A geography of case semantics: the Czech dative and the Russian instrumental

Cognitive Linguistics Research 4 Editors Rene Dirven Ronald W. Langacker

Mouton de Gruyter Berlin· New York

tive Linguistics Research 4 Editors Rene Dirven Ronald W. Langacker

Mouton de Gruyter Berlin· New York

A Geography of Case Semantics The Czech Dative and the Russian Instrumental

Laura A. Janda

1993 Mouton de Gruyter Berlin· New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin

@) Printed on acid-free paper

which falls within the guidelines of the ANSI to ensure permanence and durability

Library of Congress Cataloging in Publication Data

Janda, Laura A. A geography of case semantics: the Czech dative and the Russian instrumental/Laura A. Janda. p. em. - (Cognitive linguistics research; 4) Includes index. ISBN 3-11-012672-9 1. Grammar, Comparative and general - Case. 2. Semantics. 3. Cognitive grammar. 4. Czech language - Case.' 5. Russian language Case. 6. Russian language - Grammar, Comparative Czech. 7. Czech language Grammar, Comparative Czech. I. Title. II. Series. P240.6.J36 1993 415-dc20 93-8324

Die Deutsche Bibliothek - Cataloging in Publication Data

Janda, Laura A.: A geography of case semantics: the Czech dative and the Russian instrumental/Laura A. Janda. - Berlin; New York: Mouton de Gruyter, 1993 (Cognitive linguistics research; 4) ISBN 3-11-012672-9 NE:GT

© Copyright 1993 by Walter de Gruyter & Co., D-10785 Berlin. All rights reserved, including those oftranslation into foreign languages. No part ofthis book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printing: Werner Hildebrand, Berlin Binding: Dieter Mikolai, Berlin Printed in Germany

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin

@) Printed on acid-free paper

which falls within the guidelines of the ANSI to ensure permanence and durability

Library of Congress Cataloging in Publication Data

da, Laura A. A geography of case semantics: the Czech dative and the Rusian instrumental/Laura A. Janda. p. em. - (Cognitive linguistics research; 4) Includes index. ISBN 3-11-012672-9 1. Grammar, Comparative and general - Case. 2. Semantics. . Cognitive grammar. 4. Czech language - Case.' 5. Russian anguage Case. 6. Russian language - Grammar, Comparaive Czech. 7. Czech language Grammar, Comparative Czech. I. Title. II. Series. P240.6.J36 1993 15-dc20 93-8324

Die Deutsche Bibliothek - Cataloging in Publication Data

da, Laura A.: geography of case semantics: the Czech dative and the Russian trumental/Laura A. Janda. - Berlin; New York: Mouton de uyter, 1993 Cognitive linguistics research; 4) SBN 3-11-012672-9 E:GT

© Copyright 1993 by Walter de Gruyter & Co., D-10785 Berlin.

d, including those oftranslation into foreign languages. No part ofthis book may transmitted in any form or by any means, electronic or mechanical, including ding, or any information storage and retrieval system, without permission in writing from the publisher. Printing: Werner Hildebrand, Berlin Binding: Dieter Mikolai, Berlin Printed in Germany

Contents ix

Acknowledgements Preface

.

xi

Part I - Theory

o.

Introduction

3

1. 1.1 1.1.1. 1.1.2. 1.1.3. 1.1.4. 1.2. 1.2.1. 1.2.2. 1.2.3. 1.2.4. 1.3. 1.3.1. 1.3.2. 1.3.3.

What cognitive semantics is. . . . . . . . . . . . . . . . . . Support for cognitive semantics from other disciplines. Psychology . . . . . . . Neurobiology . . . . . Artificial Intelligence. Summary . Case as a semantic entity. Case lacks formal autonomy. Case bears subjective as well as objective meaning. Case meaning and lexical semantics. . . . . . . . . . . Summary . What a cognitive description of case should look like . Schematic representation of case meaning. Network structure . A test of the cognitive model. . . . . . . ..

4

2. 2.1. 2.1.1. 2.1.2. 2.1.3. 2.1.4.

Cognitive semantics compared with other descriptions of case Case semantics in previous traditions. . . . . . . . . . . . . .. Inventories of uses (Mrazek, Potebnja) . Semantic features (Hjelmslev, Jakobson, van Schooneveld). Semantic metalanguage (Wierzbicka) . Semantic/syntactic dichotomy (Kurylowicz, Isacenko) . Case as a purely syntactic phenomenon (Chomsky)... Symbolic logic (Sorensen, Mel'cuk) . "Deep" case (Fillmore) . Localist theories of case (Anderson, Freidhof).

2.1.5. 2.1.6. 2.1.7. 2.1.8.

7 7 8 9 10 10 10 11 13

15 15 15 17 24 25 25 25 26 27 27

29 30 31 32

vi

2.2. 2.2.1. 2.2.2. 2.2.3. 2.3. 2.3.1. 2.3.2. 2.3.3. 2.4.

How the cognitive model has been implied by various authors Relatedness of meanings . Hierarchical relations among meanings. . . . . . . . . . . . . . . Relational invariance . . Advantages of the cognitive approach A departure from empiricism . Both diversity and unity of case meaning are accommodated Many of the best insights of various scholars are incorporated An ideal description of case: Skalicka's challenge .

33 33 36 37 37 38 38 38 39

Part II - Analysis

3. 3.0.1. 3.0.2. 3.0.3. 3.1. 3.1.1. 3.1.2. 3.2. 3.2.1. 3.2.2. 3.2.3. 3.2.4. 3.3. 3.4. 3.4.1. 3.4.2. 3.5.

Why the Czech dative and the Russian instrumental? . 43 . 43 Schemas in pictorial and prose notation Indirect object serves as the prototype of the Czech dative .. 47 Cognitive network of the Czech dative . 52 Schema 1 - indirect object. . 54 Schema 1 a - governed dative . 69 Schema 1 b - dative in impersonal constructions . 79 . 81 Schema 2 and variants - free dative Schema 2 a - dative of beneficiary - transition to indirect object . 90 . 91 Schema 2 b - intransitive expressions of free dative . 92 Schema 2 c - free dative in copular sentences Further syntagmatic variants of schema 2 . 93 Dative network bound by syntagmatic variants of the schemas 95 . 97 Reflexive uses of the dative Reflexive network . 97 Reciprocal network . 107 Semantic extension via mapping - pragmatic uses ofthe dative 110

4. 4.1. 4.1.1. 4.1.2. 4.2. 4.3.

Comparison of cognitive networks - the Russian dative. Schema 1 - indirect object. . . . . . . . . . . . . . . . . . . . Schema 1 a governed dative. . . . . . . . . . . . . . . . . . Schema 1 b - impersonal dative. . . . . . . . . . . . . . . . . Schema 2 and variants - free dative. . . . . . . . . . . . . . Dative reflexives. . . . . . . . . . . . . . . . . . . . . . . . . . .

.. .. .. .. .. ..

113 113 116 128 130 134

vii

w the cognitive model has been implied by various authors tedness of meanings . archical relations among meanings. . . . . . . . . . . . . . . tional invariance . . antages of the cognitive approach eparture from empiricism . h diversity and unity of case meaning are accommodated y of the best insights of various scholars are incorporated deal description of case: Skalicka's challenge .

33 33 36 37 37 38 38 38 39

nalysis

y the Czech dative and the Russian instrumental? . 43 . 43 emas in pictorial and prose notation rect object serves as the prototype of the Czech dative .. 47 nitive network of the Czech dative . 52 ema 1 - indirect object. . 54 ema 1 a - governed dative . 69 ema 1 b - dative in impersonal constructions . 79 . 81 ema 2 and variants - free dative ema 2 a - dative of beneficiary - transition to indirect ct . 90 . 91 ema 2 b - intransitive expressions of free dative . 92 ema 2 c - free dative in copular sentences her syntagmatic variants of schema 2 . 93 ve network bound by syntagmatic variants of the schemas 95 . 97 exive uses of the dative exive network . 97 procal network . 107 antic extension via mapping - pragmatic uses ofthe dative 110

mparison of cognitive networks - the Russian dative. ema 1 - indirect object. . . . . . . . . . . . . . . . . . . . ema 1 a governed dative. . . . . . . . . . . . . . . . . . ma 1 b - impersonal dative. . . . . . . . . . . . . . . . . ema 2 and variants - free dative. . . . . . . . . . . . . . ve reflexives. . . . . . . . . . . . . . . . . . . . . . . . . . .

.. .. .. .. .. ..

113 113 116 128 130 134

5. 5.0.1. 5.0.2. 5.0.3. 5.1. 5.1.1.

5.1.2. 5.1.3. 5.1.4. 5.2. 5.2.1. 5.2.2. 5.2.3. 5.3. 5.3.1. 5.3.2. 5.3.3. 5.3.4. 5.4. 5.4.1. 5.4.2. 5.5.

. Analysis of the Russian instrumental Overview of the network. . . . . . . . . . . . . . . . . . . . . . . . Instrumental network paradigm. . . . . . . . . . . . . . . . . .. Summary of the instrumental's semantic role . Conduit instrumental - schema 1 and syntagmatic variants. Schema 1 a - instrumental of instrument and agent, verbal government . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schema 1 b - instrumental of inexplicable force . . . . . . .. Schema 1 c - instrumental of sensation . Subnetwork of schema 1 and the paradigm ofdisappearing participants . . Schema 2 - instrumental of setting Instrumental of space. . . . . . . . . . . . . . . . . . . . . . . . . . Instrumental of time . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of the instrumental of setting . . . . . . . . . . . . . . Attributive instrumental . . . . . . . . . . . . . . . . . . . . . . quantitative.......... . Attributive instrumental qualitative........... ', . . . Attributive instrumental Attributive instrumental - comparative . . Summary of the attributive instrumental . . . . . . . . . Instrumental used with a preposition. . . . . . . . . . . . . . . . Comitative instrumental. . . . . . . . . . . . . . . . . . . . . . . . Proximate instrumental . . . . . . . . . . . . . . . . . . . . . . .. Instrumental network bound by paradigm of peripherality and by alliance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

139 139 139 142 143 150 162 162 164 164 166 167 170 170 173 175 179 182 182 182 185 188

6. 6.1. 6.2. 6.3. 6.4.

The Czech instrumental . 191 Conduit instrumental. . . . . . . . . . . . . . . . . . . . . . . . .. 191 Instrumental of setting . 195 Attributive instrumental . . . . . . . . . . . . . . . . . . . . . . 196 Instrumental with a preposition . . . . . . . . . . . . . . . . . . . 197

7.

Concluding remarks and possible universals . . . . . . . . . .. 199

Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 203 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 218

Acknowledgements The initial research for this project was funded by Fulbright Grant number 8545620. Additional funding for the research and for preparation of the manuscript was provided by the University of Rochester. I would like to thank the many individuals who offered support, criticisms and commentary on drafts of this work: Henning Andersen, Leonard Babby, Greg Carlson, Charles Carlton, Rene Dirven, Michael Flier, Frank Gladney, Ron Harrington, George Lakoff, Ronald Langacker, Brygida Rudzka-Ostyn, Alla Shustorovich, and Charles Townsend. I would also like to thank my colleagues at Charles University and the Institute for Czech Language in Prague who assisted me in my research there: Prernysl Adamec, Alla Bemova, Oldfich Leska, Petr Pit'ha, and Petr Sgall. Much of the research into secondary source materials could not have been completed without the expert assistance of the Interlibrary Loans Staff at the University of Rochester. Likewise, I received invaluable help in preparing the final version of the manuscript from the Consulting Staff of the Faculty Computing Resource Center, and from Sean Singh in particular. Drawings in the figures were executed by Terri Beers. I would also like to express my gratitude to Mouton de Gruyter for making this series possible and for collecting further comments from their anonymous readers. Finally I must thank my family for traveling with me to Czechoslovakia and patiently enduring all the extra hours I put in to complete this book.

Preface

It is sometimes appalling to note how many basic linguistic terms that we use

frequently and with ease lack an operational definition. The ambiguity of concepts such as "subject" and "noun" has motivated the invention of further metalinguistic terminology, such as "subjecthood", "agentivity" and even "nouniness" in the search for adequate circumscription of these supposedly simple terms. What indeed, for example, is an "indirect object"? Most people's first reaction to this question would be to state that an indirect object is the argument of the verb instantiated by John in a sentence like Sally gave John a book. But what does the word "like" mean in this definition, and given such a definition could we sort clausal arguments gathered from natural language data into indirect objects and non-indirect objects? And is there a definition of indirect object that applies to all languages, or is this term to some extent language-specific? Let us suppose that we are dealing with a hypothetical case-marking language (and that this language has none of the exotic peculiarities associated with the English dative-shift). Here is a sampling of three-argument clauses in that language: Sally-NOM gave John-DAT Sally-NOM took John-DAT ('Sally took the book from John.') Sally-NOM bought John-DAT Sally-NOM brought John-DAT Sally-NOM baked John-DAT Sally-NOM introduced John-DAT Sally-NOM explained John-DAT Sally-NOM wrote John-DAT Sally-NOM told John-DAT Sally-NOM paid John-DAT Sally-NOM broke John-DAT ('Sally broke John's arm') Sally-NOM envied John-DAT

book-ACC book-ACC book-ACC book-ACC cake-ACC her friend-ACC problem-ACC letter-ACC her name-ACC bill-ACC arm-ACC his success-ACC

One soon finds the definition given above to be empirically inadequate. In which of the sentences above does John function as an indirect object? Certainly not in all of them. But just which of these sentences are enough "like" the sentence in the

xii

definition above to contain an indirect object and which are not; where do we draw the line? A closer look at these sentences makes us wonder whether we should restrict our view to three-argument clauses. Indeed there are likely to be many two-argument clauses that contain arguments that are very similar to those in the three-argument clauses. Some examples: Sally-NOM wrote Sally-NOM introduced-REFL ('Sally introduced herself to John. ') Sally-NOM ran-away ('Sally ran away from John.') Sally-NOM paid Sally-NOM told-off Sally-NOM nodded Sally-NOM hurt Sally-NOM helped Sally-NOM believed Sally-NOM ruled

John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT

Although our intuition tells us that the common-sense definition captures something essential about the notion indirect object, it lacks necessary specification. The prototypical case has been correctly identified, but we need to know where we can go from there and how far. A theory of the structure of grammatical categories is required, and it is just such a theory that is in the making in this book. Although all of these constructions yield grammatical sentences in Czech, I would argue that most but not all of them instantiate the indirect object, and that the observed variety among those that do results from systematic extensions of the prototypical case instantiated with the verb meaning "give". The question of what is and is not an indirect object becomes all the more pressing when we expand our view to all case usage, for indirect objects must be successfully distinguished from other constructions, including those containing dative-governing verbs. As I will show in Part two, all members of the category of indirect object are related to the central prototype by relationships of synonymy, antonymy (allowing the verb meaning "take" to participate in indirect object constructions), and metonymy (allowing the direct object to be named in the verb and thus absent from surface structure). This gives the category a clear, operational definition, making data analysis straightforward and unproblematic.

xiii

ve to contain an indirect object and which are not; where do we

ok at these sentences makes us wonder whether we should restrict ree-argument clauses. Indeed there are likely to be many two-arguhat contain arguments that are very similar to those in the three-ars. Some examples:

wrote introduced-REFL ced herself to John. ') ran-away ay from John.') paid told-off nodded hurt helped believed ruled

John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT John-DAT

our intuition tells us that the common-sense definition captures ential about the notion indirect object, it lacks necessary specificaotypical case has been correctly identified, but we need to know go from there and how far. A theory of the structure of grammatical equired, and it is just such a theory that is in the making in this

ll of these constructions yield grammatical sentences in Czech, I hat most but not all of them instantiate the indirect object, and that ariety among those that do results from systematic extensions of the ase instantiated with the verb meaning "give". The question of what n indirect object becomes all the more pressing when we expand our se usage, for indirect objects must be successfully distinguished structions, including those containing dative-governing verbs. As I art two, all members of the category of indirect object are related to totype by relationships of synonymy, antonymy (allowing the verb e" to participate in indirect object constructions), and metonymy direct object to be named in the verb and thus absent from surface s gives the category a clear, operational definition, making data htforward and unproblematic.

A quick glance at dative constructions in the languages of central and eastern Europe demonstrates the need for a certain amount of elasticity in the definition. Languages in the northeastern part of this territory (Russian and Ukrainian) do not admit extensions of the indirect object category via antonymy, and thus lack dative constructions for verbs meaning "take" and further extensions based on this one, whereas languages to the south (German, Czech, Lithuanian, Romanian, SerboCroatian, and Greek) have a more fully extended category. Clearly there must be room in the theory for the growing and pruning of peripheral branches in the development of individual languages. This book will outline the specifics of a theory of grammatical categories, as applied to the use of case in two Slavic languages. In so doing it will take up issues of: -

what determines the shape of grammatical categories to what extent diachronic development is reflected in synchronic structure syntactic and semantic uses of case and their roles in case categories the role of concepts such as passivization, subject and indirect object in the structure of case categories.

Part I -

Theory

O. Introduction In 1980, Anna Wierzbicka wrote in The case for surface case that "cases have

fallen on hard times". Indeed, in the decades that have passed since Roman Jakobson's landmark essay "Beitrag zur allgemeinen Kasuslehre" (1936 [1971]), there has been no great leap forward, although new angles have been sought by Fillmore (1968), Anderson (1971) and others. Cases have largely remained as inscrutable as they were to Bloomfield and Jespersen.! In part, they are themselves the cause of this stagnation, for cases are inherently difficult to investigate, as observed by Brecht and Levine:

"there is a general consensus that the category of case is more resistant to analysis than other grammaticized semantic domains like gender, number, person, and tense. The reason for this, it would seem, is that while these latter grammatical categories can be more or less readily identified with certain pragmatic categories of the real world, it is much more difficult to find real world correlates for the putative referents of case."2

There is, however, growing evidence that the root of the problem runs much deeper, that it in fact lies at the foundation of our theoretical framework. Investigators researching parallel themes in the fields of linguistics, psychology, philosophy, neurobiology and artificial intelligence have suggested that there is a fundamental flaw in our twentieth-century empiricist world-view of symbolic logic and set theory that renders it inadequate to account for phenomena of human cognition. A new framework, known in these various disciplines as cognitive grammar, prototype theory, tensor network theory and connectionism, is being put forth. It requires essential changes in the way we conceive of semantics. The aim of the present volume is to test the possibilities of cognitive grammar using data on the dative and instrumental in Czech and Russian. If this framework is indeed better suited to the task of describing cognitive categories, it should not only shed new light on case semantics, but also incorporate the best hypotheses of traditional research in linguistics.

1.

What cognitive semantics is

In Neurophilosophy Patricia Churchland discusses the profound influence of one's theoretical framework on the way that data is interpreted, using specific examples from the study of the human brain and cognition. She notes that the invention of symbolic logic by Frege, Russell and Whitehead "changed how people thought about mathematics, about logic, and about language. And it gave new life to empiricism by holding out the vision that the whole of science - even the whole of one's cognitive system - might be systematized by logic in the way that the whole of mathematics was systematized" (Churchland 1986: 252). This theoretical framework has since become doctrine and pervades virtually all work done in the sciences in this century. The implications this framework has for semantics are fundamental and have until recently been accepted without question. For one thing, logical empiricism entails the perfect separation of knowledge of meaning from knowledge of facts (Churchland 1986: 267), implying that the study of semantics must be entirely divorced from the study of other linguistic phenomena. In addition to being perceived as autonomous, meaning has been presumed to be composed of indivisible building blocks, symbols to be manipulated, symbols with no shape or structure more complicated than that provided by set theory. These assumptions form a common thread that runs through much of linguistic theory of the past fifty years.3 Research on cognition carried out by psychologists in the 1970s4 indicated that the presumed model could not account for the way in which human beings store and access meaning. Based on these findings and on further work carried out on natural language, a group of linguists began to develop a new theory of semantics which has come to be known as cognitive semantics.f The fundamental concepts involved are presented in Lakoff 1982 and 1986. Briefly, Lakoff states that meaning is organized in cognitive categories which have a network structure. At the center of a given category is a prototype member of the category. Other members are placed in the category according to their relationship to the prototype, which may be very close or peripheral, thus giving the category a radial structure. As a result, rather than an unanalyzable bounded set of presumably homogeneous members, cognitive semantics provides for a hierarchically structured network of interrelated members, joined by their relationships to the prototype. Langacker (1987: 49) summed up the goals of cognitive semantics quite concisely: "the only appropriate basis for natural language semantics is a subjectivist theory of meaning that successfully accommodates conceptualization, cognitive domains and the various dimensions of conventional imagery."

t cognitive semantics is

ilosophy Patricia Churchland discusses the profound influence of etical framework on the way that data is interpreted, using specific om the study of the human brain and cognition. She notes that the f symbolic logic by Frege, Russell and Whitehead "changed how ght about mathematics, about logic, and about language. And it gave empiricism by holding out the vision that the whole of science - even f one's cognitive system - might be systematized by logic in the way ole of mathematics was systematized" (Churchland 1986: 252). This ramework has since become doctrine and pervades virtually all work sciences in this century. The implications this framework has for e fundamental and have until recently been accepted without question. ng, logical empiricism entails the perfect separation of knowledge of om knowledge of facts (Churchland 1986: 267), implying that the mantics must be entirely divorced from the study of other linguistic . In addition to being perceived as autonomous, meaning has been o be composed of indivisible building blocks, symbols to be manipuols with no shape or structure more complicated than that provided by These assumptions form a common thread that runs through much of eory of the past fifty years.3 on cognition carried out by psychologists in the 1970s4 indicated that d model could not account for the way in which human beings store meaning. Based on these findings and on further work carried out on uage, a group of linguists began to develop a new theory of semantics ome to be known as cognitive semantics.f The fundamental concepts e presented in Lakoff 1982 and 1986. Briefly, Lakoff states that organized in cognitive categories which have a network structure. At f a given category is a prototype member of the category. Other memced in the category according to their relationship to the prototype, be very close or peripheral, thus giving the category a radial structure. rather than an unanalyzable bounded set of presumably homogeneous ognitive semantics provides for a hierarchically structured network of members, joined by their relationships to the prototype. Langacker ummed up the goals of cognitive semantics quite concisely: "the only basis for natural language semantics is a subjectivist theory of at successfully accommodates conceptualization, cognitive domains ous dimensions of conventional imagery."

5

Crucial to an understanding of radial categories is the fact that "the central case does not productively generate all [the] subcategories [Le., peripheral members]. Instead, the subcategories are defined by convention as variations on the central case ... and have to be learned" (Lakoff 1987: 84). Cognitive semantics therefore, unlike approaches that invoke a core plus rules, does not aim for absolute predictive power; its goal is instead to seek the principles which motivate the structures of extant categories in natural languages. It follows, therefore, that one of the important claims of cognitive semantics is that languages contain entire structured categories, rather than cores plus rules for generating such categories. This means that a given category is an integral part of a language, and not constantly and predictably generated. Without this claim it would be impossible to explain why the "same" category (i.e., one based on the same prototype ["core"] and having the same principles linking members ["rules"]) would vary from language to language, a fact that will be demonstrated in the analysis. A second essential characteristic of cognitive categories is that they are generally not built from primitives, but are defined via the prototype, which is itself a gestalt. 6 This does not mean that there are no "parts" to a prototype, but rather that the perception of the parts is secondary and must be consciously guided, whereas the perception of the whole is direct. An example of a radially-structured category is the English word mother (as explicated in Lakoff 1987: 83). At the center of the category stands the prototype, which defines mother as a woman "who gave birth to the child, suplied her half of the child's genes, nurtured the child, is married to the father, is one generation older than the child, and is the child's legal guardian". The category mother, however, also contains many non-prototypical members that stand one or more steps removed from the prototype because they lack one or more of the characteristics of the prototype listed above. Thus there are mothers who did not give birth - step-mothers, adoptive mothers, and foster mothers; mothers who did not nurture - birth mothers and surrogate mothers; and mothers who are not married to the father - unwed mothers. Modern science has produced an extreme example of mother, the "genetic mother" (who has only contributed an egg, but never bore or nurtured the child) that is a relatively peripheral member of this category. The very fact, however, that a genetic mother is referred to as a mother, demonstrates the way in which relationships to the prototype bind the network in a radial structure. The presence of category members that share no overlapping characteristics (such as birth mother vis-a-vis foster mother) rules out a feature analysis. All of these groups of women can be called "mothers" only by virtue of their relationship to the prototypical mother.

-'

6

The implications of the cognitive model for semantics are far-reaching, for this theoretical framework makes it possible to perform coherent and penetrating semantic analyses without losing sight of the integrity of a given category, and also to discuss why a category has the members it does. Under a set-theory approach one could only list subsets, and since there is no formal device for showing their interrelationships, this has produced fragmentary accounts of categories." in which one literally cannot see the forest for the trees. The only alternative was to view the undifferentiated sets as wholes, which could only be characterized in vague terms that denied the actual variety of members they contained.f By positing an internal structure for categories, cognitive semantics avoids the dilemma of choosing between unity and diversity presented by set theory. It enables the linguist to seek as much detail in his description as desired without endangering the integrity of the category. The network may become increasingly intricate, but by virtue of the fact that its structure is based on interrelationships, constant reference is made to the prototype and those members closest to it. Also, the use of conventional imagery to caption the kinesthetic image-schernas? associated with a category produces descriptions which are intuitively satisfying as well as formally elegant. Cognitive semantics is, of course, new and has to date a limited number of subscribers, yet a number of linguists who do (or did) not work within this framework have made statements indicating that they reject a symbolic logic/set theory approach and some have even advocated a network structure for meaning. 10 Sgall et al. (1986: 10) affirms that "linguistics cannot be reduced to a part of mathematics, since there is a major difference between natural language and the formal languages of logic." Potebnja ([1958]: 70, 431) agrees that symbolic logic is not adequate to describe grammatical and semantic relations and argues that it is the linguist's task to "posit formal meanings which are more concrete, to make them as distinct as possible and to show their genetic relationships", a statement which suggests that (sub )meanings are interrelated. The authors of Russkaja grammatika [Russian grammar] (Svedova 1982: 479) are even more specific in describing the structure of case meaning. They claim that cases are polysemantic, that each case has its own system of meanings and that some meanings are basic and central, whereas others are semantically peripheral. Likewise Nunberg (1979: 179) comments that "we could ... assume that all of the several uses of a form are connected by a network of referring functions". Plewes (1977: ix) set out from a Jakobsonian framework, and at a more concrete level found structure not accounted for by distinctive features: "the contextual variants of a single invariant combination, while related at an abstract level, may separate into distinct semantic groupings at intermediate levels of abstraction". Schlesinger

cations of the cognitive model for semantics are far-reaching, for this ramework makes it possible to perform coherent and penetrating alyses without losing sight of the integrity of a given category, and uss why a category has the members it does. Under a set-theory ne could only list subsets, and since there is no formal device for eir interrelationships, this has produced fragmentary accounts of in which one literally cannot see the forest for the trees. The only was to view the undifferentiated sets as wholes, which could only be d in vague terms that denied the actual variety of members they By positing an internal structure for categories, cognitive semantics dilemma of choosing between unity and diversity presented by set ables the linguist to seek as much detail in his description as desired angering the integrity of the category. The network may become intricate, but by virtue of the fact that its structure is based on ships, constant reference is made to the prototype and those members Also, the use of conventional imagery to caption the kinesthetic rnas? associated with a category produces descriptions which are tisfying as well as formally elegant. e semantics is, of course, new and has to date a limited number of yet a number of linguists who do (or did) not work within this have made statements indicating that they reject a symbolic logic/set oach and some have even advocated a network structure for Sgall et al. (1986: 10) affirms that "linguistics cannot be reduced to a ematics, since there is a major difference between natural language mal languages of logic." Potebnja ([1958]: 70, 431) agrees that ic is not adequate to describe grammatical and semantic relations and it is the linguist's task to "posit formal meanings which are more make them as distinct as possible and to show their genetic s", a statement which suggests that (sub )meanings are interrelated. of Russkaja grammatika [Russian grammar] (Svedova 1982: 479) are pecific in describing the structure of case meaning. They claim that lysemantic, that each case has its own system of meanings and that ngs are basic and central, whereas others are semantically peripheral. nberg (1979: 179) comments that "we could ... assume that all of the of a form are connected by a network of referring functions". Plewes et out from a Jakobsonian framework, and at a more concrete level ure not accounted for by distinctive features: "the contextual variants nvariant combination, while related at an abstract level, may separate semantic groupings at intermediate levels of abstraction". Schlesinger

7

investigated the semantics of the instrumental in a number of languages and concluded that the meanings comitative, instrument and manner form a continuum. He then pondered the remaining meanings of the instrumental and asked "are cognitive structures constituted of isolated continua, or perhaps these continua are components of a more complicated system?" (Schlesinger 1979: 321). These linguists were obviously groping for a model that would provide the kind of structure that cognitive semantics offers. When the discussion turns more specifically to case and the various theories that have been applied to it, there will be more mention of statements which support the cognitive framework made by linguists who did not work with this model.

1.1. Support for cognitive semantics from other disciplines 1.1.1. Psychology As mentioned above, it was work in psychology that stimulated the development of cognitive semantics in the field of linguistics. In a series of experiments probing human categorization of natural objects, Rosch (1973b: 111) found no support for the way that "psychological and linguistic research has tended to treat categories, as though they were internally unstructured - that is, as though they were composed of undifferentiated, equivalent instances - and as though category boundaries were always "well defined"." Contrary to the tenet of empiricist logic that category membership must be all or none, Rosch's work with human categorization demonstrated that category boundaries are not necessarily definite and that categories are internally structured (Mervis and Rosch 1981: 109). Their structure is based on a gradient of relationships to a prototype, which is encoded as a mental image that represents the category as a whole. 11 It should be noted that the term mental image may refer to something considerably more abstract than a picture, as noted by Shepard (1978: 130), who has also suggested that the neural representation of an image probably reflects the structure of the image, a hypothesis which has been confmned by research in neurobiology. 12 Cognitive semantics, then, has clearly adapted Rosch's prototype theory without major alteration. The central member of the cognitive category is the prototype and is captioned by conventional imagery, represented in a profile (a representation parallel to a mental image). The structure of the semantic category is determined by the nature of relationships between the profiles of various members and the prototype. As will be seen below, it is also significant that

8

Rosch's work (Mervis and Rosch 1981: 104) suggests that categories are processed holistically, i. e., that when reference is made to any member, reference is also made to the whole category, for this notion is supported by research in neurobiology and artificial intelligence. 1.1.2. Neurobiology Patricia Churchland, herself a philosopher, spent several years examining recent developments in the study of the brain in order to discoverwhat implications they might have for theories of meaning and cognition and to construct a "unified science of the mind-brain", termed "neurophilosophy" (Churchland 1986: ix). She found that the results of researchin neurobiology contradictthe "familiar and virtually doctrinalcomputer metaphorfor the mind's representations and computation" (Churchland 1986: 252) on severalcounts.It can no longer be claimed that the brain stores and processes finite bits of information in vast sequences of discrete steps. For one thing, problems of access would become astronomically difficult (Churchland 1986: 395). Also, given what is known about the speed at which neurons fire and the number of steps it takes to work through an ordinary problem (such as reaching for an object) in the linear fashion of a digital computer, it is necessary to predict a processing time several orders of magnitude greaterthan the time whichthe brain actually takes to perform such tasks (Churchland 1986: 35 and Cottrell 1985: 7). Insteadof storing information in "centers", it appears that the brain accomplishes this by using networks of neurons which map the information in the brain. Whenever any part of a network is accessed, activation spreads throughout its pattern. Problem-solving is not carried out by serial computation, but rather by matrix multiplication, i. e., the pattern of activity of one neural net is mapped onto another. This is known as Tensor Network Theory, and provides a relatively straightforward solution for the conversion of vector information, such as that necessary for catching a ball (which involves coordinating the movementof the ball with that of the hand). Scientific evidence for the validity of this theory is provided by studies of brain anatomy and by neuronal experimentation and, in addition, it gives an intuitively appealing account of the operation of the brain. After all, we certainly do not compute vectors when catching a ball, we simply match our perception of the ball's movement with the muscular movements necessary to make our hand intercept its path, or, in other words, we appear to map perceived movement onto muscular movement. The discovery that "neurons are plastic ... their informationally relevant parts.grow and shrink ... [and] this appears to be essential to their functioning as informa-

rk (Mervis and Rosch 1981: 104) suggests that categories are olistically, i. e., that when reference is made to any member, reference e to the whole category, for this notion is supported by research in y and artificial intelligence.

obiology

rchland, herself a philosopher, spent several years examining recent ts in the study of the brain in order to discoverwhat implications they for theories of meaning and cognition and to construct a "unified he mind-brain", termed "neurophilosophy" (Churchland 1986: ix). hat the results of researchin neurobiology contradictthe "familiar and ctrinalcomputer metaphorfor the mind's representations and compurchland 1986: 252) on severalcounts.It can no longer be claimed that ores and processes finite bits of information in vast sequences of ps. For one thing, problems of access would become astronomically hurchland 1986: 395). Also, given what is known about the speed at ons fire and the number of steps it takes to work through an ordinary ch as reaching for an object) in the linear fashion of a digital comnecessary to predict a processing time several orders of magnitude the time whichthe brain actually takes to perform such tasks (Church5 and Cottrell 1985: 7). Insteadof storing information in "centers", it the brain accomplishes this by using networks of neurons which map tion in the brain. Whenever any part of a network is accessed, preads throughout its pattern. Problem-solving is not carried out by utation, but rather by matrix multiplication, i. e., the pattern of activity al net is mapped onto another. This is known as Tensor Network provides a relatively straightforward solution for the conversion of mation, such as that necessary for catching a ball (which involves g the movementof the ball with that of the hand). Scientific evidence dity of this theory is provided by studies of brain anatomy and by perimentation and, in addition, it gives an intuitively appealing account tion of the brain. After all, we certainly do not compute vectors when all, we simply match our perception of the ball's movement with the ovements necessary to make our hand intercept its path, or, in other appear to map perceived movement onto muscular movement. The hat "neurons are plastic ... their informationally relevant parts.grow .. [and] this appears to be essential to their functioning as informa-

9

tion-processing units" (Churchland 1986: 35) has further ramifications for learning. A theoretical framework based on sets and symboliclogic would predict that the brain would simply add more and more bits of information. Tensor Network Theory, however, indicates that learning involves the expansion of neural nets and this entails the automatic integration of new information in established patterns, a process not readilyaccounted for by logicalempiricism. Rosch's work makes plausible a claim of psychological reality for the structure of meaning invoked by cognitivesemantics, and neurobiological research further suggests that it may have physicalreality (based on the anatomy and function of the brain) as well. It would be premature at this point to make either such claim, but it is clear that the fundamental precepts of cognitive semantics are at least compatible with the findings of neurobiologists and psychologists, and in fact they are more compatible than those of recent theories of semantics.

1.1.3. Artificial intelligence Computer scientists have also traditionally worked within the framework of symbolic logic and, as mentioned above, it was their model for information storage and processing that came to be' accepted as the metaphor for brain function. The field of artificial intelligence, which grew out of a desire to mimic the brain's abilitieswith machines, had by the 1970sreached such an impassethat Dreyfus wrote an extensive expose of its failures. In his conclusion he asks, "Is an exhaustive analysis of humanreasoninto rule-governed operations on discrete, determinate, context-free elements possible? Is an approximation to this goal of artificial reason even probable? The answer to both of these questions appears to be, No" (Dreyfus 1979: 303). Dreyfus, however, did not lay the blame with the computers, but rather with the theoretical framework assumed by their programmers. Although many computer scientists chose to reject Dreyfus' conclusions, there were some who shared his contention that fundamental theoretical changes would have to be made before the field could progress. They began to experiment with alternative meansof storing and processing information, by using systemscharacterized by massive parallelism and high connectedness. It is this feature of the resultant movement in computer science which gives it its name: Connectionism. Experiments have been conducted in the application of connectionism to simulation of phenomena as diverse as visual perception and motor control,13 but only one piece of work will be cited here because it deals specifically with case. Cottrell (1985) employed a connectionist model in writinga program which would disambiguate both word sense and case function in

10

English. He cites a desire to imitate real brain function and to avoid the pitfalls of existing theories as his motivation for choosing the connectionist model. 14 Rather than using symbol-passing (the computational equivalent of linguistic features or of discrete bits of semantic information), Cottrell's program is designed to store and access imformation in networks. When a word is referenced, the entire network of its meanings and functions is activated and then allowed to "relax" to a consistent [with other input] interpretation" (Cottrell 1985: 207). Cottrell had his program analyze a series of English sentences and found it to be efficient at this task. In addition, it was found to be comparable to the functioning of the brain in some ways: "One test of the validity of the model presented ... is to evaluate its adequacy at accounting for neurological data .... The model is shown to be adequate for explaining some of the overall effects" (Cottrell 1985: 161).

1.1.4. Summary Both psychology and neurobiology fail to provide evidence that human categorization is organized in a manner consistent with set theory and symbolic logic. In both cases we infer that information is not stored in discrete and unrelated bits, but rather in hierarchized structures which interact as wholes. The application of these principles to computer science has produced a fruitful new movement after years of stagnation, and it seems plausible that this approach would likewise be a powerful tool in advancing our linguistic understanding of semantics.

1.2. Case as a semantic entity

There are three obstacles to the semantic description of case. The first is formal and relatively minor; the second two have to do with the nature of case semantics and will be treated in more detail.

1.2.1. Case lacks formal autonomy No single case has a unique surface representation by means of which it is signaled, a fact that prevents case from fitting neatly into the structuralist schema of "one form, one meaning". Burston (1977: 51) put it succinctly: "a certain oddness remains in calling a "sign" a linguistic element which, whenever it occurs,

11

cites a desire to imitate real brain function and to avoid the pitfalls of ries as his motivation for choosing the connectionist model. 14 Rather ymbol-passing (the computational equivalent of linguistic features or its of semantic information), Cottrell's program is designed to store imformation in networks. When a word is referenced, the entire s meanings and functions is activated and then allowed to "relax" to a with other input] interpretation" (Cottrell 1985: 207). Cottrell had his lyze a series of English sentences and found it to be efficient at this ion, it was found to be comparable to the functioning of the brain in "One test of the validity of the model presented ... is to evaluate its accounting for neurological data .... The model is shown to be explaining some of the overall effects" (Cottrell 1985: 161).

mary

ology and neurobiology fail to provide evidence that human n is organized in a manner consistent with set theory and symbolic th cases we infer that information is not stored in discrete and s, but rather in hierarchized structures which interact as wholes. The f these principles to computer science has produced a fruitful new fter years of stagnation, and it seems plausible that this approach se be a powerful tool in advancing our linguistic understanding of

a semantic entity

ee obstacles to the semantic description of case. The first is formal minor; the second two have to do with the nature of case semantics eated in more detail.

acks formal autonomy

se has a unique surface representation by means of which it is ct that prevents case from fitting neatly into the structuralist schema one meaning". Burston (1977: 51) put it succinctly: "a certain oddin calling a "sign" a linguistic element which, whenever it occurs,

lacks formal autonomy". She further comments, however, that "this difficulty can be avoided if cases are envisaged not as signs, but as features of grammatical nature which exist only in combination with other features". This fact does not hinder most linguists, who are nevertheless willing to posit meanings for cases, and indeed Kilby (1982: 75) asserted that "any adequate description of the Russian case system must at some point make reference to the meanings expressed by the cases".

1.2.2. Case bears subjective as well as objective meaning A second obstacle is the permeation of case meaning by a subjective element. The fact is that case actually encodes not one, but two basic functions which are themselves related in a non-trivial way. In addition to referring to the role of a substantive in a narrated event, case also encodes the speaker's interpretation of that role. This second function is quite salient in so-called "pragmatic" uses of case, such as the ethical dative, and can be decisive in determining, for example, whether possession will be encoded by the genitive or the dative or whether a predicate substantive will be marked by the nominative or instrumental. The distinction between these two functions is dominant in the work of van Schooneveld, who terms the second of them "transmissional deixis" and attributes it to an observation by Jakobson that distinctive features can be deictic and thereby refer to the speech event itself. 15 An example of the ethical dative illustrates what van Schooneveld has termed "transmissional deixis". The addressee of the utterance in (1) has no relationship to the narrated event, he has neither witnessed it, nor was he previously aware of it, yet the speaker subjectively imposes a dative relationship that holds between the hearer and the narrated event by inserting the dative pronoun vdm "you" into his narrative. (1)

"J 0, a na universite" pokracova Yes and at university-Lac continued potichu pan Keval, "tam se quietly Mr. Keval-NOM there REFL-ACC vdm dnes seprala pnrodovedecka YOU-DAT today fought natural-science-NOM fakulta s historickou." (Capek) department-NOM with historical-INST

12

, "Yes, and at the university", continued Mr. Keval quietly, "(hey, you know what?) the natural sciences department had a fight with the history department today." , The notion that case meaning has two functions is upheld by other linguists as well, although most refer to it as an opposition of semantics to pragmatics. These two functions interact with each other in a complex manner. Adamec (1978: 35) , gives an example of a sentence in which the deictic function does more than present the speaker's point of view - it actually forces a different interpretation of the narrated event:

(2)

a.

b.

goroda. On videl ekzoticeskie cities-Ace He-NOM saw exotic 'He saw exotic cities.' Emu videlis' ekzoticeskie Him-DAT saw-refl exotic goroda. cities-NOM 'He saw exotic cities (in his dreams).'

(2a), with an ordinary nominative subject, makes a neutral statement, whereas (2b), which has an impersonal dative construction, can only be interpreted as referring to fantasy, i.e. to imagined or dreamed cities. Furthermore, Langacker (1987: 50) points out that "there are alternate ways of construing a situation, and alternate ways of symbolizing a given construal", suggesting that the deictic function of case is itself non-simple. It appears, then, that case can carry two kinds of meaning that are complex and interact with each other. Given such a situation, one might ask how the linguist is to proceed - should the functions be untangled and dealt with separately? Nunberg (1979: 143) notes that the distinction between these two functions is anything but clear-cut: "the semantics/pragmatics distinction cannot be validated even in principle". Wierzbicka (1980: 144) likewise asserts that there is no "abyss between truth-functional and other "subjective" or "pragmatic" meanings". Van Schooneveld (1978: 216) provides a clue as to why this is so when he affmns that all case features carry transmissional deixis (i.e., that the deictic function is always present). What this means is that case always makes reference to the speaker's interpretation of the narrated event, and that in fact no narrated event is ' encoded directly, unmediated by this interpretation.18

and at the university", continued Mr. Keval "(hey, you know what?) the natural scienartment had a fight with the history department ,

that case meaning has two functions is upheld by other linguists as h most refer to it as an opposition of semantics to pragmatics. These interact with each other in a complex manner. Adamec (1978: 35) mple of a sentence in which the deictic function does more than eaker's point of view - it actually forces a different interpretation of vent:

goroda. n videl ekzoticeskie cities-Ace e-NOM saw exotic He saw exotic cities.' mu videlis' ekzoticeskie im-DAT saw-refl exotic oroda. ties-NOM He saw exotic cities (in his dreams).'

ordinary nominative subject, makes a neutral statement, whereas has an impersonal dative construction, can only be interpreted as antasy, i.e. to imagined or dreamed cities. Furthermore, Langacker ints out that "there are alternate ways of construing a situation, and ys of symbolizing a given construal", suggesting that the deictic se is itself non-simple. then, that case can carry two kinds of meaning that are complex and each other. Given such a situation, one might ask how the linguist is should the functions be untangled and dealt with separately? 79: 143) notes that the distinction between these two functions is clear-cut: "the semantics/pragmatics distinction cannot be validated ple". Wierzbicka (1980: 144) likewise asserts that there is no "abyss h-functional and other "subjective" or "pragmatic" meanings". Van (1978: 216) provides a clue as to why this is so when he affmns that ures carry transmissional deixis (i.e., that the deictic function is nt). What this means is that case always makes reference to the rpretation of the narrated event, and that in fact no narrated event is ' ctly, unmediated by this interpretation.18

13

The subjective moment of case meaning is most saliently present in examples of semantic extension via mapping. It involves the application of basic meanings (here, case relationships) to domains other than perceived objective reality. Case relationships can thus be mapped onto subjective realms, such as the speech-act domain, and they can also be subjectively applied to perceived reality. This description of subjective case meaning as subjective mapping of objective (semantic) case meaning accounts for the intimate relationship between these two kinds of case meaning and easily accommodates van Schooneveld's notion of transmissional deixis and the observations of other scholars mentioned in this section. More will be said about semantic extension via mapping and its role in the structure of case categories in section 1.3.2., and abundant examples of subjective application of case meaning will be provided in the analysis.

1.2.3. Case meaning and lexical semantics The third obstacle to a description of case semantics involves the objective function of case meaning. As mentioned above in the Introduction, there is a lack of concrete, real-world correspondence for the relationships encoded by case. It seems that the kind of meaning expressed by case is different from that expressed by other morphemes. Indeed, case belongs to the realm of grammar rather than lexicon, and even by the standards of grammatical relations, it is elusive. While one can point to a real instantiation of gender, for example, it is hard to do this for a case role like the dative. Potebnja ([1958]: 35-37) states that every word contains two discrete kinds of information: 1. lexical and 2. grammatical, but then qualifies this statement by adding that these two kinds of information are inseparably combined in the minds of speakers: "the grammatical form is an element of the meaning of a word and springs from the same basis as its essential meaning". Thus it would appear to be artificial to divide these two kinds of meaning, since they do not differ in essence. In commenting on Jakobson's description of case, Sangster (1982: 114) notes that "if syntactic phenomena are sign vehicles, they should display semantic properties just like any other set of linguistic forms". Langacker (1987: 17) is even frrmer in asserting that "grammar and lexicon form a continuum" . Although Talmy (1986: 1), like Langacker, uses a cognitive framework and does not deny that lexical and grammatical information share a common basis, he views them as "two subsystems [which] have distinct semantic functions, ones that are indispensable and complementary", and thus is able to make several valuable observations about the semantic properties specific to grammatical

14

elements. He points out that grammatical, as opposed to lexical, elements are characteristically relativistic, that is, they refer to relative rather than absolute values of magnitude, shape, rate, material and other properties. They use these relativistic specifications to break up an event into parts and participants and thereby "the grammatical specifications in a sentence ... provide a conceptual framework or, imagistically, a skeletal structure or scaffolding, for the conceptual material that is lexically specified" (Talmy 1986: 1). These observations further lead him to "speculate that the cognitive function of such classification [of grammatical elementsinto small, closed classes which are relativisticrather than absolute] lies in unifying contentful material within a single conceptual systemand in rendering it manipulable - i.e., amenable to transmission, storage, and processing- and that its absence wouldrendercontentan intractable agglomeration" (Talmy 1986: 30). Talmy explains not only how the grammatical pole of the informational continuum differs from the lexicalpole, but also why this difference exists: it is crucial to the efficientfunctioning of language. Yet the fact that grammatical elements function more to organize ideas than to specify them does not entail a difference between them and lexicalelementsin basic cognitive structure. Indeed, like units of lexical meaning, "grammatically specified notions can be seen to patternin categories, and the categories, in tum, in integrated systems".19 The hypothesis that grammatical information differs in some ways from lexical information yet forms a continuum with it accounts for some of the intimateways in which these two semantic subsystems interact. Without a framework that allows for overlap between them, it would be hard to explain the following phenomena. For one thing, there appears to be a high correlation between case and noun meaning. Greenberg (1974: 33) tabulated the information given in a Russian frequency dictionary and found that "the distribution of frequency over the cases is far from random in relation to the semantic characteristics of the noun". Thus a given case is associated with certain semantic groups of nouns whose meaning is compatible with that whichthe case role entails. Popova (1970: 95) notes that when the noun's lexical meaning does not correspond to the concrete meanings of the case, metaphor is often necessarily invoked, as with the use of an abstract rather than a concrete noun in the instrumental phrase kormit' obescanijami "nourish with promises". In addition, Greenberg (1974: 34) found that the "polysemy of individual nouns [is] to a great extent disambiguated by case". Conversely it is also possible for the lexical meaning of the noun to determine which submeaning of a case is expressed, as noted by Potebnja, Pit'ha and Levine.20 In both instances we see the grammatical and lexical elements complementing each other as they engage in a semantic tug-of-war. In Russian and Czech, at least, there is evidence that the lexical end of the continuum is

points out that grammatical, as opposed to lexical, elements are cally relativistic, that is, they refer to relative rather than absolute gnitude, shape, rate, material and other properties. They use these pecifications to break up an event into parts and participants and grammatical specifications in a sentence ... provide a conceptual , imagistically, a skeletal structure or scaffolding, for the conceptual is lexically specified" (Talmy 1986: 1). These observations further "speculate that the cognitive function of such classification [of elementsinto small, closed classes which are relativisticrather than in unifying contentful material within a single conceptual systemand it manipulable - i.e., amenable to transmission, storage, and prothat its absence wouldrendercontentan intractable agglomeration" 6: 30). Talmy explains not only how the grammatical pole of the continuum differs from the lexicalpole, but also why this difference ucial to the efficientfunctioning of language. Yet the fact that graments function more to organize ideas than to specify them does not ence between them and lexicalelementsin basic cognitive structure. units of lexical meaning, "grammatically specified notions can be nin categories, and the categories, in tum, in integrated systems".19 hesis that grammatical information differs in some ways from lexical et forms a continuum with it accounts for some of the intimateways se two semantic subsystems interact. Without a framework that verlap between them, it would be hard to explain the following For one thing, there appears to be a high correlation between case aning. Greenberg (1974: 33) tabulated the information given in a uency dictionary and found that "the distribution of frequency over ar from random in relation to the semantic characteristics of the a given case is associated with certain semantic groups of nouns ng is compatible with that whichthe case role entails. Popova (1970: at when the noun's lexical meaning does not correspond to the nings of the case, metaphor is often necessarily invoked, as with the tract rather than a concrete noun in the instrumental phrase kormit' "nourish with promises". In addition, Greenberg (1974: 34) found ysemy of individual nouns [is] to a great extent disambiguated by ersely it is also possible for the lexical meaning of the noun to hich submeaning of a case is expressed, as noted by Potebnja, Pit'ha 0 In both instances we see the grammatical and lexical elements ng each other as they engage in a semantic tug-of-war. In Russian t least, there is evidence that the lexical end of the continuum is

15

encroaching on the grammatical end. In both languages the use of bare case is losing ground to prepositions.U Indeed, Apresjan (1973: 5) was quite correct in asserting that "there exists a bilateral dependence between the syntactic [i.e., grammatical] and semantic [i.e., lexical] properties of linguistic expressions." Within this framework it is clear that case always bears a meaning function, and the fact that it belongs to syntax does not diminish its semantic freight. As will be shown below, this view directlycontradictsChomsky'sGovernmentand Binding Theory, and any model that treats case as a phenomenon of syntax that can be wholly divorced from semantics. Thus I agree with Wierzbicka(1980: xi) that surface case is never arbitrary. If it were, it would be necessary to assume that case markers are semantically void at least part of the time, a notion which is incompatible with the basictenets of cognitive semantics. 22 1.2.4. Summary To sum up, the following assumptions are inherent in the cognitive framework: (i) (ii) (iii)

(iv)

case is always meaning-bearing case meaning has a constant objective moment that can be subjectively applied case meaning involves the organization of rather than the specification of information case meaning is notessentially different fromlexicalmeaning in structure.

1.3. What a cognitive description of case should look like

The present section will integrate the basic premises set out above in a specific description of how case meaning is to be approached in a cognitive framework. 1.3. 1. Schematic representation of casemeaning The remarks made by Talmy (1986) about the nature of grammatical meaning have already put some restrictions on the kind of description that case should be given:

16

(i)

(li)

it should refer to the overall organization of events, giving just their skeletal structure, and it should be relatively abstract, without reference to any absolute measure of magnitude, shape, rate, etc.

Langacker (1987) and Smith (1985 and 1987) have both presented preliminary descriptions of case that are consistent with these principles. As mentioned above, the use of conventional imagery is central to the cognitive model. The relativistic nature of grammatical meaning should not be an obstacle to captioning it imagistically, for "the fact that an internal representation is more abstract than a picture does not entail that such a representation is nonvisual" (Shepard 1978: 130). Langacker (1987: 1-18) devotes considerable discussion to developing an appropriate means of captioning the conventional imagery of case. His approach makes two assumptions about the idealized cognitive models (IeMs) of predications: (i)

(li)

human beings conceive of the world as containing discrete objects at specific locations which move and interact with each other through direct physical contact ("billiard-ball model")23 in perceiving these interactions, human beings organize scenes into participants and settings and segment clusters of action into discrete events ("stage model").

By combining these models, Langacker (1987: 5) is able to produce a caption of.the "normal observation of a prototypical action", which includes an agent that acts on a patient. The event is structured in an action chain in which circles (an arbitrary, neutral shape) symbolize participants and an arrow symbolizes interaction (movement of energy).24 Case markers "profile", or focus attention on, specific parts of the action chain. In this prototypical action chain, the "head" (leftmost participant) is the nominative and the "tail" (rightmost participant) is the accusative. The basic unit referred to here is clearly the clause; the prototypical action captioned in Figure 1 corresponds to a prototypical clause with a subject and object. Thus Langacker gives a concrete interpretation of a fundamental concept of cognitive semantics formulated by Talmy (1986: 1): that "a sentence [evokes] in the listener a particular kind of experiential complex, ... a "cognitive representation"." The assumption that a study of case must be founded on sentence structure is not unique to Langacker and Talmy, or to cognitive semantics for that matter. Serensen (1957: 31) states that "the sentence is the unit from which the analysis of case must start". Ulicny (1984: 17) likewise insisted that the relational

uld refer to the overall organization of events, giving just their sketructure, and uld be relatively abstract, without reference to any absolute measure gnitude, shape, rate, etc.

(1987) and Smith (1985 and 1987) have both presented preliminary of case that are consistent with these principles. As mentioned above, nventional imagery is central to the cognitive model. The relativistic mmatical meaning should not be an obstacle to captioning it imagishe fact that an internal representation is more abstract than a picture ail that such a representation is nonvisual" (Shepard 1978: 130). 987: 1-18) devotes considerable discussion to developing an approof captioning the conventional imagery of case. His approach makes ons about the idealized cognitive models (IeMs) of predications:

17

functions of case must be sought in the semantic structure of sentences and distinguishes nine types of sequences of beginning-middle-end ordering which are vaguely correlated to participants in Langacker's action chains.

O~E) setting

n beings conceive of the world as containing discrete objects at speocations which move and interact with each other through direct phyontact ("billiard-ball model")23 ceiving these interactions, human beings organize scenes into partis and settings and segment clusters of action into discrete events e model").

ning these models, Langacker (1987: 5) is able to produce a caption al observation of a prototypical action", which includes an agent that ent. is structured in an action chain in which circles (an arbitrary, neutral lize participants and an arrow symbolizes interaction (movement of ase markers "profile", or focus attention on, specific parts of the In this prototypical action chain, the "head" (leftmost participant) is e and the "tail" (rightmost participant) is the accusative. unit referred to here is clearly the clause; the prototypical action Figure 1 corresponds to a prototypical clause with a subject and Langacker gives a concrete interpretation of a fundamental concept of mantics formulated by Talmy (1986: 1): that "a sentence [evokes] in a particular kind of experiential complex, ... a "cognitive repreThe assumption that a study of case must be founded on sentence ot unique to Langacker and Talmy, or to cognitive semantics for that nsen (1957: 31) states that "the sentence is the unit from which the se must start". Ulicny (1984: 17) likewise insisted that the relational

viewer

Figure 1. Using his action-chain notation, Langacker (1987: 28) goes on to define six case roles which "are sufficiently fundamental and cognitively salient to be considered archetypal". His article does not examine the details of how cases are realized in specific natural languages, but rather gives general outlines for conducting such an analysis.

1.3.2. Network structure Although Langacker (1987: 39) deals only with universal prototypes, he makes it clear that he is simply outlining the frontiers; that in-depth studies of natural language cases can and should be undertaken according to this model. Smith 1985 brings us one step closer to actually realizing a cognitive description of the semantics of a case. His aim is to provide a unified account of the various uses of

18

the dative in German. He suggests the following schemas to caption "the defining property" of the nominative, accusative, and dative in German: 25

~ nom

-0 ace

dat

Figure 2 Figure 2.

Smith (1985: 393) has chosen the schema that captions the prototype of the dative because "the dative entity is viewed by the speaker as simultaneously acted upon as well as an actor in its own right", a property that he later (Smith 1987: 455) terms "bilateral involvement".26 In this schema the dative combines the properties of both the nominative (head of chain) and accusative (tail of chain). Smith then goes on to discuss various types of dative constructions in German and the schemas appropriate for them. For sentences which have nominative and accusative as well as the dative, Smith gives the schema shown in Figure 3. This schema is valid both for sentences in which the dative represents an indirect object, as well as those in which it serves other purposes, as in the sentences (dative forms and their equivalents are in roman type): (3)

Fritz offnet derDame die Ttlr, FritZ-NOM opens the lady-DAT the door-Ace 'Fritz opens the door for the lady.'

(4)

Der Arzt hat mir das Leben gerettet. The doctor-NOM has me-DAT the life-NO saved 'The doctor saved my life.'

Smith (1985: 394) notes that verbs that require the dative always contain the implication that the dative entity is the agent of some further action. In considering the verb helfen "help", for example, "a moment's thought will convince one that when one is helped, one is helped to do something, to effect some further (unspecified) action". Smith uses the schema shown in Figure 4 to represent the dative governed by verbs, in which the dotted lines symbolize an unspecifed implied action.

19

German. He suggests the following schemas to caption "the defining the nominative, accusative, and dative in German: 25

~

-0 ace

dat

Figure 2 Figure 2.

985: 393) has chosen the schema that captions the prototype of the se "the dative entity is viewed by the speaker as simultaneously acted as an actor in its own right", a property that he later (Smith 1987: "bilateral involvement".26 In this schema the dative combines the both the nominative (head of chain) and accusative (tail of chain). n goes on to discuss various types of dative constructions in German mas appropriate for them. For sentences which have nominative and s well as the dative, Smith gives the schema shown in Figure 3. This alid both for sentences in which the dative represents an indirect ell as those in which it serves other purposes, as in the sentences s and their equivalents are in roman type):

dat's sphere of control

Figure 3.

---~

nom

dat

offnet derDame die Ttlr, OM opens the lady-DAT the door-Ace opens the door for the lady.'

zt

hat mir das Leben gerettet. has me-DAT the life-NO saved octor saved my life.'

ctor-NOM

85: 394) notes that verbs that require the dative always contain the hat the dative entity is the agent of some further action. In considering en "help", for example, "a moment's thought will convince one that helped, one is helped to do something, to effect some further (unspen". Smith uses the schema shown in Figure 4 to represent the dative verbs, in which the dotted lines symbolize an unspecifed implied

unspecified implied action

Figure 4.

Finally, Smith discusses the dative used in impersonal constructions, as in (5).

nom

dat

Figure 5.

20 (5)

Es it-NOM '[ am cold.'

ist is

mir me-DAT

kall. cold

He proposes that "we can conceive of the abstract force [here, coldness - LAJ] as exerting an effect on the dative [entity] which in turn causes an internal physical reaction within the [dative entity]" (Smith 1985: 396). Smith posits the schema found in Figure 5 to caption this type of dative. The internal reaction is symbolized by a squiggly arrow in the dative landmark. Smith has put forward a series of schemas as well as a prototype for the dative. He does not outline the network that they comprise, but the notion that the submeanings should be clustered in a network structure is implicit in cognitive semantics and in other models of description as well. Vinogradov (1972: 142) stated that the meanings of the Russian dative form a whole, integrated system. In describing the uses of the Polish dative, Wierzbicka refers to its meanings as a "family" (Wierzbicka 1986: 419), and as we shall see in chapter 2, her treatment of the Russian instrumental (Wierzbicka 1980) implies a specific network structure, as does Mrazek's (1964). In a single terse sentence, Langacker (1987: 39) indicates how the schemas suggested by Smith can be arranged to form a network structure: "a case category is generally complex, comprising a network of alternate senses connected by relationships of schematicity and semantic extension". This concise description of the relationships that structure case categories is worth further scrutiny, for it contains an essential insight from which all the specific types of relationships involved can be winnowed. "Schematicity" identifies the abstract way in which the members of a category are united. "Semantic extension" refers to the system of definite positions which individual schemas occupy in the network. In conducting the extensive investigation presented in the analysis I have identified the following classification of subtypes of semantic extension: I.

Variations in the shape of schemas a) b)

paradigmatic variation - variation in the portion of the schema which represents the case-marked entity syntagmatic variation - variation in the composition of the event chain (removal of participants, replacement of the verb with a copular construction)

21

M m cold.'

ist is

mir me-DAT

kall. cold

ses that "we can conceive of the abstract force [here, coldness - LAJ] n effect on the dative [entity] which in turn causes an internal physical hin the [dative entity]" (Smith 1985: 396). Smith posits the schema ure 5 to caption this type of dative. The internal reaction is symbouiggly arrow in the dative landmark. s put forward a series of schemas as well as a prototype for the dative. outline the network that they comprise, but the notion that the subhould be clustered in a network structure is implicit in cognitive sein other models of description as well. Vinogradov (1972: 142) stated nings of the Russian dative form a whole, integrated system. In deuses of the Polish dative, Wierzbicka refers to its meanings as a "fazbicka 1986: 419), and as we shall see in chapter 2, her treatment of instrumental (Wierzbicka 1980) implies a specific network structure, zek's (1964). le terse sentence, Langacker (1987: 39) indicates how the schemas y Smith can be arranged to form a network structure: "a case category complex, comprising a network of alternate senses connected by s of schematicity and semantic extension". This concise description of ships that structure case categories is worth further scrutiny, for it essential insight from which all the specific types of relationships n be winnowed. "Schematicity" identifies the abstract way in which s of a category are united. "Semantic extension" refers to the system positions which individual schemas occupy in the network. In the extensive investigation presented in the analysis I have identified g classification of subtypes of semantic extension:

ations in the shape of schemas

paradigmatic variation - variation in the portion of the schema which represents the case-marked entity syntagmatic variation - variation in the composition of the event chain (removal of participants, replacement of the verb with a copular construction)

II.

Metaphoric extension a) b)

extension to different semantic fields - through relationships of synonymy, antonymy, and metonymy extension to different domains - through mapping of relationships onto the domain of subjective perception or the speechact domain.

The entire analysis is devoted to the presentation of examples of these types of variation, so I will give only brief explanations here. Relations of schema shape can be defined as topological variations in the event chain, and they follow two axes, the paradigmatic and the syntagmatic, involving variations either in the role of the case-marked entity or in the syntactic environment in which it finds itself, respectively. Smith's schemas, which appear in Figures 3 and 4, illustrate paradigmatic variation. In Figure 3 the characteristic mark of the dative is its sphere of control, whereas in Figure 4 the dative is recognized as an agent of an unspecified implied action. Syntagmatic variants are produced by alterations in the length and composition of the event chain with the characteristic mark of the profiled case held constant. An example would be a variation on Figure 3 realized by an intransitive verb, so that the accusative entity is removed from the schema and the verb interacts directly with the dative's sphere of control. The effect is equivalent to the difference between the two English sentences He read a book to me and He read to me. Metaphoric extension may project the relationships captioned by a schema onto a different semantic field, either by making use of relationships of synonymy, antonymy, and metonymy, or by mapping case relationships onto the domain of the speaker's subjective perception of the speech-act domain.I? The Czech indirect object provides numerous examples of the first kind of extension. Narrowly defined, the indirect object is the recipient of the verb ddt "give (to)" . Synonymy effects the spread of this case usage to clauses with other verbs that indicate the bestowal of possessions, such as koupit "buy (for)", poslat "send (to)". Antonymy further extends this usage to verbs having the opposite meaning, i.e., those which indicate the removal of possessions, e.g., vzit "take (from)", krdst "steal (from)". Extension via metonymy occurs with verbs which connote the giving of signals, money, benefit and harm, all of which take the dative in Czech, in which the direct object is already subsumed by the verb and therefore not overtly expressed. The ethical dative in (1) above illustrates the last type of variation, via mapping onto the speech-act domain. Here the speaker invokes a

22

sphere of control on the part of the hearer and makes the claim that the narrated event is in the hearer'ssphere. The four types of variation are presented in the order of their prominence in structuring case categories. Paradigmatic variants account for the basic structure of the network. Syntagmatic variants are local variations on a given paradigmatic variantand, because the syntagmatic variants are frequently syntactically parallel, they often help to build links between the paradigmatic variants. The two types of extension also produce localsubvariants, eitherof a paradigmatic variantor of one of its syntagmatic counterparts. Thus the relationships which structure case categories are arranged hierarchically, such that each successive type produces local variations on the resultsof morebasictypesof variation. Governed case in the cognitive model As Smith's work suggests, governed case (specifically the dative governed by verbs) is not to be dismissed as a mere syntactic phenomenon, for it is part of the semantic whole represented by the network of a given case. Indeed, it would be contrary to the spirit of cognitive semantics to leave any uses of case out of the scope of the description, since this would imply that they have no semantic role, that part of their network is defective, void. The importance of case semantics in verbalgovernment is supported by many scholars, and the verbs associated with this phenomenon are traditionally listed in semantic groups.28 In examining the difference between oblique and accusative objects, Plewes (1977: 100) affirms that "the syntactic processes remain the same; the difference is one of semantics". Gorbacevic (1971: 238) found that the significance of semantics is so great that it has caused some Russian verbs to switch their government. Porazat'sja "be struck (by)" governed only the instrumental in Tolstoy's time, but now, due to semantic analogy with dative-governing verbs such as udivljat'sja "be surprised (at)" and izumljat'sja "be amazed (by)" it appears most frequently in association with the dative. Even Popova (1974: 178), who at the outset of her study of verbal case government states that "the governed uses are semantically unmotivated and are maintained onlyby tradition", admitsin her conclusion that deviantuses of the dative and instrumental arise due to contamination from verbs having a similar meaning. She cannot deny the influence of semantics on case selection, even though it leadsher to an inherentcontradiction. If it is recognized that semantics contributes to the selection of cases governed by verbs, it stands to reason that the use of case with prepositions should also be included in the semantic network. Indeed, many scholars of various orientations

ntrol on the part of the hearer and makes the claim that the narrated he hearer'ssphere. types of variation are presented in the order of their prominence in ase categories. Paradigmatic variants account for the basic structure rk. Syntagmatic variants are local variations on a given paradigmatic because the syntagmatic variants are frequently syntactically parallel, elp to build links between the paradigmatic variants. The two types of so produce localsubvariants, eitherof a paradigmatic variantor of one gmatic counterparts. Thus the relationships which structure case re arranged hierarchically, such that each successive type produces ons on the resultsof morebasictypesof variation.

se in the cognitive model

work suggests, governed case (specifically the dative governed by to be dismissed as a mere syntactic phenomenon, for it is part of the hole represented by the network of a given case. Indeed, it would be he spirit of cognitive semantics to leave any uses of case out of the description, since this would imply that they have no semantic role, heir network is defective, void. rtance of case semantics in verbalgovernment is supported by many d the verbs associated with this phenomenon are traditionally listed in oups.28 In examining the difference between oblique and accusative wes (1977: 100) affirms that "the syntactic processes remain the same; e is one of semantics". Gorbacevic (1971: 238) found that the signiemantics is so great that it has caused some Russian verbs to switch ment. Porazat'sja "be struck (by)" governed only the instrumental in me, but now, due to semantic analogy with dative-governing verbs vljat'sja "be surprised (at)" and izumljat'sja "be amazed (by)" it t frequently in association with the dative. Even Popova (1974: 178), utset of her study of verbal case government states that "the governed antically unmotivated and are maintained onlyby tradition", admitsin on that deviantuses of the dative and instrumental arise due to contam verbs having a similar meaning. She cannot deny the influence of n case selection, even though it leadsher to an inherentcontradiction. ognized that semantics contributes to the selection of cases governed stands to reason that the use of case with prepositions should also be he semantic network. Indeed, many scholars of various orientations

23

agree with Kurylowicz, Isacenko and Vinogradov that prepositions serve to enhance or extendcase meaning, ratherthanreduceit to an entirelysyntactic phenomenon.j? It was this observation that inspired van Schooneveld (1978: 219) to hypothesize a single "cardinal semantic pattern" for bothcases and prepositions. If case governed by prepositions is to be included in a semantic description, then, we might ask, how shall we place it in the network? Should it occupy a special part of the network, or perhaps have a separate, related network of its own? Research on trendsin prepositional usage suggeststhat these questions are inappropriate. El'zbutas (1967) found that in Russian some phrases that previously used the bare dative nowrequire prepositions and conversely, some expressions that used to contain prepositions now appearwithjust the dative. Apparently the line which we commonly assume to divide prepositional from prepositionless case is indistinct and crossable: prepositional case interactswith the semantic network, although its role is perhaps more elusive. As noted by Isacenko (1965: 100), the fact that a preposition of necessity always brings its own semantic freight to an expression, and yet does not have an independent existence in isolation from case, makes it impossible to determine the exact semantic contribution of case in prepositional phrases. An analysis of the use of prepositions will not yield anything aboutthe "general"or "basic" meaning of cases. Thus, in cognitive terms, prepositional case belongs in the semantic network of a given case, but its placeis usually pertpheral.F' Networks and linguistic universals Questions concerning the status of the network also present themselves. What universals, if any, will it reveal? Dahl (1985) providesguidelines for answering this and related questions. He recognizes two approaches to the search for language universals: that of the naive universalist, who seeks universal categories which can be attributed wholesale to any language, and that of the sophisticated universalist, who threshes out universal principles. This implies that what we should be seeking is a set of universal strategies in the organization of case categories rather than universal definitions of given cases. A hypothesis of this sort is implicit in Langacker's (1987: 26) work, for he presents case role "archetypes" which he applies to various languages, yet asserts that "the search for all-encompassing schematic characterizations would not appear promising; case semantics is better approached in terms of language-specific families of senses organized around prototypical values". Wierzbicka (1986: 386) opens her discussion of the Polish dative with an assumption that bears a striking

24

resemblance to this hypothesis: "a case has one core meaning, on the basis of which it can be identified cross-linguistically ..., and a language-specific set of other, related meanings, which have to be specified in the grammatical description of the given language". The hierarchy of variations suggested above provides a first approximation toward a set of universals of case semantics. The implications of this hierarchy and the specific universal predictions it entails will be discussed in the conclusion, after the full weight of data has been brought to bear upon them.

1.3.3. A test of the cognitive model The aim of this book is to take the precepts of cognitive semantics to their logical conclusion; to put them to the test, with semantic data from two of the most intractable problems of Slavic linguistics: the semantics of the dative and instrumental cases in Czech and Russian, respectively. In addition to presenting a prototype and related schemas for all of the various meanings of these two cases, I will show how they are arranged in networks and state the relationships among them that give the networks their structure. The discussion will focus on the Czech dative (as compared with its Russian counterpart), and the Russian instrumental (as compared with its Czech counterpart). The comparisons will serve to show how slight network variations account for the differences in case usage observed in natural languages. This is intended as a contribution to the description of the Slavic case system as well as a test of the validity of the cognitive model.

to this hypothesis: "a case has one core meaning, on the basis of be identified cross-linguistically ..., and a language-specific set of meanings, which have to be specified in the grammatical description language". rchy of variations suggested above provides a first approximation of universals of case semantics. The implications of this hierarchy fic universal predictions it entails will be discussed in the conclusion, weight of data has been brought to bear upon them.

t of the cognitive model

his book is to take the precepts of cognitive semantics to their logical to put them to the test, with semantic data from two of the most roblems of Slavic linguistics: the semantics of the dative and instrus in Czech and Russian, respectively. In addition to presenting a d related schemas for all of the various meanings of these two cases, how they are arranged in networks and state the relationships among ve the networks their structure. The discussion will focus on the e (as compared with its Russian counterpart), and the Russian (as compared with its Czech counterpart). The comparisons will w how slight network variations account for the differences in case ved in natural languages. This is intended as a contribution to the of the Slavic case system as well as a test of the validity of the del.

2.

Cognitive semantics compared with other descriptions of case

In this chapter the cognitive model will be compared with other models that have been used to account for case usage. In addition, it will be demonstrated that many scholars have made statements which are consistent with the cognitive model, even when this entailed contradicting their own assumptions. The fact that the cognitive model can accommodate both the major contributions of most scholars as well as such apparent contradictions suggests that it is the model of choice, and indeed, it might have been chosen by most of the authors cited herein if it had been available to them when they were conducting their research.

2.1. Case semantics in previous traditions

For the sake of the present discussion the following remarks are necessarily brief and touch ony on relevant points. They are not complete summaries of the frameworks they present.

2.1.1. Inventories of uses (Mrazek, Potebnja) Perhaps the most exhaustive treatments of case have taken the form of lists of uses. Some examples of this approach include the descriptions of the Russian instrumental proposed by Mrazek (1964) and Potebnja ([1958]), as well as the inventories commonly encountered in grammars. Aside from thoroughness and attention to detail, however, this method does not have much to recommend it. It involves little more than the compilation of data. Such list-writing is superlatively atomistic and is generally frowned upon because "no synthesis follows the classification. The method consists, in fact, of splitting up into several parts what is a complex linguistic unit" (Burston 1977: 30). Thus the integrity and internal order of the semantic category are overlooked. Commentaries on the common origins or relatedness of meanings are often sprinkled in on an ad hoc (and inconsistent) basis in order to fill out the description and to reduce its choppiness, but these constitute improvements in presentation rather than alterations in theoretical framework. As mentioned in the previous chapter, the network structure of a cognitive description facilitates captioning of the unity and infrastructure of a category without

26

sacrificing any of the detail sought in the list model. It also solves one of the problems inherent in the list model, that of answering the question, how much detail is enough? Depending upon how rigorously it is applied, the list model will produce any number of submeanings. This possibility for variation is problematic for an approach the express purposes of which are to discover how many submeanings are present and to describe them. Such variation is easily incorporated in the cognitive model, since it strives instead to draw a map of semantic structure, which can be drawn to whatever scale desired, thus allowing for greater or lesser detail.

2.1.2. Semantic features (Hjelmslev, Jakobson, van Schooneveld)

Hjelmslev (1935-37) theorized that each case has a single abstract meaning, derived from a matrix of semantic features. His work inspired Jakobson's (1936 [1971]) "Beitrag zur allgemeinen Kasuslehre", which has been a subject of vigorous debate ever since. Perhaps the most ambitious author in this regard is van Schooneveld (1978), who sought a single grand semantic pattern for all morphemes that serve to structure Russian sentences. The fact that Burston (1977) succeeded in applying the very same system of features which van Schooneveld worked out for Russian to a language with a very different case system (Pali) is enough to suggest that something is amiss. It appears to be too easy to twist universal invariants around to fit the data and, conversely, this sort of approach fails to account for variation among languages. This approach has been roundly criticized, for both excessive vagueness and excessive power, and also for lack of verifiability.U Both Skalicka (1950: 135) and Isacenko (1965: 95) find this sort of description elegant, but far too abstract to be of practical value. Isacenko asks how a general meaning such as the one Jakobson assigned to the dative - "expresses the existence of the object as independent of the action" - could possibly account for meanings such as those expressed by dative subjects in expressions like mne nado "I need (lit: needed to me)" and mne xocetsja "I feel like (lit: wants-refl to me)" .32 Wierzbicka (1980: xv-xvi) sums up the major objections: "formulas like '[+peripheral, -affected]' are not self-explanatory and one could stretch them in many different ways to make them fit the facts; these formulas have limited predictive power [and are] too vague, too general to be empirically adequate." Even scholars who use Jakobson's system often criticize it and find themselves forced to hedge on some of its basic assumptions. Plewes (1977: ,263264) comments in a footnote that "although we have implicitly assumed in the

ny of the detail sought in the list model. It also solves one of the erent in the list model, that of answering the question, how much gh? Depending upon how rigorously it is applied, the list model will number of submeanings. This possibility for variation is problematic ach the express purposes of which are to discover how many subpresent and to describe them. Such variation is easily incorporated ve model, since it strives instead to draw a map of semantic strucan be drawn to whatever scale desired, thus allowing for greater or

27

discussion of feature-marking theory that semantic features like [+direction] or [+peripheral] are either present or absent (i.e., they are digital ), it is more likely that such feature-marking is analog (i.e., the features can be present in varying degrees or strengths)". There are indications, however, that Jakobson intentionally left room for such hedges, and that the apparent immutability of his system is more an artifact of the way his followers interpreted it than of how he actually conceived it.33 Necessary compromises to this approach are also predicted by the cognitive model, for when absolute invariant semantic features were posited, they created monolithic case categories with no infrastucture. Jakobson's concept of relative invariance brings his system very close to the sort of description required by cognitive semantics, and will be discussed toward the end of this chapter.

ntic features (Hjelmslev, Jakobson, van Schooneveld)

1935-37) theorized that each case has a single abstract meaning, a matrix of semantic features. His work inspired Jakobson's (1936 trag zur allgemeinen Kasuslehre", which has been a subject of vie ever since. Perhaps the most ambitious author in this regard is van (1978), who sought a single grand semantic pattern for all morserve to structure Russian sentences. The fact that Burston (1977) applying the very same system of features which van Schooneveld or Russian to a language with a very different case system (Pali) is uggest that something is amiss. It appears to be too easy to twist ariants around to fit the data and, conversely, this sort of approach nt for variation among languages. oach has been roundly criticized, for both excessive vagueness and wer, and also for lack of verifiability.U Both Skalicka (1950: 135) (1965: 95) find this sort of description elegant, but far too abstract tical value. Isacenko asks how a general meaning such as the one igned to the dative - "expresses the existence of the object as indehe action" - could possibly account for meanings such as those dative subjects in expressions like mne nado "I need (lit: needed to e xocetsja "I feel like (lit: wants-refl to me)" .32 Wierzbicka (1980: up the major objections: "formulas like '[+peripheral, -affected]' are anatory and one could stretch them in many different ways to make facts; these formulas have limited predictive power [and are] too neral to be empirically adequate." olars who use Jakobson's system often criticize it and find orced to hedge on some of its basic assumptions. Plewes (1977: ,263nts in a footnote that "although we have implicitly assumed in the

2.1.3. Semantic metalanguage (Wierzbicka)

Wierzbicka clearly recognized the drawbacks of the preceding two models: that lists were too atomistic and features too abstract. She attempted to build a fresh model to eliminate these problems34 by implementing a semantic metalanguage in her description. This would eliminate the need for features, and the expressions set in metalanguage would be self-explanatory and specific to given submeanings, rather than overly vague. Also, similarities in definitions of submeanings would point to their interrelatedness, reducing the atomistic appearance of the list of submeanings. She applied her method to the Russian instrumental, producing metalanguage defmitions of seventeen submeanings. This was a bold attempt which incorporated several of the goals of a cognitive description: it strove to reconcile the diversity of meanings with their overall unity and to demonstrate their relatedness. Hints to the structure of the category of the instrumental, however, remain diffusely scattered through the metalinguistic definitions. Relationships among submeanings are implied, but not made explicit. Still, Wierzbicka's monograph is quite significant, for it presents a very clear statement of the problems involved and provides a useful tool for the semantic description of case categories.

2.1.4. Semantic/syntactic dichotomy (Kurylowicz, Isacenko)

Another approach that has had many followers suggests that case is really two distinct phenomena which should be treated separately: semantic case and syntactic case. This distinction is clearly derived from the observation that

28

grammatical meaning differs from lexical meaning, discussed as "the third obstacle to a description of case semantics" in the preceding chapter. Kurylowicz (1960) outlined this theory,35 and even though he did not subject it to rigorous verification, it is evident that he found the syntactic and semantic components more difficult to disentangle than one might presume. Each case was stated to be either primarily syntactic (grammatical) or semantic (concrete) and to have both primary and secondary functions based on this very same distinction. In addition, the concrete cases were said to be progressing toward grammaticalization, whereas the grammatical cases were moving in the opposite direction, toward adverbialization (lexicalization). And it was claimed that concrete case occupies an intermediate position between adverbs (lexical meaning) and purely syntactic forms. Thus, even though the theoretical purpose of Kurylowicz's approach is to identify a bifurcation of semantic and syntactic functions, we observe a tendency to emphasize the relationships between these two functions as poles of a spectrum of case values rather than as a concerted effort to draw boundaries between them. 36 Kurylowicz did not, as Apresjan (1964: 33) points out, give an operational definition for distinguishing the two types of case. In spite of the lack of resolution in the supposed semantic/syntactic interface, a number of linguists have followed Kurytowicz's lead. Isacenko (1965) applied this approach to Russian, and identified many of the primary and secondary functions of the cases of that language. His work is, however, hardly an exhaustive treatment of Russian according to this model, and also contains many insights not derivable from this distinction. DeGroot (1965) makes use of the same distinction and gives an equally generalized account. Deserieva (1974) tallies up all of the semantic and syntactic uses of each case in Russian and then puts these values in a formula by means of which she arrives at figures describing the abstractness of each case. The nominative is, according to her, 100% syntactic and therefore very abstract, whereas the dative, instrumental and locative are quite concrete, as predicted by Kurylowicz. Although Deserieva is the only scholar to rigorously implement Kurylowicz's classification, it remains just that - a classification of uses into two broad categories, for she does not contribute anything new to the theoretical framework. It is interesting to note, however, that when she has identified a particular use as semantic or syntactic, even this designation is not necessarily absolute. Seven of the syntactic uses of the genitive are "partly semantic" and she admits that there are crossovers between the two categories. Perhaps the most recent revival of the syntactic/semantic dichotomy is to be found in Babby (1986), who uses it to show that Chomsky's view of case is grossly unbalanced, ignoring its semantic uses.

meaning differs from lexical meaning, discussed as "the third description of case semantics" in the preceding chapter. cz (1960) outlined this theory,35 and even though he did not subject verification, it is evident that he found the syntactic and semantic more difficult to disentangle than one might presume. Each case was ither primarily syntactic (grammatical) or semantic (concrete) and to mary and secondary functions based on this very same distinction. the concrete cases were said to be progressing toward grammatihereas the grammatical cases were moving in the opposite direction, bialization (lexicalization). And it was claimed that concrete case ntermediate position between adverbs (lexical meaning) and purely ms. Thus, even though the theoretical purpose of Kurylowicz's to identify a bifurcation of semantic and syntactic functions, we dency to emphasize the relationships between these two functions as pectrum of case values rather than as a concerted effort to draw etween them. 36 Kurylowicz did not, as Apresjan (1964: 33) points perational definition for distinguishing the two types of case. the lack of resolution in the supposed semantic/syntactic interface, a nguists have followed Kurytowicz's lead. Isacenko (1965) applied to Russian, and identified many of the primary and secondary funccases of that language. His work is, however, hardly an exhaustive Russian according to this model, and also contains many insights not m this distinction. DeGroot (1965) makes use of the same distinction equally generalized account. Deserieva (1974) tallies up all of the syntactic uses of each case in Russian and then puts these values in a means of which she arrives at figures describing the abstractness of e nominative is, according to her, 100% syntactic and therefore very ereas the dative, instrumental and locative are quite concrete, as Kurylowicz. Although Deserieva is the only scholar to rigorously Kurylowicz's classification, it remains just that - a classification of broad categories, for she does not contribute anything new to the amework. It is interesting to note, however, that when she has identilar use as semantic or syntactic, even this designation is not necessaSeven of the syntactic uses of the genitive are "partly semantic" and at there are crossovers between the two categories. Perhaps the most of the syntactic/semantic dichotomy is to be found in Babby (1986), o show that Chomsky's view of case is grossly unbalanced, ignoring uses.

29

There have been a number of attempts to isolate the syntactic function of case, which inevitably disintegrate into unwieldy and unenlightening inventories. Case uses that are essentially identical are assigned to different categories because of minor details of syntactic structure and conversely uses that are significantly different are lumped together whenever they show identical structure. This is true of the accounts found in Worth 1958 and in the 1970 and 1982 Academy grammars (Svedova et al. 1970 and 1982). It must be noted that both Academy grammars resorted to semantic captions for syntactic categories, again suggesting that the two functions are invariably intertwined. Talmy's work, as cited in section 1.2. of this monograph, shows that the assumption underlying this type of approach - that grammatical meaning is different from lexical meaning - is both true and false. Those who subscribe to the semantic/syntactic dichotomy are correct in seeing that the two kinds of meaning are not identical, but incorrect in pushing this distinction too far. Their meanings can be accessed by the same methods, although they tend to focus on different aspects of the semantic structure of sentences. It is the failure to recognize this subtle yet crucial fact that has permitted the development of this approach. Kurytowicz's deliberate omission of specific boundaries amounts to a hedge on his basic theoretical outlook, and such hedging is predicted by the cognitive model. There is one other, yet related, way in which the semantic/syntactic dichotomy approach is incompatible with the cognitive model. As Oliverius (1972: 104), points out, Kurylowicz's model must be rejected because it requires one to accept the existence of empty morphemes. Kurytowicz implies that the syntactic uses have no meaning, a corollary to his theory which is also derivable from his failure to recognize the difference between grammatical and lexical meaning. Grammatical meaning is different, but not non-existent. As noted above in section 1.2.3., the basic principles of the cognitive model suggest that a single morpheme cannot be meaningful in some contexts but devoid of meaning in others.

2.1.5. Case as a purely syntactic phenomenon (Chomsky) Chomsky is mentioned here not so much because he has made any special contribution to our understanding of case - in fact, he has said remarkably little about case 37 - but rather because his theory has been so influential that it must be reckoned with. Chomsky's treatment'f of case is restricted almost entirely to discussions of syntax; indeed, a semantic analysis of case is not one of his aims at all. 39 Jakobson (1971: 494) notes that "Chomsky has made an ingenious attempt to construct a 'completely non-semantic theory of grammatical structure' " and

30 wonders whether he is merely playing devil's advocate, since the very idea strikes Jakobson as preposterous. 40 According to Starosta (1976: 1), "Chomsky's configurational definition of grammatical functions is incorrect as a representation of surface grammatical relations." Starosta tries to develop a new notation which will "allow us to represent the fact that these relations have their own characteristic semantic content, a fact that a pure configurational representation is incapable of coping with" (Starosta 1976: 18). Criticisms of Chomsky's theory for its lack of a semantic component come not only from without, but, more importantly, from those who accept his framework as well. Bachman (1980: 47) uses Chomsky's transformations to test the status of the subjective dative in Russian, but finds it "more intuitively appealing to have morphological case assigned in accordance with basic semantic relations than by the application of a superficial mechanical operation called 'case marking' ". More remarkably, Grumet's (1983) dissertation, written entirely within the Chomskyan framework, proves exhaustively that it is impossible to provide an adequate account of the English genitive without making reference to semantics ("the lexicon") and even suggests that this may be true for all cases. She presents a diagram showing the various submeanings of the English genitive and how they overlap with each other, a figure which in many ways resembles a cognitive network. Babby (1987: 136) also follows Chomsky's method, but rejects his characterization of case as semantically vacuous: "the Russian data cast serious doubt on one of the central assumptions of Government and Binding case theory, namely, that case distribution is exhaustively determined by structural relations between the case assigner and assignee". From a cognitive standpoint, Chomsky's treatment of case is an extreme and truncated version of that put forth by Kurylowicz: not only is the semantic status of syntactic case denied, but the existence of other uses of case is altogether ignored. It comes as no surprise that some of Chomsky's followers have tried to put semantics back into the formula.

2.1.6. Symbolic logic (Serensen, Mel'cuk)

Another highly formalistic approach is the application of symbolic logic to case. Serensen (1957: 39) makes very generous use of formulas, but also makes it clear that he will "not in this work give any description of the semantic contents of the linguistic elements which are being treated". After a lengthy definition of terms, Mel'cuk (1986: 42-44) presents a two-page formula to account for case, apologizing for its "clumsy and involved construction", for he "did [his] best to come up

ther he is merely playing devil's advocate, since the very idea strikes preposterous. 40 According to Starosta (1976: 1), "Chomsky's confifinition of grammatical functions is incorrect as a representation of matical relations." Starosta tries to develop a new notation which will represent the fact that these relations have their own characteristic tent, a fact that a pure configurational representation is incapable of (Starosta 1976: 18). of Chomsky's theory for its lack of a semantic component come not thout, but, more importantly, from those who accept his framework man (1980: 47) uses Chomsky's transformations to test the status of dative in Russian, but finds it "more intuitively appealing to have l case assigned in accordance with basic semantic relations than by n of a superficial mechanical operation called 'case marking' ". More Grumet's (1983) dissertation, written entirely within the Chomskyan proves exhaustively that it is impossible to provide an adequate he English genitive without making reference to semantics ("the d even suggests that this may be true for all cases. She presents a wing the various submeanings of the English genitive and how they each other, a figure which in many ways resembles a cognitive bby (1987: 136) also follows Chomsky's method, but rejects his on of case as semantically vacuous: "the Russian data cast serious of the central assumptions of Government and Binding case theory, case distribution is exhaustively determined by structural relations ase assigner and assignee". gnitive standpoint, Chomsky's treatment of case is an extreme and sion of that put forth by Kurylowicz: not only is the semantic status case denied, but the existence of other uses of case is altogether mes as no surprise that some of Chomsky's followers have tried to back into the formula.

olic logic (Serensen, Mel'cuk)

y formalistic approach is the application of symbolic logic to case. 57: 39) makes very generous use of formulas, but also makes it clear not in this work give any description of the semantic contents of the ments which are being treated". After a lengthy definition of terms, 6: 42-44) presents a two-page formula to account for case, apologizumsy and involved construction", for he "did [his] best to come up

31

with something simpler and more digestible, but failed". Like Serensen, Mel'cuk is concerned primarily with the hows of case distribution rather than with the whys of case meaning.t! Whereas the use of features is a gesture toward mathematical reductionism, the implementation of symbolic logic is an example of the empiricist approach par excellence. From the comments made on this sort of framework in the preceding chapter, it should be clear why it is not suited to semantic inquiry and, consequently, why Serensen and Mel'cuk do not include case semantics in their formulas. Cognitive semantics invokes the use of conventional imagery to caption case. If verbal explanations such as features or metalanguage are one step removed from the use of such images (profiles), words being in some sense less immediate and more abbreviated than schemas, then the assignment of symbols takes us two steps away from the cognitivist's goal.

2.1.7. "Deep" case (Fillmore)

Of articles on case, Fillmore's (1968) "Case for Case" has perhaps provoked more response than any other. Still, his work, as well as that of the "case grammarians" who followed his lead, has little bearing on case as it is understood in this volume, for Fillmore's focus was entirely different. Indeed, what he describes is properly not called "case" at all.42 He describes instead a set of universal relations. The fact that he calls them "cases" has inspired many to criticize Fillmore of having "specifically rejected the importance of surface structure morphology".43 It appears, however, that Fillmore was misinterpreted. Fillmore (1977: 67) recognized the source of this confusion and issued a clarification, stating that "an account of the uses of surface-structure cases requires more than a theory of deep cases". In other words, his framework is not appropriate for examining the semantic structure of specific surface cases, although surface cases are used to express the relations which he calls "cases". In addition to the allegation that Fillmore ignores surface case, opponents point out some more substantive flaws in his theory. Bachman (1980: 186) finds Fillmore's proposals "thought-provoking and heuristically valuable, but unworkably abstract".44 A related problem is that of correlating the two levels that Fillmore invokes, and, particularly, "assigning a given surface structure noun phrase to one of the underlying cases".45 The most frequent criticism is also the most serious - that neither Fillmore nor any of his followers has succeeded in producing a definitive list of "cases". Given that one of the aims of the model is to arrive at a universal set of "cases", failure to do so is hard to overlook.46

32 Since Fillmore's relations are identified with case roles,47 his descriptions of them are relevant to our discussion. Indeed, the labels he uses for them, such as Agent and Experiencer, are in many cases appropriate for use in prose explanations of cognitive captions of case meanings, as will be seen in the remaining two chapters. Even more important is the way in which Fillmore (1977: 74) approaches meaning. According to him, grammatical meaning necessarily identifies both "figure" and "ground", as well as "a perspective on [this] scene". This statement could easily be used to describe cognitive schemas. The following two sentences (Fillmore 1977: 80) define the same phenomenon which Langacker refers to as "profiling": "We recognize scenes or situations and the functions of various participants in these scenes and situations. We foreground or bring into perspective some possibly quite small portion of such a scene." Certainly, Fillmore's perspective on the kind of grammatical meaning involved in case roles is not incompatible with that of cognitive semantics.48 Although Fillmore shares some of the concepts basic to a cognitive understanding of case meaning, his framework departs from our model in important ways. Within the cognitive framework, semantics is viewed as immediately relevant to all parts of language. There is no need, therefore, to subdivide language into various levels. Semantics is a part of language in all its aspects and does not require its own, separate leve1. 49 A corollary is the notion that cases form semantic wholes, all the parts of which are interrelated. Because he relegates semantics to the "deep" level, Fillmore treats it as if it were independent of surface morphemes, thus the various meanings of a surface case are scattered about among the "deep cases", with no indication of their overall unity.50 Although Fillmore's description of "deep cases" is intriguing, his system is no more straightforward, no less complex and no less ambiguous than the surface system; which begs the question: why would any language opt for a two-tiered system if it could make do with one level?

2. 1.8. Localist theories of case (Anderson, Freidhof) In spite of the fact that in some versions localism incorporates the theoretical concepts of Fillmore and Chomsky, it is essentially a feature system-! with a special theme - everything is considered derivable from physical location or movement. As Miller (1986: 296) puts it, "localism is the thesis that semantic structures should be constructed in terms of objects being located in a place or moving from one place to another". Although the rigid application of semantic features is irreconcilable with cognitive semantics for reasons already given

more's relations are identified with case roles,47 his descriptions of vant to our discussion. Indeed, the labels he uses for them, such as xperiencer, are in many cases appropriate for use in prose explanative captions of case meanings, as will be seen in the remaining two n more important is the way in which Fillmore (1977: 74) approachAccording to him, grammatical meaning necessarily identifies both "ground", as well as "a perspective on [this] scene". This statement be used to describe cognitive schemas. The following two sentences 77: 80) define the same phenomenon which Langacker refers to as We recognize scenes or situations and the functions of various parhese scenes and situations. We foreground or bring into perspective y quite small portion of such a scene." Certainly, Fillmore's perspecnd of grammatical meaning involved in case roles is not incompatible ognitive semantics.48 Fillmore shares some of the concepts basic to a cognitive g of case meaning, his framework departs from our model in importthin the cognitive framework, semantics is viewed as immediately l parts of language. There is no need, therefore, to subdivide lanarious levels. Semantics is a part of language in all its aspects and uire its own, separate leve1. 49 A corollary is the notion that cases c wholes, all the parts of which are interrelated. Because he relegates he "deep" level, Fillmore treats it as if it were independent of surface thus the various meanings of a surface case are scattered about deep cases", with no indication of their overall unity.50 Although escription of "deep cases" is intriguing, his system is no more rd, no less complex and no less ambiguous than the surface system; he question: why would any language opt for a two-tiered system if it o with one level?

ist theories of case (Anderson, Freidhof)

e fact that in some versions localism incorporates the theoretical Fillmore and Chomsky, it is essentially a feature system-! with a e - everything is considered derivable from physical location or As Miller (1986: 296) puts it, "localism is the thesis that semantic ould be constructed in terms of objects being located in a place or one place to another". Although the rigid application of semantic reconcilable with cognitive semantics for reasons already given

33 above, this basic premise concerning the centrality of location and movement, however abstractly conceived, is clearly consonant with the first of Langacker's (1986a) assumptions, that our perception is organized by means of the "billiardball model", The sentence quoted from Miller above could apply equally to the schemas drawn by Langacker and Smith, as well as to a localist description.V In most versions,53 however, localism produces vast systems of features and trees, adding little to an understanding of case as a coherent semantic structure.

2.2. How the cognitive model has been implied by various authors

Although none of the authors cited thus far in the present chapter used (or, perhaps, was even aware of) the cognitive model, many have made statements that cannot be derived from their own theoretical framework, but instead support the view of semantics proposed in this volume. It appears that these scholars found their framework flawed and tried to correct for perceived deficiencies. In every case, these corrections point to a structure that the given framework did not accommodate.

2.2. 1. Relatedness of meanings As was mentioned above, those who consider their main concern to be the cataloging of all the submeanings of a case frequently mention that certain of them are related to each other, even though from the standpoint of theory there is no reason for these authors to do so.23 This is particularly true of Mrazek (1964), who suggests some relationships with other meanings for nearly every meaning of the Russian instrumental which he identifies. Although he does so in an unsystematic fashion, Mrazek makes enough such statements to indicate that the submeanings are not as haphazard or independent as his method might otherwise imply. In fact, if one gathers together all of the statements which Mrazek has made about the relationships between meanings, and arranges his meanings according to them, the network in Figure 6 is obtained. It is curious that a listwriter actually authored a network, albeit implicitly.

34

Subject in Passive Construction Manner ......- -...... Predicate

Object

Group or Formation

Space

Time

Integral Part

Delimitation

Cause Other Concrete Means

PseudoComitative

Figure 6. Mrazek's network

Wierzbicka (1980) also produced a list of meanings for the Russian instrumental and claimed that they were all related to each other because they shared similar statements in the semantic metalanguage of their definitions. Wierzbicka does not make the system of relationships explicit either, but, like Mrazek, she leaves enough clues for a network to be derived. When analyzed into components, it is found that every meaning is related to others by sharing one, two, three or four components. Arranging these meanings according to their relatedness produces a considerably more complex network than that implied by Mrazek, as shown in Figure 7. Triple lines indicate that four components are shared, double lines indicate three shared components, and single lines indicate two shared components (inclusion of relationships based on only one shared component creates a network too complex to be easily accommodated in a two-dimensional drawing).55 '

35 Subject in Passive Construction Manner ......- -...... Predicate

Object

Space

Time

Integral Part

Delimitation

Cause Other Concrete Means

PseudoComitative

Figure 6. Mrazek's network

ka (1980) also produced a list of meanings for the Russian instruclaimed that they were all related to each other because they shared siments in the semantic metalanguage of their definitions. Wierzbicka ke the system of relationships explicit either, but, like Mrazek, she gh clues for a network to be derived. When analyzed into compofound that every meaning is related to others by sharing one, two, r components. Arranging these meanings according to their relatedes a considerably more complex network than that implied by Mrazek, n Figure 7. Triple lines indicate that four components are shared, s indicate three shared components, and single lines indicate two ponents (inclusion of relationships based on only one shared comes a network too complex to be easily accommodated in a two-dimenng).55 '

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

Instrument Object in Action Sentences Body Part Object in Reflexive Action Sentences Emotions Inexplicable Force Agent in Passive Sentences Matter in Action Sentences Matter in Process Sentences Comparison Manner Space Time Duration Transport Personal Characteristics Predicate

Figure 7. Wierzbicka's network

36

Although Deserieva (1974 & 1977) also fails to provide networks for cases, her discussion of them as "semantic fields" is even more suggestive of this sort of structure. She concludes (1977: 60) that the basis for any description of a grammatical category lies in the structure of the semantic field it commands, and that special attention must be paid to the semantic center of a category. It is difficult to conceive of a structure that would fit her description without meeting the specifications of the networks invoked by cognitive semantics. As mentioned in chapter 1, Schlesinger did a cross-linguistic study of the relationships that hold between the instrumental and two other relations which are frequently expressed by the same case: manner and comitative. In both instances he found that there is no hard and fast boundary between the meanings; "the instrumental and comitative are really only two extreme points on what is a conceptual continuum" (Schlesinger 1979: 308). Schlesinger extrapolates from his findings to infer that a category is composed of a system of continua. This is significant because cognitive semantics makes the same claim: nodes in a network are places where instantiations tend to cluster, but this does not mean that the semantic space between them is empty; on the contrary, there usually are transitional examples.

2.2.2. Hierarchical relations among meanings It is customary among both catalogers and feature-specifiers 56 to speak of "basic" meanings for cases, but this device reaches its apex in the works of Roman Jakobson. Jakobson (1936 [1971]: 35) made several distinctions relevant to the status of case meanings. In addition to the general meaning (that represented by features), he recognized both a principal meaning (comparable to the prototype of cognitive semantics) and particular meanings (the constellation of various submeanings and their instantiations). He noted that "the particular meanings ... are not a mechanical accumulation, but form rather a regular hierarchy of particular meanings" grouped about the principal meaning. This hierarchy again implies that case has a center-periphery semantic structure, precisely that which cognitive semantics proposes. Jakobson's features are also found to be apt in describing the schemas that will be presented in the analysis for the dative and instrumental, and are particularly relevant to prototype schemas. As such, Jakobson's feature specifications can largely be retained as an umbrella system for intercase organization, since doing so would not compromise the cognitive model in any way. It is interesting to note that Jakobson held the lexical environment to be responsible for generating the particular meanings (also known as "combinatory variants" for this reason) from the general one. In the cognitive model proposed

Deserieva (1974 & 1977) also fails to provide networks for cases, on of them as "semantic fields" is even more suggestive of this sort of he concludes (1977: 60) that the basis for any description of a category lies in the structure of the semantic field it commands, and attention must be paid to the semantic center of a category. It is onceive of a structure that would fit her description without meeting tions of the networks invoked by cognitive semantics. oned in chapter 1, Schlesinger did a cross-linguistic study of the relat hold between the instrumental and two other relations which are freessed by the same case: manner and comitative. In both instances he here is no hard and fast boundary between the meanings; "the instruomitative are really only two extreme points on what is a conceptual (Schlesinger 1979: 308). Schlesinger extrapolates from his findings a category is composed of a system of continua. This is significant nitive semantics makes the same claim: nodes in a network are places tiations tend to cluster, but this does not mean that the semantic space m is empty; on the contrary, there usually are transitional examples.

archical relations among meanings

ry among both catalogers and feature-specifiers 56 to speak of "basic" or cases, but this device reaches its apex in the works of Roman kobson (1936 [1971]: 35) made several distinctions relevant to the e meanings. In addition to the general meaning (that represented by recognized both a principal meaning (comparable to the prototype of mantics) and particular meanings (the constellation of various subd their instantiations). He noted that "the particular meanings ... are nical accumulation, but form rather a regular hierarchy of particular rouped about the principal meaning. This hierarchy again implies that enter-periphery semantic structure, precisely that which cognitive oposes. Jakobson's features are also found to be apt in describing the will be presented in the analysis for the dative and instrumental, and rly relevant to prototype schemas. As such, Jakobson's feature specilargely be retained as an umbrella system for intercase organization, o would not compromise the cognitive model in any way. esting to note that Jakobson held the lexical environment to be or generating the particular meanings (also known as "combinatory this reason) from the general one. In the cognitive model proposed

37

in chapters 3 and 4 the difference between some submeanings will also be attributed to choice of lexical items and to extralinguistic knowledge. Pragmatics plays an elemental role in specifying the subjective content of an expression, whose outlines have been drawn by case.

2.2.3. Relational invariance On the face of it, the notion of combinatory variants based on general meanings proposed in Jakobson 1936 [1971] seems irreconcilable with the invariant semantic features introduced in Jakobson 1958 [1971]. How can variants and invariants coexist in a single model? This contradiction in terms is only apparent, for it belies a subtler concept grasped by Jakobson, but lacking logical means for expression in the basic theoretical framework that he used. The invariant features display themselves to varying degrees in actual instantiations, a phenomenon that Jakobson referred to as relational invariance, the seeming oxymoron hailed by Sangster (1982: 78) as "probably the most powerful construct in Jakobson's linguistic arsenal". It is this hedging on invariance that allows Jakobson to posit features without making them so immutable and the categories they define so monolithic that the hierarchical structure which he previously proposed would be inconceivable. The concept of relational invariance made it possible for Jakobson to invoke a unifying force for a category without denying the existence of an array of submeanings. 57 It is, however, a clumsy term which is not integrated in an obvious way into the theoretical framework in which it appears.58 Cognitive semantics can produce the same result - accommodating both unity and diversity - without compromising its theoretical premises, by reference to a central, unifying prototype, and a surrounding network of related meanings.

2.3. Advantages ofthe cognitive approach

Various reasons for recommending the cognitive model of semantics have been mentioned here and there up until now; it would be useful at this point to gather them together for purposes of reference.

38

2.3.1. A departure from empiricism The very fact that the study of case semantics seems to have reached a dead end (indicated in the Introduction) suggests that a new approach may be needed. In the past half-century, virtually all of semantics has been couched in the settheoretical framework of logical empiricism. Cognitive semantics seeks centerperiphery structure rather than set boundaries, and thus applies a theoretical framework that has proved useful in psychology, neurology, and artificial intelligence.

·2.3.2. Both diversity and unity of case meaning are accommodated Since the structure of the semantic network of a case can be drawn as grossly or scrupulously as desired, questions of both overall coherence and differentiation among submeanings can be addressed with equal facility. The set-theory approach of semantic features and exhaustive inventories always requires that one of these aspects be emphasized at the expense of the other.

2.3.3. Many of the best insights of various scholars are incorporated Since the relevant items have been covered in previous sections, they will merely be summarized here.

- A description of case should be founded on semantics - Cases have many meanings - Case meanings are interrelated - Differentiation between case meanings is continuous rather than discrete - Case meanings are unified and arranged in a hierarchical structure rather than being haphazard - Case meanings can be explained in terms of location and movement of physical objects in a scene

(Jakobson, Starosta, Grumet, Babby) (Mrazek, Wierzbicka, Kurytowicz, Jakobson) (Mrazek, Wierzbicka) (Schlesinger) (Jakobson)

(Anderson, Miller, Fillmore)

39

parture from empiricism

ct that the study of case semantics seems to have reached a dead end n the Introduction) suggests that a new approach may be needed. In f-century, virtually all of semantics has been couched in the setramework of logical empiricism. Cognitive semantics seeks centerructure rather than set boundaries, and thus applies a theoretical that has proved useful in psychology, neurology, and artificial

- In some instances meanings are distinguished by lexical environment and extralinguistic knowledge - Case meaning is never absent, even when "governed" by verbs and prepositions

(Jakobson, Potebnja, Pit'ha, Levine) (Oliverius, Plewes, Gladney, van Schooneveld)

2.4. An ideal description ofcase: Skalicka's challenge

diversity and unity of case meaning are accommodated

In 1950, Skalicka issued a challenge that has not yet been met. He outlined the following criteria which an ideal account of case should fulfill. It should be able to:

ucture of the semantic network of a case can be drawn as grossly or y as desired, questions of both overall coherence and differentiation meanings can be addressed with equal facility. The set-theory semantic features and exhaustive inventories always requires that one ects be emphasized at the expense of the other.

c)

y of the best insights of various scholars are incorporated

I would like to argue that cognitive semantics is uniquely adapted to fulfilling these criteria.

evant items have been covered in previous sections, they will merely ed here.

tion of case should be founded tics ve many meanings

nings are interrelated ation between case meanings ous rather than discrete nings are unified and n a hierarchical structure n being haphazard nings can be explained in ocation and movement of objects in a scene

(Jakobson, Starosta, Grumet, Babby) (Mrazek, Wierzbicka, Kurytowicz, Jakobson) (Mrazek, Wierzbicka) (Schlesinger) (Jakobson)

(Anderson, Miller, Fillmore)

a) b)

(i)

describe the relation of case to reality describe the meaning of case, such that the description is both complete and apt for practical use describe the relations between case and other forms.

The idealized cognitive models captioned in schemas proposed by Smith and Langacker provide the the best approximation to a link between reality and description via reference to human perception of reality.59Cognitive semantics also recognizes the hallmarks of grammatical meaning as well as their essential identity with lexical meaning, which has more obvious realworld correlates. (ii) Cognitive semantics endeavors to embrace all meanings of a case in a single network. The schemas have an immediacy that features, lists, and symbols lack, and that makes both their meanings and their relationships to each other largely self-explanatory. The simplicity and elegance of the system give it great potential for practical applications such as language teaching. (iii) Most descriptions of case either ignore 60 governed case altogether or relegate it to a completely separate category/'! The cognitive model achieves a unified description that includes the relations of case to prepositions and verbs. Relations to other parts of the sentence (such as the subject and direct object, see the analysis in part two) are also captured in the schemas, and the focus on the internal structure or the category facilitates distinction be-

40

tween its semantic contribution and that of prepositions, verbs, and other lexical units.

its semantic contribution and that of prepositions, verbs, and other units.

Part II -

Analysis

3.

Why the Czech dative and the Russian instrumental?

Ceskjjazyk tvoiil cely ceskj ndrod v celt historii, skrze jazyk si poddvaji ruce vsechna pokolenf, v jazyce se obriiii vyvoj naseho ndroda od samycli zacatkli at do dneska, je to dilo kolektivni a tradicni. " It

[All ofthe Czech people throughout history have

worked on the creation ofthe Czech language. Through languageall generationsjoin hands, in language we see the development ofour nation from the very beginning right up to the present day. It is a collective, traditionalcreation.] Frantisek Trdvnicek

If cognitive semantics is to be challenged by the facts of natural language, it would of course be best to assign it as intricate a problem as possible. The dative and instrumental may display the lowest frequency,62 but they are semantically more complex than any other cases. Wierzbicka (1980: 1) comments that "the Russian instrumental is particularly well-known for its mind-boggling qualities", and indeed, several entire monographs have been devoted to it.63 The Czech dative has received somewhat less attention in scholarly literature. 64 Its inclusion here is motivated by a subjective observation that the dative appears to be somewhat more complex in Czech than in Russian,65 and by the desire to use material from more than one language. In addition, it will be observed that the reflexive dative pronoun si produces interesting deformations of the Czech dative network.

3.0.1. Schemas in pictorial and prose notation Figure 8 presents the elements of pictorial and prose notation that will be used in the composition of schemas to caption submeanings throughout the analysis. Although each system of notation is independently adequate for description of the

44

cognitive categories of case, they have complementary strengths. For the sake of interpreting individual schemas, the prose notation'v has an immediacy which the pictorial schemas lack, yet the pictorial schemas bring into relief the relationships that hold among schemas in a network in a way not possible in prose. Combined, these two systems of schemas resolve the dilemma of how to represent with equal force the two seemingly antithetical natures of cognitive categories - the diversity of individual members vs. the overall unity of the category. The salience of individual meanings provided by prose notation is balanced by the salience of systematic relationships of members to the prototype visible in the pictorial notation. Yet these two notations stand in a one-to-one mapping relationship to each other. Thus they are perfectly joined and can be said to be but two expressions (verbal and graphic) of a single system. The elements of pictorial and prose notation as they appear in Figure 8 and their interpretation are explained as follows: A.

B.

C.

D.

Labeled circles (an arbitrary, but neutral shape) represent case-marked noun phrases. They will be represented by their case names in prose notation, i.e., "a nominative", "an accusative", etc. A double arrow represents the expression of a non-copular verb (including constructions containing the existential be with a predicate adjective or participle, which are functionally indistinguishable from non-copular stative verbs, cf. section 5.1.1.). So that prose notation will yield grammatical English statements, it must be sensitive to the valence of the verb. Thus if the verb has both a subject and an object, it will be rendered "acts on", as in "a nominative acts on an accusative". If the verb has only a subject it will be rendered "acts", as in "a nominative acts". If the verb has only an object, it will be rendered "an action takes place on", as in "an action takes place on an accusative". Lastly, if a verb lacks both subject and object, it will be rendered merely "an action takes place". A single arrow indicates the movement of an accusative object to a dative indirect object. This element occurs only in schemas describing indirect objects. A single arrow is translated as "to bring to" in prose schemas, as in "a nominative acts on an accusative to bring it to a dative" - the prose schema of the indirect object construction. A dotted arrow indicates that the element that it marks has potential for relative independence from the action chain. This relative independence is defined disjunctively as potential agenthood and/or inviolate integrity. The element thus marked either has the ability to react to the given situation, becoming an agent of a further action; or resists actual contact or penetra-

egories of case, they have complementary strengths. For the sake of ndividual schemas, the prose notation'v has an immediacy which the mas lack, yet the pictorial schemas bring into relief the relationships ong schemas in a network in a way not possible in prose. Combined, tems of schemas resolve the dilemma of how to represent with equal seemingly antithetical natures of cognitive categories - the diversity members vs. the overall unity of the category. The salience of indings provided by prose notation is balanced by the salience of systenships of members to the prototype visible in the pictorial notation. o notations stand in a one-to-one mapping relationship to each other. e perfectly joined and can be said to be but two expressions (verbal of a single system. nts of pictorial and prose notation as they appear in Figure 8 and ation are explained as follows:

ed circles (an arbitrary, but neutral shape) represent case-marked phrases. They will be represented by their case names in prose nota.e., "a nominative", "an accusative", etc. ble arrow represents the expression of a non-copular verb (including uctions containing the existential be with a predicate adjective or par, which are functionally indistinguishable from non-copular stative cf. section 5.1.1.). So that prose notation will yield grammatical sh statements, it must be sensitive to the valence of the verb. Thus if rb has both a subject and an object, it will be rendered "acts on", as nominative acts on an accusative". If the verb has only a subject it e rendered "acts", as in "a nominative acts". If the verb has only an , it will be rendered "an action takes place on", as in "an action takes on an accusative". Lastly, if a verb lacks both subject and object, it e rendered merely "an action takes place". gle arrow indicates the movement of an accusative object to a dative ct object. This element occurs only in schemas describing indirect s. A single arrow is translated as "to bring to" in prose schemas, as nominative acts on an accusative to bring it to a dative" - the prose ma of the indirect object construction. ted arrow indicates that the element that it marks has potential for e independence from the action chain. This relative independence is d disjunctively as potential agenthood and/or inviolate integrity. The nt thus marked either has the ability to react to the given situation, ming an agent of a further action; or resists actual contact or penetra-

45

E.

F.

G.

H.

I.

1.

K.

tion. This is a typical characteristic of datives and has the prose translation "retains independent status". A line encircling the action chain and labeled "setting" represents the setting of a given clause, likewise called "a setting" in prose notation. Prose schemas describing an action chain contained in a setting will include the phrase "in a setting". Viewpoints external to the setting will be labeled with capital letters and signalled in prose by phrases such as "viewed from A". A diagram of a dative in an ellipse with single arrows radiating from it to the periphery of the ellipse represents what is known in prose as "a dative's sphere of control". The ellipse symbolizes a domain that in some sense belongs to the.dative (variously realized in context as the body, possessions, feelings, speech-act domain, etc. of the dative referent) and the arrows symbolize the dative's control over this domain. Any infringement on this domain or any action pertaining to elements located within it is indirectly experienced by the dative. A double arrow that represents verbal action and proceeds through a circle representing a case-marked (invariably instrumental) noun phrase indicates that the action takes place "via" the instrumental. The instrumental in this case acts as a conduit or vehicle for the action. As will be shown in the analysis below, context determines the precise interpretation of the instrumental as an instrument of the action, a manner of performance of the action, a path guiding the action, or as an agent of a passive action. An instrumental attached to another case-marked element indicates that the latter element is "seen as" or construed as an instrumental. The instrumental acts as an attribute - quantitative, qualitative or comparative - of the other element. Again, context disambiguates the role of the instrumental attribute, and this will be indicated in the analysis. A double line indicates the presence of a copular verb and will be translated as "is". A single line indicates a bond between two elements, translated as "together with" and interpretable as a comitative relationship. If two elements are located in close proximity to each other, but not related by any of the symbols described above, the term "close to" will appear in prose notation. This term may be negated as "not close to" if two elements are non-proximate.

46 Pictorial Schemas A.

Prose Schemas

0;0;0;0 nom

ace

dat

"a nominative" ; "an accusative" ; "a dative" ; "an instrumental"

inst

B. "acts on" ; "acts" ; "an action takes place (on)" interpretable as a (non-copular) verb

> C.

"to bring to"

• D:

- -

-

"retains independent status" interpretable as any of the following: potential agency, resistance to integration in action chain, resistance to contact/penetration

~

\.

E.

F.

A

Oetting

@

G.

"(in) a setting" "viewed from A"

"a dative sphere of control"

e

"an action takes place via an instrumental"

:>

mst

H.

Q)

>

"a nominative seen as an instrumental acts"

nom inst

I.

0=0 nom

J.

"a nominative is a nominative'" interpretable as a copular verb

nom'

ginSI

"a nominative together with an instrumental" interpetable as a comitative relationship

nom

K

0 0

"a nominative close to an instrumental" interpretable as proximity

Figure 8. Elements of the schemas

47

hemas

Prose Schemas

0;0;0

ce

dat

inst

"acts on" ; "acts" ; "an action takes place (on)" interpretable as a (non-copular) verb

>

"to bring to"

•

- -

"retains independent status" interpretable as any of the following: potential agency, resistance to integration in action chain, resistance to contact/penetration

~

A

etting

@

"a nominative" ; "an accusative" ; "a dative" ; "an instrumental"

"(in) a setting" "viewed from A"

In addition to the terms given above, prose notation contains several more items, including the self-explanatory words "horizontally", "vertically", and the conjunction "and". If one element in a pictorial schema has two labels, such as "nom/dat" or "setting/inst", this relationship of identity is captioned with the words "be" or "be one" in prose notation, as in "the nominative and the dative are one", "the setting is an instrumental". The basic network of a case category consists of its prototype and paradigmatic variants of the prototype. The prototype is labeled schema 1 and all paradigmatic variants are labeled with consecutive numerals (thus they are known as schema 2, schema 3, etc.). The prototype and each of the paradigmatic variants have a local subnetwork comprised of syntagmatic variants. Syntagmatic variants are labeled with the numeral of the paradigmatic variant followed by a letter (thus syntagmatic variants of schema 2 are labeled schema 2a, schema 2b, etc.). The presence of semantic extensions is mentioned in the text where appropriate, but not included in the schemas, since they represent variations not in the shape, but in the application of the schemas.

"a dative sphere of control"

3.0.2. Indirect object serves as the prototype of the Czech dative

e

"an action takes place via an instrumental"

:>

mst

)

>

=0

"a nominative seen as an instrumental acts"

"a nominative is a nominative'" interpretable as a copular verb

nom'

nSI

"a nominative together with an instrumental" interpetable as a comitative relationship

It will be argued here that the central meaning (also known as the prototype) of the dative rests in the indirect object construction. This claim may seem facile and obvious 67 , but it conceals a cognitive structure, the intricacy of which becomes apparent as soon as one tries to defme the term "indirect object". The term represents a skewed concept that is positively identified only in very limited circumstances. An indirect object sensu stricto can be said to exist only in the presence of the verb ddt "give" and its closest synonyms; otherwise the import of the dative quickly fades into other meanings. Observe the tension between the indirect object and other meanings (possessive, beneficiary) in the following examples, all of which have the same basic nominative-dative-accusative case structure:

(6)

nom

mu

dala

him-DAT

gave flower-Ace

kytku.

'Ludmila gave him a flower.'

"a nominative close to an instrumental" interpretable as proximity

(7)

Figure 8. Elements of the schemas

Ludmila

Ludmila-NOM

Ludmila

mu

dala

Ludmila-NOM

him-DAT

gave

pusinku kiss-Ace

na on

48

celo. forehead-ACC 'Ludmilakissed his forehead.'

(8)

Ludmila mu koupila Ludmila-NOM him-DAT bought 'Ludmilabought him a hat.'

klobouk. hat-Ace

(9)

Ludmila mu uvarila Ludmila-NOM him-DAT cooked 'Ludmilacooked kashafor him.'

kasi. kasha-ACC

(10)

Ludmila mu ukazala Ludmila-NOM him-DAT showed 'Ludmilashowed him the way home.'

cestu way-ACC

(11)

UZ jsem mu poslala Already am-AUX him-DAT sent doporucenf do Prahy. recommendation-Ace to Prague-GEN 'I already sent the recommendation to him in Praguelfor him to Prague.'

(12)

Ludmila mu rekla svojejmeno. Ludmila-NOM him-DAT told her name-Ace 'Ludmilatold him her name.'

domu. home

While (6) represents a canonical indirect object, (7) shows a drift toward expressing possession and (8), (9) and (10) present a trend in which the meaning of indirect object graduallygives way to a benefactive reading and the dative referent is less and less likely to be interpreted as a recipient. (11) can be read either way, depending upon whetherthe referent of mu is in Prague or not and whether he is the subjector the addressee of the letter. (12) is a transitional example which shows the ties that exist between the indirect object and an important group of dative-governing verbs that express communication. The fact that manyof these verysame verbscan appearin similarconstructions withoutthe accusative objectfurther demonstrates that an operative definition that would allow unequivocal identification of all indirect objects as opposed to other dative constructions simply cannot be formulated. Compare (13), (14) and ,(15) with (9), (11) and (12),respectively:

49

(13)

Ludmila mu vzdycky Ludmila-NOM him-DAT always 'Ludmila always cookedwell for him.'

dobre varila. well cooked

(14)

Uz jsem mu am-AUX him-DAT Already 'I have already written to him.'

napsala. written

(15)

Ludmila Ludmila-NOM

ead-ACC milakissed his forehead.'

mila mu koupila ila-NOM him-DAT bought milabought him a hat.'

mila mu uvarila ila-NOM him-DAT cooked milacooked kashafor him.'

mila mu ukazala ila-NOM him-DAT showed milashowed him the way home.'

klobouk. hat-Ace

kasi. kasha-ACC

mu

fekla, ze

him-DAT

told

that

chce wants

jft to-go

domu

cestu way-ACC

domu. home

jsem mu poslala dy am-AUX him-DAT sent ucenf do Prahy. mmendation-Ace to Prague-GEN ady sent the recommendation to him in Praguelfor him to Prague.'

ila mu rekla svojejmeno. ila-NOM him-DAT told her name-Ace milatold him her name.'

represents a canonical indirect object, (7) shows a drift toward ossession and (8), (9) and (10) present a trend in which the meaning bject graduallygives way to a benefactive reading and the dative reand less likely to be interpreted as a recipient. (11) can be read either ng upon whetherthe referent of mu is in Prague or not and whether ector the addressee of the letter. (12) is a transitional example which es that exist between the indirect object and an important group of ing verbs that express communication. hat manyof these verysame verbscan appearin similarconstructions ccusative objectfurther demonstrates that an operative definition that unequivocal identification of all indirect objects as opposed to other uctions simply cannot be formulated. Compare (13), (14) and ,(15) and (12),respectively:

home 'Ludmilatold him that she wanted to go home.'

Given these data, we can conclude that the purest kind of indirect object occupies just one narrow corner of the cognitive network of the dative, which elsewherespreads out to display a plethoraof meanings. If the indirectobject has such a limitedand marginal status, how can we assertthat it is central to the dative as a whole? The indirect object may not cooccur with a large number of verbs, but the frequency of these verbs is unusually high and the indirect object fulfills other requirements for recognition as the center of gravity in the cognitive network of the dative. It should be noted that the scene invokedby the verb meaning "give" is very basic in the experience of human beings. We encounterthis situation (and therefore potential "pure"indirect objects) manytimesevery day. It is a basictenet of cognitive semantics that idealized cognitive models (IeMs) are grounded in human experience. 68 Furthermore, Lakoff (1986: 4) assertsthat a "central claim" of cognitive semantics is that "meaningful conceptual structures arise from two sources: 1. from the structured nature of bodilyand socialexperience and 2. from our innatecapacity to imaginatively projectfrom certain well-structured aspects of bodily and interactional experience to abstract conceptual structures." Clearly 1. describes the positionof the indirect object and 2. accounts for the derivation of related meaningsfrom it, as demonstrated in the examplesabove. Thus, the indirect object must be realized as conceptually central to the dative's network, even though it is very narrowly defined. In order to draw a schema for the indirect object (and, subsequently, for all other meanings of the dative), it is first necessary to arrive at a definition of the roles of the nominative and accusative, since the dative interacts with both. The

50

Prose Schema

Pictorial Schema

O~:O nom

A nominative acts on an accusative in a setting.

acc

Figure 9. Prose Schema

Pictorial Schema

o

A nominative in a setting.

nom

Figure 10. Pictorial Schema

Prose Schema

Oi=====lO nom

A nominative is a nominative' in a setting.

nom'

Figure 11.

51

Prose Schema

Pictorial Schema

O~:O

nom

A nominative acts on an accusative in a setting.

acc

Figure 9. Prose Schema

Pictorial Schema

o

schema presented in Figure 9 (a modification of Langacker 1987: 5 "normal observation of a prototypical action", which appears in Figure 1, chapter 1 of this volume) will be postulated as a definition for these two cases. A comprehensive treatment of the cognitive networks of the nominative and accusative is beyond the scope of the present work, although a few comments should be made. The schema in Figure 9 captions only the prototypes of these two cases, and these are the only parts of their cognitive structure that are relevant to our definition of the dative. The prototype of the nominative is its most interactive version. It has at least two other uses: 1. a naming function (used in lists, dictionaries, signs and in some appositive expressions) which gives a bare nominative, as in Figure 10, and 2. a predicate function, in which two nominatives are joined by a bond of equation, as in Figure 11. In addition to prepositional uses, the accusative is associated with time expressions. In some prepositional uses, the preposition is not obligatory and a direct object can be used instead as in (16b) (this example and (17) below are cited by Hajicova et at 1979: 158): (16)

a.

preskocit

potok

to-jump-over stream-ACC A nominative in a setting.

b.

nom

pfeskocit

pres

potok

to-jump-over '.

across

stream-ACC

'to jump over a stream' This suggests that such prepositional uses are actually extensions of the direct object. The accusative of time likewise shows close ties with the direct object. Compare the imperfective and perfective expressions in (17):

Figure 10.

Pictorial Schema

Prose Schema

(17)

a.

plakat celou noc

b.

proplakat

celou noc

to-cry-through

all night-ACC

to-cry all night-ACC

Oi=====lO om

A nominative is a nominative' in a setting.

'to cry all night' In the first case the verb is intransitive and the time expression would normally be identified as an accusative of time. The second verb is, however, transitive and the accusative complement is obligatory and arguably a direct object, analogous to that in (18):

nom'

Figure 11.

52

(18)

precfst

celou knihu to-read-through whole book-ACC 'to read a whole book'

3.0.3. Cognitive network of the Czech dative Figure 12 contains the schemas for the major submeanings of the Czech dative and the cognitive network they comprise. It should be kept in mind that the network is in many ways a convenient artifice. It shows only major submeanings and a skeletal structure that holds them together. A truer diagram would look more like a map of the Milky Way, for it would show areas of high density, where instantiations are thickly clustered about the major submeanings, as well as sparser areas where there are only occasional transitional examples between them. 69 For all its accuracy, such a map would be difficult both to plot and to read. The network contains a brief explanation of the paradigmatic relationship between schemas 1 and 2. It is the result of an operation on the scope of the dative, which ranges from a mere potential as an independent participant in schema 1 to the maximally external scope of the sphere of control in 2. Both schemas have local subnetworks of syntagmatic variants, which together account for all of the dative's uses. Other types of syntactic variants also exist, but since they have little relevance for either the shape of schemas or the meaning of the dative, they will not be treated separately. Such syntactic variants are to be found in participial clauses, which for our purposes will be treated no differently than the finite verb phrases from which they are derived. (19) and the participial clause in (20) will therefore be treated identically. (19)

V Ufade mu dati potvrzenf, In office-LOC him-DAT gave confirmation-Ace 'They gave him a letter of confmnation at the office.'

(20)

Precetl potvrzent, dane mu v read confirmation-Ace given him-DAT in ufade, aZ doma. office-Lee when home 'He read the letter of confirmation given to him at the office when he got home.'

53

fst

celou knihu ad-through whole book-ACC ead a whole book'

nitive network of the Czech dative

ontains the schemas for the major submeanings of the Czech dative nitive network they comprise. It should be kept in mind that the netmany ways a convenient artifice. It shows only major submeanings al structure that holds them together. A truer diagram would look more of the Milky Way, for it would show areas of high density, where inare thickly clustered about the major submeanings, as well as sparser there are only occasional transitional examples between them. 69 For cy, such a map would be difficult both to plot and to read. work contains a brief explanation of the paradigmatic relationship hemas 1 and 2. It is the result of an operation on the scope of the ch ranges from a mere potential as an independent participant in o the maximally external scope of the sphere of control in 2. Both ve local subnetworks of syntagmatic variants, which together account e dative's uses. es of syntactic variants also exist, but since they have little relevance e shape of schemas or the meaning of the dative, they will not be rately. Such syntactic variants are to be found in participial clauses, ur purposes will be treated no differently than the finite verb phrases they are derived. (19) and the participial clause in (20) will therefore entically.

Ufade mu dati potvrzenf, office-LOC him-DAT gave confirmation-Ace y gave him a letter of confmnation at the office.'

etl

potvrzent, dane mu v confirmation-Ace given him-DAT in e, aZ doma. e-Lee when home read the letter of confirmation given to him at the office when he got e.'

Schemas Pictorial

Usage

Prose

A nominative acts on an accusative to bring it to a dative. The dative retains independent status.

A nominative acts on an accusative in a dative's sphere of control in a setting.

Indirect object (syntagmatic variants account for governed dative & impersonal dative)

Free Dative

Network rR---externalization

~ofdative

..

.t;\

~

Figure 12. Czech dative network

54

Another example of a trivial sort of syntactic variant is the placement of the contents of a message in a separate clause. Thus example (15) will be treated like (12). Another factor important in shaping the network is the fact that the lexical meanings of nouns as well as extralinguistic knowledge can be used to disambiguate meanings of the dative. This makes it possible for a single schema to represent more than one submeaning. Schema 2 captions a number of submeanings, many of which are distinguished by the choice of the dative's referent. The specific relevance of both syntactic variants and the role of lexical and extra-linguistic information with regard to each submeaning will be taken up indivi-dually as the schemas are discussed in turn in the remainder of this chapter.

3.1. Schema 1 - indirect object

Schema 1 in Figure 13 builds on Figure 9 by including the dative. It captions the actual transferral of an accusative direct object to the dative. Note that all of the participants in this schema are distinct. The nominative is the head of the action chain and is responsible for the movement of the accusative. The dative both receives the accusative entity and shows a potential for carrying out some further action with it (symbolized by the dotted arrow). Note that if this diagram were cut into three pieces, they would correspond quite closely to the canonical profiles of these three cases worked out by Smith (1985), cited in chapter 1, Figure 2. This schema is quite in harmony with common prose descriptions of the dative. Jakobson's feature specification for the dative is [+directional, +peripheral]. Jakobson (1936 [1971]: 48) expands this as meaning that the dative "signifies peripheral status ... and involvement in an action", and that it "denotes the existence of its referent as independent of the action". The fact that the dative does not participate directly in the prototypical action likewise indicates its peripheral status, and its potential as an agent attests to its independent existence. The schema is also an accurate caption of the two characteristics of the dative recognized by Burston (1977: 273-274): "a) the participant in question regards the dative referent as a kind of beacon exerting an attractive force, [and] b) [the dative] entity is remote, able to escape total integration inside the narrated event".

55

mple of a trivial sort of syntactic variant is the placement of the conessage in a separate clause. Thus example (15) will be treated like

actor important in shaping the network is the fact that the lexical nouns as well as extralinguistic knowledge can be used to disamnings of the dative. This makes it possible for a single schema to ree than one submeaning. Schema 2 captions a number of submeanof which are distinguished by the choice of the dative's referent. The vance of both syntactic variants and the role of lexical and extra-linmation with regard to each submeaning will be taken up indivi-dually as are discussed in turn in the remainder of this chapter.

Schema 1 and its Syntagmatic Variants Schemas Pictorial

Prose

a 1 - indirect object

Figure 13 builds on Figure 9 by including the dative. It captions the erral of an accusative direct object to the dative. Note that all of the in this schema are distinct. The nominative is the head of the action responsible for the movement of the accusative. The dative both accusative entity and shows a potential for carrying out some further t (symbolized by the dotted arrow). Note that if this diagram were cut eces, they would correspond quite closely to the canonical profiles of ases worked out by Smith (1985), cited in chapter 1, Figure 2. ema is quite in harmony with common prose descriptions of the bson's feature specification for the dative is [+directional, +periobson (1936 [1971]: 48) expands this as meaning that the dative ripheral status ... and involvement in an action", and that it "denotes e of its referent as independent of the action". The fact that the dative rticipate directly in the prototypical action likewise indicates its atus, and its potential as an agent attests to its independent existence. is also an accurate caption of the two characteristics of the dative by Burston (1977: 273-274): "a) the participant in question regards the ent as a kind of beacon exerting an attractive force, [and] b) [the y is remote, able to escape total integration inside the narrated event".

Usage

A nominative acts on an accusative to bring it to a dative in a setting. The dative retains independent status.

Indirect object

A nominative acts on a dative in a setting. The dative retains independent status.

Governed Dative

An action takes place on a dative in a setting. The dative retains independent status.

Impersonal Dative

- - - - - - setting

Local Subnetwork

~IOSSOf~IOSSOf~ ~acc

~nom

Figure 13. Schema 1 and its syntagmatic variants

"