is weaker at the ID(slot) node: Not only are the bare genitive examples better reference points at the ID(slot) node, but the overall share of genitives is considerably higher. The distribution of genitive constructions thus appears to justify the separation between the ID and the ID(slot) node. 6.4 % (25) of the occurrences at the ID(slot) node are dative constructions. Again, as with the GENUNMOD group, there is no statistically significant difference between the two ID nodes in the share of dative constructions (pvalue = 0.1580, Fisher’s exact test, two-sided), but again, I shall argue that the occurrences are qualitatively different at the ID(slot) node. The distribution of relational head nouns with the dative constructions stands out from those of the other constructions: Body part nouns are by far the largest group here, followed by deverbal nouns, kinship terms and nouns denoting other human relationships. It is easy to connect these noun types

66 ID: Reference points on instance level with observations in the literature on the semantics of the dative case. If the dative-marked noun prototypically denotes a person surrounded by a personal sphere and affected by something in that sphere, as claimed by Da˛ browska (1997),48 body parts, relatives and other significant humans are certainly parts of that personal sphere and affect the dative-marked noun (71). (71)

obratiti srżdżca otžcemż na cˇ e˛ da turn hearts fathers-DAT on children epistrepsai kardias paterˆon epi tekna ‘to turn the fathers’ hearts to the children’ Mar. Luke 1:17

It is interesting to note that neither of the four occurrences where the dativemarked NP fills a subject slot of a deverbal noun involves an agent role. The dative-marked NPs are either themes (72) or experiencers, and thus at least in the latter case conform to typical dative semantics. (72)

leˇzaaˇse mżnoˇzžstvo [. . . ] cˇ ajo˛ sˇteixż dviˇzenžˇe were-lying multitude hoping-GEN moving-GEN vodˇe water-DAT katekeito plˆethos ekdekhomenˆon tˆen tou hudatos kinˆesin ‘there lay a multitude waiting for the moving of the water’ Zogr. John 5:3

As we saw with the GENUNMOD group, the dative-marked nouns at the ID(slot) node differ in animacy and specificity from those at the ID node. Whereas the 10 ID occurrences were evenly distributed in animacy, 17 of the 25 occurrences at the ID(slot) node are human (and a further two refer to human collectives), which is not surprising given the predominance of body part and human-relation nouns. Also, whereas the possessor nouns at the ID node are mostly kind-referring or non-specific, 12 of the occurrences at the ID(slot) node are specific and accessible from context, and two are specific and accessible from world knowledge. The remaining occurrences were kindreferring or non-specific. As in the GENUNMOD group, then, it seems that the dative possessor nouns at the ID(slot) node are better reference points than those at the ID node, such as the human, contextually accessible and specific dative possessor in (73).

OCS

(73)

67

i naˇcatż umyvati nozˇe uˇcenikomż and began wash feet disciples-DAT kai eˆ rxato niptein tous podas tˆon mathˆetˆon ‘and he began to wash the disciples’ feet’ Zogr. John 13:5

Finally, it is worth noting that 17 of the occurrences are bare nouns, e.g. the alternative would have been an adjective construction. Seven of these are plural nouns denoting humans (73), reminiscent of what we found with the ADJ2 constructions and in the GENUNMOD group. However, there are also singular nouns denoting humans, even proper nouns (74). (74)

bˇe bo tžstž Kaiafˇe was PTC father-in-law Caiaphas-DAT eˆ n gar pentheros tou Kaiapha ‘for he was Caiaphas’s father-in-law’ Zogr. John 18:13

From this, we conclude that the dative construction is rare, but that it is used more freely with certain relational nouns than it is at the ID node. As with the ADJ2 constructions and the GENUNMOD group, there are signs that this was a construction that one could resort to with possessors that were non-standard in some way, e.g. human plural possessors. Finally, there are nine examples of mixed constructions, distributed quite similarly to the examples at the ID node. They are headed by kinship terms, deverbal nouns and body part nouns, and include one example of a compound adjective (75). (75)

zˇe rodstvo sice bys isxvo Jesus-Christ-OV PTC birth thus was Tou de Iˆesou Khristou hˆe genesis houtˆos eˆ n ‘Now the birth of Jesus Christ was like this’ Ass. Matt. 1:18

In conclusion, then, it seems reasonable to say that the distribution at the ID(slot) node is largely similar to that at the ID node: We find a fairly clear division of labour between adjective constructions and a restricted genitive construction. There are several indications that bare genitives can be used more freely, in the sense that they are “better”, more animate and more contextually accessible at the ID(slot) node, even though we cannot say that there is a larger share of them. There is, however, a larger share of genitives overall. We see a similar tendency for the dative construction; its share is not signi-

68 ID: Reference points on instance level ficantly larger, but it is used more freely and with “better” possessors than is the case at the ID node. This is also what our map configuration suggests: The ID node and the ID(slot) node are very close, but they are not encoded in precisely the same way.

2. 11th–14th century Old Russian 2.1. ID When we compare the distributions of constructions at the ID node in OCS and 11th–14th century Old Russian, we find clear differences. However, not unlike what we saw when we compared the ID and ID(slot) nodes in the OCS sample, it is only the proportions of ADJ1 and constructions from the GENMOD group that are significantly different in the two language samples: The share of ADJ1 constructions is smaller in the Old Russian sample (p-value = 0.005868, Fisher’s exact test, two-sided), whereas the share of GENMOD occurrences is considerably larger (p-value = 0.006202, Fisher’s exact test, two-sided). The shares of dative constructions and GENUNMOD occurrences are virtually identical. There is also a statistically significant difference in the shares of ADJ2 constructions (p-value = 0.0456, Fisher’s exact test, two-sided).

OCS OR

ADJ1 171 148

ADJ2 27 57

GENMOD 59 122

GENUNMOD 10 15

DATIVE 10 11

sum 277 353

Table 6. Constructions at the ID node, OCS and 11th–14th century Old Russian

At first glance there is nothing to indicate that our Old Russian sample should be closer to or farther from complementary distribution than our OCS one, unless there should be clear qualitative differences within the construction types. For the 148 ADJ1 constructions in the Old Russian sample, we find no such clear difference. Again, this is a group of constructions fulfilling most of the properties in Taylor’s possession gestalt (1996:340): The possessor is human, specific and mostly discourse-accessible, and the possessee is often a concrete object and does not have an argument structure of its own.

11th–14th century Old Russian

69

0

10

20

30

40

50

OCS OR

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 18. Constructions at the ID node, OCS and 11th–14th century Old Russian compared, per cent

(76)

poslanii jaˇsa korablž Glˇebovż envoys took boat Gleb-OV ‘the envoys took Gleb’s boat’ PVrL 136/17–18

(77)

cˇ stnyja gvozdi xsvi elena [. . . ] vloˇzila von Elena put-in in-there holy nails Christ-OV ‘Elena put the holy nails of Christ into it’ XID 15v/16–17

Among the ADJ2 constructions we find a fairly even distribution of human, concrete and abstract possessor referents. There are six singular and 16 plural human referents among them. As we recall, in the OCS material only plural human possessors were found in ADJ2 constructions. However, out of the 57 Old Russian occurrences, only 7 can reasonably be seen as specific and either accessible from the discourse or discourse-new (78).

70 ID: Reference points on instance level (78)

onż zˇe po maternyx dnex naˇca dˇeti prodavati he PTC after mother-N days began children sell ‘but after the mother’s days [e.g. after her death] he started selling the children’ SDZ 32/7–33/1

A further 19 cases have specific possessors, but all of them are unique and only accessible from world knowledge (79), typically proper nouns denoting persons or cities. (79)

na agarjane i vnuci izmaitelžska roda on Hagarians and grandsons Ishmael-SK lineage ‘against the Hagarians and the grandchildren of Ishmael’s breed’ PBR 356/39

The remaining 31 examples are all either kind-referring or non-specific, many of them referring to nationality groups. A kind-referring possessor is seen in (80), a nationality group in (81). (80)

nikakˇze razumż cˇ lvˇcsk ne moˇzet ispovˇedati. cˇ judes in-no-way understanding man-SK not can understand miracles tvoixż your ‘in no way can man’s understanding grasp your miracles’ PVM 244/10

(81)

zvonja Ruskymż zlatom jingling Rus-SK gold ’jingling with the Rusians’ gold’ SPI 51/7–8

32.1 % (122) of the occurrences at the ID node belong to the GENMOD group. As we saw in the OCS sample, the possessor NPs match both the types found in the ADJ1 constructions (82)and the types found in the ADJ2 construction (83), and clearly have a division of labour with both. (82)

aˇste poidu vż domż otca svoego if I-go in house father-GEN self’s-GEN ‘If I go to my father’s house’ SBG 30/10

(83)

svˇetż razumnyi krasnyja radosti light sensible beautiful-GEN joy-GEN the sensible light of the beautiful joy’ PVrL 138/22–23

11th–14th century Old Russian

71

The large majority of occurrences in the GENMOD group, however, have specific and human possessors. There are 15 occurrences of unmodified genitives. However, five of them occur in the partially lexically specific [vż NP-GEN mˇesto] construction, which may be seen as a complex PP construction, and where the genitive is clearly required, regardless of the complexity of the “possessor” NP (84). (84)

vż psa mˇesto in dog-GEN place instead of a dog’ RP 37/370

The remaining 10 occurrences all have abstract possessor nouns. Only a single occurrence is specific and accessible from discourse (85), three cases are possibly unique and accessible from world knowledge, as torˇzestva in (86), whereas the rest have kind or non-specific reference (87). (85)

zarja svˇetż zapala dawn-GEN light flared-up ‘the light of dawn flared up’ SPI 46/25

(86)

vż denž zˇe torˇzestva on day PTC triumph-GEN ˇ 363/12 ‘on the day of the triumph’ ZD

(87)

mleko bezżlobija milk innocence-GEN ‘the milk of innocence’ SBG 41/10

Thus, low as the numbers are, the GENUNMOD occurrences we find at the ID node in 11th–14th century Old Russian are even “worse” possessors than the ones we find in the OCS material, and in addition considerably less typical than the ones we find in the Old Russian ADJ2 constructions. The dative construction is also very rare at the ID node, with 11 occurrences. We find the same tendency as we found with the OCS dative constructions at the ID node, e.g. that many of them are ambiguous and appear to have benefactive shades of meaning. (88) and (89) have fairly unequivocal reference point readings, although they are far from involving prototypical possession.

72 ID: Reference points on instance level (88)

Nż uˇze, knjaˇze, Igorju utrżpˇe solncju svˇetż but already prince Igor-DAT vanished sun-DAT light ‘but already, o prince, the light of the sun had vanished for Igor’ SPI 52/26

(89)

preˇze pribliˇzenža braku cˇ istotu sżxranivˇsimż before closeness marriage-DAT purity having-preserved ˇ ‘having preserved his purity before the closeness of marriage’ ZD 364/21

(90) and (91) are rather more typical examples. They may have a reference point reading, but a benefactive reading, or at least a nuance of it, is also quite possible. 49 (90)

zercalo zˇitiju mirror life-DAT ˇ 355/12 ‘(he is) life’s mirror’ ZD

(91)

vˇenecž pobˇede crown victory-DAT ˇ 355/10 ‘(he is) victory’s crown’ ZD

These examples illustrate two other tendencies also found in OCS: Five of the examples are abstract nouns, and seven of them have kind reference. However, there are also possessor nouns denoting concretes and humans. Two of these examples have possessors with unique reference (88), and two have specific, discourse-accessible possessors (92).50 (92)

ocju i materi slezy father-DAT and mother-DAT tears ‘(my) father’s and mother’s tears’ PVM 253/25

Another interesting point is that ten out of the eleven examples contain a bare dative-marked noun. In fact, this is a general tendency for the dative constructions in 11th–14th century Old Russian, regardless of place in the possessive semantic space: About two thirds have bare dative-marked possessor nouns. This suggests that the dative construction may have been a bit of a last resort, for cases when the ADJ2 construction or a bare genitive for some reason was unavailable or undesirable. However, as we see, the choice of possessors in the dative construction is very similar to that in the ADJ2 construction and

11th–14th century Old Russian

73

the GENUNMOD group; again, low-animacy, kind-referring nouns predominate. The share of mixed constructions at the ID node is rather similar in 11th– 14th century Old Russian and OCS. There were 16 such constructions at the ID node in the Old Russian material; 5 adjective + adjective constructions and 11 adjective + genitive constructions. The adjective + adjective constructions are all paratactic (93), whereas the adjective + genitive constructions can be either paratactic, or hypotactic as in (94), always with a genitive apposition to the adjective. (93)

Bogż Avramovż i Isakovż god Abraham-OV and Isaac-OV ˇ 361/5–6 ‘Abraham’s and Isaac’s God’ ZD

(94)

predatele˛ xsva dvorż ijudin court Judas-IN betrayer-GEN Christ-OV.GEN ‘the court of Judas, the betrayer of Christ’ XID 25r/12–13

All save one example are specific, and either accessible in context or from encyclopaedic knowledge, and all save example (95) have possessors that are either human (always type 1 adjectives) or toponyms (always type 2 adjectives). (95)

na razumnyx skryˇzalˇex, srdˇcnyx, napisati heart-N write on mind-N tablets ˇ ‘to write on the mind’s, the heart’s tablets’ ZSP 1/17

We see again, then, that we are mostly dealing with deviations from the complementary distribution “rule” where the possessor is close to a paradigmatic possessor. To conclude, then, we see that, despite some frequency differences, the distributions of constructions at the ID node are strongly similar in the OCS and earliest Old Russian material. In both language samples we largely find a neat division of labour between adjective and genitive constructions, with two main deviations: When there is a high-ranking possessor (human, specific), a mixed adjective construction can be chosen even when the possessor is a complex NP; this is most typically seen with personal names with appositions. Another type of deviation concerns low-ranking possessors, which may marginally occur as bare genitives, maybe with slightly more leeway in

74 ID: Reference points on instance level

0

10

20

30

40

OCS OR

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 19. Constructions at the ID(slot) node, OCS and 11th–14th century Old Russian compared, per cent

the OCS sample than in the Old Russian one. These marginal possessors may, however, also turn up in the dative or ADJ2 construction, the latter of which is possibly slightly more flexible in the Old Russian sample than in the OCS. The differences between the two systems, then, appear to be subtle, and the similarities overwhelming.

2.2. ID(slot) Comparing the OCS and Old Russian distributions at the ID(slot) node in figure 19 and table 7, we find that the Old Russian sample has a significantly smaller share of ADJ1 constructions (p-value = 0.03375, Fisher’s exact test, two-sided) and a significantly larger share of ADJ2 constructions (p-value = 0.008443). When we compare the Old Russian distributions at the ID and

11th–14th century Old Russian

OCS OR

ADJ1 183 247

ADJ2 26 84

GENMOD 126 267

GENUNMOD 21 27

DATIVE 25 38

75

sum 381 663

Table 7. Constructions at the ID(slot) node, OCS and 11th–14th century Old Russian compared

the ID(slot) nodes in table 8, however, we find that there is no statistically significant difference between these. The difference in the shares of dative constructions is the one that comes closest to significance, with a p-value of 0.08964 (Fisher’s exact test, two-sided). ID ID(slot)

ADJ1 41.93 37.25

ADJ2 16.15 12.67

GENMOD 34.56 40.27

GENUNMOD 4.25 4.07

DATIVE 3.12 5.73

Table 8. Distribution of constructions at the ID and ID(slot) nodes in 11th–14th century Old Russian, per cent, uncertain and mixed constructions excluded

When we look at the distribution of head nouns in the ID(slot) constructions (table 9), we find that they are quite evenly distributed. We can see that GENUNMOD has an expected affinity for body part and other part nouns, but the same can be said of ADJ2. The GENUNMOD group apparently has a large share of deadjectival head nouns. With the dative constructions, we find a more even distribution of head nouns than we did in the OCS sample (cf. table 4 on page 61). ADJ1 ADJ2 DATIVE GENMOD GENUNMOD

body part 11.7 15.5 18.4 25.8 18.5

deadj. 10.1 14.3 18.4 11.6 37.0

deverbal 37.7 36.9 28.9 43.8 22.2

kin/hum.rel. 29.5 8.4 15.8 7.2 7.4

part 0.8 16.7 5.3 1.9 14.8

realm 0.4 0 0 1.1 0

other 9.7 8.3 13.2 7.5 0

Table 9. Distribution of relational head nouns by construction at the ID(slot) node, 11th–14th century Old Russian, per cent by construction

36.1 % (247) of the ID(slot) occurrences are ADJ1 constructions. Again, we find that the possessors are consistently human, singular and specific.

76 ID: Reference points on instance level (96)

ocž Mistiˇsinż Mistiˇsa’s father ‘Mistiˇsa’s father’ PVrL 55/12–13

(97)

tako volja estž knjaˇza thus will is prince-J ‘such is the prince’s will’ SBG 61/24

(98)

povˇedaja emu otˇcju sżmržtž telling him father-J death ‘telling him of his father’s death’ SBG 29/2

The most frequent types of head nouns are deverbal nouns (98) and kinship terms (96). As in OCS, we find some ADJ1 constructions that may possibly be conventionalised as patronymics (99), but most patronymics are of the -iˇc type, and the ADJ1 occurrences in question may also be read as mere appositional ID(slot) constructions. The three most convincing cases are similar to (99), with a preposed type 1 adjective; the remaining 11 candidates have postposed adjectives and are more ambiguous (100). I therefore choose to take them at face value and classify them as ID(slot) occurrences. (99)

(100)

varlamż mixailevż snż Varlam Mixail-OV son ‘Varlam Mixail’s son’ VXM 36/15–16 Vsevoloˇz Vsevolodż. snż Jurževż. vnuk Vsevolod son Jurij-OV grandson Vsevolod-J ‘Vsevolod, son of Jurij, grandson of Vsevolod’ SL 460/33

12.3 % (84) of the occurrences at the ID(slot) node are ADJ2 constructions. This is not a significantly higher proportion than at the ID node, and the distribution of animacy types is also not significantly different (table 10). ID ID(slot)

abstract 19 13

animal 0 2

concrete 16 26

human pl. 16 28

human sg. 6 15

Table 10. Possessor animacy at the ID and ID(slot) nodes, 11th–14th century Old Russian

There is a large share of constructions headed by deverbal nouns (101).

11th–14th century Old Russian

(101)

77

priemlju crkvnaja predanžja accept church-N commandments ‘I accept the church’s commandments’ PVrL 113/20

Interestingly, there is also quite a large share of constructions headed by nouns denoting inherent parts of wholes, which might have been expected to turn up with bare genitives instead. (102) is one in a series of similar examples from the same text (a description of a pilgrimage to Jerusalem). Here, the tomb is specific (the tomb of Jesus) and highly activated in the discourse. (102)

epskż paky priidˇe k dvˇerem grobnym bishop again came to doors grave-N ‘again the bishop came to the doors of the tomb’ XID 7r/3

We find that there is a significantly larger share of specific possessors (ones accessible from discourse or uniques) at the ID(slot) node (p-value = 0.01821, Fisher’s exact test, two-sided), see table 11.

ID ID(slot)

generic or non-specific 31 38

specific (discourse) 7 17

specific (unique) 19 29

Table 11. Possessor specificity at the ID and ID(slot) nodes, 11th–14th century Old Russian

None of the human singular possessors are specific and accessible from discourse, however; they are either unique (103) or have non-specific or kind reference (104). (103)

po Gospodžskomu slovu by Lord-SK word ˇ 364/4 ‘according to the Lord’s word’ ZD

(104)

priimż rabii zrakż having-taken-on slave-IJ look ‘having taken on the looks of a slave’ PVrL 113/7

Nonetheless, it seems clear that the ADJ2 construction is more freely used with “better” possessors at the ID(slot) node than at the ID node, and also that the construction is more freely used at either node in the 11th–14th century Old Russian sample than in the OCS sample.

78 ID: Reference points on instance level 39 % (267) of the occurrences at the ID(slot) node belong to the GENMOD group. The great majority of these are headed by deverbal nouns and body part nouns. Again, we find possessor nouns corresponding to those found in ADJ1 and ADJ2 constructions respectively, but largely the possessor nouns are specific. (105)

po vozvraˇstenii s pobˇedy knjazja Aleksandra after return from victory prince-GEN Aleksandr-GEN ˇ ‘after Prince Aleksandr’s return from victory’ ZAN 169/1–2

(106)

predż oˇci sžrdžca svoego before eyes heart-GEN self’s-GEN ‘before his heart’s eyes’ SBG 31/8

There are 27 occurrences of constructions with bare genitives at the ID(slot) node. The literature leads us to expect that bare genitives can be used freely with nouns denoting inherent parts of wholes, but no more than five of the GENUNMOD occurrences have such heads; the river mouth in (107) is one such example. In addition, there are four occurrences with body part head nouns (108). (107)

ustie Smjadiny mouth Smjadina-GEN ‘at Smjadina’s mouth’ SBG 40/14

(108)

ot ruky mitropolita from hand metropolite-GEN ˇ ‘from the metropolite’s hand’ ZAN 179/11–12

In fact, the majority of the occurrences are headed by deadjectival (110) or deverbal nouns (109). (109)

dˇelo sotvori blgovˇestnika deed do preacher-GEN ˇ ‘do the preacher’s deed’ ZSP 15/17

(110)

krˇepostiju razuma predrżˇza zemnoe caržstvo strength understanding-GEN ruled earthly realm ‘with the strength of his understanding he ruled his earthly realm’ ˇ 352/23–24 ZD

The nature of the possessor nouns is clearly different from what we found

11th–14th century Old Russian

79

at the ID node. At the ID node all GENUNMOD occurrences had abstract possessor nouns, apart from the ones in the partially lexically specific [vż NP-GEN mˇesto] construction. In contrast, the occurrences at the ID(slot) node span the entire animacy hierarchy. There is even one occurrence of a personal name possessor. (111)

lice zˇe ego – aky lice Iosifa face PTC his like face Joseph ˇ but his face was like Joseph’s face” ZAN 160/13

When we compare the share of abstract nouns to the rest in the two groups, the difference is statistically significant (p-value = 0.0006, Fisher’s exact test, two-sided). ID ID(slot)

human sg. 0 7

org 0 2

animal 0 0

concrete 0 8

abstract 10 10

Table 12. Animacy of possessor nouns in GENUNMOD occurrences at the ID and ID(slot) nodes, 11th–14th century Old Russian, [vż NP-GEN mˇesto] occurrences excluded

The distribution of specificity in the possessor nouns is not significantly different from what we found at the ID node, but it should still be noted that we find a full 13 examples at the ID(slot) node that are both specific and accessible from discourse, such as in (108), where the possessor noun is also human singular, and in (110). ID ID(slot)

generic/non specific 6 9

specific (discourse) 1 13

specific (unique) 3 5

Table 13. Specificity of possessor nouns in GENUNMOD occurrences at the ID and ID(slot) nodes, 11th–14th century Old Russian, [vż NP-GEN mˇesto] occurrences excluded

Thus, in the 11th–14th century Old Russian text sample we find the same tendency at the ID(slot) node as in the OCS sample: A wider range of possessors turn up as bare genitives than was the case at the ID node, and there is a larger share of “better” possessors, in the sense that there are more human singulars and maybe also more specific and discourse-accessible referents.

80 ID: Reference points on instance level 5.5 % (38) of the occurrences at the ID(slot) node are dative constructions. The distribution of head noun types found with dative constructions is not as sharply biased as in the OCS sample, where body part nouns dominate. In the Old Russian sample the main types of head nouns are (by decreasing frequency) deverbal nouns (112), body part nouns (113), deadjectival nouns (114) and kinship terms (115). As in OCS, all these noun types typically have a slot for an animate, possibly affected referent. (112)

a vˇeroju spasaemyimż sila Boˇzija estž and faith saved-ones-DAT power God-IJ is ‘but God’s power exists through the faith of the saved ones’ SBG 53/8

(113)

jakoˇze lju[bo]vniku duˇsa v tˇelˇe ljubimago like lover-DAT soul in body beloved-GEN ˇ 362/9 ‘like the lover’s soul in the body of the beloved’ ZD

(114)

podivitisja krˇeposti i muˇzestvu rezanskomu gospodstvu wonder-at strength and bravery Rjazan-SK-DAT lordship-DAT ‘wonder at the strength and bravery of the lords of Rjazan’ PBR 348/15–16

(115)

Kotjanż bˇe testž Mžstislavu Kotjan was father-in-law Mstislav-DAT ‘Kotjan was Mstislav’s father-in-law’ PBK(N) 202/11

The share of dative constructions at the ID(slot) node is not statistically significantly larger than at the ID node, but as with the ADJ2 constructions and the GENUNMOD group, we find that there is a wider range of possessor noun types. The share of abstract possessor nouns is significantly higher at the ID node than at the ID(slot) node (p-value = 0.0394, Fisher’s exact test, two-sided), and at the ID(slot) node we find that the majority of possessors are human (table 14), which is not surprising given the main types of head nouns occurring.

ID ID(slot)

human sg. 2 4

human pl. 1 15

animal 0 2

concrete 3 10

abstract 5 5

Table 14. Animacy of possessor nouns in dative constructions at the ID and ID(slot) nodes, 11th–14th century Old Russian

Further developments in the history of Russian

81

When we look at specificity, we find that there are no statistically significant differences between the two distributions. The share of kind-referring or non-specific possessor nouns is still large at the ID(slot) node (19 occurrences, as in (112) and (113)). However, there are also 12 specific and discourse-accessible possessor nouns, as in (114) and (115). Furthermore, we find far less of the ambiguities and benefactive shades of meaning seen at the ID node. Again, it seems reasonable to conclude that a reference point reading is more freely available when the dative-marked noun fills an argument slot of a relational head noun. There are only seven occurrences of mixed constructions at the ID(slot) node. Three are adjective + adjective constructions (all paratactic) (116), and four are adjective + genitive constructions (one paratactic and three hypotactic) (117). All involve type 1 adjectives and specific human possessors. (116)

prolžja krovž Borisovu i Glˇebovu spilt blood Boris-OV i Gleb-OV ‘he spilt the blood of Boris and Gleb’ PVrL 141/14–15

(117)

jego Mtre gnˇevom Bˇzžimž. i prˇcstyja wrath God-IJ and immaculate-GEN his mother-GEN ‘by the wrath of God and his immaculate mother’ PBK(L) 446/16

In conclusion, then, we see that the situation at the ID(slot) node in the 11th–14th century Old Russian sample is strongly similar to that in the OCS sample. The distribution differs subtly from the one found at the ID node. The main tendency is a fairly clear division of labour between adjective and restricted genitive constructions. However, for all the three odd constructions out, i.e. the ADJ2 construction, the dative construction and the formally unrestricted geinitive construction (as represented by the GENUNMOD group), we see that “better”, more animate and more specific possessors can occur in them when the head noun is relational.

3. Further developments in the history of Russian At the two ID nodes, then, the claims of complementary distribution are fairly well-founded, with deviations at the top and the bottom, in the sense that highly animate and specific possessors may occur in adjective constructions even when complex, and that low-animacy and non-specific/kind-referring

82 ID: Reference points on instance level possessors may turn up as bare genitives. At the ID(slot) node a wider range of possessors is permissible in the formally unrestricted genitive construction, as well as in the dative and ADJ2 constructions. In modern Russian it is not reasonable to claim that there are two separate adnominal genitive constructions. The ADJ1 construction has been marginalised,51 and the dative construction is gone, leaving the genitive with very little competition. In the diachronic survey of the constructions at the two ID nodes, then, we might therefore expect to see signs of the breakdown of the near-complementary distribution. Interestingly, what we do see is not a clear strengthening of the genitive option, but rather a weakening of the ADJ1 construction, which is clearly becoming less productive and losing ground.

3.1. The ADJ1 construction When we look at the share of ADJ1 constructions at the ID and ID(slot) nodes up to the 17th century (figure 20), we find a mild decrease at the ID node, but not at the ID(slot) node. However, when we look more closely at the occurrences, we see clear signs that the ADJ1 construction is losing productivity. The clearest indication is that the types of possessor noun stems involved in the constructions are shifting. In figure 21, we see that an ever-increasing share of the ADJ1 occurrences involve the four adjectives boˇzii “God’s”, gospodžnž “the Lord’s”, Xristovż “Christ’s” and Isusovż “Jesus”’ (grouped as “religious” in figure 21). (118)

za cerkvi boˇzija for churches God-IJ ‘for God’s churches’ PVC 37/1, 15th century

Although this increased frequency for a small group of adjectives brings the concept of type frequency to mind, we do not in fact find an overall decrease in type frequency for the ADJ1 constructions. In the 11th–14th century we have 395 occurrences and 96 types (different adjective lemmata involved in the constructions); in the 17th century material there are 139 occurrences distributed over 29 adjective lemmata. If we divide the number of adjective types by the total number of occurrences to make the type frequencies comparable, the scores are 0.24 and 0.21 respectively, e.g. very similar. If we exclude the religious adjectives, the 11th–14th century material has 214 oc-

Further developments in the history of Russian

40

50

ID ID(slot)












30










0

10

20

Per cent

83

11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 20. ADJ1 constructions at the ID nodes, diachronically, per cent

currences by 92 types (0.43) and the 17th century material has 46 occurrences by 25 types (0.54), giving a slightly higher type frequency in the 17th century material. Increasingly, the constructions with the mentioned four religious type 1 adjectives also occur in a few very frequent collocations, such as (119), which accounts for as many as 25 of the 69 ID(slot) occurrences of ADJ1 constructions with one of the four selected religious adjectives in the 17th century sample. (119)

mlstž bˇzija grace God-IJ ‘God’s grace’ GG a: 11, 17th century

It seems reasonable to see such collocations as entrenched, fully specific constructions, retrieved by language users as (complex and analysable) “chunks” rather than assembled from their component parts. In this way they are close to the constructions at the LABEL node, which have strongly conventio-

84 ID: Reference points on instance level



60

70

religious proper common

50




40





30

Per cent




20









0

10





11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 21. ADJ1 possessor stems at the ID nodes, diachronically, per cent

nalised unit status and serve as names for individual referents (see chapter 6).

11th–14th century 15th century 16th century 17th century

religious 47.42 66.46 38.00 56.15

proper 36.63 11.39 30.00 20.86

common 15.96 22.15 32.00 22.99

Table 15. Stem types in ADJ1 constructions by period, all nodes, per cent

Figure 21 shows that the share of proper noun stems in the ADJ1 occurrences decreases quite steeply over time. If we look at the overall share of proper noun stems in ADJ1 constructions at all nodes (table 15), the decrease is less sharp, since there is quite a high number of ADJ1 constructions at the LABEL node (see chapter 6), mainly due to the partially specific patronymic construction [ADJ1PROPER SON / DAUGHTER]. It is also worth noting that

Further developments in the history of Russian

85

type 1 adjectives occur on their own as surnames in the Middle Russian texts; example (120) has both. (120)

poslaˇsa [. . . ] po Borisa po Vasilževa syna Djatlova sent for Boris for Vasilij-OV son Djatlov ‘they sent for Boris Vasilij’s son Djatlov’ PBSV 554/6, 16th century

These patronymics are transparently ID(slot) constructions in origin, but have been conventionalised as complex personal names. We recall that in the 11th– 14th century material, ADJ1 constructions with a possible patronymic reading were classified as ID(slot) occurrences due to their ambiguity (see page 76). The share of regular ID and ID(slot) constructions with type 1 adjectives formed from proper noun stems is much smaller in the 15th–17th century material than in the material from the earliest attestations of Old Russian; (121) is one of only six such examples at the ID(slot) node in the 17th century material. (121)

pro Ivannovu smertž about Ivan-OV death ‘about Ivan’s death’ PMM 107/24 17th century

Finally, we see that the share of ADJ1 constructions with common noun stems at the two ID nodes is increasing. (122)

na patriarxove dvorˇe on patriarch-OV court ˇ ‘in the Patriarch’s court’ ZAvv 24/10–11, 17th century

(123)

po poveleniju deviˇcju by order maiden-J ˇ ‘by the maiden’s order’ ZPF 216/28 16th century

3.2. The ADJ2 construction For the ADJ2 constructions at the ID and ID(slot) nodes, there is no obvious tendency of development from the earliest attestations up through the 17th century in terms of frequency. The figures are so low that the data from the ID and ID(slot) nodes must

86 ID: Reference points on instance level


15

Per cent

20

25

30

ID ID(slot)







10










5




0




11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 22. ADJ2 constructions at the ID nodes, diachronically, per cent

be collapsed in the further analysis, but the tendency appears to be the same: The ADJ2 constructions at these two nodes increasingly have human, specific possessors. (124)

na voevocki dvor on commander-SK court ‘to the commander’s court’ PKS 69/12, 17th century

The share of human possessors in the 17th century sample is significantly larger than the one in the 11th–14th century sample (p-value = 0.0004), and the same holds for the share of specific possessors (p-value = 0.0028; both Fisher’s exact test, two-sided). We find that part of the increase of singular, human, specific possessors in ADJ2 constructions is accounted for by the fact that the SK adjective caržskż “tsar’s” has all but replaced the OV adjective carevż “tsar’s” in our 17th century sample (15 out of the 23 human possessor occurrences are instances of caržskż, while there is only one single occurrence of carevż).

Further developments in the history of Russian

11th–14th century 15th century 16th century 17th century

human sg. 21 3 4 23

human pl. 44 17 2 7

org 0 3 0 0

animal 2 1 0 0

concrete 42 18 0 3

87

abstract 32 4 3 5

Table 16. Animacy of possessors in ADJ2 constructions at ID and ID(slot) nodes, diachronically

11th–14th century 15th century 16th century 17th century

specific 72 29 7 30

generic/non-specific 69 17 2 8

Table 17. Specificity of possessors in ADJ2 constructions at ID and ID(slot) nodes, diachronically

(125)

a sami oˇzidajutż crskog priˇsestvie˛ and themselves await tsar-SK coming ‘and they themselves await the tsar’s coming’ RCAM 22/25–26 17th century

3.3. The genitive constructions Neither the GENMOD nor the GENUNMOD group has any obvious tendency of development when it comes to frequency up to the 17th century. The share of genitive occurrences – both GENMOD and GENUNMOD – remains stable throughout the entire period under consideration; the GENMOD share large and the GENUNMOD share small. The number of bare genitive occurrences at the ID node is really too low to tell us anything conclusive; the 9 occurrences in the 15th century material and the three occurrences in the 16th century material are all inanimate, save a single occurrence of a human proper noun (126). (126)

iz ordy Asanbega from horde Asanbeg-GEN ‘from Asanbeg’s horde’ AN 31/13, 15th century

40

50

88 ID: Reference points on instance level GENMOD, ID GENMOD, ID(slot) GENUNMOD, ID GENUNMOD, ID(slot)









30




Per cent





10

20














0




11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 23. Genitive occurrences at the ID nodes, diachronically, per cent

In the 17th century material, however, there are ten occurrences, six of which are human, singular and specific, but all six are occurrences of the same collocation from the same text, seen in (127). However, it is perfectly possible to form a type 1 adjective from the noun cˇ judotvoržcž, and in fact it does occur in the same text (128). (127)

vo obiteli cˇ judotvorca in monastery miracle-performer-GEN ‘in the miracle-performer’s monastery’ SAP 130/19, 17th century

(128)

u cˇ judotvorcovy raki by miracle-performer-OV coffin ‘by the miracle-performer’s coffin’ SAP 134/18, 17th century

At the ID(slot) node, there are slightly higher numbers of occurrences of GENUNMOD (19 in the 15th century sample, 11 in the 16th century sample and 10 in the 17th century sample). The distribution of possessor nouns, to the extent that this can be judged from such low numbers, appears to be the

Further developments in the history of Russian

89

same; there are human singulars and plurals, concrete and abstract nouns in all three samples, and also both specific and non-specific/kind-referring possessors, similar to what we found in the 11th–14th century sample. Variation between ADJ1 constructions and bare genitives can clearly be found here too – compare (123) and (129), which are both from the same 16th century text. (129)

po poveleniju devicy by order maiden ˇ ‘according to the maiden’s order’ ZPF 216/25, 16th century

3.4. The dative construction At the two ID nodes, the dative construction appears to be completely marginalised in the 15th, 16th and 17th century samples, with 3, 2 and 1 occurrences respectively. (130)

Iskoni zˇe nenavidjai dobra rodu from-beginning PTC hating goodnesses kind-DAT cˇ eloveˇcju human-DAT ˇ 211/21– ‘from the beginning hating the goodnesses of mankind’ ZPF 22, 16th century

3.5. Mixed constructions In the 11th–14th century material we found two basic types of mixed adjectivebased constructions: Constructions with multiple adjectives (131) and constructions with a mix of adjectives and genitives (132). We saw that they both could be either paratactic or hypotactic (though double adjective constructions are mostly paratactic). In the 15th–17th century material we find both these main types, in addition to a third type, which has the form [POSSESSIVE PRONOUN DA N], frequently seen in letter formulae such as in (133).52 (131)

zż gsdrva svtitelskova patriarˇsa dvara from lord-OV bishop-SK patriarch-J court ˇ ‘from his lordship the Bishop (and) Patriarch’s court’ CBK 2:10, 17th century

30

90 ID: Reference points on instance level

25

ID ID(slot)

20




15 10

Per cent







5











0




11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 24. Mixed constructions at the ID nodes, diachronically, per cent

(132)

Turki-ˇzż [. . . ] pustiˇsa surnya i trubnyja glasy i Turks-PTC let-loose flute-N and trumpet-N voices and tumbanż tmoˇcislennyxż drums-GEN countless-GEN ‘but the Turks let loose the voices of flutes and trumpets and countless drums’ PVC 9/19, 15th century

(133)

na tvoju gosudarevu zemlju against your sovereign-OV land ‘against your, the sovereign’s land’ IG/VG III:259/34, 16th century

In the 11th–14th century material the mixed constructions are not particularly frequent, but in the samples from the later stages, we see that they increase sharply both at the ID and the ID(slot) nodes (see figure 24). Thus we find that we have a growing group of constructions violating the fairly clear division of labour between simple adjective constructions and complex

Further developments in the history of Russian

91

genitive constructions found at the two ID nodes in the 11th–14th century material.

3.6. A snapshot of the 18th century: When did the genitive start expanding? As we have seen, the constructions at the two ID nodes in the 17th century sample differ quite a lot from those in the 11th–14th century one. The ADJ1 construction, though not less frequent, has shifted considerably, showing an ever-increasing share of instantiations with a small set of religious type 1 adjectives, all of which are still in use in modern Russian. Further, we saw that adjectives formed from proper noun stems are much more rarely found in the later language samples. The types of possessor stems involved in the ADJ2 constructions have also shifted. More and more of them are human, singular and specific, suggesting that the construction alternates with the ADJ1 construction for human common noun possessors, and that the “worse” possessors, such as inanimates, may increasingly be left to the genitive construction. The dative construction virtually disappears at the two ID nodes in the later samples, and is clearly no longer an option in reference point situations. Nevertheless, the situation is still far from that in modern Russian, and the main difference lies in the place and status of the genitive constructions. 17th century Russian still appears to have two genitive constructions, a formally restricted one and a semantically restricted one, with different distributions. Modern Russian, on the other hand, quite indisputably has one single genitive construction which can be used in all parts of the possessive semantic space. It meets only marginal competition from the ADJ2 construction, in addition to competition in a very restricted sphere, apparently only with reference-point ˇ semantics, from the ADJ1 construction (see Kopˇcevskaja-Tamm and Smelev 1994). Several scholars have found the expansion of the bare possessive genitive in Russian to be a quite recent change. Widn¨as (1958) says that the change starts in the 18th century, but the perceived starting point varies considerably from author to author.53 An investigation of a small 18th century sample54 shows that not even at that late stage are bare genitives frequent or remarkably different at the ID and ID(slot) nodes than in the earlier text samples. The number of GENUNMOD occurrences is very low in the sample – only 2 (out of 16) occurrences at the ID node and 6 (out of 52) at the ID(slot) node. Thus, there has cer-

92 ID: Reference points on instance level tainly been no obvious increase of bare genitives at either node, and at the ID node both occurrences have abstract, inanimate possessors. There are two bare genitive human possessors at the ID(slot) node, but of these two, one is an OV adjective surname, where the genitive was close to obligatory (134), and the other one has kind reference (135). (134)

vkljuˇcil ja raskajanie Dobroserdova included I remorse Dobroserdov-GEN ‘I included Dobroserdov’s remorse’ (Makogonenko 1970:148a/25)

(135)

no v dolˇznosti ljubovnika pokazalsja mne eˇscˇ e cˇ udnee but in duty lover-GEN seemed me even more-wonderful ‘but in the lover’s duty he seemed even more wonderful to me’ (Makogonenko 1970:189a/25)

17th century 18th century

ADJ1 34.92 22.06

ADJ2 9.30 13.24

genitive 39.45 63.24

dative 0.25 0.00

mixed 16.08 1.47

Table 18. Constructions at the ID and ID(slot) nodes, 17th and 18th century samples, per cent, n = 398 and n = 69 respectively

Apart from this, the small sample indicates quite dramatic changes (table 18): The total share of genitive constructions (both GENMOD and GENUNMOD occurrences) at the two ID nodes has increased considerably, from 39.5 % in the 17th century sample to 63.2 % in the 18th century sample. The share of ADJ1 constructions is down from 34.9 to 22.1 %. There is also only one single mixed construction occurrence. These results confirm the general trend observed in 15th–17th century samples: The changes were slow and gradual. It seems very clear that the adjective constructions had to be considerably weakened before use of the bare genitive became an acceptable option in the core domain of the ADJ1 construction, namely reference point situations with an animate reference point, particularly a personal name. In the 18th century sample, we see much more dominant genitive constructions and much diminished and more restricted adjective constructions, compared to the situation in the 17th century sample. Nonetheless, the authors still seem quite reluctant to use bare genitive-marked nouns at the ID nodes.

Chapter 5 The ELAB nodes: Intrinsic relationships

In chapter 2 we established two map nodes, ELAB and ELAB(slot), hinging on Langacker’s (2000:73–90) notion of an intrinsic relationship: If an entity X cannot be conceptualised without reference to some other entity or entities, X is conceptually dependent, and there is an intrinsic relationship between X and the other entity or entities. At the ELAB(slot) node we find occurrences where the possessor nominal fills a slot in the argument structure of a relational head noun, but where it is not reasonable to posit a reference point reading. The possessor nominal does not serve to identify the head noun, but rather to flesh it out and provide additional information. The data from early Slavic makes it necessary to distinguish between ELAB(slot) constructions headed by nouns denoting inherent parts of wholes, and those headed by other relational nouns. A separate node labelled ELAB(part) was therefore posited for these cases. At the ELAB node, on the other hand, we find cases without relational head nouns, but where it is possible to argue that the head noun cannot be conceptualised without reference to the dependent noun. In these cases a reference point reading is not reasonable, nor does the construction as a whole denote a type or constitute a conventionalised name or label for a particular referent. These occurrences fall into three main types: 1. The two elements are in semantic apposition; the head and the modifier have the same referent (that scoundrel of a man) 2. There is a salient relationship between the two elements, e.g. between an artefact and the material of which it is made (a ring of gold) 3. The modifier denotes a non-accidental quality of the head noun (a man of integrity, Joan of Arc)

1. Reference points or not? As discussed in chapter 2, it is not obvious how we should distinguish between cases where a noun fills an argument slot of a relational noun and has a reference point reading, and cases where it just fills the slot, and hence

94 The ELAB nodes: Intrinsic relationships should be classified at the ELAB(slot) node. This is particularly problematic in the kind of situation that we find in early Slavic: Most, if not all, of our five main construction types can also encode other semantics than reference point situations. Taylor (1996) argues that to serve as a reference point, a noun must be cognitively accessible. Since a corpus of Old Russian with annotation for information status is not available, I have opted for a simplified analysis. I assume that the object slot of a deverbal noun (or a two-slot deadjectival noun) (136), the slot for the whole of a noun denoting a part55 (137), the slot of a ruler noun (138) and the subject matter slot of a representation noun (139) will all normally be filled by a noun that does not serve as a reference point. (136)

na poxvalenie predobrago gospodina on praising excellent-GEN lord-GEN ˇ 363/7 ‘in order to praise the excellent lord’ ZD

(137)

pri krai morja by edge sea-GEN ˇ ‘by the edge of the sea’ ZAN 165/2

(138)

Efiopžskaja crca empress Ethiopia-SK ‘the empress of Ethiopia’ PVrL 62/9

(139)

kż ikonˇe Gospodžni to icon lord-N’ ‘at the icon of the Lord’ SBG 34/9

In cases when the slot filler is clearly cognitively accessible, the constructions have been classified as ID(slot) constructions, but in cases where there is doubt, I have opted for an ELAB(slot) analysis. Conversely, with other types of argument slots (for instance subject slots of deverbal and deadjectival nouns and the argument slots of kinship terms and body part nouns), I have given preference to the ID(slot) reading. The feasibility of such an approach is confirmed when we look at preliminary data from the OCS data in the PROIEL corpus,56 which is in the process of being annotated for information status. We see (in table 19) that dependent nouns of OCS deverbal nouns are more likely to be specific and accessible from discourse (by mention or inference) or from encyclopaedic knowledge

ELAB(slot)

95

(uniques) if they are tagged syntactically as ATR (attributes) than if they are syntactically tagged as NARG (nominal arguments). specific and accessible other

attribute 49 25

nominal argument 15 19

Table 19. Nouns dependent on deverbal nouns in the Codex Marianus, limited to occurrences that have an aligned token in Greek annotated for information status, p-value = 0.03634

2. ELAB(slot) 2.1. OCS As figure 25 on page 96 shows, the distribution at the ELAB(slot) node is radically different from the ones at the two ID nodes. In particular, we see that the share of ADJ1 constructions is small (7.7 %, against 59.2 % at ID and 46.9 % at ID(slot)). Conversely, we see that the dative construction and the GENMOD group, which were both found to be rather marginal at the ID nodes at around 5 % each, have a large share of the ELAB(slot) realisations: 20.5 % of all occurrences are dative constructions, and 17.9 % of the occurrences belong to the GENUNMOD group. ID(slot) ELAB(slot)

GENMOD 126 43

GENUNMOD 21 27

Table 20. GENMOD and GENUNMOD occurrences at the ID(slot) and ELAB(slot) nodes (occurrences with heads denoting parts of wholes excluded), OCS. P-value = 0.0001007, Fisher’s exact test, two-sided

Also, at the ELAB(slot) node, we see that there is much less imbalance between the GENUNMOD and GENMOD occurrences. There are more GENMOD occurrences (28.5 %) than GENUNMOD occurences (17.9 %), but the difference is much smaller than at the ID nodes. The difference between the ID node and ID(slot) node in this respect is highly statistically significant (see table 20).

0

5

10

15

20

25

96 The ELAB nodes: Intrinsic relationships

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 25. OCS constructions at the ELAB(slot) node, per cent, n = 151

Thus, it does not seem reasonable to assume that there are any restrictions on the use of adnominal genitives at the ELAB(slot) node, regardless of the type of head noun. If we continue positing that there are actually two adnominal genitive constructions in early Slavic (as we did in chapter 3, section 2), the occurrences at the ELAB(slot) node cannot belong to the restricted genitive construction. Since the type of head noun is a guiding criterion for deciding between ELAB(slot) and ID(slot), the range of head noun types at the ELAB(slot) node is not surprising (table 21): The main types are deverbal nouns with their object slot filled (140), representation nouns (141) and ruler nouns (142). (140)

za straxż ijudeiskż for fear Jew-SK dia ton phobon tˆon Ioudaiˆon ‘for the fear of the Jews’ Zogr. John 19:38

ELAB(slot)

ADJ1 ADJ2 GENMOD GENUNMOD DATIVE

deverbal 58.3 20.6 55.8 51.9 58.1

representation 0 0 7.0 7.4 3.2

ruler 41.7 76.5 18.6 25.9 22.6

97

other 0 2.9 18.6 14.8 16.1

Table 21. Head noun types by construction at ELAB(slot) in OCS

(141)

sżkaˇzi namż pritˇco˛ plˇevelž selżnyixż tell us parable weeds-GEN field-N.GEN diasaphˆeson hˆemin tˆen parabolˆen tˆon zizaniˆon tou agrou ‘explain to us the parable of the weeds of the field’ Ass. Matthew 13:36

(142)

vinogradu gž master vineyard-DAT ho kurios tou ampelˆonos ‘the master of the vineyard’ Savv. Matthew 21:40

Turning to the construction types at ELAB(slot), we find that ADJ1 constructions are rare – only 12 out of 151 occurrences. This fits both with our previous observations of the ADJ1 construction in OCS and with the assumptions made about the ELAB(slot) constructions: ADJ1 constructions normally have a human, specific possessor accessible from discourse or encyclopaedic knowledge, whereas the slot fillers in the ELAB(slot) occurrences are assumed not to serve as reference points, and hence normally not to be the ideal reference points that the ADJ1 possessors are. The occurrences are all headed by either deverbal nouns (143) or ruler nouns (144). In six of the cases, the slot filler is the adjective izdrailevż “Israel’s”, as in (143), and there is also one case of the similar ijudovż “Judah’s”, seen in (144). These adjectives do not refer to a single specific human being, but most plausibly to unique ethnic groups. They are therefore untypical type 1 adjectives. (143)

cˇ aje˛ utˇexy izdrvy hoping consolation Israel-OV prosdekhomenos paraklˆesin tou Israˆel ‘longing for the consolation of Israel’ Mar. Luke 2:25

98 The ELAB nodes: Intrinsic relationships (144)

vž vldkaxż ijudovaxż in rulers Judah-OV en tois hˆegemosin Iouda ‘among the rulers of Judah’ Ass. Matthew 2:6

The remaining occurrences all have type 1 adjectives referring to God or Jesus (boˇzii “God’s”, xristosovż “Christ’s”, gospodžnž “the Lord’s”). In none of the cases is the slot-filler accessible from discourse. (145)

i bˇe ob noˇstž vż molitvˇe bˇzii and was at night in prayer God-IJ kai eˆ n dianuktereuˆon en tˆei proseukhˆei tou theou ‘and all night he continued in prayer to God’ Mar. Luke 6:12

Occurrences like (146) may be seen as conventionalised units that could have been classified as LABEL or even TYPE occurrences instead. (146)

o vˇerˇe gospodžnyi of faith Lord-N’ en tˆei pistei tou kuriou ‘of the faith in the Lord’ Supr 16:206/1

ADJ1 constructions can therefore be concluded to be marginal at the ELAB(slot) node, as they are few and to some extent untypical. 22.5 % (34) of the occurrences at the ELAB(slot) node are ADJ2 constructions. However, there is an obvious bias: 26 of the examples are headed by ruler nouns and uniformly express the relationship between a ruler and his (plural or collective) subjects. The collocation in (147) accounts for most of the examples. (147)

radui se˛ crju ijudeiskż rejoice REFL king Jew-SK khaire, basileu tˆon Ioudaiˆon ‘hail, king of the Jews’ Savv. Matthew 27:29

Apart from these examples, occurrences with ADJ2 constructions are scarce at the ELAB(slot) node. Six of them have adjectives filling the object slot of deverbal nouns (148) and there is also a single occurrence of one headed by an offspring noun (149).57

ELAB(slot)

(148)

straxa radi ijudeiska fear because-of Jew-SK dia ton phobon tˆon Ioudaiˆon ‘for fear of the Jews’ Zogr. John 7:13

(149)

otż sego ploda lozžnaago from this fruit vine-N ek toutou tou genˆematos tˆes ampelou ‘of this fruit of the vine’ Ass. Matthew 26:29

99

We find the same pattern when it comes to possessor referents in ADJ2 constructions here as at the two ID nodes: The slot fillers are plural and often have kind reference if they are human, and many of the occurrences with inanimate slot fillers, too, have non-specific or kind reference. As mentioned, the main bulk of the occurrences have genitive slot fillers. 28.5 % (43) of all ELAB(slot) occurrences belong to the GENMOD group. The head nouns of these occurrences are predominantly deverbal (24 occurrences) (150), but there are also seven occurrences headed by ruler nouns (151), and a scatter of miscellaneous other types. (150)

na obliˇceniju nepokorżnaago tvojego uma on reproach disobedient-GEN your-GEN mind-GEN eis elegkhon apeithous sou gnˆomˆes ‘in reproach of your disobedient mind’ Supr. 16:194/12–13

(151)

prizżvavż zˇe episkupa solunżska grada having-sent-for PTC bishop Thessaloniki-SK.GEN city-GEN Metasteilamenos oun ton Thessalonikˆes episkopon ‘having sent for the bishop of the city of Thessaloniki’ Supr. 16:197/17

Looking at the slot fillers in these occurrences, we again find a wide range of types, from specific and discourse-accessible to non-specific and kindreferring ones, and from human singulars to abstract nouns. Not unexpectedly, we find that human singular slot fillers are relatively rare with deverbal nouns; only four out of 24 examples had human singular “objects” (152). (152)

vż vrˇeme˛ iznoˇsenija trżblaˇzenaago isakija in time out-carrying thrice-blessed-GEN Isaac-GEN Tˆoi oun kairˆoi tˆes ekdˆemias tou trismakariou Isaakiou

100 The ELAB nodes: Intrinsic relationships ‘at the time of the carrying out of the thrice-blessed Isaac’ Supr. 16:208/29–209/1 As we saw in the beginning of this section, the share of GENUNMOD occurrences is significantly larger at the ELAB(slot) node than at the two ID nodes. 17.9 % (27) of all ELAB(slot) occurrences belong to the GENUNMOD group. 14 of these occurrences are headed by deverbal nouns, as in (153) and (154), and 7 by ruler nouns (155). (153)

spż mira saviour world-GEN ho sˆotˆer tou kosmou ‘the saviour of the world’ Zogr. John 4:42

(154)

po vżzmo˛ sˇtenii vody after troubling water-GEN meta tˆen tarakhˆen tou hudatos ‘after the troubling of the water’ Zogr. John 5:5

(155)

gdnż xrama master house-GEN ho oikodespotˆes ‘the master of the house’ Ass. Matthew 24:43

The remaining six occurences have various head noun types, including kinship terms (in constructions deemed not to have a reference point reading, such as (156)), representation and source. (156)

i aˇste bo˛ detż tu snż mira and if be there son peace-GEN kai ean ekei eˆ i huios eirˆenˆes ‘and if a son of peace is there’ Mar. Luke 10:6 58

It is interesting to note that as many as 11 of the GENUNMOD occurrences at the ELAB(slot) node actually do not render a Greek genitive construction, but a compound, as in (155) and (157). One might think that the ADJ2 construction was the obvious choice for rendering compounds, cf. chapter 6. However, the examples appear to fit nicely in with the other non-compound examples of ELAB(slot) GENUNMOD, and it is not obvious that these examples would have been perceived as different, were it not for the Greek text.

ELAB(slot)

(157)

101

stroenie domu otż mene gż moi otymetż lord my takes-away ordering house-GEN from me ho kurios mou aphaireitai tˆen oikonomian ap’ emou ‘my master will take the stewardship (lit. ‘ordering of the house’) from me’ Mar. Luke 16:3

Finally, it should be noted that none of the GENUNMOD occurrences at the ELAB(node) have human singular slot fillers, and there are only two rather tenuous examples of animate slot fillers, involving demons (158) and fish (159) respectively. (158)

o velžzˇevulˇe kżne˛ zi bˇesż by Beelzebub prince demons-GEN en Beelzeboul tˆoi arxonti tˆon daimoniˆon ‘by Beelzebub, the prince of demons’ Mar. Luke 11:15

(159)

o lovitvˇe rybż about catch fish-GEN epi tˆei agrai tˆon ikhthuˆon ‘about the catch of fish’ Mar. Luke 5:9

The rest of the examples have inanimate slot fillers, concrete or abstract. 20.5 % (31) of the occurrences at the ELAB(slot) node are adnominal dative constructions. The dative construction also behaves strikingly differently at the ELAB(slot) node than at the two ID nodes; in fact it is clear that this node is the adnominal dative construction’s semantic centre of gravity, as these 31 occurrences constitute 31.6 % of all dative constructions in the OCS material. The constructions are largely headed by deverbal nouns (18 occurrences) (160), but there are also 7 headed by ruler nouns, and scattered other occurrences. None of the ruler nouns are actual agentive nouns derived from verbs, but it is easy to see that they involve the same kind of object-like relation (161). The dative occurrences thus appear to constitute a rather coherent group. (160)

i sżtvorjo˛ vy lovca cˇ komż and will-make you catchers men-DAT kai poiˆesˆo humas aleeis anthrˆopˆon ‘and I will make you catchers of men’ Ass. Matthew 4:19

102 The ELAB nodes: Intrinsic relationships (161)

molite se˛ ubo gnu zˇe˛ tvˇe pray REFL PTC lord harvest-DAT deˆethˆete oun tou kuriou tou therismou ‘pray therefore to the lord of the harvest’ Mar. Luke 10:2

If we look at the properties of the slot fillers in the dative constructions, we find a pattern rather similar to the one we found at the ID node: The dativemarked nouns tend to have non-specific or kind reference rather than specific reference (21 vs. 9 occurrences), and they tend to be inanimate (10 abstract occurrences, 15 concrete occurrences) rather than human (6 occurrences, all in the plural). The dative construction is the fixed choice for the collocations in (162)59 and (163) (4 and 8 occurrences respectively). (162)

vż otżpuˇstenie grˇexomż in forgiveness sins-DAT eis aphesin hamartiˆon ‘for the forgiveness of sins’ Mar. Luke 3:3

(163)

tu bo˛ detż plaˇcž i skržˇzetż zo˛ bomż there will-be weeping and gnashing teeth-DAT ekei estai ho klauthmos kai ho brugmos tˆon odontˆon ‘there will be weeping and gnashing of teeth’ Mar. Luke 13:28

Finally, there are three cases of mixed constructions: one paratactic double genitive construction, one paratactic adjective/genitive construction and one hypotactic adjective/genitive construction. All of these have specific slot fillers. (164)

o ustrojenii cržkżvžnˇeeˇ mż i pravye˛ vˇery about ordering church-SK and right-GEN faith-GEN epi tˆei katastasei tˆon [. . . ] ekklˆesiˆon kai tˆes orthodoxou pisteˆos ‘about the ordering of the churches and the orthodox faith’ Supr. 16:203/3–4

To summarise, at the ELAB(slot) node we find a situation that is very different from the ones at the two ID nodes. Neither of the two adjective constructions are central; the examples of ADJ1 are marginal and untypical, and the ADJ2 construction only occurs regularly with ruler nouns, to denote the relationship between a ruler and his (plural) subjects. The adnominal genitive, on the other hand, is used freely, with a large share of bare genitives. It is clear that at the

ELAB(slot)

103

0

5

10

15

20

25

OCS OR

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

genitive/dative

mixed

Figure 26. Constructions at the ELAB(slot) node in OCS (n = 151) and 11th–14th century Old Russian (n = 295), per cent

ELAB(slot) node we cannot speak of complementary distribution between adjective constructions and a restricted genitive construction. If the hypothesis that there are two genitive constructions is to be maintained, we must assume that at the ELAB(slot) node we are dealing with the non-restricted variety. We also find that the ELAB(slot) node is the dative construction’s centre of gravity; 31.6 % of all adnominal dative occurrences in the OCS sample are found here. This is a rather close-knit group of occurrences, where the dativemarked nouns mostly fill object slots of deverbal nouns or the object-like slots of ruler nouns.

104 The ELAB nodes: Intrinsic relationships 2.2. 11th–14th century Old Russian In the 11th–14th century Old Russian sample, we see a picture largely similar to the one we saw in the OCS sample (figure 26). When we compare the distribution of constructions to that at the ID(slot) node, we find the same kinds of deviations that we found in the OCS sample: The share of ADJ1 constructions is drastically smaller (5.8 % against 36.1 % at ID(slot)), whereas the GENUNMOD group and especially the dative construction have considerably larger shares (9.5 % vs. 3.9 % and 23.4 % vs. 5.6 % respectively). There is, however, a difference between the OCS and Old Russian distributions: The share of GENUNMOD occurrences is significantly smaller in the Old Russian sample than in the OCS one (p-value = 0.03762, Fisher’s exact test). However, as seen in table 22, the share of GENUNMOD occurrences is still significantly higher at the ELAB(slot) node than at the ID(slot) node (p-value = 6.896e-05, Fisher’s exact test, two-sided), indicating, again, that we are not dealing with the restricted genitive construction at the ELAB(slot) node in the 11h–14th century Old Russian sample either. ID(slot) ELAB(slot)

GENMOD 267 83

GENUNMOD 27 28

Table 22. GENMOD and GENUNMOD occurrences at the ID(slot) and ELAB(slot) nodes, 11th–14th century Old Russian

ADJ1 ADJ2 GENMOD GENUNMOD DATIVE

deverbal 82.4 6.1 65.1 78.6 88.4

power 0 1.2 2.4 0 0

representation 17.6 0 8.4 3.6 0

ruler 0 89.0 13.3 7.1 8.7

other 0 3.7 10.8 10.7 2.9

Table 23. Head noun types by construction at ELAB(slot) in 11th–14th century Old Russian, per cent

The head noun type being, again, one of the guiding criteria for distinguishing between ID(slot) and ELAB(slot), we find the same range of head noun types at the ELAB(slot) node in the Old Russian sample as we did in the OCS sample (table 23). The distribution is similar, but appears to be

ELAB(slot)

105

more polarised in the Old Russian sample: ADJ1 constructions (165), dative constructions (166) and GENUNMOD occurrences (168) are almost exclusively headed by deverbal nouns, and ADJ2 constructions by ruler nouns (167). Only the GENMOD occurrences have a fairly even distribution of head nouns, although the deverbal heads dominate.60 (165)

xsvy i potrudi se˛ so mnoju ljubve radi and troubled REFL with me love for-sake-of Christ-OV ‘and he took care of me because of his love of Christ’ XID 12r/3

(166)

i esta zastupnika Rusžstˇei zemli and are defenders Rus’-SK-DAT land-DAT ‘and they are the defenders of the land of Rus’’ PVrL 137/19–20

(167)

Efiopžskaja crca empress Ethiopia-SK ‘the Empress of Ethiopia’ PVrL 62/9

(168)

tvorimom. ot vragż i slugż zloby by enemies and servants evil-GEN done ‘done by the enemies and the servants of evil’ PVrL 41/10

As was the case in the OCS sample, ADJ1 constructions are rare at the ELAB(slot) node, constituting only 5.8 % (17) of the occurrences.61 And like in the OCS sample, most of these (14 occurrences) are headed by deverbal nouns, as in (169) and (170), whereas the remaining three occurrences are headed by representation nouns (171). (169)

pamjatž zˇe Boˇzia na dvoe razdˇeljaetsja memory PTC god-IJ in two divides-itself ˇ 363/13 ‘but the memory of God divides itself into two’ ZD

(170)

gora [. . . ] prosela se˛ v raspe˛ tie xsvo mountain cracked REFL in crucifixion Christ-OV ‘the mountain cracked at the time of the crucifixion of Christ’ XID 18r/3

(171)

pred obrazomż Xsvymż before image Christ-OV ‘before the image of Christ’ PVrL 41/23

All the ADJ1 occurrences have specific slot fillers, but in fact only one single example (172) can be said to have one that is activated in the dis-

106 The ELAB nodes: Intrinsic relationships course, since Nevrjun is also mentioned in the preceding sentence.62 Thus, this example could arguably have been classified as an ID(slot) occurrence instead. (172)

po plˇenenii zˇe Nevrjunevˇe after capture PTC Nevrjun-OV ˇ ‘after the capture of Nevrjun’ ZAN 175/1

Another example is a heading to a chronicle section (173), which might also be argued to be an ID(slot) occurrence. (173)

o ubženži Borisovˇe about killing Boris-OV ‘about the murder of Boris’ PVrL 132/1

Of the remaining occurrences, 12 involve adjectives from the group of high-frequency religious denominal adjectives isolated on page 82 (i.e. boˇzii “God’s”, gospodžnž “the Lord’s”, Xristovż “Christ’s” and Isusovż “Jesus”’); the remaining three examples have reference to Jesus, the Devil and the Old Testament Adam respectively. None of them are participants in the current discourse, and seven of the examples are found in a tale of a pilgrimage to Jerusalem (XID), where Biblical events are constantly referred to, sometimes amounting to mere place names, as in (174), which was classified as a LABEL construction, and sometimes not, as in (175). (174)

ot pupa zˇemnago do raspe˛ tže˛ gsne˛ vi. saˇzen from navel world-N to crucifixion lord-N’ 12 fathoms ‘from the Navel of the World to the Lord’s crucifixion there are 12 fathoms’ XID 20v/14

(175)

vż oltarem velicem napisano es adamovo vżzdviˇzenie great written is Adam-OV elevation in altar ‘on the great altar Adam’s ascension is painted’ XID 18v/8

To conclude: Clear, non-conventionalised examples of ELAB(slot) realised by ADJ1 are not common in the 11th–14th century Old Russian sample. 22.5 % (82) of the occurrences at the ELAB(slot) node were ADJ2 constructions. However, as in the OCS sample, the great majority of these (89 %) were headed by ruler nouns, as in (176).

ELAB(slot)

(176)

107

Togda ubienż otż Totarż velikii knjazž Kievžskii then killed by Tatars great prince Kiev-SK ‘Then the Grand Prince of Kiev was killed by the Tatars’ PBK(N) 203/1–2

In addition, there were four occurrences headed by deverbal nouns, all of which had slot fillers with plural reference, as in (177) and (178), and scattered other occurrences, such as (179). (177)

I abie uzžrˇe [. . . ] meˇcžnoe ocˇesˇtenie and immediately saw sword-N sharpening ‘And immediately he saw sharpening of swords’ SBG 35/6

(178)

drugii straxa radi zˇidovžskago otverˇzesja s kljatvoju others fear for-sake-of Jew-SK renounced with oath ‘others renounced you by oath because of their fear of the Jews’ SKT 20/26

(179)

i priimu vlastž Rusžskuju edinż and I-take power Rus’-SK alone ‘and I will take the power over Rus’ alone’ PVrL 139/28–29

Thus, apart from their specialised function with ruler nouns, ADJ2 constructions are marginal at the ELAB(slot) node, as they were in the OCS sample. As in the OCS sample, a large share of the ELAB(slot) occurrences have genitive slot fillers. 28.1 % (83) of the occurrences belong to the GENMOD group. Again as in the OCS sample, they are predominantly headed by deverbal nouns (54 occurrences), as in (180). Occasionally they are also headed by ruler nouns (11 occurrences), as in (181), representation nouns (7 occurrences), as in (182), and scattered other types. (180)

toja na vżtoroe lˇeto po ustroenie cžržkve on second year after building church-GEN that-GEN ‘in the second year after the building of that church’ SBG 64/10

(181)

igumenż Styja Gory abbot holy-GEN mountain-GEN ‘the abbot of the Holy Mountain’ PVrL 157/25–26

(182)

ot ikony svjatyja Bogorodica from icon holy-GEN mother-of-God-GEN ‘because of the icon of the Holy Mother of God’ BNS 444/2–3

108 The ELAB nodes: Intrinsic relationships As found in the OCS sample, there is a wide range of slot fillers: human and inanimate, specific and non-specific, referents that are accessible from discourse and referents that are not. We saw in the beginning of this section that GENUNMOD occurrences are much more frequent at the ELAB(slot) node than at the ID(slot) node. The difference is less striking than in the OCS sample, but still highly statistically significant. 9.5 % (28) of the ELAB(slot) occurrences belong to the GENUNMOD group. 22 of these are headed by deverbal nouns, as seen in examples (183) and (184). (183)

I bystž [. . . ] trusż ot kopij lomlenija and was noise from lances-GEN breaking ˇ ‘And there was noise from the breaking of lances’ ZAN 171/5

(184)

Iˇze kż Bogu vˇerujutż i vż nadeˇzju vżskržkržsenija who to God believe and in hope resurrection-GEN ‘Those who believe in God and in the hope of resurrection’ SBG 52/17

The remaining 6 occurrences are headed by nouns of various types, such as ruler nouns (185) and offspring nouns (186). (185)

no cˇ to reˇce gnż vinograda? but what said lord vineyard-GEN ˇ ‘but what did the lord of the vineyard say?’ ZSP 13/2

(186)

aˇste estž plodż pravednika if is fruit righteous-man-GEN ‘if he is the fruit of a righteous man’ PVM 242/23

As in the OCS sample, the share of dative constructions is strikingly large, at 23.4 %, and in the Old Russian sample as well as in the OCS one, this is the centre of gravity of the dative construction: 38.1 % of all dative constructions in the Old Russian sample are found at this node. They are predominantly headed by deverbal nouns; out of the 69 dative occurrences at the ELAB(slot) node, 61 are headed by deverbal nouns, with only scattered occurrences of other head noun types. The distribution is thus, at least at first glance, very similar to that of the GENUNMOD occurrences. 13 of the occurrences have head nouns corresponding to verbs taking da-

ELAB(slot)

109

tive arguments, such as (187), but the great majority of cases have nouns corresponding to regular transitive verbs, as in (188). (187)

cˇ lvkomż vse zˇe to dalż bż. na ugodže all PTC that gave God on satisfaction men-DAT ‘and all that God gave for the satisfaction of men’ PVM 244/30

(188)

stja elena postavila krstż kiparisen na prognanie bˇesom holy Elena placed cross cypress-N on banishing demons-DAT ‘St. Elena put up a cypress cross in order to banish demons’ XID 15v/14

Apart from the deverbal nouns, the only other discernible group is constituted by the six occurrences headed by ruler nouns. (189)

iˇze bˇeaˇse starˇeiˇsina klirikomż cžržkve toja who was head priests-DAT church-GEN that-GEN ‘who was the head of the clergy of that church’ SBG 58/19

When we look at the types of slot fillers, we find that the GENUNMOD and the dative occurrences differ. There are no GENUNMOD occurrences with human slot fillers, and most of the occurrences are abstract, whereas almost half of the dative-marked slot fillers are human (table 24). GENUNMOD DATIVE

abstract 16 18

concrete 6 18

org 0 1

human pl. 0 15

human sg. 0 9

Table 24. Slot fillers in GENUNMOD and dative occurrences at ELAB(slot) by animacy, 11th–14th century Old Russian

GENUNMOD DATIVE

generic or non-specific 13 30

specific (unique) 6 15

specific (discourse) 3 16

Table 25. Slot fillers in GENUNMOD and dative occurrences at ELAB(slot) by specificity, 11th–14th century Old Russian

110 The ELAB nodes: Intrinsic relationships When we look at referential status (table 25), there is no statistically significant difference between the two groups. Both have a propensity for kindreferring or non-specific slot fillers, as in (190),63 but specific ones also occur, as in (189). (190)

Uvy namż, [. . . ] voditelju slˇepyimż woe to-us leader blind-DAT ‘Woe unto us [. . . ], o leader of the blind’ SBG 37/1

Finally, as in the OCS sample, mixed constructions are quite marginal at the ELAB(slot) node, and there are only three such cases. One is a double adjective construction headed by a ruler noun, whereas the two others are genitive + dative constructions, bearing further witness to the affinity of dative and genitive constructions at this node. (191)

vż vremja pereneseniju svjatyima muˇcenikoma Romana i in time moving holy-DAT martyrs-DAT Roman-GEN and Davyda David-GEN ‘at the time of the moving of the holy martyrs Roman and David’ SBG 55/25–26

In sum, then, the overall picture at the ELAB(slot) node is much like what we saw in the OCS sample: The distribution contrasts quite sharply with the ones at the ID nodes, with the adjective constructions marginal and the case constructions central. ADJ1 constructions are few and tend not to have slot fillers that are activated in discourse. ADJ2 is virtually only attested with ruler nouns. Bare genitives are used more freely than at the ID nodes. This tendency is somewhat less clear than in OCS, but still strong enough to justify a claim that at this node we are probably dealing with a formally unrestricted genitive construction. As in the OCS sample, the ELAB(slot) node is the clear centre of gravity of the dative construction, which almost exclusively occurs with deverbal nouns. We find that the dative and GENUNMOD occurrences are very similar at this node, but that the dative construction is both more frequent and more versatile as to the types of slot fillers occurring, particularly when it comes to animacy.

ELAB(slot)

111

40

ADJ1 ADJ2

30




20

Per cent







10









15th cent.

16th cent.

0




11th
14th cent.

17th cent.

Figure 27. ADJ1 and ADJ2 constructions at the ELAB(slot) node, diachronically, per cent

2.3. Further developments in the history of Russian Our point of departure from the OCS and 11th–14th century Old Russian, then, is a situation where ADJ1 constructions are marginal, ADJ2 constructions specialised to the ruler nouns, and the dative construction is strong and has a function not immediately related to typical dative semantics (such as the benefactive readings found in the tenuous examples at the ID node). The genitive occurrences can arguably be seen as occurrences of a formally nonrestricted, but semantically restricted genitive construction: The slot filler can be modified or unmodified, but it has to have its interpretation dictated by the argument structure of the head noun, and animate slot fillers are apparently avoided with bare genitives. Unlike the formally restricted genitive construction at the ID nodes, it cannot reasonably be said to alternate with adjective constructions, except in the case of ruler nouns. Rather, the dative construction appears to be the more suitable alternative with animate slot fillers. When we look at the data diachronically (figure 27), we find no obvious

112 The ELAB nodes: Intrinsic relationships tendency of change for the two adjective constructions. The share of ADJ1 constructions is reasonably stable in all samples, whereas the share of ADJ2 constructions fluctuates from sample to sample; even though the share is clearly smaller in the 17th century sample than in the 11th–14th century sample, the share in the 16th century sample is much larger than both of them.64 The occurrences of ADJ1 constructions remain few in all samples, and they are generally of the same type. Mostly they are headed by deverbal or representation nouns, and the denominal adjective generally belongs to the small class of religious nouns defined on page 82. To the extent that we can judge from such a small number of occurrences, the head nouns also often belong to a small and routinised set; for instance, 4 out of 10 17th century attestations are occurrences of the collocation in (192), and 4 are headed by words meaning “icon” (193). This is the same tendency that we saw at the ID nodes. (192)

i vselisja v nju strax Boˇzij and settled in her fear God-IJ ‘and the fear of God took abode in her’ PJul 104/23, 17th century

(193)

predż obrazom g(o)s(p)odnim before image lord-N’ ˇ ‘before the image of the Lord’ ZAvv 18/38, 17th century

The occurrences of ADJ2 at the ELAB(slot) node remain specialised in all samples; we find that they are almost exclusively headed by ruler nouns, as in (194). In the 17th century sample there are only three exceptions; two occurrences with deverbal nouns (195) and one with a representation noun, which has an ADJ2 construction, although there is a competing type 1 adjective (196). (194)

Daniilu, kostromžskomu protopopu Daniil, Kostroma-SK protopope ˇ ‘to Daniil, the protopope of Kostroma’ ZAvv 23/33, 17th century

(195)

spasenija radi cˇ eloveˇceskago salvation for-sake-of man-SK ‘for the sake of the salvation of men’ PMM 106/35, 17th century

ELAB(slot)

40

GENMOD GENUNMOD DATIVE

113







30







20

Per cent







10








0





11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 28. Genitive and dative constructions at the ELAB(slot) node, diachronically, per cent

(196)

bystž emu glas ot ikony bogorodiˇcny was to-him voice from icon mother-of-god-N ‘a voice came to him from the icon of the Mother of God’ PJul 109/21, 17th century

When it comes to the three major patterns at the ELAB(slot) node, that is to say GENMOD, GENUNMOD and the adnominal dative, there are clearer trends. As we see from figure 28, the dative construction is clearly on the decrease, and the share of dative constructions is significantly lower in the 17th century sample than in the 11th–14th century sample (p-value = 0.0112, Fisher’s exact test, two-sided). Given the statements in the literature, it is actually surprising that we find the dative construction in such late texts at all, but the ELAB(slot) node appears to be its last stronghold. The examples in the late texts are very similar to the ones found in the earlier texts: 11 out of 13 17th century occurrences are headed by deverbal nouns. The examples have a range of different head

114 The ELAB nodes: Intrinsic relationships nouns and slot filler types, and thus appear to be used more freely than e.g. the ADJ1 constructions. They are also found both in religious and narrative texts. (197)

Sice ubo popusti gospodž bog [. . . ] popratelja inoˇceskomu thus for sent lord god despiser monk-SK-DAT cˇ inu rank-DAT ‘For thus the Lord God sent the despiser of the monk rank’ SAP 128/23, 17th century

(198)

na osvjaˇstenie duˇsamż i na otgnanie strastemż i on enlightenment souls-DAT and on banishing passions-DAT and ot razliˇcnyx nedug na iscelenie telesemż on healing bodies-DAT from various illnesses ‘in order to enlighten the souls and banish the passions and heal the bodies from various diseases’ PMM 111/16, 17th century

One might expect that the two groups of genitive occurrences would become more equal in size during the period under consideration in this book. However, figure 28 shows that this claim cannot be made on the basis of the text samples used. Instead, we see that the share of GENUNMOD occurrences is relatively stable throughout the period under consideration, but that the share of GENMOD occurrences is clearly on the increase.65 Thus, the picture is similar to what we saw at the ID nodes: We do not so much see a clear expansion of the genitive as a gradual weakening of its competitors, and in the case of the ELAB(slot) node, this means the dative construction. Nor do we see any changes in the types of slot fillers found with the GENUNMOD occurrences: Out of the 14 occurrences in the 17th century sample, only two have human slot fillers: example (199) and another example of the same collocation from the same text. In these two examples the slot filler is human, but plural and has non-specific reference. The remaining examples have inanimate slot fillers of various kinds (200). (199)

ot boga poslanii vo obraze inok from god sent in image monks-GEN ‘sent by God in the shape of monks’ PMM 110/26, 17th century

(200)

i ne trebovaˇse vody ej na omovenie rukż and not demanded water for-her on washing hands-GEN

ELAB(part)

115

podajuˇstago from-giver ‘and she did not demand water for her from the giver for washing her hands’ PJul 105/27 17th century Finally, we may note that whereas mixed constructions are marginal in the 11th–14th century sample, there are 13 examples in the 17th century sample, all of them adjective + genitive constructions. They are quite uniform; all save one are headed by ruler nouns and are the same as or similar to example (201). (201)

patriarxż moskovskii i vsea Rosii patriarch Moscow-SK and all-GEN Russia-GEN ‘the Patriarch of Moscow and all Russia’ DG 196/10, 17th century

To sum up, then, at the ELAB(slot) node, the main change witnessed in the period under consideration is the demise of the dative construction, which nevertheless keeps this node as its final stronghold when it is virtually ousted everywhere else. The overall share of genitive constructions increases, but there is no obvious change to the proportion of GENUNMOD occurrences, or to the types of slot fillers they can have.

3. ELAB(part) At the ELAB(part) node we find constructions headed by a noun denoting an inherent part of a whole. The nouns have a slot for the whole of which the head noun is a part, and a nominal fills this slot, but does not serve as a reference point.

3.1. ELAB(part) in OCS and 11th–14th century Old Russian The OCS and 11th–14th century Old Russian data seen in figure 29 confirm the consensus in the literature that genitives are used freely with partitive meanings: In the overwhelming majority of cases, the slot is filled by a genitive-marked NP, and there is no formal restriction on that NP. There is a considerably larger share of GENUNMOD than GENMOD occurrences at ELAB(part) in the OCS sample, and fairly equal shares in the Old Russian

50

116 The ELAB nodes: Intrinsic relationships

0

10

20

30

40

OCS OR

ADJ1

ADJ2

DATIVE

GENMOD

GENUNMOD

genitive/dative

mixed

Figure 29. Constructions at the ELAB(part) node in OCS (n = 70) and 11th–14th century Old Russian (n = 62), per cent

sample. Clearly there is no restriction on the use of bare genitives in either sample. (202)

kosno˛ se˛ vžskrilii rizy ego touched REFL fringe garment-GEN he-GEN hˆepsato tou kraspedou tou imatiou autou ‘she touched the fringe of his garment’ Mar. Luke 8:44, OCS

(203)

svoego vse˛ dni zˇivota all days life-GEN his-GEN ˇ 13/18, 11th–14th century Old Russian ‘all the days of his life’ ZSP

(204)

vˇese˛ i do vrżxu gory lead him to top mountain-GEN

ELAB(part)

117

eˆ gagon auton heˆos ophruos tou orous ‘they lead him to the top of the mountain’ Mar. Luke 4:29, OCS (205)

ob onż polż goroda at that half town-GEN ‘at the other part of the town’ PVrL 109/4, 11th–14th century Old Russian

All other constructions are marginal at this node. In the OCS sample there are three ADJ2 occurrences, such as (206), where it does not seem reasonable to posit a TYPE reading. None are found in the Old Russian sample. (206)

imżˇsa za obˇe polˇe rizžnˇei having-grasped at both halves robe-N sphiggonta to rhˆegma tou kolobiou ‘grasping the two parts of the robe (Greek: the rift of the robe)’ Supr. 16:187/7, OCS

Both samples have a small number of dative occurrences at ELAB(part), two in the Old Russian sample and seven in the OCS sample. In most cases, the head nouns in these constructions have a double existence as deverbal (207). We recall from section 2 that most dative constructions at ELAB(slot) are headed by deverbal nouns. The ELAB(part) occurrences thus appear to fit in with the predominant pattern for the dative construction. (207)

doide konžca zˇitiju reached end life-DAT efthasen tˆen hesperan tou telous66 ‘he reached the end of his life’ Supr. 16:208/4

3.2. Further developments in the history of Russian As figure 30 shows, to the extent that the distribution at ELAB(part) undergoes changes during the period under consideration in this book, they amount to an increase in the share of GENUNMOD occurrences. We also see that the dative construction is not attested in the later samples.

118 The ELAB nodes: Intrinsic relationships

60

GENMOD GENUNMOD DATIVE










40







20

Per cent







0




11th
14th cent.










15th cent.

16th cent.

17th cent.

Figure 30. Constructions at the ELAB(part) node, diachronically

4. ELAB As we recall from section 4.5 in chapter 2, at the ELAB node we find cases of possessive-like constructions that neither involve reference points nor relational head nouns, nor serve as a whole to denote a type or constitute a conventionalised name or label for a particular referent (cf. page 93). For these constructions we posit only an intrinsic relationship between head and modifier.

4.1. ELAB in OCS and 11th–14th century Old Russian As figure 31 shows, the distribution of constructions at the ELAB node is rather similar to what we found at the ELAB(slot) node, and different from the distributions at the ID nodes, both in the OCS and the 11th–14th century Old Russian sample. In particular, the near-absence of ADJ1 constructions is striking.

ELAB

119

0

10

20

30

40

50

OCS OR

ADJ1

ADJ2

DATIVE

GENMOD

GENUNMOD

genitive/dative

mixed

Figure 31. Constructions at the ELAB node, OCS (n = 173) and 11th–14th century Old Russian (n = 229), per cent

In the Old Russian sample, there are no occurrences of ADJ1 constructions at the ELAB node. In the OCS sample, there are five possible examples. Three of these are patronymics without a kinship term, as in (208), which might as well belong to the LABEL node. (208)

simone ioninż Simon John-IN Simon Iˆoannou ‘Simon, son of John’ Zogr. John 21:17

However, we also find (209), where Judah does not seem to be a reference point, and (210), which has an adjective not formed from a human noun, as well as a purpose reading.

120 The ELAB nodes: Intrinsic relationships (209)

vż gradż ijudovż in city Judah-OV eis polin Iouda ‘to a city of Judah’ Mar. Luke 1:39

(210)

pokaˇzeˇ te mi skžle˛ zž kinosovy show me money tax-OV epideixate moi to nomisma tou kˆensou ‘show me the money for the tax’ Savv. Matthew 22:19

We may safely conclude, then, that the ADJ1 construction is quite marginal at the ELAB node. In both samples, the ADJ2 constructions constitute the largest group at the ELAB node; these account for 59.5 % (103) of the OCS occurrences and 34.9 % (80) of the Old Russian occurrences. In both samples these constructions can be divided into three clear types. The first is the appositional type, where the head and the adjective denote the same referent, literally, as in (211), or metaphorically, as in (212). There are 34 such examples in the OCS sample and 24 in the Old Russian sample. (211)

vż rimžstˇe gradˇe in Rome-SK city epi tˆei polei Rˆomˆei ‘in the city of Rome’ Supr. 16:186/10–11, OCS

(212)

vvergżˇse i v propastž smrtnuju having-thrown him in abyss death-N ‘having thrown him into the abyss of death’ PVrL 175/3

The second is the locative type, where the adjective denotes the location of the head, as in (213) and (214). There are 53 occurrences of this type in the OCS sample and 16 in the Old Russian one. (213)

otż nazarefa galileiskago from Nazareth Galilee-SK apo Nazareth tˆes Galilaias ‘from Nazareth in Galilee’ Savv. Matthew 21:11, OCS

(214)

estž znamenže nbsnago Ba heaven-N god is sign ‘it is a sign of God in heaven’ PVrL 179/19

ELAB

121

The third is the origin type, where the adjective denotes the place of origin of the head, as in (215) and (216). This type is clearly closely related to the ELAB(slot) occurrences headed by ruler nouns, and generally expressed in the same way. There are 13 occurrences of the origin type in the OCS sample and 33 in the Old Russian one. (215)

sż isomż galileiskyimż with Jesus Galilee-SK meta Iˆesou tou Galilaiou ‘with Jesus of Galilee’ Ass. Matthew 26:69, OCS

(216)

k velikomu cˇ judotvorcu Nikole Korsunskomu to great miracle-performer Nikola Korsun-SK ‘to the great miracle-performer Nikola of Korsun’ PBR 358/15

There are only scattered occurrences of other relation types. It should be noted that ADJ2 occurrences from all these three types, especially when they involve names, often appear to be conventionalised units, and thus are close to the examples classified at the LABEL node, as we shall see in section 2, chapter 6. The discrepancy between the OCS and Old Russian samples when it comes to the share of ADJ2 constructions at the ELAB node is due to the large shares of locative and appositional occurrences in the OCS sample. This is probably due to subject matter rather than to a real difference between the two languages; the OCS texts contain a large number of foreign toponyms and needs to clarify whether they are cities, rivers, districts etc., as in (211), and where they are, as in (213), and therefore resort to this type of “complex toponyms”. GENMOD is the second largest group at the ELAB node in both languages, constituting 19.7 % (34 occurrences) in the OCS sample and 28 % (64 occurrences) in the Old Russian sample. A large range of relations between heads and genitive-marked nouns are found in both samples, but the three clearest groups are the origin type, the appositional type and a group of relations akin to partitives: constructions with genitive-marked NPs denoting wholes, material, constituent parts, content or source. In the Old Russian sample, the origin type is the most common, with 27 occurrences, whereas there are only 4 in the OCS sample. Thus, at least in Old Russian, the GENMOD occurrences appear to have a fairly clean division of labour with the ADJ2 constructions with this type of semantics: If the place of origin is not modified in any way, the ADJ2 construction is chosen; if

122 The ELAB nodes: Intrinsic relationships it is complex, the choice is normally the genitive. Thus it appears that the GENMOD occurrences here are instances of the formally restricted genitive construction. (217)

kž ovcamż pogybżˇsiimż domu izlva to sheep lost house-GEN Israel-OV-GEN pros ta probata ta apolˆolota oikou Israˆel ‘to the lost sheep of the house of Israel’ Ass. Matthew 10:6, OCS

(218)

se estž novyi Koste˛ ntinż velikogo Rima behold is new Constantine great-GEN Rome-GEN ‘behold, he is a new Constantine of the great Rome’ PVrL 130/31

The appositional type is common in both samples (9 occurrences in the OCS sample, 12 occurrences in the Old Russian one). Unlike the ADJ2 appositionals, these occurrences rarely involve toponyms, but are often abstract metaphoric expressions, as in (219), where the rock is clearly a metaphorical expression of the faith, or plainer appositions, as in (220), where the deed clearly is coreferential with the murder. (219)

na kameni. vˇery gospodžne˛ on stone faith-GEN lord-N’ en tˆei petrai tˆes pisteˆos tou kuriou ‘on the rock of the faith in the Lord’ Supr. 16:208/8, OCS

(220)

aˇste bo dosžde ostavlju dˇelo ubiistva moego if for now renounce deed murder-GEN my-GEN ‘for if I now renounce the deed of my murder’ SBG 38/18

The partitive-like type accounts for most of the remaining occurrences (15 in the Old Russian sample and 13 in the OCS one). Most of these occurrences have genitives denoting wholes, as in example (221), where the Tatar force is the whole to which the killed men belong. There are also several occurrences with the genitive denoting constituent parts, as in (222). (221)

vidja svoeja sily tatarskyja mnoˇzestvo pobženyx seeing self-GEN force-GEN Tatar-GEN multitude killed-GEN ‘seeing a multitude of killed men from his Tatar force’ PBN 348/27

ELAB

(222)

123

i bˇe narodż mżnogż mytarž i inˇexż and was crowd big tax-collectors-GEN and others-GEN kai eˆ n okhlos polus telˆonˆon kai allˆon ‘and there was a large crowd of tax collectors and others’ Mar. Luke 5:29, OCS

Of the scattered remaining occurrences, various types of relationships hold between the genitive-marked element and the head noun, for instance location, purpose and description, such as in (223). (223)

vžsˇekż mladenecż mo˛ zˇżska polu every infant male-GEN sex-GEN 67 pan arsen ‘any infant of the male sex’ Mar. Luke 2:23, OCS

GENUNMOD occurrences make up a larger group at the ELAB node in the Old Russian sample than in the OCS one, constituting 17.9 % (41) of all 11th–14th century Old Russian ELAB occurrences, but only 8.1 % (14) in the OCS sample. When we look at the relation types involved, we see that the situation differs clearly from the ELAB GENMOD occurrences, in that the origin type is missing entirely in both samples, thus supporting the hypothesis that the GENMOD origin occurrences are instances of the formally restricted genitive construction and that they are more or less in complementary distribution with the ADJ2 construction. In the OCS sample, the partitive-like type is the most frequent one (8 occurrences), as in (224). (224)

i razido˛ tż se˛ ovecž stada and will-scatter REFL sheep flock-GEN kai diaskorpisthˆesontai ta probata tˆes poimnˆes ‘and the sheep of the flock will be scattered’ Ass. Matthew 26:31, OCS

In the Old Russian sample, on the other hand, the appositional type appears to be very frequent, and is found in 27 of the occurrences (against two in the OCS sample). Several of these occurrences are of the metaphorical type seen in example (219), for instance (225). 21 of the examples, however, are occurrences of the collocation in (226), or ones very like it, in the Russkaja

124 The ELAB nodes: Intrinsic relationships Pravda. This group of examples accounts for the discrepancy between the two language samples. (225)

oblecyse˛ vż brone˛ pravdy dress-self in armour truth-GEN ˇ ‘dress yourself in the armour of truth’ ZSP 15/14

(226)

.g. grvne. prodaˇze three grivna fine-GEN ‘three grivna as fine’ RP 53/912–913

Both samples also have examples of the emphatic construction seen in (227) and (228) (227)

i zˇivi v vˇeky vˇeka and live in ages age-GEN ‘and live in all eternity’ PVM 242/15

(228)

i bo˛ detż na xvalo˛ xvalenija and will-be on praise praise-GEN kai estai eis epainon kaukhˆeseˆos ‘and it will be to the praise of praise’ Supr. 16:194/18, OCS

There is a sizeable share of dative constructions classified as ELAB occurrences in both samples, 6.9 % (12) in the OCS sample and 17.9 % (41) in the Old Russian sample, but many of them have typical dative semantics, such as benefactive, addressee/recipient and purpose.68 This is particularly clear in the Old Russian sample, where 30 out of the 41 occurrences have such semantics (4 out of 12 in the OCS sample), and many of these have a rather uncertain interpretation (229). Consequently, there is only a relatively small share of the dative occurrences that are actually in competition with genitive and adjective constructions. When we look at the remaining examples in the OCS sample, they are rather scattered. The two recurring types are the appositional type, as in (230), and occurrences where the relation between head and dative-marked noun is a temporal one. The latter type could arguably be counted as ID occurrences instead. The sample contains four such occurrences, all similar to (231). (229)

poloˇziˇsa trupžja rabż tvoixż braˇsno pticam put corpses servants-GEN your-GEN food birds-DAT

ELAB

125

nbsnym heaven-N.DAT ‘they laid out the corpses of your servants as food for the birds in the sky’ SL 463/35 (230)

prˇezˇde zˇe prazdžnika pascˇe before PTC feast passover-DAT Pro de tˆes heortˆes tou paskha ‘Now before the feast of the Passover’ Zogr. John 13:1, OCS

(231)

iˇze bˇe arxierei lˇetu tomu who was high-priest year-DAT that-DAT hos eˆ n arkhiereus tou eniautou ekeinou ‘who was the high priest of that year’ Zogr. John 18:13, OCS

In the Old Russian sample, on the other hand, there appears to be only one dative construction usage at the ELAB node that is really in competition with other constructions, namely the emphatic construction (10 occurrences), as seen in (232). There is one such occurrence in the OCS sample (233), which is also the most frequent collocation in the 11th–14th century Old Russian sample. In these occurrences, the dative construction clearly competes with the genitive, cf. examples (227) and (228). (232)

Po istinˇe vy cesarja cesaremż i knjazja kżnjazemż in truth you emperors emperors-DAT and princes princes-DAT ‘In truth you are the emperors of emperors and princes of princes’ SBG 49/16

(233)

vž vˇeky vˇekomż in ages ages-DAT eis tous aiˆonas tˆon aiˆonˆon ‘in all eternity’ Supr. 16:209/19–20, OCS

Finally, the OCS sample has four occurrences of adjective + adjective constructions at the ELAB node, all involving type 2 adjectives. All are paratactic and fit the pattern of the ADJ2 occurrences in general; they belong to the appositional, locative and partitive-like types. (234)

otż vżsˇekoje˛ vžsi galileisky i ijudeisky from every village Galilee-SK and Judea-SK

126 The ELAB nodes: Intrinsic relationships ek pasˆes kˆomˆes tˆes Galilaias kai Ioudaias ‘from every village in Galilee and Judea’ Mar. Luke 5:17, OCS (235)

nesy sżmˇesˇenie zmżrno i olżguino carrying mixture myrrh-N and aloe-N pherˆon migma smurnˆes kai aloˆes ‘bringing a mixture of myrrh and aloe’ Zogr. John 19:39, OCS

To sum up, then, we find that the ADJ1 construction is virtually absent from the ELAB node in OCS and 11th–14th century Old Russian. The ADJ2 and GENMOD occurrences appear to have a rather clean division of labour when the relation between head and modifier is one of origin, and many of these occurrences also tend towards being conventionalised as complex names. The dative construction mostly has benefactive, recipient or purpose semantics at this node, and does not really compete with the other constructions, except for in the emphatic constructions seen in examples (232) and (233). The dative occurrences at the ELAB node are possibly a more real competitor in OCS than in 11th–14th century Old Russian. Apart from the origin type, the genitive appears to be used freely, perhaps especially when the construction is of the appositional type.

4.2. Further developments in the history of Russian When we look at figures 32 and 33 and table 26, we see few obvious trends of development at the ELAB node.

11th–14th century 15th century 16th century 17th century

ADJ1 0 0 0 1

ADJ2 80 30 20 40

DATIVE 41 8 5 6

GENMOD 64 39 7 44

GENUNMOD 41 12 6 17

mixed 0 2 1 1

Table 26. Constructions at the ELAB node, diachronically

It is clear that the share of dative occurrences decreases steadily, but apart from that, the shares are stable or fluctuate unpredictably. When we look at the semantic subtypes, we also find little change. The ADJ1 construction is only attested in one single benefactive 17th century occurrence.

50

ELAB ADJ1 ADJ2

127

40








10

20

Per cent

30










11th
14th cent.

15th cent.

16th cent.

0







17th cent.

Figure 32. ADJ1 and ADJ2 constructions at the ELAB node, diachronically, per cent

(236)

v slavu boˇziju in glory God-IJ ‘for the glory of God’ PMM 108/19–20, 17th century

The ADJ2 construction remains attested in the three main types identified in the 11th–14th century Old Russian sample: the origin type (237), the locative type (238) and the appositional type (239). In the 17th century sample, this construction it is nearly exclusively attested with adjectives formed from toponyms in all three types, and most examples tend toward being conventionalised complex names. (237)

jako i Stefana Permžskago as and Stefan Perm-SK ˇ ‘as also Stefan of Perm’ ZAvv 21/16, 17th century

(238)

v kadomskuju votˇciny to Kadom-SK property ‘to the property in Kadom’ DG 194/4, 17th century

50

128 The ELAB nodes: Intrinsic relationships GENMOD GENUNMOD DATIVE

40




30




20

Per cent














10










0




11th
14th cent.

15th cent.

16th cent.

17th cent.

Figure 33. Genitive and dative constructions at the ELAB node, diachronically, per cent

(239)

v Muromstem grade in Murom-SK city ‘in the city of Murom’ PMM 111/23, 17th century

In addition, there are still attestations of the abstract metaphorical subtype of the appositionals. (240)

svˇetom razumnymż prosveˇstajutsja light understanding-N enlighten-REFL ‘they are enlightened by the light of understanding’ PMM 111/33– 34, 17th century

The GENMOD group also retains much the same pattern as we saw in the 11th–14th century sample. There is still a clear division of labour with the ADJ2 construction when it comes to the origin type.

ELAB

(241)

129

slobody xolopż tvoi Ogorodnoi servant your Ogorod-N-GEN village-GEN ˇ 2–3, 17th century ‘your servant of the village of Ogorod’ CL

Most of the remaining occurrences in the 17th century sample are of the partitive-like type, whereas the appositional type is rare (2 out of 17). The 17th century GENUNMOD occurrences are more uniform than the 11th–14th century ones. Most of them are partitive or partitive-related, and they are mostly found in occurrences like (242). We also, however, find one locative and two appositional occurrences, as in the metaphorical (243), where the darkness and the ignorance are coreferential. (242)

on [. . . ] dastž tebe denegż sto rublevż he will-give you money-GEN hundred roubles ‘he will give you a hundred roubles of money’ PKS 66/29, 17th century

(243)

Ne k tomu bo oˇci myslennii nevˇedenija mrakom not to that for eyes thought-N ignorance-GEN darkness pokryvajuca cover-REFL ‘For the mind’s eyes are not covered by the darkness of ignorance to that’ PMM 111/33

The 17th century sample has only six dative occurrences, and two of these are found in the same quotation from Romans 9:33, where the relation is probably one of purpose.69 (244)

Se polagaju v Sionˇe kamen(ž) pretykaniju i kamenž behold lay in Sion stone stumbling-DAT and stone soblaznu offence-DAT ˇ ‘Behold, I lay in Sion a stumblingstone and a rock of offence’ ZAvv 28/8–9, 17th century

The remaining occurrences have clear benefactive readings, except one single example of the emphatic type.

130 The ELAB nodes: Intrinsic relationships (245)

vo vˇeki vekom in ages ages-DAT ‘in all eternity’ PMM 111/44, 17th century

5. Conclusions To conclude, both at the ELAB(slot) and the ELAB node we find a picture that is quite different from that at the two ID nodes. When there is no reference point reading, the ADJ1 construction is marginal both in the OCS and 11th– 14th century samples, and the genitive is dominant and more freely used. In both samples, there appears to be a near complementary distribution between the ADJ2 construction on one side and GENMOD occurrences the relation expressed is one of origin, and there is a similar (but much weaker) tendency with constructions headed by ruler nouns. These two types are obviously closely related. Apart from these occurrences, this appears to be the domain of the formally non-restricted genitive constructions. The ELAB(slot) node is the centre of gravity of the dative construction in the two early samples. In both, the primary function of the adnominal dative is to encode objects of deverbal nouns, and to some extent other object-like argument slots, such as those of ruler nouns. We also note that although the dative construction decreases steadily during the period under consideration in this book, the ELAB(slot) node is its last stronghold, and a fair number of dative occurrences are still found in 17th century texts. At the ELAB node, the dative mostly has typical dative semantics, encoding benefactive, recipient and purpose meanings. At the ELAB(part) node, genitive constructions dominate in all samples, and there is no change of importance to be detected.

Chapter 6 Types and conventionalised units: TYPE and LABEL

As we recall from the map of the possessive semantic space, the TYPE and LABEL nodes are set well apart from the rest of the nodes, and also well apart from each other. Nevertheless, these two nodes have much in common: Together, they constitute the main domain of the ADJ2 construction in all samples. Furthermore, it is characteristic for constructions at both nodes that the whole is more important than the constituent parts. The constituent parts may or may not be referential in themselves, but a TYPE or LABEL construction will serve as a whole to name a type or a referent. 1. TYPE As mentioned on page 21, Taylor (1996:293) expands the reference point analysis by subsuming cases where the possessor is not a discourse referent, but denotes a type or kind, and where it does not serve to identify a referent or group of referents, but picks out a subcategory of the head noun’s category instead. In these cases the resulting construction does not involve reference to two discourse referents, and the construction is not a grounding device, it just creates a “complex noun” which is then available for grounding and further modification. To put it in the terms of Gundel, Hedberg and Zacharski (1993), the reference point or “possessor” nominal does no more than render the whole complex construction type identifiable,70 and assign it to a smaller category than the head noun does alone, whereas the reference point/possessor in an ID construction anchors the head noun, making a particular instance uniquely identifiable. For English, there are good diagnostics (cf. Taylor 1996:288–291) to separate “real” possessives from possessive compounds, but most of these are based on the fixed word order of English, making them inapplicable to the early Slavic data of this book. The classification therefore has to rely on the semantics alone. On the present analysis, occurrences classified as belonging to the TYPE node must fulfill the following three criteria:

132 Types and conventionalised units: TYPE and LABEL 1. The possessor nominal cannot be more than type identifiable 2. The contribution of the possessor nominal cannot be more than to make the construction as a whole type identifiable 3. The possessor nominal must assign the whole construction to a category that is smaller than (or in some cases entirely different from) that of the head noun Naturally, it is not always easy to distinguish between ID and TYPE occurrences, since the ADJ2 construction is clearly present at both nodes. The number of constructions judged as having uncertain semantics is therefore quite high just in the transitional zone between ID and TYPE constructions, and this will be dealt with in section 1.2.

1.1. TYPE in OCS and 11th–14th century Old Russian At the TYPE node we find almost exclusively ADJ2 constructions both in the OCS and the 11th–14th century Old Russian sample. 97.8 % (46) of the TYPE occurrences in the OCS sample are ADJ2 constructions, and 98.2 % (218) of those in the Old Russian sample. The adjective in all cases is formed from a type identifiable noun, and does not refer to any particular referent in the discourse. The entire construction denotes a type that is narrower than (or different from) that of the head noun on its own. A trading house (246) is a narrower subtype of house, whereas a sheepfold (247) is a narrower type than dvorż “court”, and probably not even its subtype. (246)

ne tvorite domu oca moego. domu kuplžnaego not make house father-GEN my-GEN house trade-N mˆe poieite ton oikon tou patros mou oikon emporiou ‘do not make my father’s house a trading house’ Zogr. John 2:16, OCS

(247)

vż dvorż ovžˇcii in court sheep-IJ eis tˆen aulˆen tˆon probatˆon ‘into the sheepfold’ Zogr. John 10:1, OCS

The relationship between head and modifier may be of many kinds; often the adjective denotes the function or purpose of the head noun, as in (246) and (247), or, frequently, the material of the head noun, as in (248).

TYPE

(248)

133

bljudo serebržno plate silver-N ‘a silver plate’ MG 33/16

There are many other relation types to be found. When the head noun is deverbal, the modifier may, for instance, classify the type of action by its time or location of occurrence, both seen in (249) or its subject, as in (250). (249)

straˇza noˇstnaja morskaja guard night-SK sea-SK ˇ ‘night sea guard’ ZAN 164/8–9

(250)

i paky pˇenija angelžskaja slyˇsaaxu [. . . ] gostie and again singings angel-SK heard strangers ‘and again strangers heard angel song’ SBG 44/1

Types of realisation other than ADJ2 are very rare in both samples. In the OCS sample there is only one single example.71 In (251), the Greek noun prosthˆesis may reasonably be translated as “presentation”, as may the OCS prˇedżloˇzenie, and hence the bread appears to be sorted into a subcategory by its function, “presentation bread”. Nevertheless the modifier is a bare genitive noun, not a type 2 adjective.72 (251)

xlˇeby prˇedżloˇzeniˇe prije˛ i eˇ stż breads presentation-GEN took and ate tous artous tˆes prostheseˆos labˆon ephagen ‘he took and ate the show bread’ Mar. Luke 6:4, OCS

In the Old Russian sample, there is one single example of an ADJ1 construction with an -žn’- suffix, which should arguably be classified at the TYPE node: (252)

a zadniˇca jei mouˇzne˛ ne nadobˇe but inheritance to-her husband-N’- not necessary ‘but the husband’s inheritance is not necessary to her’ RP 46/659– 660

The husband’s inheritance may be interpreted as a conventionalised type of inheritance and a legal term. This could also just be a case where the possessor is a short-term discourse referent (non-specific), in which case this should be counted as a regular ID(slot) or possibly ELAB(slot) occurrence.

134 Types and conventionalised units: TYPE and LABEL There are also three occurrences of mixed constructions, all paratactic double adjective constructions with type 2 adjectives. (253)

oˇze dubż peretnetž znamenny. ili meˇznyi if oak hews sign-N or border-N ‘if someone hews up a sign or border oak’ RP 44/596–599

1.2. Borderline cases In the 11th–14th century Old Russian sample, a large number of ADJ2 occurrences were deemed to be ambiguous, many of them between a TYPE and an ID/ID(slot) reading. This is less of a problem in the OCS sample, where the Greek original can often be used to settle which of the readings was the intended one. Typically, the examples have type 2 adjectives formed from a noun that is also used for a discourse-accessible referent, and it is not clear whether the author is actually picking up the discourse referent or just referring to the complex type. In example (254) there is a church accessible in the context, but it is not clear whether the adjective crkvnyja picks up this particular church or just serves to classify the door as a member of the subtype “church door”. (254)

Tatarove zˇe siloju otvoriˇsa dveri crkvnyja Tatars PTC with-force opened doors church-N ‘But the Tatars opened the church(’s) doors’ SL 463/22

In (255) the head noun may mean either “woman” or “wife”, and it is not clear whether the adjective refers to the women’s actual husbands or just subclassifies the women as Boyar women. (255)

i zˇeny bole˛ rskye˛ muˇzei svoixż i ospodarevż ostali and women Boyar-SK husbands their and masters lost ‘and the Boyars’ wives / the Boyar women lost their husbands and masters’ Z 32/257

1.3. Further developments in the history of Russian In the further history of Russian, the distribution of the constructions at the TYPE node is stable; the ADJ2 construction is virtually the only choice in all

LABEL

135

samples.73 This is not surprising, since this construction is virtually the only possible choice in modern Russian. In the 17th century sample, typical examples are various military and civilian ranks. (256)

strˇeleckoi golova strelets-SK head ˇ ‘strelets commander’ CBK 1:18, 17th century

(257)

dumnoi dvorjaninż duma-N nobleman ‘duma nobleman’ DG 192/2, 17th century

In addition to the ADJ2 constructions, there are only sporadic attestations of ADJ1 and mixed adjective occurrences. For instance, gosudarevż dolgż is a type of tax, not a particular debt to a particular sovereign. (258)

dolgi a budetż kakie gsdrvy and will-be some sovereign-OV debts ‘and there will be some sovereign’s debts’ DG 194/17, 17th century

2. LABEL We saw in chapter 2, section 4.3 that the constructions at the LABEL node are similar to the ones at the TYPE node in that the modifier does not serve to identify a particular instance of what the head noun denotes; instead, the construction as a whole names something. At the TYPE node, that “something” is a type itself; at the LABEL node this is not necessarily so. As stated on page 132, to be classified at TYPE, the possessor nominal has to be no more than type identifiable, and serve to do no more than render the whole construction type identifiable. At the LABEL node, on the other hand, the possessor nominals are often uniquely identifiable, and the whole construction may be so, too. The occurrences also differ from ID occurrences, in the following ways (repeated from page 22): 1. The possessor does not need to be cognitively accessible, which indicates that it is not entirely referential (you can talk about Occam’s razor without knowing anything about William of Ockham) 2. The expressions are conventionalised to the extent that they are often

136 Types and conventionalised units: TYPE and LABEL listed in dictionaries (Parkinson’s disease) or serve as complex proper nouns 3. At least in English, there is sometimes clear syntactic evidence that such constructions occur as complex nouns, not as NPs (the usual Hobson’s choice), and semantically it is clear that in some of the cases the possessor nominal contributes to type identifiability only As seen in chapter 2, the LABEL constructions fall into three main groups. The first group has a uniquely identifiable possessor which picks out a subtype. This group is very close to the TYPE constructions, since the construction as a whole serves to name a subtype of the head noun’s type, as in Hobson’s choice, or a different type entirely, as in Adam’s apple. The second group has a kind-referring or type identifiable possessor, but the entire construction names something uniquely identifiable. This is typically found in the early Slavic country name construction, which appears to be a productive partially lexically specific construction in its own right. In the early Slavic material, we see this especially in constructions naming countries, languages (259) and religions, where the possessor nominal is the name of an ethnic or religious group, and the head noun is the word for “country”, “tongue” or “faith”. The third group has a uniquely identifiable possessor that picks out a uniquely identifiable referent. It is thus very close to the ID nodes, but differs from them in that the entire construction serves as a name for the referent in question, and in that the possessor nominal needs not be cognitively accessible, such as in Halley’s comet or St. Basil’s Cathedral. (259)

ne razumˇeemż ni Greˇcsku jazyku. ni Latynžsku not understand neither Greek-SK language nor Latin-SK ‘we do not understand neither the Greek nor the Latin language’ PVrL 26/10–11

2.1. OCS At the LABEL node in the OCS sample, we see immediately that, as was the case at the TYPE node, the predominant construction is the ADJ2 construction, accounting for 83 % (88) of all the constructions. However, the distribution of construction types varies considerably between the three main LABEL subtypes identified above. There are only 10 cases where the possessor nominal is a uniquely identifi-

137

0

20

40

60

80

LABEL

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

Figure 34. Constructions at the LABEL node, OCS, per cent, n=106

able possessor nominal and the construction as a whole denotes a type. Three of them are ADJ1 constructions and seven are ADJ2 occurrences. There are five examples like (260), two like (261), all ADJ2, and three like (262). In these three cases, the possessor nominal appears to serve as a constant epithet to the noun in question; “birds of the sky” just means birds, and “God’s churches” means no more than “churches”. (260)

ptice˛ nebskyje˛ birds heaven-SK ta peteina tou ouranou ‘the birds of the sky’ Mar. Luke 9:58

(261)

niktoˇze ne vˇestž ni agli nbscii nobody not know not angels heaven-SK oudeis oiden. oude hoi aggeloi tˆon ouranˆon ‘no one knows, not even the angels of heaven’ Savv. Matthew 24:36

138 Types and conventionalised units: TYPE and LABEL (262)

na sve˛ tye˛ boˇzie˛ crżkvi on holy god-IJ churches kata tˆon hagiˆon tou theou ekklˆesiˆon ‘against the holy churches of God’ Supr. 16:190/21–22

There are 71 cases where the possessor nominal is type identifiable or kind-referring, while the whole construction denotes a uniquely identifiable referent. As many as 68 of these are ADJ2 occurrences, which fits our observations on the reference of type 2 adjectives so far, and almost all of these again (64) are occurrences of “the son of man”, uniquely identifiable as Jesus (or the Messiah).74 (263)

snż cˇ sky son man-SK ho huios tou anthrˆopou ‘the son of man’ Zogr. John 6:27

The remaining four occurrences are instances of the pattern where the possessor nominal has kind reference to an ethnic group and the entire construction names a country or a city. (264)

na zemljo˛ gadarinżsko˛ [jo˛ ] in country Gadarene-SK eis tˆen khˆoran tˆon Gerasˆenˆon ‘to the country of the Gadarenes/Gerasenes’75 Mar. Luke 8:26

There are two ADJ1 occurrences, both of the same pattern as (264), but with an adjective formed from the collective noun Izdrailž rather than corresponding to a plural noun.76 (265)

vż zemljo˛ izlvo˛ in land Israel-OV eis gˆen Israˆel ‘to the land of Israel’ Ass. Matthew 2:20

Finally, there is one single dative construction of this type. (266)

tˇemžˇze nareˇce se˛ selo to selo kržvi by-this called REFL field that field blood-DAT

LABEL

139

dio eklˆethˆe ho agros ekeinos agros haimatos ‘therefore that field has been called the Field of Blood’ Savv. Matthew 27:8 The two first types of LABEL constructions are both close to TYPE constructions; they either denote types or have a type-level possessor nominal. This is not the case for group three, where the possessor nominal is uniquely identifiable, and where the whole construction also denotes a uniquely identifiable referent. Since they are essentially ID constructions conventionalised as names, it is not surprising to find that this group of occurrences has a distribution that resembles those at the ID nodes more than the two first groups do. There are 25 such occurrences in the OCS sample. 10 of these are ADJ1 constructions, 13 are ADJ2 constructions and 2 belong to the GENMOD group. All the ADJ1 occurrences are established toponyms, such as (267) and (268), and the possessor nominals in all cases refer to uniquely identifiable human beings which are not activated in the context. (267)

vż gradż konžstatinž in city Constantine-J epi tˆen Kˆonstantinopolin ‘to Constantinople’ Supr. 16:191/7

(268)

klade˛ zž iakovlž well Jacob-J pˆegˆe tou Iakˆob ‘Jacob’s well’ Zogr. John 4:6

The same is the case for the two GENMOD occurrences, which are both church names. (269)

pržvomo˛ cˇ enika stefana vż cżrkżvi staago in church holy-GEN first-martyr-GEN Stephanos-GEN en tˆoi marturiˆoi tou agiou prˆotomarturos Stephanou ‘to the church of the holy first martyr Stephanos’ Supr. 16:209/8–9

The ADJ2 occurrences are more of a mixed bag. The adjectives are mostly not formed from nouns denoting singular human beings, except the three occurrences referring to Arianism, such as (270). Note that in all three cases, the Greek has a denominal adjective as well.

140 Types and conventionalised units: TYPE and LABEL (270)

arijanżskaja xula Areios-SK blasphemy he Areianikˆe blasphˆemia ‘Areios’ blasphemy / the Arian blasphemy’ Supr. 16:197/20–21

In the other examples, the possessor noun is a toponym or some other inanimate with unique reference, and the whole construction serves as an established name for something, e.g. a historical period, as in (271). (271)

po prˇeselenii zˇe vavilonstˇe after deportation PTC Babylon-SK meta de tˆen metoikesian Babulˆonos ‘after the deportation to Babylon’ Ass. Matthew 1:12

2.2. 11th–14th century Old Russian The overall picture in the 11th–14th century Old Russian sample is similar to the one in the OCS sample: ADJ2 constructions dominate, constituting 76.4 % (265) of the occurrences, but there are also 7.5 % (26) ADJ1 constructions and 15.6 % (54) GENMOD occurrences. Even more than in the OCS sample, we see that the distribution of constructions varies between the three LABEL types. As in the OCS sample, there are few occurrences where the possessor nominal is uniquely identifiable, but where the construction as a whole denotes a type. There are 20 such occurrences; five are ADJ1 constructions and 15 are ADJ2 constructions. All the five ADJ1 occurrences are instances with the meaning “fear of God”, such as in (272), which appear to be strongly conventionalised and to name a type of fear. (272)

v srdci svoemž strax imˇeite Bii fear have god-IJ in heart self’s ‘have the fear of God in your heart’ PVM 241/11

The ADJ2 occurrences fall into three clear types: In the OCS sample, there are eight occurrences meaning “birds of the sky” or “beasts of the earth”, as in (260), three occurrences naming the inhabitants of a particular place (273), and four occurrences naming the monetary unit of a particular city (274).

141

0

10

20

30

40

50

60

70

LABEL

ADJ1

ADJ2

GENMOD

GENUNMOD

DATIVE

Figure 35. Constructions at the LABEL node, 11th–14th century Old Russian, per cent, n=347

(273)

Vyˇsegorodžskyˇe muˇzeˇ Vyshegorod-SK men ‘Vyshegorod men’ SBG 32/4

(274)

kuna. smolžneskaja kuna Smolensk-SK ‘a Smolensk kuna’ SRG 49/7

There are 195 cases where the possessor nominal is type identifiable or kindreferring, whereas the whole construction denotes a uniquely identifiable referent. All of these are ADJ2 constructions, and the largest group by far consists of constructions on the form [ETHNIC GROUP-SK, zemlja] “the land of X”, as already seen in example (264) in the OCS sample. The most frequent collocation is the ubiquitous name for the land of Rus’ itself (83 occurrences), but there are also 29 occurrences naming various other countries in the same way. This is clearly a productive, partially lexically specific construction.

142 Types and conventionalised units: TYPE and LABEL (275)

Ruskaja zemlja Rus’-SK land ˇ 366/6 ‘the land of Rus’’ ZD

(276)

do Gržˇcžsky zemlja to Greek-SK land ‘to Greece’ SBG 61/28–29

There are 41 other occurrences of toponyms with a similar pattern: [ETHNIC GROUP -SK, NATURAL FORMATION], as in (277). (277)

ˇ sžskago lˇesa do Ceˇ to Czech-SK forest ‘to the Czech Forest’ PVM 247/26

Similarly, there are 20 occurrences naming languages on the pattern [ETH and 20 occurrences naming religions on the pattern [ETHNIC GROUP-SK, vˇera]. NIC GROUP -SK, LANGUAGE ]

(278)

a Slovenžskyi jazykż i Ruskyi: odno es but Slav-SK language and Rus-SK one is ‘but the Slavonic language and the Rusian [language] are one’ PVrL 28/21–22

(279)

vˇera xrejanžska urodžstvo estž faith Christian-SK abomination is ‘the Christian faith is an abomination’ PVrL 63/9

Thus, the second group of LABEL constructions, as found in the 11th– 14th century Old Russian sample, appears to be a close-knit family of more or less productive partially specific constructions used to name places, languages, religions and other uniques by reference to the ethnic (or other stable) group they belong to. Typically, the referents of these constructions occur in sets, e.g. there is one country and one language per ethnic group, and this makes them similar to TYPE constructions, even though they are not themselves types. The third group, LABEL constructions with a uniquely identifiable possessor nominal where the whole construction names something uniquely identifiable, is also a large one in the Old Russian sample, with 130 occurrences. More clearly than in the OCS sample, the distribution of constructions is reminiscent of those at the two ID nodes: 16.2 % (21 occurrences) are ADJ1

LABEL

143

constructions, 42.3 % (55 occurrences) are ADJ2 constructions, and as much as 41.5 % (54 occurrences) belong to the GENMOD group. The ADJ1 occurrences all have adjectives referring to uniquely identifiable singular human beings. Most of them are complex toponyms. There are seven occurrences naming various buildings (and one street), as in (280), one naming a city (Constantinople, cf. example (267)) and 11 occurrences of the particular type found in (281). These constructions do not denote the event of the crucifixion of Jesus, but name the place where it occurred. (280)

ta sutž vrata venže˛ minova those are gates Benjamin-OV ‘those are the Gates of Benjamin’ XID 18r/10–11

(281)

gsne˛ do sne˛ tja gsne˛ estž. e a ot raspe˛ te˛ and from crucifixion lord-N’ to taking-down lord-N’ is five saˇzen fathoms ‘and from the crucifixion of the Lord to the taking-down of the Lord there are five fathoms’ XID 21v/16

There is also one occurrence naming a church holiday (282) and one naming a battle (283). (282)

na Gsˇzinż dnž on lady-IN day ‘on the day of Our Lady’ PVM 248/38

(283)

do Mamaeva poboiˇsta to Mamaj-OV battle ‘until the battle against Mamaj’ Z 21/32

The GENMOD occurrences are very similar. The genitive-marked NP is in most cases a uniquely identifiable human being in the singular, and the construction as a whole names either a building (284) or a church holiday (285). (284)

Vasilžja v cerkvi stgo in church holy-GEN Vasilij-GEN ‘in St. Vasilij’s church’ PVrL 111/18

(285)

vż džnž svjatago Nikoly in day holy-GEN Nikola-GEN ‘on St. Nikola’s day’ SBG 58/6

144 Types and conventionalised units: TYPE and LABEL When we look at the ADJ2 occurrences, we find that the adjectives are all formed from inanimate nouns, and that most of them fit the patterns found in group 2: In 43 of the cases the construction as a whole names a place, mostly countries or districts (286), and in seven cases it names a language or an ethnic group (287). The only difference is that the possessor nominal is not an ethnic group, but a city or other toponym instead. (286)

vż novżgorodžskoi volosti in Novgorod-SK district ‘in the Novgorod district’ DTN 63/11

(287)

jazykż permžskyi pokuˇsaˇsese˛ isuˇciti language Perm-SK tried-REFL learn ˇ ‘he tried to learn the Permian language’ ZSP 8/21

In the third group, then, much more clearly than in the OCS sample, we see complementary distribution again: ADJ1 for unmodified singular human possessors, ADJ2 for unmodified inanimate possessors, and GENMOD for the modified possessors. The patterns found in the 11th–14th century Old Russian sample may seem quite different from those found in the OCS sample, but this is probably mostly due to subject matter: The early Old Russian texts are from a Christian society, and so naturally there is much reference to particular churches and church holidays, and also to various countries and languages. The subject matter of the Gospels is much more confined to a small set of places, and the texts focus on a small set of main participants rather than on sweeping historical events. In this respect, the vitae of the Codex Suprasliensis are closer to the Old Russian texts, and many of the LABEL examples are from the Suprasliensis excerpts.

2.3. Further developments in the history of Russian When we look at the further development in Russian, we find rather clear tendencies at the LABEL node. As we see from figure 36, the share of GENMOD occurrences remains fairly stable, but there is a clear decrease in the share of ADJ2 constructions, a corresponding increase of ADJ1 constructions, and a surge of mixed constructions in the 17th century sample. However, when we compare table 27 and table 28, we discover that the variation and change is

LABEL

11th–14th century 15th century 16th century 17th century

ADJ1 26 10 27 35

ADJ2 265 49 17 50

DATIVE 0 0 1 0

GENMOD 54 10 12 9

145

GENUNMOD 1 1 0 0

mixed 0 0 1 13

GENUNMOD 0 0 0 0

mixed 0 0 1 13

Table 27. Constructions at the LABEL node, diachronically

11th–14th century 15th century 16th century 17th century

ADJ1 21 10 26 30

ADJ2 55 9 11 34

DATIVE 0 0 1 0

GENMOD 54 10 12 9

Table 28. Constructions at the LABEL node, diachronically, type 3 only

almost exclusively confined to the third group of LABEL constructions, the ones that have a uniquely identifiable possessor nominal and name something unique. In the first two groups, ADJ2 constructions maintain a stable dominance in all periods. Looking at the developments in group 3 alone in figure 37, we get a more realistic view. The main tendency is an overall increase in the share of ADJ1 constructions, making the shares of ADJ1 and ADJ2 constructions more or less equal in the 17th century. The reason behind the increase is immediately apparent: Unlike the occurrences in the 11th–14th century sample, the largest group of ADJ1 constructions of group 3 in the 17th century sample consists of patronymics such as (288). This group accounts for for 23 out of 30 examples. (288)

sż Nastžkoju Petrovoju doˇcerju with Nastka Petr-OV daughter ‘with Nastka Petr’s daughter’ PGMK 358/10–11, 17th century

In addition, we still find the familiar types seen in the earliest texts, that is, constructions naming church holidays (three occurrences), such as (289) and churches, monasteries and other buildings (four occurrences), such as (290).

146 Types and conventionalised units: TYPE and LABEL ADJ1 ADJ2 GENMOD mixed




60







40

Per cent







20


















11th
14th cent.

15th cent.

0




16th cent.

17th cent.

Figure 36. Constructions at the LABEL node, diachronically, per cent

(289)

otż Jurževa dnja do Jurževa dnja from Jurij-OV day to Jurij-OV day ‘from Jurij’s day to Jurij’s day’ PGMK 358/14, 17th century

(290)

do Andron(ž)eva m(o)n(a)st(y)rja to Andronij-OV monastery ˇ ‘to the monastery of Andronij’ ZAvv 23/13–14, 17th century

We recall from page 76 that in the 11th–14th century sample, possible patronymics of this form were taken at face value as ID(slot) constructions; however, no more than three occurrences were particularly convincing examples. At least in the 16th and 17th century samples, the ADJ1 patronymic appears to be a productive, partially lexically specific construction in its own right. The apparent increase of ADJ1 constructions at the LABEL node is thus no more than a reflection of the rise of the patronymic subconstruction. For other types of ADJ1 label constructions, we actually see a sharp decrease. As we have already noticed, there was an overall weakening of the schematic ADJ1

LABEL

147

60

ADJ1 ADJ2 GENMOD mixed

40

Per cent













20


















11th
14th cent.

15th cent.

16th cent.

17th cent.

0




Figure 37. Constructions at the LABEL node, diachronically, per cent, group 3 only

construction in the possessive semantic space in progress. Filtering out the patronymic constructions, we see that this tendency holds for the LABEL node as well. The 17th century GENMOD occurrences in this group, however, are all of the building-name type. There are no GENMOD occurrences to match the frequent ADJ1 patronymics. Thus, the decline of GENMOD in this group is only an apparent one. The share of ADJ2 constructions in the 17th century sample is approximately the same as in the 11th–14th century sample, but here, they are less uniform. Out of 34 occurrences, 17 are of the familiar “country/district of X” type, where X is typically a city name. (291)

vż muromskomż uˇezde in Murom-SK district ‘in the Murom district’ DG 194/16, 17th century

However, there are also 15 occurrences naming buildings, constructions and

148 Types and conventionalised units: TYPE and LABEL streets. The occurrences either have inanimate possessor nominals or are similar to (292), but there is also one single example with a singular human possessor nominal (293). (292)

Troickoj monastyrž trinity-SK monastery ‘the Trinity monastery’ SAP 136/4, 17th century

(293)

za Nikickimi voroty behind Nikita-SK gates ‘behind the Nikita gates’ DG 194/28–29, 17th century

Thus, in the 17th century material the ADJ2 construction appears to be more versatile and more productive at the LABEL node than the ADJ1 construction, and to be entering into competition with it in constructions with human singular possessor nominals. In the 17th century sample there are also 13 mixed constructions in group 3, 11 of these are double adjective constructions, and two are adjective + genitive constructions. All but one example, however, are varieties of the name of the Trinity-Sergius Monastery, as in (294) and (295).77 Still, the examples fit the general trend for mixed constructions to be on the rise in the 16th and 17th century samples. (294)

Troickoj Sergiev monastyrž trinity-SK Sergij-OV monastery ‘the Monastery of the Trinity and Sergij’ SAP 135/18, 17th century

(295)

ot Troica zˇivonaˇcalžnyja Sergieva monastyrja from trinity-GEN life-beginning-GEN Sergij-OV monastery ‘from the monastery of the life-beginning trinity and Sergij’ SAP 129/10, 17th century

3. Conclusions As we saw in the outset of this chapter, TYPE and LABEL constructions are similar in that the whole is more important than the constituent parts; these constructions are names for types and individuals. They are also similar in that the ADJ2 construction is the predominant one. This construction is virtually the only choice at the TYPE node, and at the LABEL node it

Conclusions

149

predominates in the two subgroups that are the most similar to the TYPE node: constructions that name types (though their possessor nominals are uniquely identifiable) and constructions with kind/type possessor nominals (though they name uniquely identifiable referents). In the third group, however, we find a complementary distribution-like pattern between ADJ1, ADJ2 and GENMOD occurrences, reminiscent of the pattern found at the closely related ID nodes. It is only in this group that we see signs of change. The other LABEL groups, as well as TYPE, remain the domain of ADJ2 constructions. In the third group we see the rise of ADJ1 patronymics, but there are indications of a tendency for ADJ2 constructions to take over for ADJ1 constructions even in this group. This is certainly the pattern in modern Russian.

Chapter 7 Synchrony and diachrony

After the close examination of the map of the possessive semantic space node by node and language stage by language stage, it is time to return to the overall map (figure 38, repeated from page 33). This chapter summarises the synchronic and diachronic findings from chapters 2–6, and interprets them with the semantic map as the point of departure.

1. Synchrony: Division of labour vs. complementary distribution In this section I will draw some synchronic conclusions based on the two earliest text samples: the OCS sample and the 11th–14th century Old Russian sample.

ID ID(slot) LABEL

ELAB(slot)

ELAB

ELAB(part)

Figure 38. A map of the possessive semantic space

TYPE

152 Synchrony and diachrony 1.1. Complementary distribution? As we saw in chapter 1, section 2.1, most authors assume some level of complementary distribution between genitive and adjective constructions in expressing possession: In some functions, adjective constructions are used if the possessor on its own would be an unmodified noun, while the genitive is used elsewhere (e.g. when the possessor has an attribute or apposition, with coordinated nouns, or when the possessor is a nominalised adjective or participle, cf. the formulation in Makarova 1954). The exact nature of this complementary distribution has in many accounts been obscured by the failure to define the concept of possession properly, and also by the failure to clarify which adjective constructions are actually under consideration (e.g. Makarova 1954). An account such as Bratishenko’s (1998, 2005), on the other hand, with a better-defined notion of possession and a clear account of which constructions are actually under consideration, inevitably concludes that the “complementary distribution” is far from perfect. It is violated both by complex adjective constructions and unmodified genitive constructions, but these “violations” are clearly conditioned by semantic factors. In terms of our map of the possessive semantic space, we can conclude that a situation approaching complementary distribution is only found at some of the nodes. The best case for complementary distribution between a formally restricted genitive construction and both ADJ1 and ADJ2 can be made for the ID node, as seen in chapter 4, sections 1.1 and 2.1. Both in the OCS and the 11th– 14th century Old Russian sample, very few GENUNMOD occurrences are found, and to the extent that they are found, they are clearly competitors to the ADJ2 construction, not to the ADJ1 construction. In the Old Russian sample, all occurrences have abstract possessor nouns, commonly kind-referring or non-specific. In the OCS sample, a slightly wider range of possessors is found: abstract nouns, but also inanimate concretes and human plurals, all of which more commonly occur in ADJ2 constructions at the ID node. None of the possessors are both specific and discourse-accessible. A more common violation of the complementary distribution is the occurrence of complex adjective constructions, most of them expanded versions of the ADJ1 construction, where occurrences of the restricted genitive construction are expected. In these cases the possessor is normally human and specific. At the ID(slot) node, we also find a situation fairly close to complementary distribution between the restricted genitive construction and the two adjective constructions, as seen in chapter 4, sections 1.2 and 2.2. Violations

Synchrony: Division of labour vs. complementary distribution

153

in the shape of GENUNMOD occurrences are still few, but both in the Old Russian and the OCS sample, a wider range of possessors is found. Both text samples even have cases of human singular possessors, indicating that the GENUNMOD occurrences can, marginally, compete with the ADJ1 construction at the ID(slot) node. This impression is supported by the fact that the share of GENMOD occurrences in both samples is significantly higher at ID(slot) than at ID. Furthermore, we see, as we did at the ID node, mixed adjective constructions competing with the formally restricted genitive construction, usually when the possessor is human and specific. At the other nodes, however, we do not find complementary distribution between adjective and genitive constructions in the two oldest text samples. TYPE and ELAB(part) are both nodes virtually reserved for a single construction; ADJ2 and the formally unrestricted genitive construction respectively. At LABEL we find complementary distribution only to the extent that the constructions there are conventionalised ID and ID(slot) constructions. At ELAB, we see a fairly stable complementary distribution between genitives and ADJ2 constructions when the relationship between the head noun and the modifier is one of (place of) origin (see chapter 5, section 4). Apart from this, GENUNMOD occurrences are much more frequent at ELAB than at the ID (and LABEL) nodes.

1.2. Division of labour The fairly strong complementarity between genitive and adjective constructions at the ID nodes is interesting in that it appears to be a primarily syntactically motivated phenomenon. It is mainly the complexity of the possessor that decides whether it should be encoded by a genitive construction or not, although we see that even at the ID node, where the complementary distribution is at its most convincing, this formal criterion may still be overridden by semantic factors. However, the main impression from the exploration of early Slavic form– meaning correspondences in the possessive semantic space is rather that there are such a lot of semantically conditioned interactions and divisions of labour. As mentioned, there are two nodes reserved for one single construction (but not the same construction): TYPE and ELAB(part). At the other nodes, one construction dominates in terms of frequency, but this construction always has various interactions with other construction types.

154 Synchrony and diachrony At the two ID nodes, there is certainly such a relationship between the ADJ1 and the ADJ2 construction both in the OCS and 11th–14th century Old Russian samples. Whereas the ADJ1 construction is for “prototypical”78 possessors – human, singular, specific and usually discourse-accessible referents – the ADJ2 construction is used for the less typical possessors. The text samples suggest that OCS and Old Russian differ somewhat in this respect. In the OCS sample we found that ADJ2 possessors at both ID nodes must have plural reference if they are human, and that they often have kind reference as well (cf. page 55). Non-human possessors may easily have singular reference, but they are very rarely both specific and accessible from discourse. Rather they have either non-specific, kind or unique reference. In the Old Russian sample, we find all animacy types represented with ADJ2 constructions, including a fair share of human singulars. However, these human singular possessors have either non-specific, kind or unique reference; they are not accessible from discourse, even though the uniques do, of course, have specific reference. Non-human possessors were found to be different at the two ID nodes: At the ID node they were rarely accessible from discourse, but at the ID(slot) node this was the case significantly more frequently (see page 76). Thus, we find that the ADJ2 is, typically, the choice for possessors that are inferior reference points in some sense. It is a possible choice for human possessors that are not singular (a group is inferior to an individual as a reference point), not discourse-prominent (uniques) or not individuals at all (non-specifics and kinds), and it is the usual choice for non-human possessors, which are all inferior to human possessors as reference points. The dative construction also has interesting interactions with other constructions. It interacts with the adjective-and-genitive conglomerate at the ID(slot) node, and especially with the formally unrestricted genitive construction (represented by GENUNMOD occurrences) at the ELAB(slot) node. At the ID(slot) node, both samples have fairly small shares of dative constructions, but the OCS sample in particular has an interesting pattern, as seen in chapter 4, section 1.2. The dative construction clearly has a strong preference for body part head nouns, and the other preferred types of head nouns are kinship terms and other human relationship nouns and deverbal nouns with a non-agentive subject slot. The generalisation seems to be that the dative construction competes with genitive and adjective constructions at this node when the head noun has a slot for a human being who is in control of and/or is affected by the head noun. The Old Russian sample has a similar

Synchrony: Division of labour vs. complementary distribution

155

tendency, but the bias for body part head nouns is less clear, as seen in chapter 4, section 2.2. Apparently, judging by the limited number of findings, the Old Russian construction has also specialised with kind-referring and nonspecific possessors. As seen in chapter 5, section 2, the ELAB(slot) node is the centre of gravity of the dative construction both in the OCS sample and in the 11th– 14th century Old Russian sample, and the occurrences are almost exclusively headed by deverbal nouns and fill an object slot. This is also an important function of the GENUNMOD occurrences at the ELAB(slot) node in both languages. What we see in both samples is that the dative construction is more permissive than the formally unrestricted genitive construction (as represented by the GENUNMOD occurrences) when it comes to animacy. In fact, there are no such human GENUNMOD occurrences in either sample, whereas all animacy types are found in the dative construction. However, slot fillers in both constructions tend to have non-specific or kind reference. In general, then, we see that when several constructions compete at a single node, the choice is normally conditioned by the properties of the possessor (animacy, referentiality, degree of activation in discourse).

1.3. Map and territory As we saw in chapter 3, the map of the possessive semantic space clearly captures important features of most of the constructions posited in this book. For the ADJ1 construction, a reference point situation seems close to obligatory. It may be conventionalised so that it must be classified at the LABEL node, but at all other nodes, it is marginal. Conversely, for the dative construction and the formally unrestricted genitive construction (as represented by the GENUNMOD occurrences) an intrinsic relationship appears to be close to obligatory; most of the occurrences have relational head nouns or are found in non-reference point situations involving non-argumental intrinsic relationships. The ADJ2 construction has a very clear stronghold in cases where the construction as a whole names a category or an individual. Although e.g. the 11th–14th century Old Russian ADJ2 construction has a neat and confined centre of gravity in the map on page 39, it is still the construction least accounted for by our map of the possessive semantic space: When we plot the construction’s boundaries on the map, we fail to describe some of its more salient features. No other construction is distributed that

156 Synchrony and diachrony equally across the nodes. This is probably due to the fact that the map does not capture one of the important features of the ADJ2 construction’s semantics, namely, that it appears to be strongly driven by the properties of the possessor nominal, which is normally non-prominent in one of several possible ways: It may be merely type identifiable, and not an individual at all, or it may be low in referentiality in other ways; it is usually low in animacy, and if high in animacy, then at least low in accessibility, or plural. This appears to hold fairly consistently across the map of the possessive semantic space. Such possessors are the only choice at the TYPE node, and a frequent choice at the LABEL node, and they are still not a marginal choice at ELAB(slot) and the ID nodes.

2. Diachrony 2.1. OCS vs. Old Russian One of the goals of this book was to examine whether OCS and Old Russian differ in the use of adnominal possessive constructions. The general conclusion must be that the two languages are remarkably similar in this respect. To the extent that there are differences, these mostly concern cases where both languages have the same tendency, but where this tendency is stronger in one of the languages. The most important of these cases are discussed in this section. Firstly, the ADJ2 construction is more frequently used and more flexible in the 11th–14th century Old Russian sample than in the OCS one. In the Old Russian sample, there are more ADJ2 constructions at the ID nodes, and we see that the construction can e.g. be used for human singulars and thus potentially be in competition with the ADJ1 construction, whereas this is not attested in the OCS sample. As seen in chapter 1, section 2.1, the dative construction has been pointed out as a potential difference between the two languages, several authors even claiming that the possessive dative is a mere loan from Church Slavonic into Old Russian. However, in terms of frequency, there is no discernible difference between the two language samples. We also find that the two dative constructions cover the same nodes in the map of the semantic space and have the same centre of gravity. What we do see, however, is that the OCS dative construction to a stronger extent specialises with relational nouns typ-

Diachrony

157

ically requiring affected human beings, whereas in the Old Russian sample it is much used for abstract, often kind-referring slot fillers. Furthermore, we noted that there have been claims that unmodified genitives are more frequently found in OCS than in Old Russian. Again, our investigation does not confirm this, as the shares of GENUNMOD occurrences appear to be remarkably similar in both language samples. As seen on page 103, there is a clearer tendency for GENUNMOD occurrences at the ELAB(slot) in OCS, but the same tendency is present in the Old Russian sample, even though here deverbal nouns with object slots are more frequently found with dative constructions. To conclude, then, the two languages have only subtle differences in this respect. It is also difficult to identify any clear diachronic trend from OCS to Old Russian, even though the OCS sample predates the Old Russian one by several centuries.

2.2. Development trends Let us again examine the correspondence analysis plot based on the three selected language stage samples: OCS, 11th–14th century Old Russian and 17th century Russian (figure 39, page 158, repeated from page 31). We find that it reveals three important diachronic trends. Firstly, there is the obvious movement of the adnominal dative construction from a fairly central position in OCS, between the ELAB nodes and the ID nodes, to a position on the very outskirts of the plot in 17th century Russian. Thus, we see that the 17th century Russian dative construction is much less similar to the other constructions in the possessive semantic space than the OCS and OES dative constructions are. Secondly, we note the movement of the ADJ2 construction from the centre of the plot in OCS up to the TYPE node in 17th century Russian. Thus, the ADJ2 construction has also become less similar to the other constructions in the possessive semantic space. Thirdly, and less conspicuously, we see that in the 17th century Russian sample, the ADJ1 construction has moved from the corner above the two ID nodes to a slightly more central position closer to the LABEL node, although it is still firmly situated in the ID section of the map. The changes to the two adjective constructions are also related to a fourth change, which is not plotted on the map: the rise of mixed constructions at the ID nodes.

158 Synchrony and diachrony

0.5

OCS.ADJ1 OR.ADJ1 17cent.ADJ1

ID IDslot

LABEL

0.0

OR.ADJ2


0.5

OCS.GENMOD 17cent.GENMOD

OCS.ADJ2

ELABslotELAB

OCS.DATIVE OR.DATIVE


1.0

Factor 2 (27.1 %)

TYPE

17cent.ADJ2

OR.GENMOD

OR.GENUNMOD 17cent.DATIVE


1.5

OCS.GENUNMOD


2.0

17cent.GENUNMOD ELABpart


1.0


0.5

0.0

0.5

1.0

1.5

2.0

Factor 1 (52 %)

Figure 39. A correspondence analysis plot of the seven nodes in OCS, 11–14th century Old Russian (OES) and Middle Russian (17th century)

However, the plot does not show any clear trend concerning the development we would expect to see in the diachronic material, namely the expansion of the unrestricted genitive construction and the end of complementary distribution. If anything, the map indicates the opposite trend; the GENMOD and GENUNMOD occurrences appear to become more polarised in the course of the period under consideration.

2.3. The demise of the dative construction In chapters 4 and 5,79 we saw that, except at the ELAB(slot) node, the dative construction diminished radically or disappeared at every node during the period under consideration in this book. This is also made clear when we plot the 17th century distribution of the dative construction onto the map of

Diachrony

159

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 40. Old Russian adnominal dative on the map of the possessive semantic space

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 41. 17th century Russian adnominal dative on the map of the possessive semantic space

50

160 Synchrony and diachrony

0

10

20

30

40

11th
14th cent. 17th cent.

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 42. Dative constructions by node, 11th–14th century Old Russian (n=181) and 17th century Old Russian (n=28), per cent

the possessive semantic space (figure 41) and compare it to the 11th–14th century one (figure 40, repeated from page 159), cf. also figure 42. In doing so, we see that the dative construction has withdrawn towards the node that was found to be the most frequent one both in the OCS and the 11th–14th century Old Russian sample, namely ELAB(slot). The dative construction now only covers ELAB(slot) and ELAB, and these two together constitute its centre of gravity: 50 % of the 18 occurrences are found at ELAB(slot) and 21.4 % are found at ELAB. In addition, there is one single ID example and as many as 25 % uncertain cases. In the 17th century, the dative construction is likely to be a highly literary construction, reserved for the typically literary purpose of expressing the object arguments of deverbal nouns. It is interesting to note that this central function is the last one to go – apparently, this is the end of the adnominal dative’s diachronic path.

Diachrony

161

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 43. Old Russian ADJ2 on the map of the possessive semantic space

2.4. The withdrawal of the ADJ2 construction From the data presented in chapters 4–6 it is not obvious that any great changes happened to the ADJ2 construction. Except for at the LABEL node, there were no discernible trends in terms of frequency. At the LABEL node (see chapter 6, section 2.3), we did find that there was a clear decrease in the share of ADJ2 constructions; however, closer scrutiny showed that this was not so much a question of a decrease in use of ADJ2 constructions as an increase in the use of ADJ1 constructions to form patronymics. The apparent decrease was found only in the subgroup of ELAB where a uniquely identifiable referent identifies a uniquely identifiable referent, and in the other two subgroups, the share of ADJ2 constructions was stable. There were, however, some discernible changes in the choice of possessor referents at LABEL and the two ID nodes. At the ID nodes, we saw that ADJ2 possessors were increasingly human and specific (chapter 4, section 3). At the ELAB node, ADJ2 seemed to specialise with toponyms (chapter 5, section 4.2). However, when we compare the overall distribution of ADJ2 in the 11th– 14th century sample and the 17th century sample, we note another clear

162 Synchrony and diachrony

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 44. 17th century Russian ADJ2 on the map of the possessive semantic space

change: A much larger share of the ADJ2 constructions is now found at the TYPE node. When we add the boundaries of the 17th century construction to the map (compare figure 43, repeated from page 39, and figure 44), we find that it still covers the same area, but that TYPE is now the centre of gravity alone, accounting for 55.2 % of the occurrences (vs. 25.7 % in the Old Russian sample). LABEL, on the other hand, now only accounts for 13.4 % (vs. 31.3 % in the Old Russian sample). Thus we see a similar trend as in the history of the dative construction: The construction withdraws to its semantic centre of gravity.

2.5. The weakening of the ADJ1 construction A first glance of the overall frequency of the ADJ1 construction in all text samples suggests that there was no change. The constructions remains stable and accounts for around 20 % of all occurrences in each sample. But as we saw in chapters 4–6, there are clear signs that the ADJ1 construction is losing ground. We saw that the ADJ1 construction had a mild decrease in frequency at the

Diachrony

163

0

10

20

30

40

50

11th
14th cent. 17th cent.

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 45. ADJ2 constructions by node, 11th–14th century Old Russian (n = 848) and 17th century Old Russian (n = 373), per cent

ID node, but not at the ID(slot) node (see chapter 4, section 3). However, we also noted that an increasing share of the occurrences had adjectives formed from a small set of religious nouns, and, furthermore, that many of these occurred in fixed, frequently recurring expressions (see page 82). There was also a very clear decrease in the use of ADJ1 with proper noun stems at the ID nodes, as seen in table 15 on page 84. In other words, ADJ1 was losing ground with the most typical of all possessors: human, uniquely identifiable possessors. At the same time, we see that the ADJ1 construction has a clear increase at the LABEL node, almost entirely due to its increasing use in patronymics on the form [ADJ1 SON / DAUGHTER]. To the extent that type 1 adjectives were found to be formed from proper nouns in the 17th century sample, they were primarily found in these occurrences. The increase in ADJ1 constructions at the LABEL node was sufficient to shift the construction’s semantic centre of

164 Synchrony and diachrony

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 46. Old Russian ADJ1 on the map of the possessive semantic space

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 47. 17th century Russian ADJ1 on the map of the possessive semantic space

Diachrony

165

0

10

20

30

40

50

11th
14th cent. 17th cent.

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 48. Distribution of ADJ1, 11th–14th and 17th century sample, per cent

gravity to comprise the LABEL node as well as the two ID nodes, see figures 46 (repeated from page 37) and 47. In the 11th–14th century sample, LABEL constructions accounted for 5.8 % of all ADJ1 constructions, but it comprised as much as 18.2 % in the 17th century material (figure 48). It is reasonable to see all these developments as symptoms of the construction losing productivity (cf. Eckhoff 2009). In OCS and 11th–14th century Old Russian, ADJ1 must have been a fairly generalised construction schema for all unmodified, human singular possessors at ID and ID(slot), as illustrated in figure 49, where frequency is indicated by the thickness of lines (cf. figure 21 on page 84). The topmost box represents the maximal generalisation over the construction’s usages. With the distributions we see in the early samples, there is little reason to believe that this generalisation is not a productive and strong schema. (The posited productivity is also indicated by a bold line in the figure.) This is clearly no longer the case in 17th century Old Russian. Instead, as illustrated in figure 50, there appear to be productive, more specific

166 Synchrony and diachrony

[Adj1, N]

[Adj1, N]

[Adj1, N]

[Adj1, N]

Figure 49. Tentative schematic network of the ADJ1 construction, 11th–14th century

subconstructions for common nouns and for the small set of religious nouns, and also a number of fixed expressions commonly used as formulae in letterwriting (exemplified by straxż boˇzii “the fear of God” and boˇzija volja “God’s will” in the figure), which are best seen as fully lexically specific constructions. In addition, there is the highly productive patronymic subconstruction. Taking the usage-based perspective of Croft (2001) and BarDdal (2008), it is much less obvious from the 17th century data that the maximal generalisation should be represented in the schematic network. In modern Russian, the maximal generalisation clearly does not hold at all, since the ADJ1 construction is limited to colloquial use in intimate settings, and then only with a-stems. In addition, the religious adjectives are still (optionally) in use, and there is a series of fully specific ADJ1 constructions, many of them belonging to the LABEL node, which must be considered relics of the previous system, for instance toponyms such as Barencovo more “the Barents sea”. Why, then, is the ADJ1 construction so poorly attested with proper noun stems outside the patronymic construction? It seems that this may mainly be a cultural development, as suggested by Makarova (1954:29), leaning on Borkovskij (1949:355): In the period under consideration, people are increasingly being referred to by complex phrases in the public sphere, including written texts; their first names are more and more often embellished with patronymics, last names and often titles as well, and thus they will tend to occur in the genitive or in a mixed adjective construction. As we saw in section 3.5 in chapter 4, there is a sharp rise in the use of

Diachrony

167

[Adj1, N]

[Adj1, N]

božija volja “God’s will”

[Adj1, N]

straxъ božij “fear of God”

[Adj1, N]

[Adj1, N] LABEL

Figure 50. Tentative schematic network of the ADJ1 construction, 17th century

mixed constructions in the period under consideration in this book. It seems reasonable to interpret this as a response to the increased use of complex names and titles; the impulse to use ADJ1 constructions with specific singular persons is so strong that either a genitive construction or a mixed construction (296) may be the outcome. (296)

Ivanka Mixajlovż cˇ elovˇekż Kaznakova Ivanka Mikhail-OV man Kaznakov-GEN ‘Ivanka, Mikhail Kaznakov’s man’ PGMK 358/2–3

The retraction of the ADJ1 construction to the informal, intimate sphere may thus perhaps best be seen as a result of changes in naming traditions in the public sphere.

40

168 Synchrony and diachrony

0

10

20

30

11th
14th cent. 17th cent.

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 51. Distribution of GENMOD occurrences, 11th–14th and 17th century sample, per cent

2.6. From two genitive constructions to one? Looking at the genitive data alone, there is very little evidence to indicate that the two posited genitive constructions have really merged into one by the 17th century. The distributions of the two constructions change little over time; if anything, they become more different than in the 11th–14th century sample. If we compare the 17th century situation in figures 53 and 54 to the maps on page 44 and 45, we see that the outer boundary of each construction is the same as in the 11th–14th century sample. However, the semantic centre of gravity of the GENMOD occurrences has expanded to include the ELAB(slot) and ELAB nodes, whereas the centre of gravity for the GENUNMOD occurrences has retracted and no longer includes ID(slot). We recall from chapters 4–6 that the node-by-node survey yielded much the same results. There were no frequency trends of obvious importance. At

169

40

Diachrony

0

10

20

30

11th
14th cent. 17th cent.

TYPE

LABEL

ID

ID(slot)

ELAB(slot)

ELAB

ELAB(part)

uncertain

Figure 52. Distribution of GENUNMOD occurrences, 11th–14th and 17th century sample, per cent

the ID nodes, however, there were slight indications of change, in that attestations of variation between GENUNMOD and ADJ1 constructions were found when the possessor was a human common noun, as attested for the ID(slot) node in examples (297) and (298), which are from the same 16th century text. (297)

po poveleniju deviˇcju by order maiden-J ˇ ‘by the maiden’s order’ ZPF 216/28 16th century

(298)

po poveleniju devicy by order maiden-GEN ˇ ‘by the maiden’s order’ ZPF 216/25, 16th century

Such examples are also found at the ID node, as in (299) and (300) (repeated from page 88).

170 Synchrony and diachrony

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 53. 17th century Russian GENMOD on the map of the possessive semantic space

ID ID(slot) LABEL

ELAB(slot)

TYPE

ELAB

ELAB(part)

Figure 54. 17th century Russian GENUNMOD on the map of the possessive semantic space

Diachrony

171

(299)

vo obiteli cˇ judotvorca in monastery miracle-performer-GEN ‘in the miracle-performer’s monastery’ SAP 130/19, 17th century

(300)

u cˇ judotvorcovy raki by miracle-performer-OV coffin ‘by the miracle-performer’s coffin’ SAP 134/18, 17th century

However, in the small study of ID occurrences in an 18th century sample in chapter 4, section 3.6, we do see a tendency to change: The GENUNMOD occurrences are still few and low in animacy, but we notice that the share of genitives overall is much higher than in the 17th century sample, and that there is a clear decrease in the share of ADJ1 (and mixed) constructions. Overall, then, it is not so much the genitive constructions that have changed during the period under consideration in this book, but their competition. The dative construction has all but disappeared, and has withdrawn almost completely to the ELAB(slot) node. The ADJ1 construction is weakened, and meets competition from the ADJ2 construction to some extent, since that construction can also be used for human, specific common noun possessors. The complementary distribution at the ID nodes is muddled by the increasing presence of mixed constructions up to the 17th century. The expansion of the genitive, which makes it fairly easy to find such examples as (301) in the Russian National Corpus80 , must have been a later development, maybe as late as the 19th century. (301)

Poetomu mat’ Leny snaˇcala zanimalas’ poiskami Therefore mother Lena-GEN first busied-herself searching doˇceri samostojatel’no daughter alone ‘Therefore Lena’s mother first did the looking for her daughter alone’ Moskovskij komsomolec v Niˇznem Novgorode, 30.07.2004

2.7. A brief note on origins and causes As we saw in chapter 1, section 2.1, there are many points of disagreement in the previous literature concerning both the origins and the further development of Slavic adnominal possessive constructions. In this section, some of

172 Synchrony and diachrony these points of disagreement are briefly reviewed in light of the findings in this book. We saw that some authors posit a Common Slavic possessive system which was almost entirely the realm of the denominal adjective constructions (Marojevi´c 1989, Uryson 1980), while other authors assume a Common Slavic state that is very similar to the one attested in OCS (Richards 1976). The findings in the present book appear to support the latter position. The use of the formally restricted genitive construction is prevalent both in OCS and Old Russian. The mixed constructions, which are assumed by some authors to have been used instead of formally restricted genitive constructions in Common Slavic, are very rare both in the OCS and 11th–14th century Old Russian samples (but they have a clear increase in the subsequent Old Russian samples). The remarkable freedom with which Greek genitives are translated into the various OCS possessive constructions, and the striking similarities between the OCS and 11th–14th Old Russian systems also supports the hypothesis that Common Slavic had a very similar system, at least in its later stages. There is also independent support for this point of view from Baltic: Lithuanian and particularly Latvian have very broad adnominal genitive constructions, covering the entire possessive semantic space from ELAB(part) via the ID nodes to TYPE. Both Lithuanian and Latvian have denominal adjectives as well, but they are much less used than what we see both in early and modern Slavic, and marginal in Latvian. As argued in Eckhoff and Berg-Olsen (2002) (see also Vaillant 1958:601–602), this is striking, considering that the Slavic and Baltic branches are closely related in an Indo-European perspective. This contrast would suggest that the languages have roots in a system where both the adnominal genitive and denominal adjectives were available in the possessive semantic space, but have then evolved in radically opposite directions. In Slavic, the denominal adjectives expanded at the expense of the genitive, whereas the Baltic genitive expanded at the expense of the adjectives. Given the earliest attested Slavic, it seems doubtful that the genitive should have been marginalised i.e. at the ID nodes, as Marojevi´c (1989) suggests. In the subsequent history of Russian, the pendulum has swung back, but not to such extremes as in Latvian. Finally, a brief remark on the adnominal dative is in place. As we saw in chapter 1, section 2.1, there is much disagreement on its status in East Slavic. The findings in this book show that there is no statistically significant differences in frequency and distribution of the adnominal dative in the OCS and 11th–14th century Old Russian samples. There are slight suggestions that the

Diachrony

Legal Narrative Religious

TYPE 0 0 0

LABEL 0 0 0.0

ID 0 10.0 6.1

ID(slot) 0 24.0 26.3

ELAB(slot) 16.7 44.0 45.4

ELAB 83.3 20.0 21.2

173

ELAB(part) 0 2.0 1.0

Table 29. Distribution of dative constructions across genres in 11th–14th century Old Russian. 13 occurrences in business/legal texts, 66 occurrences in narrative text, 102 occurrences in religious text

constructions have specialised in somewhat different ways (cf. section 2.1), but overall, the dative construction displays remarkably similar behaviour in both samples. As we see in table 29, there is clear genre variation in the use of dative constructions in the Old Russian sample, but even though the dative construction is used almost twice as frequently in the religious texts as in the narrative ones, the distribution across functions is very similar. In business and legal texts, the dative construction is very rare, but that may easily be due to subject matter. To conclude, the findings of this book support the hypothesis that Old Russian at the time of the earliest attestations had a living and independent adnominal dative construction, very similar to that found in OCS.

2.8. Diachronic paths Many languages have a single device that covers the entire possessive semantic space as represented in this book. This is the case for the genitive in several Indo-European case languages, such as Ancient Greek and the Baltic languages. However, many other languages are more like early Slavic, in that they carve the semantic possessive space up in various ways. In this section I will propose a traditional-style semantic map based on the semantic centres of gravity and observed paths of development in the current data set. Given that the data is from a relatively short span in the history of a single language, and from another, very closely related language, this map cannot pretend to universality. However, since the situation in early Slavic is particularly complex, I believe such a map can provide an interesting set of working hypotheses for further work towards a detailed map of the possessive semantic space. Let us look at each of our five constructions. ADJ1 clearly has ID and ID(slot) as its semantic centre of gravity in all

174 Synchrony and diachrony

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 55. Proposed connecting lines for the ADJ1 construction

samples. The two other nodes that can be expressed by ADJ1 are LABEL and ELAB(slot). For this construction, it seems very clear that the two latter nodes must be due to extensions from the central meaning; ID and ID(slot) constructions can be conventionalised as LABEL constructions. As for ELAB(slot), to the extent that ADJ1 is really present at that node, it would seem to be a feasible extension from the ID(slot) node: If a construction is suitable for a relational noun and its slot filler, then the next step may be that it becomes suitable in that situation even when no reference point situation is involved. This reasoning thus indicates that we should have connecting lines between ID and ID(slot), between both ID nodes and LABEL, and between ID(slot) and ELAB(slot), as illustrated in figure 55. The ADJ2 construction’s clear centre of gravity is the TYPE node, but it is also very frequently attested at LABEL, and then typically in the LABEL subtype where the possessor is no more than type identifiable, but where the construction as a whole denotes an individual (see chapter 6, section 2). It seems reasonable to take this as the path of extension to the two ID nodes: If an ADJ2 construction as a whole may be used as a label on an individual, the next step may be to let the possessor nominal identify that individual in a reference point situation. ADJ2 is also found at the ELAB and ELAB(slot)

Diachrony

175

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 56. Proposed connecting lines for the ADJ2 construction

nodes. We recall that the predominant types of attestation here are origin modifiers at the ELAB node and ruler nouns with a slot for the ruler’s subjects at the ELAB(slot) node. Again, it makes sense to make a connection to to LABEL; the ADJ2 construction denotes an individual, and bestows it with a title or permanent epithet. For other relational nouns than the ruler nouns, the path to ELAB(slot) may perhaps be via ID(slot), as proposed for the ADJ1 construction. A possible solution is seen in figure 56, which indicates that if ID, ID(slot), ELAB(slot) or ELAB are coded the same way as TYPE, it must also be possible to code LABEL the same way. However, as figure 56 is drawn, it indicates that there are three different paths to ELAB, ELAB(slot) and ID(slot), and so the proposed connections are not very informative. The difficulty in outlining ADJ2 on the map ties in with the question raised about the suitability of the map nodes for this construction in section 1.3. For the GENUNMOD group, our certain attestations of the non-restricted genitive constructions, the three ELAB nodes are clearly the centre of gravity.81 It would not be unreasonable to posit that the unrestricted genitive construction’s ability to occur at ELAB(slot) is an extension from the ELAB(part) node; if the construction is suitable for part nouns with slot fillers, it may also be suitable for other relational nouns with slot fillers. Following the spirit of

176 Synchrony and diachrony

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 57. Proposed connecting lines for the unrestricted genitive construction, represented by the GENUNMOD group

Langacker’s intrinsic relationship analysis, we should also posit a connection between ELAB(slot) and ELAB. If the construction is suitable for expressing the argumental relationship between a relational noun and its slot filler, it could be extended to other intrinsic or necessary relationships between nominals. Many of the ELAB occurrences are clearly related to the ELAB(part) node, since they often carry partitive-like semantics, so it is reasonable to also posit a connection between these two nodes. It seems clear that the construction’s infrequent occurrences at ID(slot) should be seen as extensions from the ELAB(slot) node, and that the marginal attestations at ID are extensions from ID(slot) again. We thus end up with the graph in figure 57. The dative construction is different from the other constructions in the possessive semantic space in that it has strong links with other dative constructions82 and with the pervading and homogeneous dative semantics associated with all of them, where most dative-marked NPs are affected target( person)s in some way (Da˛ browska 1997). It has a strong flavour of being the result of a semantic extension of such typical dative meanings, which led the construction to inadvertently end up in the possessive semantic space, and to find itself in competition with the other possessive constructions. In the possessive semantic space, however, ELAB(slot) is clearly the con-

Diachrony

177

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 58. Proposed connecting lines for the dative construction

struction’s centre of gravity. It is the most frequently attested node in all samples, and we note that in the course of history, it retracts to this node. The construction’s ability to accommodate the relationship between a relational noun and a slot filler must be seen as an extention from the dative case’s typical semantics of denoting an affected target. The occurrences at the ID(slot) node can reasonably be seen as an extension from this node. However, we recall that at this node, too, the attested relational head nouns are typically very suitable to the general target person semantics; kinship terms and body part nouns are very common, and relatives and body parts are typically within a dative-marked human being’s sphere of control. To the extent that ID occurrences exist at all, they must be seen as extensions from the ID(slot) node again: If a construction can be used for expressing a reference point situation between a relational head noun and its slot filler, then it may be extended to other reference point situations, too. Finally, the occurrences at the ELAB node are a mixed bag, but at least some of them can be seen as extensions from the ELAB(slot) node. The connecting lines proposed for the dative construction are found in figure 58. If we combine all these connections in a single map (figure 59), then, we have the beginnings of an implicational semantic map enriched by the dis-

178 Synchrony and diachrony

ID ID(slot) TYPE LABEL

ELAB(slot) ELAB

ELAB(part)

Figure 59. The beginnings of an implicational map of the semantic possessive space

tances from the correspondence analysis plot. The strength of each connection is indicated by the weight of the line. A dotted line indicates a connection due to one single construction, a dashed line indicates a connection shared by two constructions, a fully drawn line indicates a connection shared by three constructions, and a bold line indicates a connection shared by all four constructions. Again, given the small and uniform data set, these are very small beginnings. They may nevertheless give rise to some interesting predictions. One will be that we will expect many languages with shared coding of ID(slot) and ELAB(slot). Another will be that if TYPE is coded like ID and ELAB(slot), it must be possible to code LABEL that way, too. A third prediction could be that if ID is coded like ELAB(part), ELAB(slot) and ID(slot) should be coded that way too. Whether those predictions bear up remains to be tested.

Chapter 8 Concluding remarks

In this book we have established a fine-grained map of the possessive semantic space based on previous literature on adnominal possession, with seven map nodes based on a cross-classification of constructional meanings and properties of head nouns. In order to be able to express semantic distance, the nodes were configured by way of a correspondence analysis of the data set from the case study of early Slavic possessives on which this book is based. The resulting map does thus not pretend to universality, but is rather to be understood as a modest beginning to a better-supported map of the possessive semantic space. This book was intended as a close, empirically driven study of the relationship between form and meaning as exhibited by the Slavic possessive data. To achieve this end, the semantic map was exploited in several ways. In chapter 3, the boundaries and semantic centres of gravity of the five main possessive constructions in 11th–14th century Old Russian were plotted onto the map. We saw that although all five constructions overlapped considerably, none of them overlapped completely, and each construction had a distinct semantic centre of gravity. In chapters 4–6, the map nodes were examined one by one, with synchronic surveys of the OCS and 11th–14th century samples, and a diachronic overview of the further development in the history of Russian. Here, we noted that the distributions at each node differed considerably in the early text samples; whereas the ADJ1 construction was dominant in reference point situations, i.e. at the two ID nodes, datives and unmodified genitives were wellrepresented at ELAB(slot), as non-reference point arguments of relational nouns. Only two nodes were dominated by a single construction: For the categorising semantics at the TYPE node, the ADJ2 construction was virtually the only option, whereas intrinsic relationships between a part noun and a slot filler at the ELAB(part) node were uniformly expressed by a formally non-restricted genitive construction. Evaluating the map as a tool for synchronic semantic analysis in chapter 7, we saw that it captures important features of most of the constructions posited. It does, however, fail to capture one of the major characteristics of the ADJ2 construction, namely the fact that it is strongly driven by the properties of the possessor. The ADJ2 con-

180 Concluding remarks struction is thus found scattered across most of the possessive semantic space, but it does have a clear centre of gravity at the TYPE node. We also found that variation at each node was mainly driven by properties of the possessor. In chapter 7, the map was used as a diachronic tool. The diachronic changes to the constructions’ boundaries and semantic centres of gravity were plotted on the map, and diachronic paths for each construction were suggested. The diachronic paths were then combined to draw a tentative, weighted implicational map of the possessive semantic space. As a diachronic tool, the semantic map was able to capture slow, semantic drift in a set of competing syntactic constructions, which is characteristic of the changes that happened to the Old Russian possessive constructions. This semantic drift was accompanied by a loss of schematicity in one of the major constructions in the semantic space, namely the ADJ1 construction, which can be accounted for in a constructionist approach to syntactic productivity. I believe this type of approach may be suitable for many cases of syntactic change.

Appendix: Corpus and method

For the purposes of both the synchronic and diachronic aims of this book, a fairly extensive and representative text sample was deemed necessary. However, there is no such thing as a standard corpus that should be used for any diachronic study of Old Russian. The size and composition of the corpus needs to be selected in accordance with the nature and frequency of the specific phenomena under scrutiny. The Old Russian and OCS possessive constructions being studied in the present book are highly frequent in all styles of writing, but they have different distributions in different genres. This justifies the selection of a medium-sized, but stylistically varied text sample in order to provide a fair picture of the first documented situation in Old Russian and in OCS, as well as of the further development of the Old Russian possessive constructions.

1. Text selection principles 1.1. Fair representation of each period The size of the text samples was adjusted to the time span of each selected period. A longer time span thus requires a larger text sample, distributed evenly over the entire period. The Old Russian texts were split into four samples. The 11th–14th century sample, which spans 400 years, is naturally considerably larger than those of the 15th, 16th and 17th centuries. The 17th century text sample is larger than those of the 15th and 16th centuries, to make the final conclusions more certain, and because many of the changes become more pronounced in the 17th century. The canonical OCS sources span a much shorter period of time than the Old Russian material, and the sample is about one third of its size. To be able to state the extent of each text clearly, the texts were measured by standard pages of 2000 characters (including blank spaces). For the text volume and number of occurrences in each sample, see table 30.

182 Appendix: Corpus and method OCS 11th–14th century Old Russian 15th century Russian 16th century Russian 17th century Russian

standard pages 130.3 213.0 76.0 59.9 81.9

number of occurrences 1275 2351 713 520 1053

Table 30. Page volume and number of excerpted occurrences, all text samples

1.2.

Geography

In the case of OCS, geographical considerations make it necessary to select texts with extant manuscripts from both the Ohrid (Macedonian) and Preslav (East Bulgarian) schools, which, although linguistically very close, do have some differences (cf. e.g. Schenker 1995:188). In the case of Old Russian, the oldest texts are naturally mostly Kievan in origin. From the 13th century onwards, the selected texts are generally from central/northern Russia, in order to avoid Belarusian and Ukrainian linguistic features. Ideally, the author of each text should be a native speaker of the language in which he writes. This is fairly easily achieved with the Old Russian original texts, but is probably not always the case with the OCS translations.

1.3. Literary genres The text samples were selected to give fair representation of the main literary genres in the language and period in question. The language in general, and in particular the frequency of the constructions sought, may vary greatly from genre to genre. This is particularly important in Old Russian, where the proportion of Church Slavic linguistic elements is highly genre-dependent. The OCS text sample, on the other hand, must include both biblical (selections from the Gospels) and non-biblical texts (vitae), as the translation technique may differ.

1.4. Text editions and manuscripts Whenever possible, linguistic editions of the texts were selected, preferably with textological comments. If no such edition was available, editions made

Text selection principles

183

primarily for e.g. historians were used. These editions normally have simplified orthography, but this is unlikely to affect the syntax much. Some charters and documents were taken from Obnorskij and Barxudarov’s excellent collection of Old Russian texts (1999), rather than from specialised editions. Editions where corrections have been introduced without comment have been avoided. Where there was a choice, the edition using the older and/or more reliable manuscript was preferred.

1.5. Late copies of early manuscripts A pervasive problem in dealing with early Slavic is that many texts exist only in quite late copies, although it is well substantiated that they were originally composed centuries earlier. Most of the earliest Old Russian texts are preserved in copies that are several centuries younger than the lost original. Likewise, the canonical OCS texts are mostly extant in 11th century copies, but the original translations date back to the ninth and tenth centuries. Is it, then, justifiable to use such late copies as sources to the syntax of the language of the time when the text was composed? Uspenskij (1987:57) takes a maximally strict position when it comes to the Old Russian data, claiming that we should not use these late copies at all. His main argument is that the notion of personal authorship did not exist in Russian mediaeval literature; there was no clear distinction between copying and revising texts. One must assume that quite a lot of changes may have found their way into the copies, and it is a fact that such changes sometimes affect syntax as well as phonology and morphology. We know e.g. that non-prepositional locatives in the Codex Laurentianus (1377) are replaced by prepositional locatives in the only slightly later Codex Hypatianus (ca. 1425). Nonetheless, I have chosen to select texts by the time of creation rather by the date of the manuscript. There are several reasons for this. Firstly, we must ask whether scribal changes are likely to have had much impact on the constructions in question. We have seen that the scribes sometimes did change syntax and were not unlikely to change case constructions. A comparison of older and newer copies of various texts shows, however, that possessive constructions were little affected: The textual variants from 16th century copies in the edition of Russkaja pravda (RP) based on a manuscript from 1282 (Karskij 1930) provide very few variants involving possessive constructions in a wide sense, and these variants are almost exclusively found in

184 Appendix: Corpus and method new headings added in the later copies. A look at the textual variants from later copies in Abramoviˇc’s edition of Skazanie o Borise i Glebe (SBG) based on the Uspenskij sbornik of the 12th–13th century (M¨uller 1967, reprint edition with M¨uller’s corrections and comments in extensive footnotes) gives almost identical results. Of course, early texts only extant in late manuscripts cannot be checked in the same way, but it seems unlikely that the situation should be very different. It is certainly a possibility that late copyists might have felt tempted to substitute adnominal datives with genitives, and one should be very cautious of drawing far-reaching conclusions from unusual examples found in texts extant only in late copies. Nevertheless, I believe that these texts provide an (admittedly somewhat blurred) picture of the syntax of the original author, far more than of that of the copyist. Secondly, the scarcity of early manuscripts of Old Russian original texts is an important argument in favour of including late copies of early texts. From the 11th century, very little material exists: a few inscriptions, marginal notes and a few birch bark letters. From the 12th century there is some more, including charters (gramoty) and the important Uspenskij sbornik. From the 13th and 14th centuries there is a fair number of manuscripts, such as the Codex Laurentianus. Many of the 13th and 14th century manuscripts actually render 11th and 12th century texts, and, more importantly, many of the very earliest texts known to us are extant only in even later manuscripts. The extreme case is the 12th century Slovo o polku Igoreve (SPI), of which our only sources are actually a manuscript copy from 179383 and a printed edition from 1800 – a late 15th/early 16th century manuscript was lost when Moscow burnt in 1812. I have chosen to include some of the most important texts, even when they were extant only in quite late manuscripts, since a lot of text otherwise would have to be excluded from consideration altogether. After all, a 15th century manuscript of an 11th century text can hardly be considered a reliable source of the language of the 15th century. The problems inevitably connected with such a practice are lessened by the fact that the 11th through 14th centuries are treated as one synchronic stage of Old Russian.

The Old Russian text samples

185

2. The Old Russian text samples 2.1. Text genres Old Russian literature is traditionally divided into three main genres, or rather groups of genres (see for instance Vlasto 1988:346–349), which may be termed the religious genre, the narrative genre and the business/legal genre respectively. It is essential to have all three genres represented in a balanced way, as they correlate neatly with the degree of Church Slavic linguistic influence on Old Russian. The religious texts – not only the translated liturgical texts, but also to a lesser extent original texts such as saints’ vitae and homilies – are written in comparatively pure Church Slavic, which to some degree shows signs of East Slavic influence. The business/legal texts, on the other hand, are written in relatively pure East Slavic, though traces of Church Slavic are evident even here. These texts are primarily law codes, donation charters and treaties. The third genre, broadly called narrative texts and mainly consisting of chronicles, occupy an intermediate position between the two other genres. These texts are written in a fluctuating mixture of East Slavic and Church Slavic – even within a single text the language can vary considerably, depending on the subject matter. Two important questions should be asked, pertaining to the two extremes of this linguistic continuum: a) Should the religious texts be used at all, being so strongly permeated by Church Slavic features? b) Ought not the business/legal texts be used as the main body of evidence, since this apparently is the purest extant source of Old Russian? The first question is one to which different scholars have different answers. Many have excluded a large part of the religious text from consideration in their studies of Old Russian syntax, while others have included them. Unsurprisingly, this practice has influenced the results. Scholars working on historical syntax who have kept such texts in their corpora include Bratishenko (1998) and Krys’ko (1997), and this is my choice as well. The answer to the second question is much simpler. The language of the business/legal texts is rigid, brief and formulaic, and many important constructions are very rare in the texts – not because these constructions were rare in Old Russian, but simply because of the style and subject matter of this genre. No matter how large the text sample, it is unlikely that a sufficient number of good examples of, for instance, genitive constructions headed by deverbal nouns would be found. Therefore the proportion of business/legal

186 Appendix: Corpus and method texts in the text samples is smaller than the proportion of religious and narrative texts, simply because very few interesting examples pertaining to the main questions of this book will be found no matter how many pages of such texts are read. For the same reasons, the Novgorod birch bark letters are excluded altogether, since the constructions examined in this book are so rarely found in them. In the later Russian samples, the three main genres dissolve and change. Particularly in the 17th century, all sorts of new types of texts appear (see e.g. Kuskov 1989:14). Linguistically, these texts are often written in very colloquial Russian, and are therefore most closely related to the business/legal texts. Texts that might be described as publicistic are ascribed to this genre, whereas the fictional tales that start appearing are taken to be narrative texts. Even the religious texts are far more colloquial in this period.

2.2. Periodisation of Old Russian texts The 11th–14th centuries are considered a single language stage and therefore represented by a single text sample. This is the usual practice, this period traditionally being labeled drevnerusskij jazyk (literally Ancient Russian), whereas the 15th through 17th centuries are labeled either starorusskij jazyk (literally Old Russian) or srednerusskij jazyk (Middle Russian). This periodisation does away with some of the problems connected with late copies of early texts (see section 1.5), since many (but not all) of the late copies used fall within its boundaries. In my opinion, this is a sufficient argument in favour of this grouping. The main disadvantage of such a periodisation is the risk of constructing “a grammar which generates too many ‘generations’ of the language, and as a consequence treats chronologically separate forms as synchronic variation” (Faarlund 1990:16). There are indeed good arguments to expect linguistic differences between the first and the second half of this period (the Kievan period ends, the Tatar invasions are followed by consolidation under Moscow; Old East Slavic splits into Russian, Belarusian and Ukrainian), but the source problems make these differences difficult to discern in a reliable way. This first, long period is represented by eight text samples from each genre. As the business/legal texts are naturally shorter, the bulk of those texts (22.3 standard pages) is considerably smaller than the bulk of religious texts (111.3 standard pages) and that of narrative texts (79.4 standard pages).

The OCS text sample

187

The 15th and 16th centuries are both represented by 1–3 texts from each genre, whereas the 17th century is analysed in more depth, with three religious texts, three narrative texts and five business/legal/publicistic texts. This overrepresentation is justified by the many interesting features of the 17th century: the appearance of more colloquial texts and new genres, the increased influence of Western European languages, and the Nikonian writing reforms, where possessive dative and adjective constructions were actively corrected into genitive ones (Uspenskij 1987:302–306).

3. The OCS text sample The main focus of this book is not OCS, but Old Russian possessive constructions. The OCS text sample is therefore smaller than it would have been if this were an independent study of OCS possessives. The aim of the analysis is to establish similarities and differences between the OCS system and the Old Russian system as far as possible, in order to assess the possible influence of OCS on Old Russian, and also the diachronic relationship between the two. OCS is generally considered to be closer to Common Slavic than Old Russian, unsurprisingly, as the extant OCS texts from the 10th and 11th centuries (copies of 9th century texts) are considerably older than most of the Old Russian (original) texts. The OCS sample is about two-thirds of the size of the 11th–14th century Old Russian sample. It is smaller not only because of the contrastive aim of the analysis, but also because the OCS texts cover a shorter period of time, and because they are less varied linguistically and stylistically, and because all of them belong to the religious genre. The 1275 occurrences document all the main construction types sufficiently to get a good picture of the OCS situation. All the texts chosen are from the OCS canon, and all texts are translations from Greek. All texts have been read in parallel in Greek and OCS, and the features of the corresponding Greek construction were registered in the database. Care has been taken to include both texts with extant manuscripts of the Preslav (East Bulgarian) school and texts with extant manuscripts of the Ochrid (Macedonian) school, as these differ to some extent. An important point to consider here is that the possessive dative is often considered a Bulgarianism, and that it is known to be more frequent in the East Bulgarian

188 Appendix: Corpus and method Savvina kniga than in Codex Marianus, Codex Zographensis and Codex Assemanianus, which all have manuscripts of Macedonian origin (see Veˇcerka 1963:200 for figures). This is a strong argument for including Savvina kniga, which might otherwise have been excluded on the grounds that the translation has clearly been much modified in the copying process. Both biblical and non-biblical texts have been included, as it has been suggested that biblical texts are translated more literally than other texts. The biblical texts are all excerpts from the Gospels. As far as possible, different excerpts have been chosen from each manuscript, but there are some overlaps. The non-biblical texts are two vitae from the Codex Suprasliensis. It has been suggested that some texts in the Codex Suprasliensis have more archaic features than the others (Blagova 1980:120–121), and this has been taken into consideration when texts were selected: Vita nr. 11 (Passion of St. Sabinos) is one of the texts claimed to be more archaic, whereas vita nr. 16 (Life of Isaac) has been assigned to the newer layer of texts.

4. Excerpation, data registration and citation All instances of noun phrases with dependent genitives, datives or denominal adjectives were excerpted, with the following restrictions: – Only constructions with an overt head were excerpted – Constructions headed by numerals or quantifier nouns (such as “multitude”) were excluded (since dependent nouns in such constructions are consistently genitive-marked) – Constructions with bare genitive- or dative-marked prounoun dependents were excluded – The selection of denominal adjectives was limited as described in 3, section 1.1 and 1.2 – Clearly qualitative adjectives were excluded Texts of the Old Russian religious genre particularly, but also texts of the narrative genre, quite often contain quotations from other texts, usually religious Church Slavic texts. Such quotations are treated like regular text, but very special features in obvious quotations are of course treated with caution. All occurrences were registered in a FileMaker database. Each record provides enough context to be intelligible, as well as an English translation. The entry types are described in table 31. After registration, the data were ex-

Excerpation, data registration and citation

189

ported as comma-separated files before being imported and further processed in the statistical software package R. construction type source reference construction meaning construction submeaning period noun type relation type excerpt translation comment genre possessor animacy possessor specificity possessor stem type LABEL subtype Greek construction type Greek excerpt long or short form Greek head definite? Greek possessor definite?

Old Russian base yes yes page and line yes yes yes yes yes yes yes yes yes ADJ2, GENUNMOD, dative ADJ2, GENUNMOD, dative for ADJ1 LABEL occurrences no no no no no

OCS base yes yes book, chapter and verse yes yes no yes yes yes yes yes yes ADJ2, GENUNMOD, dative ADJ2, GENUNMOD, dative no LABEL occurrences yes yes a subset of the occurrences a subset of the occurrences a subset of the occurrences

Table 31. Database fields per record, Old Russian vs. OCS database

The examples cited in the book are all transliterated, but otherwise given in the orthography of the edition that was used. Supralinear letters, letters with abbreviation marks and renditions of these in the editions are all indicated by underlining in the examples. Examples have simplified glossing, with only highly relevant morphological features glossed (see the list of abbreviations). Other features are indicated by rough translation in the glosses. References to texts give the abbreviated title (see section 5), the page number of the edition and the number of the line on that page, divided by a slash (e.g. AN 35/10). If there are two text columns on a page, the left one will be referred to as a and the right one as b, unless the columns are numbered separately (e.g. 24a/15). When the edition gives line numbers in accordance with the manuscript, these numbers are used in the reference. The 17th century letters and documents taken from Kotkov, Oreˇsnikov and Filippova (1968) had their lines numbered text by text rather than page by page (e.g. GG a:2). References

190 Appendix: Corpus and method to New Testament texts are given by manuscript, gospel, chapter and verse (e.g. Zogr. John 5:12). The examples are always given with some context if this is necessary to understand them properly.

5. Text excerpts 5.1. Old Russian AN

Xoˇzenie za tri morja Afanasija Nikitina. In B.D. Grekov and V.P. AdrianovaPeretc (eds.) 1948: Xoˇzenie za tri morja Afanasija Nikitina 1466–1472 gg. Moskva: Izdatel’stvo Akademii Nauk SSSR. 15th century. 21.3 standard pages.

BNS

Skazanie o bitve novgorodcev s suzdal’cami. In D.S. Lichaˇcev, L.A. Dmitriev, A.A. Alekseev and N.V. Ponyrko (eds.) 1999: Biblioteka literatury Drevnej Rusi. Tom 6. Saint Petersburg: Nauka, 444–446. 11th–14th century. 2.6 standard pages.

ˇ CBK

ˇ Celobitnaja A. Bermackogo na streleckogo golovu P. Krasnogo, kotoryj klepal ˇ ego poˇzogom i oglaˇsal naprasno. 7 June 1633./ Celobitnaja streleckogo golovy P. Krasnogo na A. Bermackogo, obvinjavˇsego ego v zaˇzoge dvora. 7 June ˇ 1633./ Celobitnaja A. Bermackogo ob oborone ot vorovskogo umyslu P. Krasnogo. June 1633. In S.I. Kotkov, A.S. Oreˇsnikov and I.S. Filippova (eds.) 1968: Moskovskaja delovaja i bytovaja pismennost’ XVII veka. Moscow: Nauka, 48–51. 17th century. 4.8 standard pages.

ˇ CL

ˇ Celobitnaja tjagleca Ogorodnoj slobody F. Larionova na A. i R. Michajlovyx v boe, uveˇc’e i lomanii okonˇciny. 10 August 1698. In S.I. Kotkov, A.S. Oreˇsnikov and I.S. Filippova (eds.) 1968: Moskovskaja delovaja i bytovaja pismennost’ XVII veka. Moskva: Nauka, 123–124. 17th century. 1.5 standard pages.

D

ˇ svadebnoj”) from Domostroj. In Domostroj 1971. Rarity reprints Excerpt (“Cin No. 18. Letchworth: Bradda Books, 170/6–193/23. 16th century. 14.3 standard pages.

DG

Duxovnaja V.Ja. Goloxvastova. 6 December 1678. In S.I. Kotkov, A.S. Oreˇsnikov and I.S. Filippova (eds.) 1968: Moskovskaja delovaja i bytovaja pismennost’ XVII veka. Moskva: Nauka, 192–197. 17th century. 9.2 standard pages.

DIG

Duxovnaja Ivana Jur’eviˇca Grjaznogo 1579–1580 gg. In S.P. Obnorskij and ˇ S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii russkogo jazyka. Cast’

Text excerpts

191

pervaja. 3-e izdanie. Moscow: Aspekt Press, 243–244. 16th century. 1.3 standard pages. DIK

Duxovnaja gramota Moskovskogo knjazja Ivana Daniloviˇca Kality. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii russkogo ˇ jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 97–100. 11th–14th century. 2.5 standard pages.

DSG

Duxovnaja gramota velikogo knjazja Simeona Gordogo. In S.P. Obnorskij and ˇ S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 102–104. 11th–14th century. 1.1 standard pages.

DTN

Dogovor Tverskogo knjazja Mixaila Jaroslaviˇca s novgorodcami okolo 1294– 1301 gg. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: Xrestomatija ˇ po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 69–70. 11th–14th century. 0.9 standard pages.

GG

T.I. Golicyna V.V. Golicynu. 6 August, 13 August–22 September 1677. In S.I. Kotkov, A.S. Oreˇsnikov and I.S. Filippova (eds.) 1968: Moskovskaja delovaja i bytovaja pismennost’ XVII veka. Moskva: Nauka, 16–21. 17th century. 7.1 standard pages.

IG/VG

Perepiska Ivana Groznogo s Vasiliem Grjaznym. Letters I, II and III. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii russkogo ˇ jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 260–268. 16th century. 7.6 standard pages.

MG

Mstislavova gramota. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: ˇ Xrestomatija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, p. 40. 11th–14th century. 0.4 standard pages.

PBKL

Povest’ o bitve na reke Kalke. In Codex Laurentianus 1962 = Polnoe sobranie russkix letopisej. Tom pervyj: Lavrent’evskaja letopis’ i Suzdal’skaja letopis’ po Akademiˇceskomu spisku. Moscow: Izdatel’stvo Vostoˇcnoj literatury, 445/31–447/9. 11th–14th century. 1.1 standard pages.

PBKN

Povest’ o bitve na reke Kalke. In Novgorodskaja cˇ etvertaja letopis’ 1915 = Polnoe sobranie russkix letopisej. Tom 4: Novgorodskaja cˇ etvertaja letopis’ ˇ Cast’ 1, vypusk 1. Petrograd: M.A. Aleksandrov, 201/11–203/6. 11th–14th century. 1.3 standard pages.

PBR

Povest’ o razorenii Batyem Rjazani. In L.A. Dmitriev and D.S. Lixaˇcev (eds.) 1969: Sbornik proizvedenij literatury drevnej Rusi. Moscow: Xudoˇzestvennaja literatura, 344–360. 11th–14th century. 9.4 standard pages.

192 Appendix: Corpus and method PBSV

Povest’ o bolezni i smerti Vasilija III. In Novgorodskaja cˇ etvertaja letopis’ 2000 = Polnoe sobranie russkix letopisej. Tom IV. Novgorodskaja cˇ etvertaja letopis’. Moscow: Jazyki russkoj kul’tury, 552/18–564/21. 16th century. 18.3 standard pages.

PGMK

Pravaja gramota Mixailu Kaznakovu po kabale v dvux s polovinoju rubljax. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii ˇ russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 358–359. 17th century. 2.1 standard pages.

PJul

Povest’ o Julianii Lazarevskoj. In T.R. Rudi and R.P. Dmitrieva (eds.) 1996: ˇ Zitie Julianii Lazarevskoj (Povest’ ob Ul’janii Osor’inoj). Saint Petersburg: Nauka, 103/12–115/2 (heading and tale of miracles omitted). 17th century. 7.9 standard pages.

PKS

Povest’ o Karpe Sutulove. In L.A. Dmitriev and D.S. Lixaˇcev (eds.) 1988: Pamjatniki literatury Drevnej Rusi. XVII vek. Kniga pervaja. Moscow: Xudozˇ estvennaja literatura, 65–70. 17th century. 7 standard pages.

PMM

Povest’ o Marfe i Marii. In L.A. Dmitriev and D.S. Lixaˇcev (eds.) 1988: Pamjatniki literatury Drevnej Rusi. XVII vek. Kniga pervaja. Moscow: Xudoˇzestvennaja literatura, 105–111. 17th century. 7.9 standard pages.

PoslMD

Poslanie mitropolita Daniila. In Grigorij Kuˇselev-Bezborodko (ed.) 1862: Pamjatniki starinnoj russkoj literatury. Vypusk 3. Saint Petersburg, 194–199. 16th century. 8.9 standard pages.

PSG

Pskovskaja sudnaja gramota. In S.P. Obnorskij and S.G. Barxudarov (eds.) ˇ 1999: Xrestomatija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 215–218. 15th century. 1.9 standard pages.

PVC

Povest’ o vzjatii Car’grada. In Arximandrit Leonid 1886: Povest’ o Car’gradˇe (ego osnovanii i vzjatii Turkami v 1453 godu). Saint Petersburg. 15th century. 30.6 standard pages.

PVM

Pouˇcenie Vladimira Monomaxa. In Codex Laurentianus 1962 = Polnoe sobranie russkix letopisej. Tom pervyj: Lavrent’evskaja letopis’ i Suzdal’skaja letopis’ po Akademiˇceskomu spisku. Moscow: Izdatel’stvo Vostoˇcnoj literatury, 240/23–256/23. 11th–14th century. 11.8 standard pages.

PVrL

Excerpts from Povest’ vremennyx let. In Codex Laurentianus 1962 = Polnoe sobranie russkix letopisej. Tom pervyj: Lavrent’evskaja letopis’ i Suzdal’skaja

Text excerpts

193

letopis’ po Akademiˇceskomu spisku. Moscow: Izdatel’stvo Vostoˇcnoj literatury, 19/14–21/6, 25/10–29/2, 32/24–42/2, 54/16–60/17, 60/25–64/13, 109/1– 122/3, 130/12–131/29, 132/1–141/21, 155/26–160/25, 174/18–181/17, 214/25– 215/26. 11th–14th century. 35 standard pages. PXV

Prodaˇznaja Xon’ki Vaskovoj na Kalenikov monastyr’. In S.P. Obnorskij and ˇ S.G. Barxudarov (eds.) 1999: Xrestomatija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 144–145. 11th–14th century. 1 standard page.

RCAM

Excerpt from Grigorij Kotoˇsixin: O Rossii v carstvovanie Alekseja Mixajloviˇca. In Grigorij Kotoˇsixin 1980: O Rossii v carstvovanie Alekseja Mixajloviˇca. Text and commentary. Ed. A.E. Pennington. Oxford: Clarendon Press, 15/1–26/14. 17th century. 11.3 standard pages.

RP

Russkaja pravda. In E.F. Karskij (ed.) 1930: Russkaja pravda po drevnejˇsemu spisku. Leningrad: Izdatel’stvo Akademii nauk. 11th–14th century. 11.2 standard pages.

SAP

Excerpt (chapters 8–13) from Skazanie Avraamija Palicyna. In O.A. Derˇzavina and E.V. Kolosova (eds.) 1955: Skazanie Avraamija Palicyna. Moscow: Izdatel’stvo Akademii nauk. 17th century. 10.2 standard pages.

SBG

Skazanie o Borise i Glebe. In Ludolf M¨uller (ed.) 1967: Die altrussischen hagiographischen Erz¨ahlungen und liturgischen Dichtungen u¨ ber die heiligen Boris und Gleb. Munich: Wilhelm Fink Verlag. 11th–14th century. 28.7 standard pages.

SDZ

Slovo Daniila Zatoˇcnika. In N.N. Zarubin 1932: Slovo Daniila Zatoˇcnika po redakcijam XII i XIII vv. i ix peredelkam. Leningrad: Izdatel’stvo Akademii Nauk SSSR, 4–35. 11th–14th century. 6.9 standard pages.

SKT

ˇ Kirill Turovskij: Slovo vo svjatuju velikuju subotu. In I.S. Sljapkin 1889: Slovo Kirilla Turovskogo po vsem izvestnym spiskam. Pamjatniki drevnej pis’mennosti i iskusstva. Saint Petersburg. 11th–14th century. 3.8 standard pages.

SL

Excerpt from Suzdal’skaja letopis’. In Codex Laurentianus 1962 = Polnoe sobranie russkix letopisej. Tom pervyj: Lavrent’evskaja letopis’ i Suzdal’skaja letopis’ po Akademiˇceskomu spisku. Moscow: Izdatel’stvo Vostoˇcnoj literatury, 460/22–467/20. 11th–14th century. 5.8 standard pages.

SPI

Slovo o polku Igoreve. In L.A. Dmitriev and D.S. Lixaˇcev (eds.) 1967: Slovo o polku Igoreve. In the series Biblioteka po˙eta, bol’ˇsaja serija. Leningrad: Sovetskij pisatel’. 11th–14th century. 9 standard pages.

194 Appendix: Corpus and method SRG

Dogovornaja gramota Smolenskogo knjazja Mstislava Davidoviˇca s Rigoju i Gotskim beregom. In S.P. Obnorskij and S.G. Barxudarov (eds.) 1999: Xrestoˇ matija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, 51–57. 11th–14th century. 4.9 standard pages.

VXM

Vkladnaja Varlaama Xutynskomu monastyrju. In S.P. Obnorskij and S.G. Barxˇ udarov (eds.) 1999: Xrestomatija po istorii russkogo jazyka. Cast’ pervaja. 3-e izdanie. Moscow: Aspekt Press, p. 43. 11th–14th century. 0.4 standard pages.

XID

Xoˇzdenie igumena Daniila v svjatuju zemlju. In: Elisabeth Lfstrand (ed.) 1993: Xoˇzdenie igumena Daniila v svjatuju zemlju. Rukopis’ Stokgol’mskoj Korolevskoj biblioteki. In the series Acta Universitatis Stockholmensis. Stockholm Slavic Studies, 22. Stockholm: Almqvist & Wicksell. 11th–14th century. 13.9 standard pages.

Z

Zadonˇscˇ ina. In Pavel Simoni (ed.) 1922: Zadonˇscˇ ina po spiskam XV–XVIII stolˇetij. Vypusk III. In the series Pamjatniki starinnago russkago jazyka i slovesnosti. Petrograd: Rossijskaja gosudarstvennaja akademiˇceskaja tipografija. 11th–14th century. 3.5 standard pages.

ˇ ZAN

ˇ Zitie Aleksandra Nevskogo. In Ju.K. Begunov (ed.) 1965: Pamjatnik russkoj literatury XIII veka. “Slovo o pogibeli russkoj zemli”. Moscow: Nauka, 159– 180. 11th–14th century. 8.3 standard pages.

ˇ ZAvv

ˇ Excerpt from Zitie protopopa Avvakuma. In V.P Malyˇsev, I.S. Demkova and L.A. Dmitrieva (eds.) 1975: Pustozerskij sbornik. Avtografy soˇcinenij Avvakuma i Epifanija. Leningrad: Nauka, 17/30–31/3. 17th century. 12.9 standard pages.

ˇ ZD

Slovo o zˇ itii i o prestavlenii velikogo knjazja Dmitrija, carja Rus’skago. In Novgorodskaja cˇ etvertaja lˇetopis’ 1925 = Polnoe sobranie russkix lˇetopisej. ˇ Tom 4: Novgorodskaja cˇ etvertaja lˇetopis’. Cast’ 1, vypusk 2. Leningrad, 351/16– 366/22. 11th–14th century. 14.6 standard pages.

ˇ ZPF

ˇ Zitie Petra i Fevronii. In R.P. Dmitrieva (ed.) 1979: Povest’ o Petre i Fevronii. Leningrad: Nauka, 211/18–222/11. 16th century. 9.8 standard pages.

ˇ ZSP

ˇ ˇ Excerpt from Zitie Stefana Permskogo. In V. Druˇzinin (ed.) 1959: Zitie sv. Stefana episkopa Permskogo. Photomechanic reprint with an introduction by ˇ zevskij. In the series Apophoreta Slavica, II. The Hague: Mouton, Dmitrij Ciˇ 1/1–22/7. 11th–14th century. 16.6 standard pages.

Text excerpts ˇ ZZS

195

ˇ Excerpt from Zitie i cˇ judesa prep. Zosimy i Savvatija Soloveckix cˇ judotvorcev. In S.V. Mineeva (ed.) 2001: Rukopisnaja tradicija zˇitija prep. Zosimy i Savvatija Soloveckix (XVI–XVIII vv.). Teksty. Moscow: Jazyki slavjanskoj kul’tury, 7/13–22/38. 15th century. 22.2 standard pages.

5.2. OCS Ass.

Codex Assemanianus. In Josef Vajs and Josef Kurz (eds.) 1929/1955: Evangeliarium Assemani. Codex Vaticanus 3. slavicus glagoliticus. Prague. Excerpt: Gospel according to Matthew, chapters 1–26. 30.5 standard pages.

Mar.

Codex Marianus. In Vatroslav Jagi´c (ed.) 1960: Quattuor evangeliorum versionis Palaeslovenicae Codex Marianus glagoliticus, characteribus Cyrillicis transcriptum. Reprint. Graz: Akademische Druck- u. Verlagsanstalt. Excerpt: Gospel according to Luke, chapters 1–16. 31 standard pages.

Savv.

ˇ cepkin (ed.) 1959: Savvina kniga. Reprint. Graz: Savvina kniga. In Vjaˇceslav Sˇ Akademische Druck- u. Verlagsanstalt. Excerpts: Gospel according to Mark (the selection found in the manuscript, which is an aprakos; 5.6 standard pages); Gospel according to Luke, chapters 17–23 (3.4 standard pages); Gospel according to Matthew, chapters 20–28. 3.9 standard pages.

Supr.

Codex Suprasliensis. In Jordan Zaimov and Mario Capaldo (eds.) 1982–83: Supras˘ulski ili Retkov sbornik. Sofia: Bulgarska akademija na naukite. Excerpts: Passion of St. Sabinos, Codex Suprasliensis 11, 5.5 standard pages; Life of Isaac, Codex Suprasliensis 16, 10.7 standard pages.

Zogr.

Codex Zographensis: Vatroslav Jagi´c (ed.) 1954: Quattuor evangeliorum Codex glagoliticus, olim Zographensis nunc Petropolitanus: characteribus Cyrillicis transcriptum. Reprint. Graz: Akademische Druck- u. Verlagsanstalt. Excerpt: Gospel according to John. 29.66 standard pages.

Notes

1. I will use the term “Old Russian” throughout this book, although Old East Slavic is certainly a better term for the 11th–12th century texts, which are mostly Kievan in origin. 2. I.e. ADJ1 constructions in this book’s terminology. 3. E.g. Buslaev (1881/1959:421–422, 459–460, 464), Potebnja (1899/1968:383–390). 4. Borkovskij and Kuznecov (1963:422–432), Ivanov (1990:380–381), Lomtev (1956:438– 440, 453–470, 474–478), Sprinˇcak (1960:118–122, 131–140), Borkovskij (1968:79–89, 164–178, 197–204), Borkovskij (1978:149–159). 5. Miklosich (1883:7–17, 447–474, 605–611), Vondr´ak (1928:229–234, 319–320), Meillet (1934/2000:374–375), Vaillant (1958:595–605), Vaillant (1977:51–52, 87–88). 6. Flier (1974), Huntley (1984), Huntley (1993), Veˇcerka (1993:186–216). 7. Pravdin’s term is not “possessive”, but opredelitel’nyj “determinative”, which is rather close in spirit to the terminology used in this book. 8. This is a very brief and selective survey of the literature on Slavic possessives. For a detailed survey, see Eckhoff (2007:chapter 2). 9. See, however, the alienable:inalienable split in Nikolaeva and Spencer’s possession– modification scale (2010). 10. Cf. the classical treatment in Pustejovsky (1995), Partee’s many publications on possession; a Cognitive Grammar formulation in Taylor (1996:239), similarly in Stefanowitsch (2003), see also the overview in Barker (forthc.) and Barker and Dowty’s work on nonverbal proto-roles (1993). 11. The work of Pustejovsky (1995) on qualia structure and type shifting is designed for accounting for such shifts. For an account of possessives using an adapted version of qualia structure, see Jensen and Vikner (2004). See also Stefanowitsch (2003:435), Delsing (1993:147–148). 12. Jensen and Vikner (2004) would ascribe these cases to the constitutive qualia role. 13. In dealing with inalienable possession, Heine (1997:17) gives a German contrastive example: Ich zerriß meine Hose vs. Ich zerriß mir die Hose (both sentences are translated into English “I tore my pants”). The first example has a possessive pronoun, and does not tell us anything about whether the trousers were worn or not when they were ripped. The second example is a dative construction, the so-called “external possession construction”, which is reserved for inalienable possession: The trousers are conceptualised as a body part, and they must have been worn by the subject of the sentence in order for the construction to be possible. 14. Nouns denoting uniques must also be considered non-relational, but will rarely occur in possessive constructions. 15. The profile is defined by Langacker (1987:491) as “The entity designated by a semantic structure”, i.e. a particular facet of the semantic base of the word which is given particular prominence by the word. For instance, the semantic base of hypotenuse would be a rightangled triangle (cf. Taylor 2002:193). 16. For citation principles, see the Appendix section 4. 17. Taylor also assumes an aspectual distinction with these nouns, distinguishing between episodic and process nouns. Episodic nouns are “perfective” in the sense that they reify

198 Notes

18. 19.

20.

21.

22. 23. 24.

25.

a single instance or episode of the process. These nouns, like types a–e, tend to be established lexical items with idiosyncratic morphology and semantics. The so-called process nouns, on the other hand, are claimed to be “imperfective”, nominalising “an internally homogeneous process, rather than a completed event with its successive temporal stages” (Taylor 1996:269). They are derived with the suffix -ing, which is fully productive and completely regular, in contrast to the idiosyncratic morphology of the episodic nouns. Aspectual differences can also be found between deverbal nouns in early Slavic, but the distinction does not follow the dividing line between idiosyncratic and regular derivational morphology, since nouns can be derived from both perfective and imperfective verbs with the highly regular suffix -ie (cf. Dickey 2011), which is productive to the extent that Nilsson (1972), in an early generative account, takes these nouns and their modifiers to be derived directly from underlying sentences. I will not go further into aspectual differences between deverbal nouns. For an interesting discussion of transitional types between deverbal nouns and other relational nouns, see Barker and Dowty (1993). The treatment in this book differs specifically from that in Jensen and Vikner (2004) in that nouns that are considered to have an agentive qualia role on that account are considered relational in this book. For type c nouns, this means that I assume them to have an agent slot and a subject matter slot. For a different constructionist approach to possession in English, see Stefanowitsch (2003). Here the prenominal possessive is assumed to have the schematic meaning “possession”, and to impose the role “possessed” on the non-relational head noun and “possessor” on the modifier, rather similar to the control role in the approach of Jensen and Vikner (2004). When there is a relational head noun involved, however, these roles are overridden by those in the semantic frame of the relational noun (Stefanowitsch 2003:430). However, Stefanowitsch must modify this claim by specifying that in constructions with non-relational head nouns denoting persons, the interpretation of a construction will be one of kinship, not one of possession, such as in Martha’s girl. He also does not take into account the many cases where there is a non-relational, inanimate head noun, but where the meaning is not one of possession in a prototypical sense. Thus, Taylor’s account seems to cover the facts better. When a reference point is used to access a non-relational noun, the construction is given the interpretation that is most likely from the context – often one of possession in a strict sense, but certainly not always. In the figure, tr stands for “trajector” and lm for “landmark”. A trajector is understood as the more prominent participant in a relation, whereas the landmark is a salient substructure other than the trajector of a relational predication (Langacker 1987:490, 494). Langacker terms the argument slots “elaboration sites in the noun’s semantic base”. Cf. the control relation in Jensen and Vikner (2004) and the possessive semantic role in Stefanowitsch (2003). The semantics of the ID node thus correspond pretty closely to what Koptjevskaja-Tamm (2004:155) refers to as anchoring relations, where possessors act “as anchors or reference point entities for identification of the head”. Anchoring relations are opposed to non-anchoring relations, “in which the nominal dependent is to classify, describe or qualify the class of entities denoted by the head”. At least the TYPE and non-relational ELAB node in the present account can be subsumed under her non-anchoring relations. For an extensive discussion of the facts in English, see Taylor (1996:184–195).

Notes

199

26. In this book I do not take an epistemic approach to specificity, as Taylor (1996:185) does. A non-specific NP is not deemed to be one that is unidentifiable to the speaker, but rather one that exists only inside certain embeddings, such as negation, modality, quantification etc. They cannot be referred to from outside this embedding. In the thief example, the strong man cannot be referred to outside the conditional clause; the example cannot be continued thus: *The strong man bought a new lock yesterday. 27. By kind-referring NP I understand an NP that “does not refer to an ‘ordinary’ individual or object, but instead refers to a kind”, a genus (Krifka et al. 1995:2). Krifka et al. (1995:11) go on to problematise what “a well-established kind” might be; especially for plurals, kinds and non-specifics are difficult to distinguish, and I have not attempted to make a sharp distinction in my analysis, but have often collapsed kind-referring and nonspecific possessors. 28. Note that the head noun may still be relational, but in such cases the inherent relation has been overridden by context. 29. “The addressee is able to access a representation of the type of object described by the expression.” (Gundel, Hedberg and Zacharski 1993:276). To denote an individual, an expression must be at least referential on Gundel et al.’s givenness hierarchy. 30. The drawback of maps generated without semantic preanalysis is that they may be hard to interpret; the dimensions in the map may correspond to no established linguistic category, and sometimes to complex combinations of such categories. 31. Experiments with including the Greek correspondences to the OCS data set in the correspondence analysis produced essentially the same plot, even when cross-classifying the Greek data by definiteness marking on the head and possessor nominals respectively. 32. The correspondence analysis was performed in R with the corres.fnc function in the languageR package, as described in Baayen (2008:128–134). 33. The two plotted factors account for 79.1 % of the variation in the data set. 34. Note that these lines are not implicational; they do not carry hypotheses about whether function B must be encoded in the same way as function A if function C is, in the style of classical hand-drawn maps. An attempt at an implicational map is made in chapter 7. 35. The connecting lines do not take diachrony into consideration, it is enough that e.g. the dative construction can code two nodes in one of the text samples. To reduce noise, a construction had to have more than three attestations at each of the nodes in at least one period sample to count as a connection. There were no node pairs connected by two or four constructions. 36. The Greek text used for the Gospel texts is the standard Nestle-Aland New Testament edition (Aland et al. 1998), which is not an entirely unproblematic choice. To compensate for this, whenever the OCS text clearly coincided with one of the Greek variant readings instead of the main text, that reading was chosen instead. In most of these cases, the OCS text rendered a passage that had been dropped from the Nestle-Aland. For the Suprasliensis vitae, the Greek texts of Zaimov and Capaldo (1982–83) were used. Since these Greek texts are amalgamations of various Greek manuscripts tailored to fit the OCS translation, this is even more problematic. We must thus be aware that these data come with a considerable amount of noise. 37. For a further problematisation of the distinction between possessive and relative adjectives in Old Russian, see Zverkovskaja (1986:4–5) and Flier (1974:73–82). 38. There is in fact a single, somewhat spurious TYPE example with an -žn’- suffix, see page

200 Notes

39.

40.

41. 42. 43.

44.

45. 46.

47. 48.

49. 50. 51. 52.

133, but I will stick to the principle that there must be at least three occurrences of a construction at a node to include that node withinin the boundaries of the construction. Its derivates in Old Russian include those with stems in d, t, s, z, b, m, v that were retained without iota alternations. In cases where there were stem-final velars or the sonorants r, l, n (džjaˇcii, dˇeviˇcii, igumenii), the derivates may be considered either long (pronominal) adjective forms with the suffix -j- or forms with -žj- (see Zverkovskaja 1986:10). A typical formulation of the complementary distribution rule is that given by Makarova (1954:11–15), who says that the genitive could be used a) when it had an attribute, b) when it had an apposition, c) when realised by a proper name consisting of more than one word, c) with substantivised adjectives or participles, and d) in cases where something belonged to two or more persons. The commas indicate free constituent order. See Payne and Barshi (1999). There were two dative occurrences at ELAB(part) in the 11th–14th century Old Russian sample. In the OCS sample, however, there were seven occurrences, which is why ELAB(part) is connected to the ID nodes and the other ELAB nodes by three constructions, not two. All the constructions in question were headed by nouns that could be perceived either as deverbal nouns or nouns denoting inherent parts of wholes, such as beginning, end, see section 2. This supports Borkovskij’s (1968:199) notion that a “dative of relation” (datel’nyj otnoˇsenija) may be distinguished in the Old Russian text material, expressing “kinship, relationships of friendship or enmity, domination or submission etc.” Including God and other divinities. Interestingly, the adjective cˇesare˛ “king’s” clearly agrees in case and gender with uzdy “bridle”, even though it is quite clear from the Greek original that it actually modifies “horse”. The shares of ADJ1 constructions vs. the totals are approaching significance, however, p-value = 0.06322. This analysis borrows the notion of a personal sphere from Wierzbicka (1988), and is rather similar to Fillmore’s definition of the dative case role as an animate being that is conscious of being affected by the state or event expressed by the verb (1968:24–25) ˇ Both examples come from a chain of epithets praising Dmitrij Donskoj in his vita (ZD), with a very high proportion of datives, which is typical for such effusions. It is, however, easy to read tear as a relational noun, as tears are necessarily the product of crying, and therefore normally invoke reference to a subject. For an analysis of ADJ1 constructions and their interaction with other possessives in ˇ modern Russian, see Kopˇcevskaja-Tamm and Smelev (1994). Most of the pronoun-adjective constructions in the material are clearly letter formulae referring to high-ranking persons (see Borkovskij 1968:88–89), but there are also sporadic other examples.

(302)

na moi Vasilevż dvorż on my Vasilij-OV court ‘to my, Vasilij’s court’ DG 194/33, 17th century

53. Borkovskij (1968:166) reports an increased tendency to use genitives instead of denominal adjectives in the 18th–19th centuries. Richards (1976) sees a clear increase in the

Notes

54.

55. 56.

57. 58.

59. 60.

61. 62. 63.

64.

65. 66. 67. 68.

69.

201

share of bare genitives in the 18th century, and finds denominal adjectives and bare genitives in stylistic variation throughout the 19th century. Only Richards backs her claims with quantified data. None of them really define the concept of possession, making it hard to compare their claims to the findings of the present book. The preface to V.I. Lukin’s comedy Mot, ljuboviju ispravlennyj (1765; Makogonenko ˇ (1970:145–151) 13.7 standard pages), and an excerpt from M.D. Culkov’s tale “Prigoˇzaja povarixa” (1770; Makogonenko (1970:185–191), 14.5 standard pages). Recall that non-reference point occurrences headed by part nouns are classified at the ELAB(part) node. http://foni.uio.no:3000/; http://www.hf.uio.no/ifikk/proiel/ The corpus has a fully annotated text of the Codex Marianus, which is tagged for morphology, syntax, animacy, noun types (relationality) and adjective stem types. The text is automatically aligned with the Greek NT, which is also annotated for morphology, syntax, animacy and several other semantic features, and is currently in the process of being annotated for information status and anaphoric relations. The information status data for OCS are therefore sourced from the Greek annotations via the token alignments. This last example is clearly bordering on TYPE. Note that this construction reflects a known Hebraism in the Greek text: persons who have a quality are described as children of that quality (Moule 1953:174). Depending on how literally we take that metaphor, the three kinship term examples could alternatively have been classified as ELAB occurrences. One of the occurrences has ostavlenie instead of otżpuˇstenie (Ass. Matthew 26:28). Given table 23, one might consider looking at deverbal nouns alone, since deverbal nouns with object slots are the most obvious exception from the complementary distribution “rule” at the ELAB(slot) node. I have, however, chosen to proceed as in the analysis of OCS. There are, however, a fair number of occurrences like these deemed to belong to the LABEL node, see chapter 6, section 2. i posla voevodu svoego Nevrunja povoevati zemlju Suˇzdalžskuju “and sent his commanˇ der Nevrjun to fight against the land of Suzdal”’ ZAN 174/17–175/1. Example (190) is from a string of invocations, several of which are deverbal agentive nouns with a dative slot filler. Such strings are rather typical in religious style, and there are seven such occurrences among the dative constructions at the ELAB(slot) node in the Old Russian sample. The fluctuations may be a result of the size of the samples, they may simply be too small to capture a trend for this subgroup. A related possibility is that they are due to subject matter differences. 83 out of 295 in the 11th–14th century sample vs. 57 out of 131 in the 17th century sample, p-value = 0.0383, Fisher’s exact test, two-sided. The OCS text is not particularly close to the Greek original, which is to be translated “he reached the evening of the end”. The Greek text has just “any male”. By rights these occurrences might have been entitled to a map node (or several!) of their own, which could have indicated the connection to other dative meanings. However, I chose to subsume them under the ELAB node to keep the map uncluttered. The Romans passage is an indirect quotation from Isaiah. The Greek original has adnom-

202 Notes inal genitives: (303)

idou tithˆemi en Siˆon lithon proskommatos kai petran skandalou behold lay in Sion stone stumbling-GEN and stone offence-GEN ‘Behold, I lay in Sion a stumbling stone and a rock of offence’ Romans 9:33

70. Defined thus: “The addressee is able to access a representation of the type of object described by the expression. This status is necessary for appropriate use of any nominal expression, and it is sufficient for use of the indefinite article a in English”. 71. There were in fact six OCS examples of TYPE constructions with adjectives formed with type 1 adjective suffixes, but all of them were formed from animal or plant noun stems and thus excluded from the count, since only adjectives derived from nouns denoting persons, personifications or words with metonymic reference to persons were counted, as stated in section 1.1, chapter 3. A typical example: (304)

iˇste˛ diˇe exidżnova offspring viper-OV gennˆemata ekhidnˆon ‘viper’s offspring’ Mar. Luke 3:7, OCS

72. It seems likely that the translator was reluctant to form a type 2 adjective from a noun stem in -enie. According to Kurz (2006:vol. 3, 429), both Zogr., Mar. and Ass. have a genitive construction in the same place, and the same construction is also found in Matthew 12:4 in Zogr. and Mar., and in Mark 2:26 in Zogr., Mar. and Ass. Mark 2:26 in Savv. has xleby prˇedloˇzenye˛ , where prˇedloˇzenye˛ is interpreted as a participle in Kurz (2006), but it might possibly be taken to be a denominal adjective instead. 73. 98.2 % in the 11th–14th century sample, 98.7 % in the 15th century sample, 98.6 % in the 16th century sample and 96.2 % in the 17th century sample. 74. The collocation is special in that it is clearly calqued from Hebrew, where there is a partially lexically specific construction on the form “son of [abstract property]”, which simply means “human being with said property”. In Hebrew, “son of man” just means “human being”, but both in Greek and OCS it is clearly a fixed “name” for Jesus / the Messiah. 75. The two words are used indiscriminately about the same ethnic group in the Gospels. 76. Excluded from attention were four examples of the collocation kranievo mˇesto “Skull Place”, which have adjectives derived with the suffix -ov-, but are derived from an inanimate noun. All occur in explications of the toponym Golgotha, and there is no indication that this name involves any reference to a specific skull. 77. The final example is a paratactic country-name construction. 78. There are several formulations of a possessor prototype in the literature. Taylor (1996:340) sets up a possession gestalt, where the typical possessor is described as “a specific human being”. Bratishenko (2005:362) sets up a continuum of noun types ranging from typical possessors (personal proper nouns, personal common nouns) to typical possessions (inanimates). 79. As we have seen, the dative construction is virtually unattested at the TYPE and LABEL nodes in any sample. 80. http://ruscorpora.ru

Notes

203

81. I will not propose a separate set of connecting lines for the restricted genitive construction, since it is hard to determine its exact boundaries, as its existence is dependent on its interaction with other constructions. The GENMOD group contains occurrences from both constructions, and would be expected to have a combination of the connection lines of ADJ1 and GENUNMOD. 82. In Slavic, it is not difficult to argue in favour of strong semantic links between adnominal genitives and adverbal genitives via the partitive meaning, cf. Berg-Olsen (2005) and Toft (2010) for discussions on this connection in Baltic and in Old Norse, where it is clearly more tenuous. It is less straightforward to make the connection to genitive usages with prepositions. Overall, however, the semantics of the Slavic (and more generally, the Indo-European) genitive is clearly much less homogeneous than the dative’s semantics. 83. According to Svane (1989:221); Kuskov (1989) claims it to be from 1795–96

References

Aland, Barbara, Kurt Aland, Johannes Karavidopolous, Carlo M. Martini and Bruce M. Metzger (eds.) Greek-English New Testament. Stuttgart: Deutsche Bibelgesellschaft. 1998 Baayen, R.H. Analyzing Linguistic Data: A Practical Introduction to Statistics using R. 2008 Cambridge: Cambridge University Press. Barker, Chris Possessives and relational nouns. In Semantics: An International Handbook forthc. of Natural Language Meaning, Claudia Maienborn, Klaus von Heusinger and Paul Portner (eds.). Berlin: de Gruyter. Barker, Chris and David Dowty Non-verbal Thematic Proto-Roles. In Proceedings of NELS, 49–62. 1993 BarDdal, J´ohanna Productivity. Evidence from Case and Argument Structure in Icelandic. Ams2008 terdam: John Benjamins. Berg-Olsen, Sturla The Latvian dative and genitive: A Cognitive Grammar account. Disserta2005 tion submitted for the degree of Doctor artium, University of Oslo, Faculty of Humanities. Blagova, Emilija Leksika Supraslskoj rukopisi i leksika Ioanna Eksarxa. In Prouˇcvanija v˘archu 1980 Supras˘alskija sbornik, Jordan Zaimov (ed.). Sofia: Izdatelstvo na B˘algarskata Akademija na Naukite. Borkovskij, V. I. Sintaksis drevnerusskix gramot (prostoe predloˇzenie). Lvov: Izdanie 1949 L’vovskogo gosudarstvennogo universiteta. Borkovskij, V. I. (ed.) ˇ Sravnitel’no-istoriˇceskij sintaksis vostoˇcnoslavjanskix jazykov: Cleny 1968 predloˇzenija. Moscow: Nauka. Istoriˇceskaja grammatika russkogo jazyka: Sintaksis prostogo predloˇzenija. 1978 Moscow: Nauka. Borkovskij, V. I. and P. S. Kuznecov Istoriˇceskaja grammatika russkogo jazyka. Moscow: Izdatel’stvo Akademii 1963 nauk SSSR. Bratishenko, Elena Morphosyntactic variation in possessive constructions and the accusative in 1998 Old East Slavic texts. PhD thesis, University of Toronto, Department of Slavic Languages and Literatures. Genitive-Accusative and Possessive Adjective in Old East Slavic. Scando2003 Slavica 49: 83–103. Denominal adjective formation and suffix hierarchy in Old East Slavic. Slavic 2005 and East European Journal 49 (3): 361–377.

206 References Brown, C. 1983

Topic continuity in written English narrative. In Topic Continuity in Discourse: A Quanttitative Cross-Language Study, Talmy Giv´on (ed.), 315–341. Amsterdam: Benjamins.

Buslaev, F. I. Moscow: Gosudarstvennoe 1881/1959 Istoriˇceskaja grammatika russkogo jazyka. uˇcebno-pedagogiˇceskoe izdatel’stvo Ministerstva prosveˇscˇ enija RSFSR. Chomsky, Noam Remarks on nominalization. In Studies on semantics in generative grammar, 1970 11–61. The Hague: Mouton. Clancy, Steven J. The Topology of Slavic Case: Semantic Maps and Multidimensional Scaling. 2006 Glossos 7. Comrie, Bernard The syntax of action nominals: A cross-language study. Lingua 40: 177–201. 1976 Corbett, Greville G. The morphology/syntax interface: Evidence from possessive adjectives in 1987 Slavonic. Language 63: 299–345. Croft, William Radical Construction Grammar. Syntactic Theory in Typological Perspective. 2001 Oxford: Oxford University Press. Croft, William and Keith T. Poole Inferring universals from grammatical variation: multidimensional scaling for 2008 typological analysis. Theoretical Linguistics 34 (1): 1–37. Cysouw, Michael Building semantic maps: the case of person marking. In New Challenges 2007 in Typology, Matti Miestamo and Bernhard W¨alchli (eds.), 225–248. Berlin: Mouton de Gruyter. Delsing, Lars-Olof The Internal Structure of Noun Phrases in the Scandinavian Languages: A 1993 Comparative Study. Lund: Department of Scandinavian Languages. Dickey, Stephen M. Metonymy and Verbal Nouns in Old Church Slavic. University of Tromsø: 2011 Paper presented at the conference Time and Variation in Cognitive Linguistics. Da˛ browska, Ewa Cognitive Semantics and the Polish Dative. Berlin: Mouton de Gruyter. 1997 Eckhoff, Hanne M. Old Russian possessive constructions. A construction grammar account. Dis2007 sertation submitted for the degree of doctor artium, University of Oslo, Faculty of Humanities. A usage-based approach to change: Old Russian possessive constructions. In 2009 The Role of Semantic, Pragmatic, and Discourse Factors in the Development of Case, J´ohanna BarDdal and Shobhana L Chelliah (eds.). Amsterdam: John Benjamins.

References

207

Eckhoff, Hanne M. and Sturla Berg-Olsen Adnominal genitives and denominal adjectives in Latvian and Old Russian – 2002 a typological and historical approach. Balto-Slavjanskie issledovanija XV. Faarlund, Jan Terje Syntactic Change. Toward a Theory of Historical Syntax. Berlin: Mouton de 1990 Gruyter. Fedriani, Chiara Experiental Constructions in Latin. A Synchronic and Diachronic Study. Ph.D. forthc. thesis, University of Pavia. Fillmore, Charles The Case for Case. In Universals in Linguistic Theory, Emmon Bach and 1968 R. Harms (eds.). New York: Holt, Rinehart, and Winston. Flier, Michael Aspects of Nominal Determination in Old Church Slavic. The Hague: Mou1974 ton. Goldberg, Adele E. Constructions: A Construction Grammar Approach to Argument Structure. 1995 Chicago: The University of Chicago Press. Constructions at Work: The Nature of Generalization in Language. Oxford: 2006 Oxford University Press. Grimshaw, Jane Argument Structure. Cambridge: The MIT Press. 1990 Gundel, Jeanette K., Nancy Hedberg and Ron Zacharski Cognitive Status and the Form of Referring Expressions in Discourse. Lan1993 guage 69 (2). Haspelmath, Martin Indefinite pronouns. Oxford: Oxford University Press. 1997 The geometry of grammatical meaning: Semantic maps and cross-linguistic 2003 comparison. In The new psychology of language, Michael Tomasello (ed.), vol. 2, 211–242. Mahwah, NJ: Lawrence Erlbaum. Heine, Bernd Possession: Cognitive sources, forces and grammaticalization. Cambridge: 1997 Cambridge University Press. Huntley, David The distribution of the denominative adjective and the adnominal genitive in 1984 Old Church Slavonic. In Historical Syntax, J. Fisiak (ed.). Berlin: Mouton. Old Church Slavonic. In The Slavonic Languages, Bernard Comrie and Gre1993 ville G. Corbett (eds.). London/New York: Routledge. Ivanov, V. V. Kategorija posessivnosti v slavjanskix i balkanskix jazykax. Moscow: Nauka. 1989 Istoriˇceskaja grammatika russkogo jazyka. Moscow: Prosveˇscˇ enie. 1990 Jensen, Per Anker and Carl Vikner The English Pre-nominal Genitive and Lexical Semantics. In Possessives and 2004 Beyond: Semantics and Syntax, Ji yung Kim, Yury A. Lander and Barbara H. Partee (eds.). (UMOP) Amherst, Mass.: GLSA Publications.

208 References Karskij, E. F. 1930

Russkaja Pravda po drevnejˇsemu spisku. Leningrad: Izdatel’stvo Akademii nauk. Karttunen, Lauri Discourse referents. Syntax and Semantics 7: 363–385. 1976 Koptjevskaja-Tamm, Maria Maria’s ring of gold. Adnominal possession and non-anchoring relations in 2004 European languages. In Possessives and Beyond: Semantics and Syntax, Ji yung Kim, Yury A. Lander and Barbara H. Partee (eds.), 155–181. (UMOP 29) Amherst, Mass.: GLSA Publications. ˇ Kopˇcevskaja-Tamm, Marija and Aleksej Smelev Aleˇ s ina s Maˇ s ej stat’ja (o nekotoryx svojstvax russkix “pritjaˇzatel’nyx prila1994 gatel’nyx”. Scando-Slavica 40. Kotkov, S. I., A. S. Oreˇsnikov and I. S. Filippova (eds.) Moskovskaja delovaja i bytovaja pismennost’ XVII veka. Moscow: Nauka. 1968 Krifka, Manfred, Francis J. Pelletier, Gregory N. Carlson, Alice Ter meulen, Gennaro Chierchia and Godehard Link Genericity: An Introduction. In The Generic Book, Gregory N. Carlson and 1995 Francis J. Pelletier (eds.), 1–124. Chicago: University of Chicago Press. Krys’ko, V. B. Istoriˇceskij sintaksis russkogo jazyka: Ob”ekt i perechodnost’. Moscow: In1997 drik. Kurz, Josef (ed.) Issledovanija po sintaksisu staroslavjanskogo jazyka. Prague: Izdatel’stvo 1963 ˇ Cexoslovackoj Akademii Nauk. Slovar’ staroslavjanskogo jazyka. St. Petersburg: Izdatel’stvo Sankt-Peter2006 burgskogo universiteta. Kuskov, V. V. Istorija drevnerusskoj literatury. Moscow: Vysˇsaja sˇkola. 1989 Lambrecht, Knud Information structure and sentence form: Topic, focus, and the mental repre1994 sentation of discourse referents. Cambridge: Cambridge University Press. Langacker, Ronald W. Foundations of Cognitive Grammar, vol. I. Stanford: Stanford University 1987 Press. Foundations of Cognitive Grammar, vol. II. Stanford: Stanford University 1991 Press. Grammar and Conceptualization. Berlin: Mouton de Gruyter. 2000 Lomtev, T. P. Oˇcerki po istoriˇceskomu sintaksisu russkogo jazyka. Moscow: Izdatel’stvo 1956 Moskovskogo universiteta. Luraghi, Silvia On the Meaning of Prepositions and Cases: The Expression of Semantic Roles 2003 in Ancient Greek. Amsterdam: John Benjamins.

References

209

Makarova, S. Ja. Roditel’nyj padeˇz prinadleˇznosti v russkom jazyke XI-XVII vv. Trudy Insti1954 tuta jazykoznanija AN SSSR . Makogonenko, G. P. (ed.) Russkaja literatura XVIII veka. Leningrad: Prosveˇscˇ enie. 1970 Malchukov, Andrej and Heiko Narrog Case polysemy. In The Oxford handbook of case, Andrej Malchukov and 2009 Andrew Spencer (eds.), 518–535. Oxford: Oxford University Press. Marojevi´c, Radmilo Posesivne kategorije u ruskom jeziku (u svome istorijskom razvitku i danas). 1983a Belgrade: Filoloˇzski Fakultet Beogradskog Universiteta. Posessivnye kategorii v russkom jazyke (istorija razvitija i sovremennoe sos1983b tojanie). Nauˇcnye doklady vysˇsej sˇkoly. Filologiˇceskie nauki 5. K rekonstrukcii praslavjanskoj sistemy posessivnyx kategorij i posessivnyx 1989 proizvodnyx. Etimologija 1986–1987. Meillet, Antoine 1934/2000 Obˇscˇ eslavjanskij jazyk (Le Slave commun). Moscow: Izdatel’skaja gruppa “Progress”. Miklosich, Franz Vergleichende Syntax der slavischen Sprachen. Vienna: Wilhelm Braum¨uller. 1883 Moule, C. F. D. An Idiom Book of New Testament Greek. Cambridge: Cambridge University 1953 Press. Mrazek, Roman Datel’nyj padeˇz v staroslavjanskom jazyke. In Kurz (1963). 1963 M¨uller, Ludolf (ed.) Die altrussischen hagiograpischen Erz¨ahlungen und liturgischen Dichtungen 1967 u¨ ber die heiligen Boris und Gleb. (Slavische Propyl¨aen 14) Munich: Wilhelm Fink Verlag. Narrog, Heiko A Diachronic Dimension in Maps of Case Functions. Linguistic Discovery 8 2010 (1). Nikolaeva, Irina and Andrew Spencer The Possession-Modification Scale: A universal of nominal morphosyntax. 2010 manuscript. Nilsson, Barbro Old Russian Derived Nominals in -nie, -tie: A Syntactical Study. Stockholm: 1972 Almqvist & Wiksell. Obnorskij, S. P. and S. G. Barxudarov (eds.) Xrestomatija po istorii russkogo jazyka. Moskva: Aspekt Press, 3rd ed. 1999 Partee, Barbara H. and Vladimir Borschev Possessives, favorite, and Coercion. In Proceedings of ESCOL99, Anastasia 2000 Riehl and Rebecca Daly (eds.), 173–190. Cornell University: CLC Publications.

210 References 2001

Some Puzzles of Predicate Possessives. In Perspectives on Semantics, Pragmatics, and Discourse. A Festschrift for Ferenc Kiefer, I. Kenesei and R. M. Harnish (eds.). Amsterdam: John Benjamins.

Payne, Doris L. and Immanuel Barshi (eds.) External possession. Amsterdam: John Benjamins. 1999 Potebnja, A. A. 1899/1968 Iz zapisok po russkoj grammatike. Tom III: Ob izmenenii znaˇcenija i zamenax suˇscˇ estvitel’nogo. Moskva: Prosveˇscˇ enie. Pravdin, A. B. K voprosu o vyraˇzenii prinadleˇznosti v russkom jazyke. Uˇcenye zapiski Tar1957 tuskogo gosudarstvennogo universiteta 51. Prince, Ellen F. Toward a taxonomy of given-new information: The two genitives of English. 1981 In Radical Pragmatics, P. Cole (ed.), 223–256. New York: Academic Press. Pustejovsky, James The Generative Lexicon. Cambridge: MIT Press. 1995 Richards, Karen Rondestvedt Towards a history of the possessive in literary Russian. The demise of the 1976 possessive adjective. In Papers from the parasession on diachronic syntax, 260–273. Chicago: Chicago Linguistic Society. Schenker, Alexander M. The Dawn of Slavic. An Introduction to Slavic Philology. New Haven and 1995 London: Yale University Press. Sprinˇcak, Ja. A. Oˇcerk russkogo literaturnogo sintaksisa., vol. 1: Prostoe predloˇzenie. Kiev: 1960 Radjans’ka sˇkola. Stecenko, A. N. Istoriˇceskij sintaksis russkogo jazyka. Moscow: Vysˇsaja sˇkola. 1977 Stefanowitsch, Anatol Constructional semantics as a limit to grammatical alternation: The two gen2003 itives of English. In Determinants of Grammatical Variation in English, G¨unther Rohdenburg and Britta Mohndorf (eds.). Berlin: Mouton de Gruyter. Svane, Gunnar O. ˚ Ældre kirkeslavisk litteratur 9.–12. a˚ rhundrede. Arhus: Aarhus universitets1989 forlag. Taylor, John R. Possessives in English. An Exploration in Cognitive Grammar. Oxford: Ox1996 ford University Press. Cognitive Grammar. Oxford: Oxford University Press. 2002 Toft, Ellen Hellebostad Adnominal and adverbal genitive constructions in Old Norse. A cognitive con2010 struction grammar account. PhD dissertation, University of Oslo, Faculty of Humanities.

References

211

Uryson, E. V. Otnositel’noe prilagatel’noe v paradigme drevnerusskogo imeni. (Na materi1980 ale Uspenskogo sbornika XII–XIII vv.). In Drevnerusskij jazyk. Leksikologija i leksikografija, R. I. Avanesov (ed.). Moscow: Nauka. Uspenskij, B. A. Istorija russkogo literaturnogo jazyka (XI-XVII vv.). Munich: Verlag Otto Sag1987 ner. Vaillant, Andr´e Grammaire compar´ee des langues slaves, vol. II: Morphologie, Deuxi`eme 1958 partie: Flexion pronominale. Lyon: IAC. Grammaire compar´ee des langues slaves, vol. V: La syntaxe. Paris: Editions 1977 Klincksieck. van der Auwera, Johan In defense of classical semantic maps. Theoretical Linguistics 34 (1). 2008 Veˇcerka, Radoslav Sintaksis bespredloˇznogo roditel’nogo padeˇza v staroslavjanskom jazyke. In 1963 Kurz (1963). Altkirchenslavische (altbulgarische) Syntax, vol. II: Die innere Satzstruktur. 1993 Freiburg i. Br.: U.W. Weiher. Vlasto, A.P. A Linguistic History of Russia to the End of the Eighteenth Century. Oxford: 1988 Clarendon Press. Vondr´ak, Wenzel Vergleichende Slavische Grammatik:, vol. II: Formenlehre und Syntax. 1928 G¨ottingen: Vandenhoeck & Ruprecht. Wackernagel, J. Genetiv und Adjektiv. In M´elanges de linguistique offerts a` M.Ferdinand de 1908 Saussure. (Collection linguistique publi´ee par La soci´et´e de linguistique de Paris II) Paris: Libraire ancienne Honor´e Champion. W¨alchli, Bernhard Similarity Semantics and Building Probabilistic Semantic Maps from Parallel 2010 Texts. Linguistic Discovery 8 (1). Widn¨as, Maria O vyraˇzenii prinadleˇznosti v russkom jazyke XVIII-XIX vv. Scando-Slavica 1958 4. Wierzbicka, Anna The semantics of grammar. Amsterdam: John Benjamins. 1988 Xodova, K. I. Sistema padeˇzej staroslavjanskogo jazyka. Moscow: Izdatelstvo Akademii 1963 Nauk SSSR. Zaimov, Jordan and Mario Capaldo (eds.) 1982–83 Supras˘ulski ili Retkov sbornik. Sofia: Bulgarska akademija na naukite. Zverkovskaja, N. P. Suffiksal’noe slovoobrazovanie russkix prilagatel’nyx XI–XVII vv. Moscow: 1986 Nauka.

Index

aboutness topics, 20 adjectives type 1, 36, 42, 58 delimitation, 36–38, 202 type 2, 36, 39, 42, 58 delimitation, 39, 40 affectedness, 66, 157, 176, 177 agent, 66 agent/possessor hierarchy, 42, 59 anchoring, 21, 131, 198 animacy, 3, 12, 65, 66, 73, 76, 81, 110, 111, 154–156, 171, 201 animacy hierarchy, 42, 79 apposition, 19, 25, 93, 120–124, 126–129 benefactive, 11, 57, 71, 72, 81, 111, 124, 126, 129, 130 complementary distribution, 4, 41, 49–51, 54, 56, 58, 59, 65, 68, 73, 81, 82, 90, 103, 123, 130, 144, 149, 151–153, 158, 171, 200, 201 complex name, 126, 127, 143, 144, 147 complex toponyms, 121 construction grammar, 6–8, 166, 180, 198 constructional polysemy, 1, 7, 52 constructional synonymy, 1, 7, 8, 52, 179 conventionalisation, 61, 76, 84, 85, 93, 98, 106, 112, 118, 121, 133, 140, 153, 155 correspondence analysis, 11, 28–30, 157, 178, 179 cue validity, 20, 24 dative construction, 4, 46, 48, 50 dative semantics, 66 definiteness, 3, 42 description, 123 discourse referent, 21, 131 ELAB subtypes, 25 ELAB(part)

definition, 25, 93 ELAB(slot) definition, 23, 93 or ID(slot), 93, 94, 201 emphatic construction, 125, 129 English of construction, 1, 18, 19, 21, 24, 25 onomastic compounds, 22 possessive compounds, 21, 131 prenominal possessive, 1, 17, 19, 21, 24, 198 excerpation principles, 6, 188 experiencer, 66 genitive construction formally restricted, 44, 51, 58, 65, 67, 81, 91, 96, 104, 111, 123, 152, 153, 168, 203 semantically restricted, 44, 45, 48, 51, 82, 91, 96, 103, 110, 111, 116, 130, 153–155, 158, 168, 175, 179 Greek, 35, 50, 57, 100, 199, 201 ID criteria, 20, 53 idiom, 7 information status, 20, 94, 201 inheritance network, 7 intrinsic relationship, 18, 25, 30, 93, 155, 176, 179 LABEL criteria, 22, 135 subtypes, 23, 136, 139, 140, 145, 148, 149 landmark, 198 lexically specific construction, 7, 56, 64, 71, 79, 83, 136, 141, 142, 146, 166, 202 locative, 120, 121, 123, 127, 129 mixed construction, 110

214 Index mixed constructions, 49, 58, 67, 73, 81, 89, 90, 102, 115, 134, 148, 152, 167, 171 non-relational nouns, 13, 20, 197 non-specific NP, 199 object-like nominal arguments, 94, 130 origin, 121, 123, 126–128, 153, 175 partitive, 41, 43, 60, 115, 121–123, 129, 176, 203 patronymics, 61, 76, 84, 85, 119, 145–147, 149, 161, 163, 166 personal sphere, 66, 176, 177 possession definition, 1, 53, 68, 152 external, 46, 197, 200 inalienable, 21, 197 prototypical, 71 possessive adjectives, 35 possessive relation, 1, 3, 17 possessor generic, 21 indefinite, 21 kind-referring, 20, 21, 53, 55, 56, 58, 59, 62, 65, 66, 70–72, 77, 81, 82, 89, 92, 99, 102, 110, 131, 136, 138, 141, 149, 154, 155, 157 non-specific, 20, 21, 56, 58, 59, 62, 65, 66, 70, 71, 77, 81, 82, 89, 99, 102, 110, 133, 154, 155 prototypical, 59, 154 unique, 154 productivity, 7, 8, 35, 37, 40, 82, 136, 141, 142, 146, 148, 165, 166, 180, 198 purpose, 119, 123, 124, 126, 129, 130, 132 qualia structure, 26, 197, 198 recipient, 124, 126, 130 reference point, 18–21, 30, 53, 65, 66, 71, 81, 91–94, 97, 118, 119, 130, 131, 154, 155, 174, 177, 179

relational nouns, 3, 11–13, 18, 20, 21, 24, 26, 42, 48, 59, 65, 67, 81, 93, 96, 104, 118, 155, 156, 175– 177, 179 body parts, 13, 16, 60, 62, 65–67, 75, 78, 80, 154, 155, 177, 197 control nouns, 17 deadjectival nouns, 16, 19, 60, 62, 75, 78, 80, 94 deverbal nouns, 3, 12–14, 16, 18, 48, 60, 62, 65–67, 76, 78, 80, 94, 96, 97, 99–101, 103, 105, 107– 109, 112, 117, 130, 133, 154, 155, 157, 160, 200, 201 aspect, 197 classification, 14–16 human relationship nouns, 60, 65, 66 information nouns, 16 kinship terms, 12, 15, 60, 65, 67, 80, 100, 154, 177, 201 part nouns, 13, 16, 19, 25, 43, 60, 63, 75, 77, 78, 93, 94, 102, 117, 175, 179, 200 realm nouns, 60, 61 representation nouns, 16, 94, 96, 100, 105, 107, 112 ruler nouns, 94, 96, 98, 100–103, 106, 107, 111, 112, 115, 130, 175 source nouns, 100 relative adjectives, 36 relatum, 24 schematic network, 166 semantic centre of gravity, 38, 40, 45, 46, 48, 51, 52, 101, 103, 110, 130, 155, 160, 162, 165, 168, 173– 175, 177, 179, 180 semantic map, 1, 8, 11, 26–28, 37–39, 44, 45, 48, 151, 152, 155, 179 computer-generated, 27–29, 199 diachrony, 30, 158, 160, 162, 165, 168, 173–178, 180, 203 hand-drawn, 27 implicational, 30, 173–178, 180, 203 Semantic Map Connectivity Hypothesis, 27 semantic roles, 197

Index semantic space, 1, 3, 4, 7, 8, 27, 52, 147, 152, 153, 155, 157, 173, 176, 179 similarity semantics, 27 sortal nouns, 13 source, 121 specificity, 3 syntactic loan, 4, 50, 156 temporal, 124 text samples, 5 theme, 66 topicality, 20, 24 trajector, 198

215

TYPE criteria, 22, 132 or ID, 132, 134 type frequency, 82, 83 type identifiable, 131, 132, 135, 138, 141, 156, 174 unique, 55, 56, 58, 62, 65, 70–72, 77, 97, 140, 142, 154, 197 uniquely identifiable, 135–143, 145, 149, 161, 163 usage-based model, 6, 7, 166 valency, 12