Constructing a Lexicon of English Verbs [Reprint 2012 ed.] 9783110800623, 9783110164169

Citation preview

Functional Grammar Series 23

Editors

A. Machtelt Bolkestein Casper de Groot J. Lachlan Mackenzie

Mouton de Gruyter Berlin · New York

Constructing a Lexicon of English Verbs

I

1999

Constructing a Lexicon of English Verbs

by

Pamela B. Faber Ricardo Mairal Uson

W DE Mouton de Gruyter G Berlin · New York

1999

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

Faber, Pamela B., 1950Constructing a lexicon of English verbs / by Pamela B. Faber, Ricardo Mairal Uson. p. cm. - (Functional grammar series ; 23) Includes bibliographical references (p. ) and indexes. ISBN 3-11-016416-7 (cloth : alk. paper) 1. English language - Verb. 2. English language - Lexicology) I. Mairal Usön, Ricardo, 1965II. Title. III. Series. PE1271.F3 1999 425-dc21 99-33327 CIP

Die Deutsche Bibliothek -

Cataloging-in-Publication-Data

Faber, Pamela B.: Constructing a lexicon of English verbs / by Pamela B. Faber ; Ricardo Mairal Uson. - Berlin ; New York : Mouton de Gruyter, 1999 (Functional grammar series ; 23) ISBN 3-11-016416-7

© Copyright 1999 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced 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: Lüderitz & Bauer-GmbH, Berlin. Printed in Germany.

To the memory ofLeocadio Martin Mingorance

Preface by Christopher Butler In July 1995, the functional linguistics community was shocked and deeply saddened to hear of the sudden and untimely death of Professor Leocadio Martin Mingorance, of the University of Cordoba, Spain. Martin Mingorance's work, combining the Functional Grammar of Simon Dik with the lexematics of Eugene Coseriu into the lexically-based Functional Lexematic Model, began the process of developing the Functional Grammar conception of the lexicon into a model which integrates semantic, syntactic and pragmatic aspects of lexemes within a framework in which both paradigmatic and syntagmatic patterning find their place. Prominent among Martin Mingorance's collaborators were Pamela Faber and Ricardo Mairal Uson, whose determination to carry on and develop the line of research pioneered by their friend and mentor has resulted in the present volume. Their aim in this book is impressively ambitious: to give an account of the English verbal lexicon which not only systematises the meanings of lexemes within a hierarchical framework, but also demonstrates the principled connections between meaning and, on the one hand, the syntactic complementation patterns of verbs, and on the other hand, patterns of conceptualization in the human mind. Such an endeavor is entirely compatible with the tendency towards lexically-based approaches in modem grammatical theory. This shift in paradigm is explored in the first part of Chapter 1 of the book, where developments in lexicology and lexicography are reviewed in relation to their impact on linguistic theorising. Matters of psychological adequacy and computational implementation are also discussed. Chapter 2 provides an overview of Functional Grammar, concentrating on the role of the lexicon, as a background to the integrated onomasiological model of lexical description represented by the Functional Lexematic Model, within which Faber and Mairal's own proposals are situated.

viii

Preface

Chapter 3 provides a more detailed picture of the Functional Lexematic Model and begins to outline Faber and Mairal's own proposals. The scene is set by discussion of the often-criticised concept of 'semantic field' (relabelled as 'domain' in the Functional Lexematic Model), and its relationship with cognitive and generative linguistics, and with a number of approaches to lexical semantics. The aim in the Functional Lexematic Model is to find a kind of onomasiological organization of the lexicon which will be consonant with what is known of the organization of the mental lexicon. Faber and Mairal argue that a hierarchical paradigmatic model comes closest to this goal, but also that the paradigmatic organization of the lexicon into domains and subdomains is the basis for predicting the syntagmatic behaviour of lexemes, insofar as this is represented in syntactic complementation patterns. Hierarchies are developed through detailed, bottom-up analysis of entries in a range of monolingual dictionaries, the guiding principles being those of definitional analysis and lexical decomposition. Faber and Mairal's aim is to develop hierarchies, and their associated complementation patterns, for the whole of the verbal lexicon, in order to achieve a global characterisation of the design of the English vocabulary. Here, as throughout the book, methods are explained, and arguments made, through the analysis of an impressively wide range of semantic domains and their attendant syntax. In Chapter 4, the relationships between the semantic and syntactic behaviour of verbs are explored in much greater depth, through discussion of three types of lexically-realized parameters which play a part in the generation of clause structures. Grammatical parameters are those which have a direct effect on complementation, and include duration, temporal sequence, iteration, inception, achievement, cessation, causation, conation and factivity: optional parameters are those which are semantically present, but not necessarily syntactically realized; while contextual parameters act as clues for contextual setting. An important claim is that the greater the semantic scope of a lexeme, the greater is its variation in complementation behaviour. Chapter 5 completes Faber and Mairal's extension of the Functional Lexematic Model, by providing an account of the relationship between lexical structure and cognition. A key concept

Preface

ix

here is that of the predicate schema, which integrates paradigmatic and syntagmatic information about an individual lexeme, a lexical subdomain, or a whole domain. Predicate schemas at the subdomain level are accorded particular importance in the model, as they represent the links between the lexical macrostructure and individual lexemes. The authors demonstrate the power of the predicate schema approach in explicating the systematic relationships which underlie many metaphorical extensions of meaning. Chapter 6 concludes the book and presents an overview of the main claims. Faber and Mairal's book is most welcome at this stage in the development of functional linguistics, in presenting a thoughtprovoking, challenging and at times controversial account of lexical patterning and its relationship with meaning, syntax and cognition. It is sure to give rise to discussion which will take us even further forward on the road to a functional account of language which responds to the criteria of pragmatic, psychological and typological adequacy so strongly advocated by Dik. It also represents a fitting tribute to the pioneering work of a scholar who was able to take an aspect of Dik's theory and develop its potential: Leocadio Martin Mingorance. January 1999

Contents

List of tables and figures List of abbreviations 1. The lexicon and linguistic theory 1.1. Introduction 1.2. Lexical revival 1.2.1. Linguistic theory 1.2.2. Lexicology and lexicography 1.2.3. Psychology 1.2.4. Computational linguistics 1.3. Lexicon and grammar 1.3.1. Event structure and lexical representation 1.4. Summary 2. Simon Dik's Functional Grammar 2.1. An overview 2.2. FG lexical representation 2.2.1. Representational parameters 2.3. FLM lexical description 2.3.1. The FG lexicon and onomasiological structure 2.4. Summary 3. Lexical organization and the FLM 3 .1. Introduction: the concept of semantic field 3.1.1. Cognitive Linguistics 3.1.2. Generative Linguistics 3.1.3. Pustejovsky's Generative Lexicon 3.1.4. Fillmore's Frame Semantics 3.1.5. WordNet 3.1.6. Natural Semantic Metalanguage (NSM) 3.1.7. Meaning Text Theory and the Explanatory Combinatorial Dictionary 3.2. The FLM: paradigmatic axis 3.2.1. Definitional analysis and lexical decomposition 3.2.2. Lexical hierarchies

xv xix 1 1 4 6 8 15 18 22 23 34 37 37 46 47 56 58 65 67 67 69 71 73 73 75 76 77 79 89 101

xii

Contents

3.2.3. Differentiae: semantic parameters 3.3. The FLM: syntagmatic axis 3 .3 .1. The semantic characterization of complement selection 3 .3.2. Complementation and FG 3.3.3. Complementation and the FLM 3.4. Summary 4. Towards a semantic syntax 4.1. Introduction 4.2. Meaning definitions and syntactic projections 4.2.1. Lexically-realized grammatical parameters 4.2.2. Lexically-realized optional parameters 4.2.3. Lexically-realized contextual parameters 4.2.4. Position 4.3. Lexical iconicity 4.3.1. The genus and its scope 4.3 .2. The interaction of semantic and syntactic hierarchies 4.4. Summary 5. The cognitive axis 5.1. Introduction 5.2. Language and brain 5.3. Language and memory 5.4. Lexical structure and conceptual structure 5.5. Predicate schemas in the FLM lexicon 5.5.1. Towards a typology of predicate schemas in FG 5.6. Form and representation 5.7. A semantic macronet 5.7.1. Connections and dependencies 5 .7.2. The nature and type of lexical interconnections 5.8. Summary 6. Conclusion Appendix 1: Lexical domains Appendix 2: Semantic network Appendix 3: Possession Appendix 4: Synsem interface of to come to have

107 114 115 118 121 140 143 143 144 145 176 177 179 186 187 193 200 203 203 204 207 210 212 217 248 251 251 254 271 275 279 294 298 301

Contents

Notes References Author index Subject index

xiii 305 308 336 339

List of tables and figures

Tables Table 1. Table 2. Table 3. Table 4. Table 5. Table 6. Table 7. Table 8. Table 9. Table 10. Table 11. Table 12. Table 13. Table 14. Table 15. Table 16. Table 17. Table 18. Table 19. Table 20. Table 21. Table 22. Table 23.

Vocabulary matrix SoAs in RRG Typology of RRG lexical representations The structure of the clause in FG Symbols used in FG representations Overview of operators Overview of satellites First argument semantic functions Second and third argument semantic functions Segmentation of the dictionary definitions of embezzle Meaning components of steal Acquisition phase ofpossession Possession : inventory of complementation patterns of get verbs Interaction of semantic distinctions and syntactic complementation Semantic and perceptual parameters in manner-of-walking verbs Relations between complements and matrix predicates Shatter: inventory of selection restrictions Semantic and morphosyntactic typology of complements of to think carefully about something Typology of syntactic frames: plan and troponyms Semantic and morphosyntactic typology of complements of understand Typology of syntactic frames: to complain continuously Affected arguments in resultative constructions Movement in/on liquid

21 29 30 39 41 42 43 51 52 93 96 102 104 108 113 120 127 138 139 140 156 170 171

xvi

List of tables and figures

Table 24. Verbs of contact, lexically-realized contextual parameters Table 25. Typology of syntactic frames: liquid consumption Table 26. Typology of syntactic frames: visual perception Table 27. Typology of syntactic frames: to think something is true Table 28. Typology of syntactic frames: to say that something is certain Table 29. Domain of cognition: interrelations with other domains Table 30. Domain schema Table 31. Sense components of bawl Table 32. Definitional constituents of bawl and warble Table 33. Typology of syntactic frames: to think carefully about something Table 34. Typology of arguments: to think carefully about something Table 35. Contemplate: argument structure and meaning correlations Table 36. Lexical distinctions in perception Table 37. Phasal distinctions in existence Table 38. Phasal distinctions in cognition Table 39. Variational features Table 40. Institutionalized cultural models Table 41. First-level axiological polarizations in sound Table 42. Second-level axiological polarizations in feeling Table 43. Third-level axiological polarizations in cognition and speech Table 44. Scale of intensity in verbs of damage Table 45. Intensity scale of light Table 46. Verbs of stable light Table 47. Verbs of unstable light Table 48. Glare: argument structure and meaning correlations Table 49. Vision as an event and process

185 189 191 192 194 214 218 224 227 230 231 232 237 238 239 240 241 244 247 248 257 261 261 261 263 266

List of tables andfigures

xvii

Figures Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure

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

Paradigmatic lexical functions Outline of FG modules Macrostructural organization of a lexical domain To come to have Synsem interface ofpromise and its troponyms Functional divisions of memory Representation of the domain of existence Discrimination net for verbs of contact Image schema: visual image of light sources Container schema and feeling Scientific model of vision Visual perception (physical) Manner-of-staring verbs Visual perception with a parallel mental process Mental perception To see something in one's mind

13 38 64 103 200 209 219 221 259 260 264 265 267 268 269 270

List of abbreviations Theories CG FG FLM GB LFG MTT NSM RRG TG

Construction Grammar Functional Grammar Functional-Lexematic Model Government and Binding Theory Lexical Functional Grammar Meaning Text Theory Natural Semantic Metalanguage Role and Reference Grammar Transformational Grammar

Theoretical concepts CA ICM LCS LDB LF LKB LS LTM MRD MTM NLP NP SLD SoA SoW STM ΤΑΜ VP

Componential Analysis Idealized Cognitive Model Lexical Conceptual Structure Lexical Database Lexical Function Lexical Knowledge Base Logical Structure Long Term Memory Machine-Readable Dictionary Meaning Text Model Natural Language Processing Noun Phrase Stepwise Lexical Decomposition State of Affairs State of the World Short Term Memory Tense-Aspect-Modality Verb Phrase

xx

List of abbreviations

Dictionaries

CC ECD

The American Heritage Dictionary of the English Language Collins Cobuild English Language Dictionary Explanatory-Combinatorial Dictionary

LDCE

Longman Dictionary of Contemporary English

LLCE OALD RHD

Longman Lexicon of Contemporary English Oxford Advanced Learners' Dictionary The Random House Dictionary of the English Language Webster's New World Dictionary for Young Readers

AHD

WDYR

Abbreviations used in FG-representations Word classes A Adjective Ν Noun V Verb Syntactic functions Obj Object Subj subject Pragmatic functions Foe focus Top topic Layers f χ e X Ε

predicate term predication proposition clause

Semantic functions zero 0 agent Ag Ben beneficiary Dir direction Exp experiencer Fo force Go goal (patient) Instr instrument Loc location Man Manner Po Positioner Proc Ree Ref So Temp

Processed Recipient Reference Source Time

List of abbreviations

Satellites σι G2 03 σ4

any predicate satellite any predication satellite any proposition satellite any illocutionary satellite

Π-operators Π,

any predicate operator

Π2 Π3

any predication operator any proposition operator

Π4 Poss Post Sim

any illocutionary operator Possibility Posterior Simultaneous

Term operators Ω 0 1 d I

any term operator zero quantifier singular definite Indefinite

xxi

1.

The lexicon and linguistic theory

1.1.

Introduction

One of the major points of agreement in contemporary grammatical theory is the centrality of the lexicon as an important source of information about sentence structure. In recent years, many linguistic theories have undergone a gradual change in paradigm, discarding syntactically oriented postulates in favor of more lexically-based ones. The fact that the lexicon is presently in the spotlight signifies that lexemes (and the different types of information they convey) are the building blocks of language in all senses. According to Fellbaum (1998b: 3), the perception of the lexicon as a central component of grammar is due largely to "the discovery that the lexicon is a highly structured repository of rules and principles that give it status and prominence previously accorded only to syntax". Lexical knowledge has a central role because it is now assumed that much of the structural information of a sentence is best encoded from a lexical perspective. This panlexicalist perspective has logically brought the message home that semantics cannot be divorced from syntax in any sense. Langacker (1991b: 275) affirms that it is ultimately as pointless to analyze grammatical units without reference to their semantic value as to write a dictionary which omits the meanings of lexical items. Semantic properties must be accounted for, regardless of where the semantics is encoded or what one's assumptions about the lexicon and syntax are. As to the place where semantics is encoded, most linguistic models agree that it is in predicate-argument structure and that the structure of the clause is predictable from the semantics of the predicates. This premise is now implicit in a wide range of models, such as Role and Reference Grammar (RRG), Government and Binding Theory (GB), and Simon Dik's Functional Grammar (FG). Other theoretical frameworks, such as Construction Grammar (CG), argue that lexicon and grammar form a continuum, and encode semantics in constructions, or pairings of form-meaning

2

The lexicon and linguistic

theory

correspondences. Syntactic-semantic mappings are thus explained by the association of verb meanings and constructions (Goldberg 1995, 1996; Kay 1997). As the convergence point for syntax and semantics, the lexicon also becomes the interface of the grammatical and the conceptual component. According to Pustejovsky (1995: 6), the meaning of words should somehow reflect the deeper conceptual structures in the cognitive system and the domain it operates in. In consonance with this, cognitive approaches study meaning in terms of a speaker's construal of situations and how that meaning is embedded in a particular background, scene or frame. As shall be seen, there are different views on the extent to which the lexicon is a map of conceptual structure, but no one can deny that nonlinguistic organizing principles are reflected in the semantics of natural language. Our study of the English verbal lexicon has been carried out over the last decade within the framework of the Functional-Lexematic Model elaborated by Martin Mingorance (1984, 1990, 1995). We have chosen to examine the organization of the verbal lexicon because of the decisive role that verbs play in sentence structure. Since, to a great extent, verbs determine the number as well as the semantic characteristics of the other elements in the linguistic expression, we believe that their role is just as decisive in the paradigmatic structure of the lexicon. We have found that the analysis of verbs in terms of their semantic and conceptual components can reveal many of their syntactic properties, and that such regularities also are present in other European languages. If a way can be found to organize the verbal lexicon in terms of both semantic and syntactic properties, that would provide the basis for a viable model of lexical representation. Although there are many lexical semantic analyses of one or two verbs, considerably less has been written about the lexicon as a whole. Those studies with a wider scope invariably seek to explain semantic organization through syntactic behavior. However, in lexical semantic analysis, semantic considerations are all-important because syntactic distinctions in themselves are not sufficient as a basis for the establishment of an inventory of semantic categories. The evident importance of semantic considerations in any study of the lexicon led us to choose Simon Dik's Functional Grammar (FG) as the framework for our analysis. FG is a dictionary-based grammar in

Introduction

3

which the lexicon has always been the most important component (Dik 1997ab). Within this framework, each predicate is described in terms of its form, syntactic category, quantitative valence, qualitative valence, and meaning definition. Proof of the relevance of these parameters is the fact that they also appear in many other grammatical models, both formal and functional. By themselves, however, they are insufficient to account for the full scope of information to be included in a lexical entry. Although the structure of the information within a lexical entry is crucial, provision must also be made to include information about how lexemes are related to others in the lexicon. More specifically, it is necessary to come to grips with meaning on both the microstructural and macrostructural level. Microstructurally, this means examining the role meaning definitions play in the development of an interface between syntax and semantics. Macrostructurally, this signifies situating lexemes within the larger context of their lexical domain, and specifying their relations with lexemes in other areas of meaning. In this way, lexemes are not conceived as a frozen list of items, but rather as dynamic representations within a conceptual network. Nevertheless, the specification of the macrostructural design of the lexicon, or how lexical items are related through the meaning they convey, cannot be done by using one's intuition to first create a set of ad hoc meaning areas, and then trying to distribute lexemes a posteriori in pre-ordained semantic classes. An inventory of lexical domains is best arrived at by working upward from word meaning through the systematic application of theoretical premises. In the 1980s, Martin Mingorance (1984, 1985ab, 1987, 1990, 1995, 1998ab) advanced the theoretical claim that lexical representations are the key to the determination and explanation of the syntactic representations of predicates. The result was the elaboration of the Functional-Lexematic Model (FLM), which integrates Dik's FG and an adapted version of Coseriu's Lexematic Theory. The chapters that follow describe how this model develops the FG lexicon component by configuring it in lexical domains, according to principles of definitional analysis implicit in Stepwise Lexical Decomposition (Dik 1978b). The lexical organization obtained is the result of the convergence of both paradigmatic and syntagmatic information. As result, both the micro- and macrostructure of these domains are a reflection of the interface of syntax and semantics. In a parallel way, it

4

The lexicon and linguistic theory

also provides insights into the encoding of lexical meaning as a base for knowledge representation. This means that an accurate linguistic theory should be able to account for the interrelationships between syntax and semantics. One of the research goals in lexical semantics is the development of semantic representations that could serve as a basis for the formulation of a set of Unking rules, which will ultimately determine the syntactic realization of arguments. As a result, the design of the lexicon is allimportant because it has become the core of most grammatical models. Jackendoff (1975: 639-640) states that a theory of the lexicon must meet the following three levels of adequacy: (i) Observational adequacy, by means of which a theory of the lexicon supplies each lexical entry with sufficient information to account for the full potentiality of the speaker's lexical competence. (ii) Descriptive adequacy, which maps out the type of relations, regularities and generalizations which hold between lexical items; (iii) Explanatory adequacy, which shows how and why the set of regularities and particular relationships in the lexicon are chosen, and moreover, how these relations affect grammar. Since it is beyond the scope of this book to discuss all of these levels in detail, we shall focus principally on the levels of explanatory and descriptive adequacy. However, first we shall examine how different models of lexical analysis have tried to explain lexical regularities and relations, and how they affect grammar.

1.2. Lexical revival The lexicon did not always hold the privileged position in linguistic theory that it now enjoys. For many years, it was relegated to the background, and regarded as a wastebasket for peripheral information and irregularities which syntax could not explain. Giving clear priority to formal grammatical description, Bloomfield (1933: 274) considered the lexicon merely to be "an appendix of grammar". For this reason, he

Lexical revival

5

is often blamed for shoving meaning into the closet, but as Wierzbicka (1996: 4) points out, the banishment of meaning from linguistics can primarily be attributed to Bloomfield's followers. Although Bloomfield himself wished to eliminate semantic considerations from linguistic analysis, he did not reject meaning in the sense of avoiding any mention of it in linguistic description. As syntax was easier to study objectively, linguists seemed to think that it should be dealt with first, and that meaning could be put on the back-burner until all the mysteries of syntax had been resolved. This is one of the basic premises of the first version of Transformational Grammar. Wierzbicka (1996: 5-6) mentions the fact that although Chomsky was one of the main actors in the "cognitive revolution" of the late fifties and sixties, he is still Bloomfieldian in the sense that he is also reticent about dealing with meaning. This was most evident in Syntactic Structures, in which Chomsky (1957) makes no provision for either a lexicon component, or specific principles to determine the contextual restrictions that a language imposes on lexical insertion. The majority of linguistic schools at that time were not very interested in lexical items, and chose to disregard the fact that the lexicon contains regularities that are meaningful for syntax. There was a marked tendency to minimize lexical information to the benefit of a more algorithmic syntactic component. Even semantically oriented linguists seemed to be more concerned with explaining how to obtain the meaning of sentences recursively from syntactic structure, as well as from the semantic value of those constituents with an autonomous meaning. Nevertheless, it soon became evident that the emphasis on autonomous syntax was not especially conducive to understanding language, and that certain syntax-related problems could only be resolved by taking a closer look at meaning. Important factors in the process were the evolution of Chomskyan linguistics and the increasingly important role of the lexicon within its framework. In fact, the latest versions of Chomskyan linguistics now include references to meaning though statements in this direction have been largely programmatic. Another contributing factor was the appearance of other competing frameworks, which were even more lexically centered. Relevant examples are Hudson's (1976: 1-26) Word Grammar, which

6

The lexicon and linguistic theory

dispenses with transformations altogether; the grammar proposed by Montague (1970ab, 1973) with its semantic component, which formalizes the relations that transformations signal at the level of semantics; and Lexical-Functional Grammar (Bresnan 1979, 1982), which treats syntactic phenomena as operations on lexical forms. The specific reasons why linguists began to look towards semantics again are various, but it is no coincidence that this "lexical reconciliation" (Grimshaw 1994) is closely linked to major advances in linguistic theory, lexicology and lexicography, psychology, and computational linguistics.

1.2.1. Linguistic theory One of the things that contributed greatly to this lexical revival was the progressive theoretical reorganization that took place in Transformational Grammar (TG). In Aspects of the Theory of Syntax (Chomsky 1965), a lexicon containing context-sensitive lexicalinsertion rules is introduced as a subcomponent of the base. Phrase structure rules, together with other components and principles, generate a deep structure, which is a sort of skeleton with preterminal nodes. The function of the lexicon is to insert a posteriori lexical items into the nodes provided by the syntactic rules of the categorial component (Katz—Fodor 1963). In the TG lexicon, each lexical entry has a subcategorization frame, specifying the type of complements a lexeme can co-occur with. The semantic nature of the complements is encoded by means of a series of selection restrictions (e.g. animate, inanimate, abstract, concrete). In this sense, Aspects paved the way for the more lexicallybased approaches that followed. At the end of the sixties, serious attempts were made to give TG a more important semantic component. Generative Semanticists formulated underlying semantic representations on the basis of what they believed to be an inventory of semantic primitives (McCawley 1968). Of special interest here is the use of semantic decomposition, a process by means of which lexical items are analyzed in terms of a finite set of semantic components. A case in point is McCawley's (1968) decomposition of the verb kill, which Fodor (1970) criticizes.

Lexical revival

7

Despite the fact that this conception of semantic decomposition came to be considered inadequate for semantic representation, it was never completely abandoned, and has now been incorporated in various semantic theories of the lexicon (Jackendoff 1983, 1987; Talmy 1985; Ravin 1992; Van Valin 1993b). Generative Semantics was harshly criticized by Interpretative Semantics, a rival theory which extended the format proposed in Aspects (Chomsky 1970; Jackendoff 1972). The new proposal for the subcategorization frames of lexical entries included information, such as category, syntactic pattern, and thematic roles, introduced as variable positions in the semantic representations of predicates (Jackendoff 1972: 38-43). The debate between Generative and Interpretative Semantics had the positive effect of causing linguists to reconsider the lexicon and of including linguistic phenomena previously regarded as syntactic in lexical entries. For example, Chomsky's Lexicalist Hypothesis postulates that more and more syntactic phenomena (including transformations) need semantic explanations (Kiparsky— Kiparsky 1970; Katz—Postal 1964). However, after the fall of Generative Semantics, linguistic models became extremely complex and abstract. As a result, they had difficulty in coping with morphology, word-formation, lexical semantics, and pragmatic factors such as presupposition and discourse analysis (Aronoff in Nowakowski 1990: 4; Sanchez de Zavala 1976: 95-103). In an effort to make generative models more explanatory, transformations were virtually eliminated by reducing all of them to move a (Chomsky 1973, 1981b; Newmeyer 1980). The rise in importance of the lexicon brought with it a corresponding simplification of algorithmic rules, and in some cases, the categorial component (e.g. phrase structure rules) was eliminated altogether.1 Chomsky (1986: 86-87) questions the validity of the classical theory, which postulates that both categorial selection (c-selection) and semantic selection (s-selection) should be included in each lexical entry. In his opinion, the classical analysis contains redundant information, and he comes to the conclusion that U(niversal G(rammar) does not need an independent theory of subcategorization since these properties are largely predictable from other principles such as Lexical Projection and Case Theory. He suggests that the internal make-up of the lexicon should be restricted to s-selection since subcategorization

8

The lexicon and linguistic

theory

can be derived from s-selection, but not vice versa. An important consequence of this radical change of perspective in Chomskyan linguistics was the formulation of lexical theories in which the lexicon is conceived as a repository of the syntactic properties of argumenttaking predicates.

1.2.2. Lexicology and lexicography As meaning began to come out of the closet in linguistic theory, more importance was logically given to lexicology and lexicography. This led to the development of lexicological frameworks with a clear semantic focus, one of the aims of which was the elaboration of enriched lexical representations. Two of these frameworks, which have certain aspects in common with the FLM, are Wierzbicka's Natural Semantic Metalanguage and Mel'cuk and Zholkovskij's Meaning Text Theory.

1.2.2.1. Natural Semantic Metalanguage The Reductive Paraphrase or Natural Semantic Metalanguage (NSM) approach elaborated by Wierzbicka (1987, 1988, 1996) combines the philosophical and logical tradition in the study of meaning with a typological approach to the study of language, and with empirical crosslinguistic investigations. Wierzbicka's research has focused on the identification of the shared core of all natural language in order to formulate the NSM, an intermediary language of semantic description, which uses natural language expressions, thus avoiding abstract features, terms or logical symbols (Goddard 1998: 56-64). The NSM is the result of semantic analysis, the underlying principle of which is explication through reductive paraphrase. Wierzbicka (1995: 149) compares the meanings of words to "objects constructed out of various Lego blocks". Her approach to lexical semantics analyzes word meaning in order to find its most basic components. With this inventory of semantic primitives, she claims that it is theoretically possible to define all of the other words in the

Lexical revival

9

language. In this way, the meaning of a word is formulated in the simplest possible terms, and circularity is avoided. Nevertheless, in order for such a description to be considered descriptively adequate, it must be able to replace the original expression without change of meaning. For example, the NSM definition of mother is the following: (1)

X is Y'smother at some time before now, Y was very small at this time Y was inside X's body, Y was like part of X because of this, people can think something like this: X wants to do good things for Y X doesn't want bad things to happen to Y (Goddard 1998: 62).

One of the main objectives of this semantic analysis is the elaboration of an inventory of a semantically minimal core, consisting of primitives that cannot be defined in simpler terms. According to Wierzbicka, such primitives are not directly observable, but can only be established by trial and error. Over the years, her original list of fourteen semantic primitives has grown to nearly sixty. These include predicates from various semantic categories: mental [think, want, know, feel, see, hear], speech [say], action and event [do, happen, move], existence and life [be, live], relational [like, part of], space [in, under, above, on, side, inside, far, near], evaluation [good, bad], and description [big, small]. Such semantic primitives are accorded the status of lexical universale, which are the symbolic representation of universal conceptual primitives. The sets of semantic primitives identified in this way across languages are regarded as language-specific manifestations of a universal set of fundamental human concepts, and are the result of a limited isomorphism in the lexicon as well as grammar. Wierzbicka (1995: 154) affirms that these concepts are the basis of the tertium comparationis that necessarily underlies successful interlinguistic communication. Within a particular language, every element belongs to a unique network of elements, and occupies a particular

10

The lexicon and linguistic

theory

place in a unique network of relationships. When we compare two or more languages, we cannot expect to find identical networks of relationships. We can, none the less, expect to find corresponding sets of indefinables (Wierzbicka 1996: 15). The set of semantic primitives is supposed to be a complete lexicon for semantic analysis. As a kind of "mini-language" or subset of a full natural language, it naturally has its own syntax as well. Wierzbicka (1996: 144) characterizes the syntax of the NSM as analogous in some respects to, but much simpler than, the syntax of natural languages with a rudimentary parts-of-speech system. Goddard (1998: 329) describes it as primarily combinatorial in nature. In fact, the latest research in this area is focused on discovering universal patterns of co-occurrence among primitives, which constitute canonical context sentences or sentence fragments. Another concept within this theory is that of valency options in relation to semantic primitives. For example, in the NSM say has four valency options: (2)

SOMEONE SAID SOMETHING SOMEONE SAID SOMETHING ΤΟ SOMEONE

[optional addressee slot]

SOMEONE SAID SOMETHING ABOUT SOMETHING [optional topic slot] SOMEONE SAID SOMETHING TO SOMEONE ABOUT SOMETHING [both optional slots filled]

Since universal semantic primitives cannot be defined because more basic words than themselves do not exist, they are identified through universal syntactic patterns. This is extremely important, given that such primitives are frequently polysemous, and their different senses vary from one language to another. Goddard (1998: 335) affirms that a universal set of participant roles can be formulated by tying them to the argument slots of semantic primitives like DO, HAPPEN, MOVE, KNOW, THINK, FEEL, and WANT. In this sense, Wierzbicka's proposal runs parallel to Jackendoff (1990), who also posits that a complex or non-primitive predicate can impose

Lexical revival

11

multiple participant roles upon an argument because of its compositional nature. Since the NSM is intended to be a model of the innate and universal lingua mentalis, the NSM grammar styles itself as a hypothesis about the grammar of human cognition. The basic claim is that there exists a set of correspondences between semantic primitives of all languages and also between certain combinations of primitives, so that essentially anything that can be said in one NSM can be accurately translated into another.

1.2.2.2. Meaning Text Theory Meaning Text Theory (MTT), first proposed by Mel'cuk— Zholkovskij (1970), is one of the most influential meaning-based theories in contemporary linguistics. Though it is somewhat different from the FLM approach, there are also many areas of convergence. MTT is based on the fact that any act of linguistic communication involves content (meaning), an oral or written signal (text), and a mapping (a set of correspondences between meanings and texts). One of its postulates is that a natural language L is a finite set of many-to-many correspondences between a specific set of meanings and a specific set of texts. Though meanings and texts are accessible to the linguist, the correspondence between them is not. As a result, a natural language can only be described as a functional model, or a system of formal rules which simulates the linguistic behavior of L's speakers. MTT describes a set of Meaning Text Models (MTMs) or possible models, which relate texts to their meaning representations through the description of mappings between meanings and natural language texts. It specifies the following seven levels of description: (i) (ii) (iii) (iv) (v) (vi)

Semantic representations (SemR) Deep syntactic representations (DsyntR) Surface syntactic representations (SsyntR) Deep morphological representations (DmorphR) Surface morphological representations (SmorphR) Deep phonetic representations (DphonR)

12

The lexicon and linguistic theory (vii)

Surface phonetic representations (SphonR)

An utterance is simultaneously characterized by representations at all levels (Wanner 1996b: 23). Lexical information is encoded in the Explanatory Combinatorial Dictionary (ECD) (Mel'cuk 1981, 1988, 1989, 1996; Mel'cuk— Zholkovskij 1970; Mel'cuk et al. 1984). The purpose of the ECD is to cover all linguistic knowledge and capture lexical regularities at all levels. Lexical entries are divided into three zones: the semantic zone, the syntactic zone and the lexical combinatorial zone. The semantic zone specifies a semantic network, which defines the meaning of the lexical entry in terms of simpler meaningful elements. The syntactic zone of an entry contains the government pattern, which specifies for each semantic actant of the headword, the corresponding deep-syntactic actant, and all surface representations of the latter in the text. The lexical combinatorial zone specifies semantically related lexemes as the values of lexical functions. MTT also proposes an inventory of lexical functions (LFs), which codify different types of semantic and syntactic relations. According to Mel'cuk (1996: 39), a lexical function / is a correspondence that associates a given lexical expression L with a set of lexical items Lj which express a specific meaning associated with/. This can be represented by the following formula: / ( L ) = Li. Mel'cuk defines close to sixty lexical functions and using Saussure's dichotomy, divides them into two types: paradigmatic LF relations and syntagmatic LF relations. Paradigmatic LFs associate with a keyword a set of lexical items that share a non-trivial semantic component, and include all contrast and substitution relations between lexical items in specific contexts. A few examples of such functions, some of which coincide with sense relations, can be seen below: (3)

a. b. c.

Gener (republic) = state Anti (joy) = grief Contr (fire) = ice

The lexical function Gener (3a) not only covers hyponymic institutionalized lexical relations, but also the relation between a lexeme and its closest generic concept, which can be used as an

Lexical revival

13

attributive modifier. Mel'cuk specifies different types of antonymy, but generally this relation in MTT is the result of the semantic decomposition of lexical units, as seen in their respective definitions. In (3b), joy is considered antonymous to grief because the former is defined as a pleasant emotion and the latter, as an unpleasant emotion. The opposition between pleasant and unpleasant is thus the basis for antonymy. The contrastiveness relation is different from traditional antonymy in that it does not necessarily entail a logical opposition between the lexical units opposed. Example (3c) shows a contrastiveness relation in the sense of Apresjan's modal antonyms (Wanner 1996b: 9). Although various classifications for LFs have been proposed, one of the most coherent proposals is that of Alonso Ramos, Tutin, and Lapalme (1995: 353), who have summarized paradigmatic LFs in Figure 1: Paradigmatic LFs

Substitutive LFs

Anti C o n v

Contr

Semantic Derivations

Syntactic Derivations

Syn Gener

A„

TYPICAL

TYPICAL

CATEGORIES

N O U N S FOR

FOR ACTANTS

ADVERBIALS

TYPICAL QUALIFIERS

Adv.

A A ^TV. Quail,

Able

AdV|

A

Si

Smed Smod

Sres

Sloe

Sinsr

Figure 1. Paradigmatic lexical functions

Syntagmatic LFs formalize a semantic relation between two lexemes Li and L2, which is instantiated in the textual string in a nonpredictable way. Such a relation is nonpredictable when the cooccurrence of one cannot be derived from the semantic selection restrictions of the other, but rather has to be learned as an instantiation

14

The lexicon and linguistic theory

of a specific syntagmatic relation. Syntagmatic LFs may only be semantically constrained, or they may have a lexical bias. Examples of such functions can be seen in the following: (4)

a. b. c.

Magn (naked) = stark Instr (satellite) = via Funco (SHOWN) = fails

For example, the lexical function Magn (4a) is an attributive relation between two lexemes, and codifies the highest degree of the semantic content in question. Nakedness always means not having any clothes on, but being stark naked is the total absence of any covering/clothing whatsoever, probably within a context that makes this state seem even more surprising than usual. Instr (4b) signifies by means of In the case of satellite, the instrument relation is indicated by via since television programs are more often described as being received via satellite, rather than by/by means of satellite. Another example of a syntagmatic lexical function is that of Func (4c), which codifies a type of noun-verb collocation indicating performance. In this sense, the verb that generally collocates with snow is fall because this is the action usually associated with it. Such relations are extremely widespread across language and languages, and have proved to be extremely useful in computational linguistics. LFs are a means for a systematic description of what Wanner (1996b: 1) describes as "institutionalized" lexical relations, and can be used in the representation of lexical semantic relations in computer applications, such as text generation (Alonso Ramos— Tutin—Lapalme 1995).

1.2.2.3. The formalization of lexical knowledge Both the NSM and the MTT, each in its own way, underline the fact that conceptual knowledge can be derived from fully specified lexical entries, and that such entries constitute mini-knowledge representations. The models described are similar to the FLM in that both of them are based on meaning and lexical analysis. Another similarity is that they initially arose from a general dissatisfaction

Lexical revival

15

with the traditional concept of a dictionary and the desire to produce a better one by enriching it with more information, structured according to a set of underlying theoretical principles. This new type of dictionary endeavors to meet various standards of adequacy. It is envisioned as being a lexicological as well as a lexicographic product, whose dynamic character accounts for the actual lexical competence of speakers. This dynamism is manifest in an enriched entry, which formalizes, at least to a certain degree, different types of relations between words and aspects of meaning previously not accounted for. Such lexically-based approaches to linguistic theory point to a more mentalistic approach, in which language is conceived as a blueprint of the mind of the Natural Language User.

1.2.3. Psychology Enhanced lexical representations go hand in hand with psychological adequacy. The objective of including more information about words and their interrelations in lexical entries is to make their configuration more similar to the organization within our mental lexicon. In line with this search for psychological (and neurological) adequacy, developments in psychology have contributed to a better understanding of meaning, as well as the structure of the mental lexicon. The word lexicon is used in psycholinguistics to refer to the mental lexicon of the natural language user. Emmorey—Fromkin (1988: 12) define the mental lexicon as a component in the grammar in which information about individual words and/or morphemes is entered. Consequently, the mental lexicon contains what a speaker/hearer of a language knows about the form of the entry (its phonology), its meaning (its semantic representation), and its combinatorial properties (its syntax). The lexicon is thus conceived as lying at the interface between those conceptual structures generated by our cognitive systems and their syntax (Gonzalez Escribano 1991: 213). Two central issues in psycholinguistics are lexical acquisition and the configuration of lexical knowledge in memory. The fact that a speaker can mentally find the word that he/she wants in less than 200

16

The lexicon and linguistic

theory

milliseconds, and in certain cases, even before it is heard, is proof that the mental lexicon is organized in such a way as to facilitate access and retrieval. In order to arrive at the meaning of an utterance, the hearer must, among other things, determine the semantic relationships between rapidly arriving words. This is done by analyzing the acoustic information received in order to access the lexicon for both semantic and syntactic information in lexical representations. The ultimate goal is to develop the propositional representation of the utterance or the relationship among the entities, processes and events described (Wingfield—Titone 1998: 241-242). Studies of sentence processing suggest that normally the hearer/text receiver tries to understand the meaning as quickly as possible and then discards the surface structure (Bransford—Franks 1971; Bransford—Barclay—Franks 1972). Wingfield—Titone (1998: 243) point out that if this is true, then it is the conceptual representation of the utterance that is the primary focus of the memory system. The logical conclusion is that surface form is subordinate to meaning and not the other way around. In a parallel way, the complexity of lexical memory has fascinated many psycholinguists, who have proposed different ways of exploring and analyzing the cognitive processes involved in acquisition and retrieval, as well as the configuration of concepts. It is not easy to gain access to the "black box" of the mind, and for this reason, methods have been principally analogical, or according to Miller—Fellbaum (1991: 198), only centered on a tiny part of the lexicon. Researchers often have used language structure to obtain a better understanding of conceptual organization, despite the fact that the exact relationship between lexical structure and conceptualization is not known. Pederson—Nuyts (1997: 4—6) affirm that linguistic behavior (language comprehension and production) has a special status because it provides the richest possibilities for investigating conceptualization. Furthermore, theories which consider conceptualization to be closest to language are generally much further ahead in terms of formal modeling of conceptual structures and its relationship with linguistic structure (as in the case of Dik or Jackendoff) than are views which maintain that conceptualization is more abstract or at least less directly languagebased. However, Pederson—Nuyts are careful to indicate that this does not mean that the more abstract or less language-related views are mistaken since the language view obviously cannot account for the

Lexical revival

17

relationship between conceptualization and other non-linguistic types of behavior (Nuyts 1990, 1992). Though language-based research has provided and will continue to provide many insights, a true characterization of conceptualization is evidently not possible on the basis of only one type of behavior. The search for a psychologically adequate model of the lexicon led to the integration of Rosch's (1978) theory of categorization in various semantic theories. Experiments carried out by Rosch and other cognitive psychologists show that the structure of nominal concepts is built around their most prototypical member and covers a set of more peripheral category exemplars. Such theories can be said to fall within the more general framework of cognitive semantics (Fillmore 1985; Lakoff 1987; Langacker 1987, 1991a). Despite certain differences, all of them basically agree that categorization is an essential aspect of human cognition, and that grasping a meaning is an event of understanding in which human perceptual experience relates language to some version of the projected world. For example, Fillmore develops the notion of frames (also termed scripts, schemata)·, Lakoff (1987) proposes idealized cognitive models (ICMs); and finally Jackendoff (1983,1991) formulates lexical conceptual structures (LCS) in terms of necessary, centrality and typicality conditions. Jackendoff s basic hypothesis is that there is a form of mental representation called conceptual structure that is common to all natural languages. His theory contains no notion of reference in the standard sense, but rather is based on the mind's construal of the world, or how the speaker is at the moment inviting the hearer to view the world (Jackendoff 1991: 12).

Since cognitive semanticists believe that there is no access to a reality independent of human categorization, the structure of reality is necessarily reflected in language as a product of the human mind. In line with this, Lakoff (1987) and Johnson (1987) posit the existence of image schemas such as CONTAINER, PATH, FORCE, etc, which represent the pre-conceptual structuring of bodily experience. Image schemas organize our understanding of very broad concepts (e.g. time, cognition and causation) across a wide range of more abstract domains. They are represented by means of diagrams and are often deduced on the basis of a corpus of phrases, which are different specifications of the same ontological metaphor.

18

The lexicon and linguistic

theory

The model of categorization in cognitive semantics is derived from Rosch's theory and based on typicality. At first, it would seem to be in conflict with a model of categorization based on hierarchies such as the one that will be described here. However, in reference to nominal concepts, Miller (1998: 33) states that it is not necessary to choose between typicality and hierarchical organization, and points out that hierarchical organization appears to be a necessary feature of the mental dictionary: Perhaps typicality and hierarchy coexist. That is to say, more information must be stored with every lexicalized concept than is required to establish its position in a lexical hierarchy. Imagery or other mental models might be associated with lexicalized nouns without disrupting their hierarchical organization. In our opinion, Miller's assertion can be extended to apply to the verbal lexicon as well. The basic (or most cognitively salient) level of categorization is not the most abstract or superordinate term, but rather appears at the second or third level. Mental models more related to fuzzy meaning could be associated with this type of representation in order to enrich word meaning. However, this does not exclude the use of hierarchies as a configurational pattern. Though hierarchical organization is evidently not the whole story behind paradigmatic structure, it has been shown to be psychologically relevant, and extremely useful for the representation of ontologies of concepts.

1.2.4. Computational linguistics The close interrelation between psychology and language also has had a direct repercussion on computational linguistics. One of the goals of cognitive science has always been the elaboration of rules as well as representations which will explain both the way humans think and the way computers work. The purpose of natural language processing (NLP) is the automatization of linguistic processes such

Lexical revival

19

as language comprehension, production, and acquisition, all of which imply an extensive knowledge of vocabulary. Oddly enough, before the so-called lexical revival, this type of lexical knowledge was not considered important. The first NLP systems designed had lexicons of around a hundred words, but it soon became clear that the first step in the development of a robust computational prototype was the construction of a powerful lexicon with a rich network of semantic relations. In fact, there is now general agreement that the viability of a computational system is in direct relation to the delicacy of its lexical organization (Moreno Ortiz 1998: 35). Calzolari (1994: 267) observes that it is almost a tautology to affirm that a good computational lexicon is an essential component of any linguistic application. The design of such a lexicon means trying to partially simulate the working of the mental lexicon on the computer. Although the computational lexicon and the mental lexicon obviously differ in that one functions with chips and the other with neurons, insofar as possible, the operations and procedures they perform should try to achieve a certain degree of similarity. The ultimate goal in NLP is the development of a totally explicit theory of language processing. As a result, research in this area has become a convergence point for a variety of disciplines such as psycholinguistics, computational linguistics, and computer science. Although such a theory is still a long way from being formulated, certain NLP systems have been elaborated on the basis of a partial understanding of some of those processes (e.g. text-to-speech synthesizers and speech recognition systems). Considerably less progress, however, has been made in the areas of language comprehension and generation. The majority of NLP systems now use a knowledge-based approach since in order to attain their primary objective, explicit linguistic knowledge is necessary, along with more general knowledge. One of the most important research projects of this type is Acquilex, the objective of which is the extraction of multilingual lexical information from various machine-readable dictionaries (MRDs) and its integration in a lexical knowledge base (LKB). According to Ingria et al. (1992: 360), an LKB is "a large-scale repository of lexical information, which incorporates more than just static descriptions of

20

The lexicon and linguistic

theory

words, e.g. by means of clusters of properties and associated values". Such a repository specifies constraints on word behavior, dependence of word interpretation, and distribution of linguistic generalizations. The type of representation in an LKB is thus much more dynamic than the static representation found in a lexical database (LDB). Generally speaking, such knowledge can be divided into five subtypes, the majority of which is stored in the lexicon: (i)

Phonological information about the sound system and the structure of words and expressions, stress, and intonation (ii) Morphological information about the structure of words (iii) Syntactic information about the configuration of words in phrases and sentences (iv) Semantic information about the meaning of words and how those meanings combine to form the meaning of sentences (v) Pragmatic information such as the retrieval of pronoun referents, the analysis of presuppositions or communicative intentions. The representation of all these different types of knowledge is such a complex problem that it has become a stumbling block for NLP systems. B. Levin (1991: 205) states that the lexicon has "often proved to be a bottleneck in the design of large-scale natural language systems, given the tremendous number of words in the English lexicon, coupled with the constant coinage of new words and shifts in the meaning of existing words". Various proposals have been made for the codification of lexical information as knowledge. One of the best known is that of Miller (1986, 1990; Miller et al. 1993). Miller and his research team at Princeton have elaborated WordNet, an electronic lexical database, whose design is inspired by psycholinguistic and computational theories of human lexical memory. English nouns, verbs, adjectives, and adverbs are organized in synonym sets (synsets), each representing one underlying lexicalized concept. One of the design features of WordNet is the organization of the lexicon according to grammatical category, as well as semantic field. This means that WordNet contains a large quantity of redundant information that obviously would not

Lexical revival

21

appear in a traditional dictionary, especially in those cases where a word belongs to more than one category. This type of organization has the advantage of not imposing the same representational schema on each syntactic category, and so the most appropriate type of organization can be found for each one. For example, nouns are organized according to conceptual hierarchies, verbs according to different kinds of entailment, and adjectives and adverbs as Ndimensional hyperspaces. The theoretical basis of the system is the idea of vocabulary matrix (Miller 1986). Miller uses the term word form to refer to the actual phonological or graphological expression, and word meaning to refer to the lexicalized concept that is expressed by means of the word form. He argues that any paper dictionary can be reduced to the projection of its meanings onto lexical entries, the convergence of which can be represented by a matrix. The columns of this matrix would contain all of the word forms of a language, while the rows would contain all of its meanings. Each cell is thus the convergence point of a word form and a word meaning. Table 1.

Vocabulary matrix Word forms

Word meanings

FI

F2

M, M2 M3

Ε,.,

ELZ E2.2

MN

F3

F„

E3.3 EN.N

As shown in Table 1, an entry (E) thus signifies that a word form (F) can be used (in the appropriate context) to express a particular meaning (M). The problem with the vocabulary matrix is the same one that plagues all models of conceptual organization. In the normal order of things, natural language speakers have no difficulty in freeing themselves from language and working directly with concepts. However, when it is a question of trying to implement a

22

The lexicon and linguistic theory

system of conceptual organization on the computer, there is no escape from language. Conceptual representation in WordNet is based on the vocabulary matrix. Each row of the matrix constitutes a set of synonyms (synset), and is assigned a number, which thus identifies the concept designated by the set of word forms. The synsets do not explain what the concept is, but only signify that such a concept exists. Furthermore, in contrast to traditional thesauri, a synset does not have a headword. Synonymy is thus the most basic lexical relation in WordNet, but certainly not the only one. Lexical relations such as antonymy, hyponymy, and meronymy are also reflected. Given that semantic relations are meaning relations, and that meanings are presented by means of synsets, such relations are conceived of as pointers between synsets. WordNet has been used for many different applications, such as word sense identification, information retrieval, selectional preferences of verbs, and lexical chains. One of the reasons for its success is the fact that it is much easier to deal with extensive lexical databases rather than with complex algorithmic operations. In other words, computational linguists have found that it is preferable to store huge amounts of pre-computerized information rather than gradually synthesizing and computerizing this information through a set of complex algorithms. It is no accident that this realization also has coincided with developments in linguistic theory, such as the progressive elimination of transformations in Chomskyan linguistics, an enriched lexicon component, and the development of more semantically-oriented linguistic theories such as Simon Dik's FG.

1.3. Lexicon and grammar This rise in prominence of the lexicon has logically generated a parallel debate about how lexical information should be represented in linguistic theory. One of the issues is precisely what type of information a lexical entry should contain, and how this information should be configured. A variety of proposals have been made advocating fully-specified lexical entries with a rich relational structure. Despite certain differences, they all coincide in the use of

Lexicon and grammar

23

predicate argument structure in the representation of meaning. Each argument appears with its corresponding role (Agent, Patient, Instrument, etc.) which describes the way in which the argument participates in the state of affairs designated by the predicate. The argument slots also have selection restrictions, which describe the semantic and pragmatic characteristics of possible arguments. This naturally places constraints on the entities that can fill the slots. Where linguistic models diverge is in the type of meaning components used in the meaning representation of a verb, the inventory as well as the explanatory force of semantic roles, and finally the explicitness of selection restrictions.

1.3.1. Event structure and lexical representation Levin—Rappaport (1996: 488) classify approaches to lexical representation as either role-centered (e.g. GB) or predicate-centered (e.g. Role and Reference Grammar (RRG), Dik's FG). We have added a third category to this classification. This would include constructionist approaches such as Construction Grammar (CG), which have a non-modular view of language (Goldberg 1995; Kay 1997).

1.3.1.1. Role-centered approaches Role-centered approaches have a universal set of thematic roles, such as AGENT, RECIPIENT, GOAL, and THEME, which capture generalizations concerning the relation between syntax and semantics. The arguments of a verb are classified according to the roles that the participants associated with each of the arguments play in the state of affairs (SoA) depicted in the clause. (5)

break (Agent, Patient)

For example, a two-place predicate like break has an Agent and a Patient (5), and these arguments are mapped onto the syntactic properties of the verb by a set of rules. In Government and Binding

24

The lexicon and linguistic theory

Theory (GB), for example, such mapping rules are absolute and are stated in the following terms. (6) (7)

Agents cannot be mapped to syntactic objects. Agents cannot be mapped to NPs dominated by the VP at Dstructure.2

Alternatively, instead of absolute mapping rules, thematic roles can be arranged "relativistically" along a hierarchy of prominence which determines argument selection. Grimshaw (1990: 2) proposes the following proto-argument structure: (8)

(Agent (Experiencer (Goal/Source/Location (Theme / Patient)

At one end of the hierarchy in (8) the most salient element is the Agent, followed by Experiencer. At the other end, the Patient/Theme has several roles at the same level (Goal/Source/Location). As the first argument type of an event is the most salient, it receives the semantic function of Agent, and is then mapped onto the grammatical function of Subject. One problem with role-centered approaches in general is the fact that they reduce the meaning of a word to a set of thematic roles, which in most cases have no precise definition. When verbs are defined in this way, kill and break have the same meaning because their thematic roles coincide. Moreover, there does not seem to be any sound criteria to establish the kind of entity that can fill each role. Still another problematic aspect is that it is often necessary to enlarge the inventory of roles in those cases where none of the labels fits the entity in question. Obviously, this type of ad hoc extension is not an adequate solution when the inventory does not include the appropriate role.

1.3.1.2. Predicate-centered approaches In predicate-centered approaches, such as Rappaport and Levin's lexical templates and Role and Reference Grammar (RRG), predicates are decomposed into a set of primitive elements, which

Lexicon and grammar

25

form part of a well-defined semantic metalanguage. In this regard, Jackendoff (1990: 37-41) shows the viability of such a procedure, and convincingly argues in favor of systems of lexical decomposition. The most obvious difficulty lies in the specification of an appropriate set of primitives capable of capturing syntacticsemantic generalizations. Where predicate-centered theories differ is in the type of semantic primitives used as well as in the formalism adopted for their representation. Their goal is the development of a system of lexical representation that permits the formulation of semantic generalizations that are valid for a word as well as a class of words, and which are also relevant to syntax. It presumably allows the concentration of multiple verb senses within a meta-lexical representation and is based on a decomposition system, which makes use of primitives and constants. In this type of approach, the assignment of thematic roles depends on the meaning components as well as the event structure depicted in the lexical representation. In certain models, this has led to the development of macro-roles, which are used as mapping devices between the information contained in the lexical representation and the syntactic properties of the predicate. The resulting representations encode syntactically relevant aspects of meaning and act as an interface between the grammatical and the cognitive components.

1.3.1.2.1.

Rappaport and Levin's lexical templates

One way of grouping verbs into semantic classes is on the basis of syntactic alternations (Rappaport—Levin 1988, 1998). B. Levin (1993) analyzes the semantic characteristics of English verbs by examining their grammatical behavior. For example, she argues that the alternations typical of a verb like whistle are ultimately determined by its membership in the semantic class of verbs of sound emission: (9)

a. b. c. d.

Norm Norm Norm Norm

whistled. whistled at the dog. whistled a tune. whistled a warning.

26

The lexicon and linguistic theory e. f. g. h. i.

Norm whistled her disgust. The boy whistled that the meeting was over. The rope whistled through the air. The air whistled with bullets. John whistled to the boy to come.

By relating the full set of the meanings of a verb to its semantic class, she claims that it is possible to account for principles that govern variation in meaning. In this case, the whole range of meanings in (9) is shared by all of the verbs of sound emission. This signifies that verbs with shared meaning components also share a similar set of syntactic frames. In this regard, Rappaport—Levin (1998) develop a theory of semantic representation in which the meaning of a verb is divided into structural and idiosyncratic aspects. Structural aspects of meaning are the criteria upon which semantic classes of verbs are established that are relevant to argument realization. They correspond to event categories, such as states, achievements, accomplishments, and actions. In contrast, idiosyncratic aspects are devoid of any grammatical significance in terms of argument projection, and capture the differences that distinguish one verb from others within the same class. These two aspects partly coincide with the notions of primitives and constants used in decompositional approaches to meaning. Structural aspects of meaning are identified with a set of primitive predicates called lexical semantic templates, whereas idiosyncratic aspects of meaning are similar to constants. The meaning of a verb is the association of a constant with a corresponding semantic template. Consequently, each semantic class of verbs has a lexical template from which it is possible to derive the different predicates through the instantiation of the constant. (10)

a. b.

[[x ACT ] CAUSE [ BECOME [ Y ]]] [[x ACT ] CAUSE [ BECOME [ Y ]]]

For example, causative change of state verbs share the representation in (10a) and the various members of the class are derived by specifying the slot represented by the constant . When this slot is filled by more specific information, it results in the representation of

Lexicon and grammar

27

individual predicates that belong to that class. Example (10b) is the lexical template for melt. Rappaport—Levin (1998:108) distinguish the following inventory of lexical templates: (11) a. [x ACT ] (activity) b. [x ] (state) c. [BECOME [x ] (achievement) d. [x ACT ] CAUSE [BECOME [y ]]] (accomplishment) e. [x CAUSE [BECOME [y ]]] (accomplishment) This inventory is closed in contrast to the open-ended nature of constants (e.g. , ). These constants are derived from a fixed set of types such as state, thing, manner, place, etc. The ontological type of a constant determines its association with one of the lexical templates through canonical realization rules such as (12ab): (12)

a. b.

manner [x ACT ] instrument [x ACT ] (Rappaport—Levin 1998: 109)

Each rule is the pairing of a constant (manner, instrument) with an event structure. The associations of event structures and constants as triggered in the appropriate canonical realization rules form the basic meaning of a verb. The lexical entry of a verb is the name given by the constant together with the meaning represented by the event structure. For example, sweep has the following representation: (13)

a. [x ACT Y. ] b. [x ACT 1 CAUSE [BECOME [y ]]]

—» He swept the floor.

He swept the floor clean. c. [x ACT ] CAUSE -> He swept the dust [BECOME [z ]]} under the rug.

28

The lexicon and linguistic theory

Sweep is a constant, which is derived from another constant indicating a type of surface contact. It is associated with an activity event structure through the canonical rule (13a), which represents the activity sense of the verb. However, Rappaport—Levin (1998) argue that in order to account for multiple verb senses, it is necessary to create a mechanism that allows basic structure templates to be expanded into more complex ones. In this sense, they propose Template Augmentation, which explains the event structures (13b) and (13c) that represent the two other senses of the verb. As shall be seen, this proposal is similar to that of RRG in many respects because it is able to accommodate the full range of verb senses into a unified lexical entry by deriving them from a small set of templates.

1.3.1.2.2.

Role and Reference Grammar: logical structures

In RRG, states of affairs can be states, events, processes or actions, each of which is associated with a corresponding class of predicates (states, achievements, accomplishments, and actions). Verbs are thus divided into aspectual types or according to their Aktionsart (Vendler 1967).3 Each of these four basic verb classes has a causative counterpart: (14)

a. b. c. d.

The tiger scares Ron. (Causative state) John popped the balloon. (Causative achievement) The heat melted the ice. (Causative accomplishment) The girl bounced the ball around the room. (Causative action)

Examples (14abcd) show the different states of affairs in RRG, their corresponding verb class, defining features, as well as an examples of predicates belonging to each class:

Lexicon and grammar Table 2.

29

SoAs in RRG

SoA

Aktionsart

Defining parameters

Examples

Situation

State

[+static], [—telic] [-punctual]

know, have, be sick, love

Event

Achievement

Process

Accomplishment

[-static], [+telic] [+punctual] [-static], [+telic] [+punctual]

pop, explode, shatter melt, freeze, dry, learn

Action

Activity

[-static], [-telic] [+punctual]

march, walk, roll, think

In order to capture the distinctions established in this typology of states of affairs (Table 2), RRG adopts a system of lexical decomposition, which associates each verb with a logical structure (LS), composed of predicates and variables. Predicates appear in boldface and are followed by a prime. They are an integral part of the semantic metalanguage used in decomposition, whereas variables (x, y) represent words from any language. (15) (16) (17)

shatter'(x) hear' (x, y) do' (x, [walk'(x)]

Example (15) is a one-place state predicate. It is represented by a primitive element shatter' and a variable (x), which can be any word of a language that fits the selection restrictions for this slot (e.g. glass, window, mirror, etc.). Example (16) is a two-place perception predicate, composed of a primitive element hear' and two arguments (x, y). Finally, (17) is an activity predicate (activities are signaled by the primitive do'), which takes a primitive element walk' and one argument (x). State and activity predicates are primitives, whereas achievements and accomplishments are defined in terms of these primitives plus the addition of operators: INGR (for achievements) and BECOME (for accomplishments). The resulting inventory of general

30

The lexicon and linguistic theory

lexical representations is summarized in Table 3 (Van Valin—LaPolla 1997: 109): Table 3.

Typology of RRG lexical representations

Verb class

Logical structure

State

predicate' (x) or (x, y)

Activity

do' (x, [predicate' (x) or (x, y)]

Achievement

INGR INGR

Accomplishment

BECOME BECOME

Active accomplishment

do' (x, [predicatei' (x, (y))] & predicate2' (z, x) or (y)

Causative

α

predicate' (x) or (x, y), or do' (x, [predicate' (x) or (x, y)] predicate' (x) or (x,y), or do' (x, [predicate' (x) or (x,y)]

CAUSES

BECOME

ß where α, ß are LS of any type

In this system, only primitive elements (states and activities) require a detailed definition because other predicates can be derived from them through the use of operators. For example, kill can be derived from the primitive dead by using the operators CAUSE and BECOME, along with the activity marker do' (x, 0): (18)

kill

[do' (x, 0 )

CAUSE [BECOME

dead' (y)]

Van Valin—LaPolla (1997: 111) also postulate representations for entire semantic sets of verbs. For example, the class of speech verbs has an underlying representation from which it is possible to derive the more specific representations of each member through the assignment of different interpretations to the variables in the logical structure. In the same way as Rappaport and Levin's lexical templates, this type of representation also concentrates various lexemes into a single meta-entry, (cf. Van Valin—LaPolla 1997: 117-118):

Lexicon and grammar

(19)

31

Verbs of saying: do' (x, [express' (a).to.(ß).in.language.(T)' (x,y)])

The more general entry for speech verbs in (19) is the basis for the entries of more specific speech verbs. For example, due to its causative and telic character, tell has the representation in (20): (20)

do' (x, [express' (α ).to.(ß).in.language.(Y)' (χ, y)]) [BECOME aware.of' (y, z)] where y = ß and ζ = α

CAUSE

Thematic relations in RRG are derived from the argument positions in logical structures for a verbal predicate, and thus have no independent status. Each argument position defines a thematic relation, and the labels mark the first argument (x), the second argument (y) or both arguments (x, y) in the logical structure. Thematic roles are used to indicate the subclass of the state or activity predicate that the argument co-occurs with. Accordingly, POSSESSOR means the first argument of a two-place state predicate of possession; EXPERIENCER, the first argument of a one-place state predicate of experience; OBSERVER, the first argument of a two-place activity predicate of directed perception; and PERFORMANCE, the second argument of a twoplace activity predicate of performance. Thematic roles are arranged along a semantic continuum with Agent (volitional, instigating) at one end, and Patient (non-instigating, maximally affected) at the other, whereas the remaining groups are classed according to how agent-like or patient-like they are (Van Valin—LaPolla 1997: 116). Although logical structures have the advantage of capturing generalizations across different lexemes within the same verb class, they do not account for factors such as axiological evaluation, or the metaphoric extensions of lexemes to other domains. No mention is made of the extent to which logical structures encode conceptual structure or the criteria that determine which elements fall within the group of primitives and which do not.

32

The lexicon and linguistic

1.3.1.2.3.

theory

Construction Grammar

Construction Grammar (CG) contrasts with RRG in that it conceives grammar as composed of conventional associations of form and meaning, and in this sense can be considered an extended lexicon, or a structured "construction" (Goldberg 1996: 8). Within this framework, lexicon and grammar are unified to the extent that morphemes, lexemes and clause structures are described within the wider framework of a construction (or form-meaning correspondence). Goldberg (1995: 25-31) rejects the idea of defining words in terms of lexical decomposition since this does not adequately capture all of the features contained within the meaning of a word. She claims that lexical decompositions of the type X ACTS, which would be the canonical formula for a verb like sneeze, do not adequately explain an expression like (21). (21)

Paul sneezed the napkin off the table.

In (21) sneeze designates an activity where forceful expulsion of air is required. In order to account for this meaning, a richer inventory of features is obviously necessary. Goldberg convincingly demonstrates that certain aspects of the meaning of a word, which are not visible to syntax, play an important role in a system of lexical representation. CG draws on Fillmore's Frame Semantics, and defines predicates relative to a frame, which is a repository of both lexical and encyclopedic structure. The meaning of a word is analyzed within the larger framework of the construction it occurs in, and consequently, syntactic-semantic mappings are not based on predicates. Constructions are semantically- based, and their format is identical to that used in decompositional models: (22)

X CAUSES Y TO RECEIVE

Ζ

Example (22) shows a ditransitive construction with x, y, and ζ as the argument roles. Arguments are said to be profiled if they are linked

Lexicon and grammar

33

to a direct grammatical function (SUBJ, OBJ, or OBJ2). Profiled arguments are represented in boldface: (23)

CAUSE-RECEIVE

The semantics of the ditransitive construction is represented in (23). Three arguments are profiled, an agent, recipient, and patient. The association of this verb meaning with a construction is the following: (24) Sem R: instance Means

CAUSE-RECEIVE

OBJECT

OBJECT2

PRED

t Syn

SUBJECT

CAUSE-RECEIVE specifies the semantics associated with this construction, and PRED is the predicate which

instantiates it. Obligatory roles are indicated by a solid line, and optional roles with a dotted line. The variable R specifies the means by which the verb is integrated into the construction. The representations of all of the verbs in the same class can be derived from (24). For example, mail is a type of causing to receive, which has a three-argument potential. However in (25), only two arguments of its three-argument potential are realized:

34 (25)

The lexicon and linguistic theory Paul mailed a letter. Sem R: instance Means

CAUSE-RECEIVE

MAIL

entailer

mailee

mailed>

Τ Syn

ν

SUBJECT

I

OBJECT

i

OBJECT2

Mail instantiates the ditransitive construction Χ CAUSES Υ TO RECEIVE z. All three of the participant roles are profiled since each is associated with a different grammatical relation. However, in this case, the recipient and the mailee are joined by a dotted line because this argument is not syntactically realized in (25). Such CG constructions are the basic representations by means of which multiple verb senses are accounted for (Goldberg 1995: 21). They avoid the creation of a separate lexical entry for each syntactic configuration. Goldberg argues that this does not mean that the meaning of a verb is devoid of function and form, but rather that constructions are interrelated to verb meanings in such a way that there is always an interaction between the two. In a parallel way, the FLM places great emphasis on verb meaning, and also adopts the notion of schema for the lexical representation of predicates, which has certain parallelisms with the notion of construction presented here. However, unlike CG, the syntactic properties of FLM predicates are projected from the lexicon.

1.4.

Summary

The centrality of the lexicon in contemporary linguistic theory has led linguists to take a closer look at lexical items and the information that they contain. Advances in lexicology, lexicography, psychology, and computational linguistics offer a partial explanation why many linguistic models have adopted a lexicalist paradigm. This has led to

Summary

35

a debate on how information should be configured in a lexical entry. Though many different proposals have been put forward, all of them coincide in basing representations on predicate-argument structure. Basically, approaches can be classified as role-centered, predicate-centered or constructionist. The different focuses evidently affect the type of lexical representation. In line with this, we have taken a closer look at Rappaport and Levin's lexical templates, Role and Reference Grammar, and Construction Grammar, which present evident parallelisms with both the FG model and FLM. Chapter 2 gives an overview of Simon Dik's Functional Grammar. Of the different grammatical models, Dik's FG is one of the most lexically-based, but as yet, no type of lexical macrostructure has been proposed within this framework. The Functional-Lexematic Model is a way to enhance the FG lexicon by giving it a more meaningful configuration. In the chapters that follow, we shall describe in greater detail a lexicon structured according to FLM principles. Chapter 3 describes the internal structure of the FLM lexicon in terms of its paradigmatic and syntagmatic axes. The paradigmatic axis configures verbal predicates in lexical domains and subdomains, which are highly informative as to the semantic mapping of perceived reality. The syntagmatic axis encodes the syntactic potential of each predicate in terms of FG clause structure. Chapter 4 deals with the way paradigmatic and syntagmatic descriptions interact. In this regard, we claim that verbs classified within the same lexical domain not only share the semantic content of their meaning definitions, but also have similar syntactic behavior. On the basis of this, we have proposed a set of synsem parameters that are used to exemplify the extent to which syntactic projections are semantically motivated. Chapter 5 develops what we have termed the cognitive axis, the third axis of a FLM lexical structure. Lexical subdomains are theoretical constructs in which sets of semantic, syntactic and pragmatic constituents converge. Each is the representation of a predicate conceptual schema, which is in itself a microstructural representation of a lexical grammar.

2.

Simon Dik's Functional Grammar

2.1. An overview Simon Dik's FG is representative of the functional paradigm in linguistic theory, in which language is seen as an integrated part of the communicative competence of the Natural Language User. In line with this, Dik (1997a: 3) states that language is "an instrument of social interaction among human beings, used with the intention of establishing communicative relationships". Precisely for this reason, he sees pragmatics as the general framework within which semantics and syntax must be studied. Semantics is regarded as instrumental with respect to pragmatics, and syntax as instrumental with respect to semantics (Dik 1997a: 8). FG categorically rejects the formal conception of an autonomous syntax. In fact, one of its basic methodological tenets is that the properties of clause structure are predictable from the semantics of predicates. As has been shown in the first chapter, this premise is now implicit in many other grammatical models as well, with the exception of those with a clearly constructivist orientation. The fact that FG conceives language as an instrument of communication integrated into a more general activity of verbal and social interaction means that linguistic phenomena are explained in terms of their functionality with respect to the ways they are used and to the ultimate purposes of these uses (Dik 1997a: 3). Linguistic expressions are represented in terms of underlying clause structures, which encode the syntactic, semantic, and pragmatic functions holding between the participants in the state of affairs. The modular organization of FG differs from that of cognitive models such as CG, in which lexicon and grammar form a continuum. As shown in Figure 2, FG has three modules: (i) the fund; (ii) the representation of underlying clause structures; (iii) an expression rule component.

38

Simon Dik's Functional Grammar

Figure 2. Outline of FG modules The fund consists of a lexicon and two rule components. In FG the lexicon is the ultimate source of the syntactic and morphosyntactic properties since the underlying clause structure is derived from predicate frames through the addition of operators. The FG lexicon contains basic terms (terms referring to entities in some world), as well as basic nominal, adjectival, and verbal predicates. Each predicate is provided with a predicate frame, a notational device that formalizes each predicate in terms of arguments, semantic roles, selection restrictions, and meaning definition. In contrast, derived predicates are formed through predicate formation rules, which are meant to introduce changes over a given predicate frame. For example, agentive nouns can be derived from the predicate frame corresponding to that verb through the application of the following predicate formation rule:

An overview (26)

39

Agent Noun Formation input: pred [V, +contr] (xi)Ag/Pos · • .(xn) [n>l] output: {Ag pred} [N] (xi) 0 .... (x„) meaning: "a person who has the property of being (habitually) involved in the action of pred-ing" (Dik 1997b: 3).

The input to this rule (26) must be a [+contr] predicate (whose first argument is either Agent or Positioner). Its quantitative valency is undefined as there is no specification as the number of arguments it takes. The output is a nominal predicate that has the same quantitative valency as that of the input, and which is characterized by the semantic function Agent. Predicate formation rules can account for other changes in a predicate as well, such as the extension or reduction of its valency, a semantic function shift in its arguments, semantic modification, or a change in its syntactic category (De Groot 1987: 14). As a result, this type of rule offers an explanation for certain syntactic alternations. Another possible way of dealing with this type of phenomenon would be through a unified semantic description of each predicate. This offers the advantage of giving the lexicon a more economic organization. With regard to the FG conception of clause structure, Dik (1997ab) and Hengeveld (1987, 1989ab, 1990a), propose a multilayered hierarchical clause. This proposal is similar to the layered structure of Role and Reference Grammar, though with a clearly semantic orientation. In fact, Van Valin (1990:199) observes that the layers as envisaged in Functional Grammar respond to a semantic or pragmatic conception of the clause, and thus none appears to be syntactically founded. Table 4.

The structure of the clause in FG

Variable Restrictor

Clause unit

Reference

Ε X Ε χ

(E,:[Clause](Ei)) (Xi:[Proposition](Xi)) (ei: [predication] (e,)) (Xl:[PredN](xi))

Speech act Potential fact State of affairs Individual

Clause Proposition Predication PredN

40

Simon Dik's Functional Grammar

As shown in Table 4, each of the layers or levels of semantic organization of the FG clause refers to a particular entity and is represented by a structural unit (Hengeveld 1989a: 130).4 The underlying structure of linguistic expressions consists of four hierarchically ordered layers, each designating a different type of entity. The first three types of entity correspond to the classification made by Lyons (1977: 442-447). First-order entities are physical objects that can be located in space and time. Second-order entities refer to processes, events, and states of affairs, which are said to take place rather than to exist; and third-order entities are abstract entities outside space and time, which can be asserted in terms of their truth. FG even extends Lyons' classification to include a fourth-order entity, referring to a speech act, and which can be evaluated in terms of its felicity (Austin 1962). With regard to the representation of utterances, two main levels are distinguished in the analysis of a clause, the representational and the interpersonal level, each of which is modified by different kinds of operators and satellites. In making this distinction, Hengeveld (1990a) is indebted to Bühler (1934), Foley—Van Valin (1984), and also to Halliday (1994). Halliday's metafunctional hypothesis distinguishes a textual, interpersonal, and ideational function. The textual function is similar in some ways (though not equivalent) to Dik's pragmatic function assignment; the interpersonal function bears a certain resemblance to Hengeveld's interpersonal layer (though with certain obvious differences); and the ideational function corresponds to the representational level.5 Generally speaking, the higher (or interpersonal) level is concerned with those linguistic means used by the speaker to produce a certain effect in the addressee, whereas the lower (or representational) level is concerned with the linguistic means used to provide the addressee with a description of a state of affairs. The representational level has two layers: (i) a predicate frame (Layer 1), which specifies a predicate plus the number of arguments required, (x); (ii) a predication (Layer 2), which describes the state of affairs in terms of its time, frequency and actuality (e). The interpersonal level is also composed of two layers: (i) the propositional layer (Layer 3), which represents the sentence as a possible fact that can be verified or denied (X); (ii) an illocutionary frame (Layer 4), which codifies the speech act (E):

An overview

41

(27) The representation of utterances (En [π4ΠΧ:σ4 (S) (Α) (π3Χι: [, J (Χι): σ 3 (Χχ))] (Ει): σ 5 (Ε1)) π2βι: fripredß: σι (axi:pred N (χΟ) ...(χ η )] (ei): σ 2 (ei) Underlying clause structures thus have the form in (27). This representation corresponds to Hengeveld's (1990a: 12) proposal, which differs from Dik's in the status of ILL. Dik (1997a: 292) treats ILL as an operator (π*) of E, whereas Hengeveld treats it as an abstract predicate which specifies a relation between S and A and some prepositional content. The symbols used in FG clause representations are listed in Table 5: Table 5.

Symbols used in FG representations

Layers and frames Xi pred ei Xi ILL

Term Predicate frame Predication Proposition niocution

Operators

Satellites

term operators predicate op. πα predication op. π3 proposition op. 7C4 illocution op.

σι σ2 σ3 σ4 σ5

Ω πι

predicate satellites predication satellites proposition satellites illocution satellites clause satellites

At the interpersonal level the clause consists of a proposition inserted within an illocutionary frame (ILL) together with a set of operators modifying the proposition (π3) and the illocutionary frame (714). The proposition at the representational level consists of a predication plus a variable (e), which signals the space and time parameters together with predication operators (π2). The predication in turn consists of a predicate, its arguments as well as a set of predicate operators (πι). Each layer is provided with its own set of operators and satellites. Operators capture possible modifications and modulations effected at the relevant layer by grammatical means, whereas satellites (σ) capture possible modifications brought about by lexical means.6 Operators can be of the types shown in Table 6:7

42

Simon Dik's Functional Grammar

Table 6. πι π2 π3

Jt4

Overview of operators

Operators that specify additional properties of the state of affairs [e.g. imperfective/perfective, phasal aspect, predicate negation] Operators concerned with the setting of the SoA [e.g. tense, quantificational aspect, objective modality, polarity] Operators that specify the grammatical means through which the speaker expresses his commitment towards the validity of the propositional content [e.g. epistemological modality (subjective and evidential modality)] Operators concerned with the communicative strategy of the speaker [e.g. procedures that mitigate and reinforce the illocutionary force of the speech act]

Hengeveld (1988:142) formulates the hypothesis that operators of the class 7U„ may impose restrictions on the selection of operators of the class 7Vi. This explains why the complement of see in (28), which in one of its readings designates a second-order entity (e), cannot occur with a subjective modality adverb like certainly. (28)

?I saw him certainly walk down the street.

Ultimately, this hypothesis is a determining factor in the explanation of the form and behavior of certain complements. Satellites codify the properties and setting of an SoA, validity of propositional content, as well as the communicative strategy of the speaker:

An overview Table 7. σι σ2 σ3 σ4 σ5

43

Overview of satellites Additional properties of the SoA [e.g. manner, speed and instrument] setting of the SoA [e.g. time, place, frequency and probability satellites] Validity of the prepositional content [e.g. epistemological modality, attitudinal and evidential satellites] Communicative strategy of the speaker [illocutionary force satellites] setting of the utterance; location of utterance within the context of discourse

Two modifications have been made to the FG layered model, and both play a key role in the delineation of complementation patterns. The first arises from Bolkestein's (1990, 1992) observation that certain indirect speech reports and complements of thought, which designate third order entities, can occur with illocutionary satellites, thus invalidating to a certain extent Hengeveld's hypothesis regarding the scope of application of operators: (29)

Benjamin said that since the lights are off, Peter must have left.

In (29), the third-order entity and goal argument of say (i.e. Peter must have left) should only be modified by third-order operators, or impose restrictions on second or first-order entities. However, in this case, Bolkestein observes that it is modified by a fourth-order satellite (since the lights are off). She thus suggests that there should be an extra layer between Ε and X, which would be symbolized by (F). The second proposal stems from the work of Keizer (1992) and Hengeveld (1992), who, drawing on Dik's addition of a fifth layer to the clause model, propose the addition of a predicate variable (f) at the representational level. The resulting model has the format in (30):

44

Simon Dik's Functional Grammar

(30)

(Ei:[Fi:ILL(Fi))(S)(A)(Xi:

](Xi))l (Ei))

(e i: [(fi: Predß (fO) ( Xl )] (eO) In this new model, a speech act (E) is performed with an illocutionary force (F), in which a propositional content (X) is transmitted by a speaker (S) to an addressee (A). Moreover, the propositional content consists of a predication (e) in which several individuals/entities (x) are engaged in a relation or have a property (f) (Hengeveld 1996:120): (31)

Frankly, Peter certainly sent the book yesterday.

The FG representation for (31) is the following: (32) (E,: [F,: DECL (FO (σ4: frankly) (S) (A) (X i: [ ril (Χι): (σ3: certainly) (Χ,))] (Ει): σ 5 (EO) Past ei: [send [V] (xj: Peter (eO: (σ2: yesterday}nme

[N]) Ag SubjTo P

(dlx2: book

[N])GoObjFoc]Action

If we examine this representation in (32) step by step, we can see how it has been built up. The predicate frame of send is at the innermost layer: (33)

send [V] ( X] : human)Ag

(X2)GO

The insertion of terms in the slots of each of the argument variables produces the following nuclear predication: (34)

[send [V] (xi: Peter [N])Ag (dlx2: book [N])Go]Acti,

This nuclear predication in (34) is located in time by means of an operator Past and a σ 2 satellite yesterday, and thus gives rise to the extended predication:

An overview (35)

45

Past ei: [send [V] ( X l : Peter [ N ] ) ^ (dlx2: book [N])Go]Action (ei): (σ 2 : yesterday)^

The assignment of syntactic functions supplies the means for interpreting the predication. In this case, the subject function is assigned to Peter, the agent, whereas the object function is assigned to book, the goal argument: (36)

Past ei: [send [V] (xf Peter [N])AgsUbj (dlx2: book [N])Goobj]Action (ei): (σ 2 : yesterday)^

This core predication acquires the status of proposition, adding the subjective attitudinal adverb, certainly (37)

[Xi: [Past ex: \[send[V] (xi: Peter [N])AgSubj (dlx2: book [N])GoObj]Actio„(e,): (σ2: yesterday)rime] (σ 3 : certainly)Attitude (Xi)]

This proposition (37) becomes a clause structure through the addition of its illocutionary force, which in this case, is that of a declarative sentence plus the 04 frankly. (38)

[Ei: DECL [Xi: [Past ei: [send [V] (xi: Peter [N])Agsubj (dlx2: book[N])Goobj] Action(ei): (σ 2 : yesterday)^*] (σ 3 : certainly)Abtäte (Xi)] (a4:frankly)u anner (Ei)]

The assignment of pragmatic functions places the utterance within the wider setting of the discourse in which it occurs by indicating the informational status of the different participants:

(39)

[Ei: DECL [Xi: [Past ei: [send [V] (x,: Peter [N])AgSubjTop (dlx2: 6oo£[N])GoObjFoc]Action (ei): (σ 2 : yesterday)Tims] (σ 3 : certainly)Attitude (Xi)] (a+Jrankfy)M anner (Ei)]

As example (39) is decontextualized, there is no discourse setting satellite (05), but if such a satellite were part of the text, it would be

46

Simon Dik's Functional Grammar

added at this point. Finally, underlying clause structures are mapped onto the actual linguistic expression by means of a set of expression rules. The expression rule component serves as a bridge between the structure and the actual expression of the clause. Expression rules are of three types: (i) trigger rules (form of constituents) (De Groot 1989: 1314) (ii) placement rules (order of constituents) (iii) prosodic rules (prosodic features of the clause) The expression rule component would be the most directly affected by a structured lexicon because information currently explained by expression rules could be included as part of the lexical representation of a semantic class of predicates. A reorganization of the lexicon component in FG offers an enhanced representation of lexical structure, as well as a new conception of the interface between syntax and semantics.

2.2. FG lexical representation Despite the obvious centrality of the lexicon in FG, this component has not received as much attention as others such as the pragmatic component, expression rule component, and the layers within the organization of the clause.8 As a result, the basic organizational parameters of the lexicon have not been modified since they were first formulated by Dik (1978a). The modifications introduced by Dik (1997a: 77-104) are slight, and primarily concern typographical conventions. However, the essence of the theory is much the same as in the earlier versions of the model. Predicate frames and meaning definitions are the basis of lexical representation in FG. Like RRG representations and Rappaport and Levin's lexical templates, predicate frames are formalizations of predicate argument structure, but they differ from these models in that abstract predicates are not used as the basis for semantic decomposition. FG also differs from the predicate-centered theories that we have described in its linking algorithm. In this respect, FG does

FG lexical representation 47 not have configurational representations of utterances because the structure defined in the predicate frame is the underlying structure.9 An important issue regarding predicate frames is the fact that little or nothing is said about the way they interact with meaning definitions. Dik (1997a: 97) appears to rule out the possibility that meaning definitions have a role in determining syntactic expressions. However, in our opinion, this is highly debatable, and in this respect, our proposal is closer to that of RRG.

2.2.1. Representational parameters In FG, lexical items are structured representations, encoded in predicate frames, which encapsulate information concerning predicate variable, predicate form, syntactic category, redundancy rules, quantitative valency, qualitative valency, selection restrictions, and meaning definitions.

2.2.1.1. Predicate variable Drawing on the work of Keizer (1992) and Hengeveld (1992), Dik (1997a: 82-83) proposes a predicate variable for each predicate. The evidence to support this distinction comes from the fact that the properties designated or the relations referred to by the predicates can actually function as antecedents in anaphoric expressions. For example, according to Hengeveld (1992: 31), the property designated by the predicate smart serves as the antecedent for a relative pronoun as shown in (40): (40)

Peter is smart, which you are not. (e,: [fi: smart [A] (f,)) (x,: Peter [N] ( Xi )) 0 ] (eO) (Neg ej: [(RfO (dl Xj : [-S, +A] ( Xj )) 0 ] ( ej ))

This then signifies that each predicate has a predicate variable associated with it as part of its lexical entry: (41)

(fj: give) [V] (xj: )Ag

(X2)GO

(X3: )Rec

48

Simon Dik's Functional Grammar

In this connection, Dik (1997a: 83) observes that the encoding of such a variable is not always necessary, and thus we have used a more simplified notation: (42)

give [V] (XI: ) Ag (X2)GO fe ) Rec

2.2.1.2. Predicate form The predicate frame contains information concerning the phonological representation of the predicate together with specifications of tone and characteristic accent position, if applicable. With regard to those forms that cannot be predicted by predicate formation rules, and thus must be stored in the lexicon, Dik (1997a: 84) claims that they are stored in paradigms, which are related to the predicate: (43)

paradigm [sing] = {Past sang, Past Participle sung}

Such sets are evidently based on phonological and syntactic variation. Our proposal extends this primitive notion of paradigm to include both semantic and syntactic components.

2.2.1.3. Syntactic category Information as to syntactic category is given in the form of a label (e.g. Ν, V, Adj, etc.). In FG, basic predicates can be verbal (V), adjectival (A) or nominal (N). Dik (1997a: 194) also suggests that categories should be defined according to the different functions they perform in the predication. This view contrasts with the more simplistic systemic notion of predicator captured on purely syntactic grounds, such as the verbal complex minus certain verbal operators. In this sense, the adjectival and nominal expressions are conceived as bearing semantic functions concerned with attribution or identification, and are thus not assimilated in any way to verbal predicating elements (Butler 1990: 26). The FG lexicon does not contain other categories such as prepositions, articles, and

FG lexical representation 49 conjunctions because their form is determined by expression rules, and as a result, they are derived through the use of operators. In addition, Dik (1997a: 84) introduces variations concerning predicate type. He argues that certain properties of the type of predicate can be read off from the information in the predicate frame. However, other type elements are not derivable, and thus should be encoded together with the label indicating the type of predicate. For example, in the case of nouns the predicate frame contains information regarding gender in order to account for the correct reflexivization: (44)

boy [N, masc]

This representation (44) accounts for the fact that the boy requires a masculine reflexive: (45)

The boy shaved himself.

Dik also makes an initial proposal for the inclusion of semantic information, which also influences syntactic form. For example, since only predicates of movement can be combined with directional satellites, he suggests the addition of a feature [move] to the predicate frame: (46)

walk [V, move, atelic] (xj: )Ag

This proposal (46) seems to point the way to the possible organization of predicates into coherent semantic classes, which encode both syntactic and semantic regularities. As shall be seen, our study takes Dik's proposal further in this direction.

2.2.1.4. Redundancy rules Dik (1997a: 86) claims that some of the information in a predicate frame can be encoded in terms of a set of lexical redundancy rules. For example, the interaction of SoAs and semantic functions yields the following rules:

50

(47)

Simon Dik's Functional Grammar

a. b.

All action predicates have the Type features [+control] and [+ dynamic], All action predicates have a first argument with semantic function Agent.

This means that predicate frames can be further specified to include more information: (48)

(fi: give [V, action, +contr, +dyn] (xj: )A g (x2)go (Χ3: )Rec

Although the rules in (47ab) show that the predicate frame in (48) contains redundant information, Dik (1997a: 86) claims that such redundancy structures can be useful at certain points in the grammar. He also suggests that some sort of non-redundancy information should be stated in the lexicon: (49)

a. b. c.

The Type contains "action". The Type contains "+contr, +dyn". The semantic function of the first argument is Agent.

This seems to indicate that a deeper analysis of lexical regularities will provide the basis for a more extensive inventory of rules. In our opinion, the end result will be the simplification of lexical entries in the FG lexicon with the progressive formulation of rules that can adequately capture syntactic and semantic generalizations of the kind described.

2.2.1.5. Quantitative valency Quantitative valency refers to the number of arguments subcategorized by a given predicate (xi ...xn). Predicates can have one, two, or three arguments. Dik (1997a: 79) states that the maximum number of arguments in a natural language is four though such predicates are extremely rare. In contrast to approaches like GB, FG makes no distinction between internal arguments (those which are theta-marked by the VP) and external arguments (those which are

FG lexical representation

51

theta-marked by the entire predication (Williams 1981)). The order imposed by the variables (xj x n ) does not reflect the linear order in which the constituents are finally expressed. Dik (1997a: 80) states that predicate frames "are order-free structures, the constituents of which will finally be linearized by the placement rules". Obviously, this is evidence of the cross-linguistic validity of predicate frames in the sense that representations of same SoA in two languages with different constituent ordering will be similar. Problems can arise when it comes to differentiating arguments and satellites because the boundary between argument and satellite is not always clear. In this respect, our proposal offers a way to differentiate the two inventories more clearly.

2.2.1.6. Qualitative valency Qualitative valency refers to the semantic role of the arguments in the SoA designated by the predication. FG semantic functions are determined by the SoA. However, no distinction is made between the arguments of the verb and the participants in the SoA, as in models like RRG, CG, or Rappaport and Levin. Dik (1997a: 119) classifies the semantic functions according to argument position. As shown in Table 8, first-argument semantic functions (Al) are those assigned to the only argument in one-place predicates or to the most central argument in many-place predications. Table 8.

First argument semantic functions

Agent (Ag)

entity controlling an action (= Activity or Accomplishment) Positioner (Po) entity controlling a position Force (Fo) non-controlling entity instigating a process (= Dynamism/ Change) entity that undergoes a Process Processed (Proc) entity primarily involved in a State Zero(0) Processed Experiencer entity that experiences a [+exp] Process (ProcExp) entity that experiences a [+exp] State Zero Experiencer (ZeroExp)

52

Simon Dik's Functional Grammar

He expands this inventory of first argument functions with two new ones (ProcExp, ZeroExp), which come about as the result of treating the parameter [+/- experienced as a secondary semantic function or a footnote to other first argument semantic functions. Second (A2) and third-argument (A3) functions are those assigned to the second and third arguments of many-place predicates. Table 9.

Second and third argument semantic functions

Goal (Go) Recipient (Ree) Location (Loc) Direction (Dir) Source(So) Reference (Ref)

entity affected or effected by the operation of some controller (Agent/Positioner or Force) entity into whose possession something is transferred place where something is located entity towards which something moves/is moved entity from which something moves/is moved second or third term of a relation with reference to which the relation is said to hold.

FG semantic functions are not primitives, but are derived from other notions such as the type of SoA designated by the predication. This view is similar to that advocated by other predicate-centered approaches. However, FG semantic functions are different in that they are not mnemonics in a pure sense because no distinction is made between the participant roles in an SoA and the thematic functions typical of a given verbal predicate. In this regard, Nuyts (1992: 200) claims that it is necessary to distinguish between a set of conceptual roles, which categorize the position that entities can have in a speaker or hearer's knowledge of the world, and a more limited set of case categories, which function in predicate frames and canalize the more subtle conceptual roles. FG semantic functions seem to be floating somewhere in-between conceptual roles and case categories. Schack Rasmussen (1994: 53) claims that there should exist some sort of correspondence between semantic functions as encoded in meaning definitions and the semantic functions of the predicate frame. She argues that semantic arguments are notions that form part of the action schemes, whereas the semantic functions reflect the type of SoA

FG lexical representation 53 in the predicate frame and can be defined as "manifest Semantic Arguments". Another problem with FG semantic functions is the question of whether or not they form a complete inventory. According to Dik (1997a: 122), the catalogue of semantic functions is far from definitive: None of the distinctions made here is definitive, nor is it clear which and how many nuclear semantic functions would suffice to capture the crosslinguistic inventory of semantic functions. In general we will try to find those semantic functions which are necessary and sufficient to capture both the semantics, the grammatical behaviour, and the formal expression of term structures in the nuclear predication. In fact, in our analysis of the English verbal lexicon, we have come across several cases which do not seem to fit any of the labels in the inventory.10 In our opinion, a distinction should be made between those cases which operate within the SoA and those which are derived from the semantics of the predicate. This would avoid the generation of an ad hoc list of semantic roles, which at some point would cease to be explanatory, and moreover, it would offer more viable solutions for those problematic cases.

2.2.1.7. Selection restrictions Selection restrictions specify the nature of the arguments which can appear in complement and subject positions. In FG this information is part of the subcategorization frame of each lexical item. For example, eat prototypically subcategorizes two arguments. The first argument is characterized as animate, and the second as solid food. These selection restrictions block, at least to a certain extent, the generation of anomalous sentences such as the following: (50)

a.

The refrigerator [inanimate] ate the pastrami sandwich. [Violation of the selection restrictions of the first argument]

54

Simon Dik's Functional

b. c.

Grammar

The football player ate a beer [liquid]. [Violation of the selection restrictions of the second argument] The refrigerator [inanimate] ate a beer [liquid], [Violation of the selection restrictions of both arguments].

Nevertheless, the exact nature and function of selection restrictions are questions that are far from being resolved. Some linguists have suggested that they may not be linguistic at all, but rather form part of our knowledge of the world (Johnson-Laird 1983). Dik is of the opinion that they are linguistic because as speakers of a language, we intuitively know how many arguments each predicate has, as well as what their characteristics are. Yet, even among those who agree that selection restrictions are essentially linguistic, there is some dispute as to whether they are syntactic, semantic, logical or pragmatic (Fodor 1977; Keenan 1970). Traditionally, selection restrictions have been considered to be constraints that yield absolute judgments of the acceptability of a given linguistic expression (Chomsky 1965; Katz—Fodor 1963). Accordingly, a set of unconditional features allows or prohibits the insertion of terms. This view has been criticized, because the determination of the full set of selection restrictions would involve the impractical task of testing every conceivable context within certain grammatical constructions (Siewierska 1991: 24). A more viable alternative is to regard selection restrictions as constraints that mark the most compatible readings of a predicate. As Dik (1997a: 94—97) remarks, selection restrictions should not be seen as bans on term insertion in those cases in which the semantic nature of the term does not conform to the selection restriction imposed on the argument slot. In fact, he affirms that violations of selection restrictions are often a source of metaphorical expressions. In his view, selection restrictions, as well as predicates, do not belong to any sort of abstract metalanguage, but are a part of language.

FG lexical representation

55

2.2.1.8. Meaning definitions Although some linguistic models such as GB and LFG make no use of meaning definitions, other theories have incorporated them in their representation of lexical entries. FG describes non-nuclear semantic predicates in terms of a more general one, while specifications in terms of arguments and satellites represent the properties that differentiate one predicate from the others. Dik (1997a: 101) gives the following examples: (51)

assassinate [V] (xj: )Ag fe )Go murder [V] (xi)Ag (x2)go(x3: treacherous [A])Manner

(52)

murder [V] (xj: < h u m a n > ) A g fe )c 0 kill [ V ] ( x i ) ( x 2 ) g o ( X 3 : intentional [ A ] ) n n e r A g

(53)

(54)

Ma

kill [V] (xi)Ag/Fo fe ) Go cause [V] (xi) Ag /Fo (ei: [die [V] (x2))proc die [V] (x^ )Proc come about [V] (ei: [dead [V] ( x 0 ) o

])go

])ptoc

In (51), the predicate frame for assassinate shows that it has two human arguments, and is defined in terms of murder with the manner specification of treacherous. In a parallel way, murder has two human arguments, and is defined in terms of kill with the manner specification of intentional (52). Kill has two arguments, the first of which can be an agent or a force, and the second is specified as a goal. It is defined as to cause to die (53). Die is decomposed in terms of come about and dead (54) The idea behind this Stepwise Lexical Decomposition (SLD) is that meaning definitions constitute a web in which more specific predicates are defined in terms of more general ones. Given the important role meaning plays in FG, it is surprising that there is no connection between the information contained in meaning definitions and the actual predicate frame. In fact, according to Dik (1997a: 101— 102), meaning definitions do not have a direct role in the construction

56

Simon Dik's Functional Grammar

of the underlying clause structure,11 although he admits that under certain circumstances meaning definitions may be "unpacked" when necessary. The idea is that SLD eventually leads to a set of predicates which cannot be defined any further. Whether these undefinables have the status of concealed primitives, which can be used in the construction of a more cognitive-conceptual structure of meaning, is still an open question.12 Although predicate frames encode the microstructure of a lexical entry, the macrostructure of the FG lexicon is not specified. Vossen (1995: 22) observes that FG follows a semasiological approach to meaning because the semantics of predicates is defined languageinternally and there is no way of relating it to a non-verbal indication of the denotation. Unfortunately, this type of order is not very informative about semantic structure. To overcome this difficulty, he extends the FG lexicon with a cognitive layer in which he distinguishes analytic and perceptual knowledge, as well as different dependencies between concepts (subtype-category relations, part-whole relations, and analogy relations). He claims that this system of conceptual knowledge can be used to define what is meant by denotation, and provide the criteria to get at the anchoring of language to an independent cognitive system. Vossen (1995) has worked primarily with nouns. Nevertheless, the same premise is just as valid for the verbal lexicon. The only way to get at denotational information is to enrich the verbal lexicon with an onomasiological structure in which predicates are grouped in semantic classes or domains. In consonance with this, FLM lexical organization includes a paradigmatic description of lexical-syntactic relations through which mapping relations between syntax and semantics can be predicted, and which accounts for pragmatic, semantic, and syntactic generalizations across different sets of predicates.

2.3. FLM lexical description One of the goals of the FLM is to expand the FG lexicon component by making its relational structure explicit. This is also a valid way of making a theoretical contribution to lexicography. In this regard, Martin Mingorance (1990) observes that lexicography is very slow in accepting advances made in lexical semantics. This often results in

FLM lexical description

57

an oversimplification of the information found in dictionaries, which do not take into account factors such as syntactic and pragmatic features, axiological distinctions and metaphor. In the FLM the lexicon of a language is seen as a grammar, which can account for a speaker's lexical competence. The central unit of description in a fully specified FLM lexicon is the word, which appears with all its syntactic, morphological, semantic and pragmatic properties. The lexicon is thus conceived as a dynamic, text-oriented network of information about words and their contexts. Microstructurally, lexemes are characterized as complexes of phonological, morphological, syntactic, semantic, and pragmatic information. It is no accident that these are the five types of explicit linguistic knowledge in NLP. On the level of macrostructure, lexical entries are interconnected by cohesive, associative and encyclopedic functions. The FLM lexicon thus contains information for the speaker and hearer relevant to the act of communication. It integrates a modified and extended version of Coseriu's Lexematics into Dik's Functional Grammar. One of the points of convergence of these two models is that the methodological paradigm adopted for the description of lexical entries in both is functional. Lexematics is structurally functional, in that the principle of functional oppositions regulates the system of a language (Coseriu 1981: chapter VIII), while FG is teleologically functional because it is based on the assumption that language is essentially a means of verbal interaction. The FLM traces lexical structure through the analysis of dictionary entries as knowledge representation. The basis for our structural description of the lexicon is the distinction between syntagmatic and paradigmatic relations, or the complementary principles of combination and selection (Saussure 1916 [1990]; Lyons 1977: 241). According to Martin Mingorance (1990: 229), the construction of a formalized grammatical lexicon organized onomasiologically in semantic hierarchies entails the following four stages: (i)

The distinction between the primary and derived lexicon. Lexical units forming the primary vocabulary of the lexicon are those which cannot be derived by word-formation rules. In

58

Simon Dik's Functional Grammar

contrast, the derived lexicon constitutes a separate component and runs parallel to the grammatical one.13 (ii) The organization of this vocabulary in lexical domains. Generic terms are specified through the factorization of meaning components, and lexical domains are organized in a network of subdomains of lexical hierarchies in order to account for both inter- and intralinguistic lexical relations. (iii) The analysis of the complementation patterns of each lexeme, using predicate frames as integrated formulae. The location of each lexeme in a lexical domain/subdomain is directly related to its inventory of syntactic information; (iv) The establishment of a hierarchy of semantic and syntactic prototypes for each lexical domain. Semantic and syntactic prototypes are configurations of the meaning constituents shared by all the members of the same lexical set. The resulting configurational pattern is data-driven, and is based on meaning structure as codified in dictionary definitions.

2.3.1. The FG lexicon and onomasiological structure The onomasiological structure proposed in the FLM is implicit in Dik's embryonic description of Stepwise Lexical Decomposition. The development of his subsequent proposal is carried out by adapting and extending lexematic structures.1 Lexematic structures are divided into paradigmatic structures, which establish basic lexical organization, and syntagmatic structures, which indicate the combinatorial properties of lexical items. Although we have used lexematics as a starting point, we have taken its principles, adapted them to our needs, and integrated new concepts. As a result, our method of lexical analysis is somewhat different from Coseriu's framework, and the terminology that we have used to describe the different levels of paradigmatic structure does not correspond with that traditionally used in lexematic theory. As shall be seen in Chapter 3, the following terms play an important role in lexical structure:

FLM lexical description

59

(i)

Lexical domain·. The set of lexemes which together lexicalize all or part of a conceptual domain. (ii) Subdomain: A subdivision of a lexical domain which is a meaningful set of lexemes within a lexical domain and forms the basis of its internal structure. Each subdomain focuses on a particular area of meaning, and is a different specification of its content. (iii) Genus: The superordinate term of the domain or subdomain by means of which the other lexemes are directly or indirectly defined. (iv) Differentiae: The semantic information in the meaning definition of a lexeme which distinguishes it from others in the same lexical domain. (v) Synsem parameter: Recurrent semantic information which appears throughout the lexicon in a wide range of lexical domains, and which influences both semantic and syntactic coding. Martin Mingorance (1990: 233) states that since Dik's FG belongs to the synthetic type of grammar (and thus focuses on the syntagmatic axis), it is complementary with Coseriu's analytic model, which organizes the core vocabulary of a language in lexical domains, and thus develops the paradigmatic axis. Although Dik does not explicitly mention lexical domains in his grammar, both the principles underlying the structure of meaning definitions as well as the procedure of Stepwise Lexical Decomposition are similar to the basic premises of lexematics: Although the view of lexical analysis fits in nicely with the model of FG, the assumptions embodied in it do not necessarily follow from this model. That is, FG would also be compatible with other conceptions of lexical meaning definition (Dik 1978a: 46). As previously stated, the meaning definition in each lexical entry (or predicate frame) in the FG lexicon has the format in (55):

60 (55)

Simon Dik's Functional Grammar a = def π (a being a predicate frame and π some combination of predicate frames of more generic semantic nature) (Dik 1978a: 46).

Dik (1978b: 21) establishes the following corollaries for Stepwise Lexical Decomposition: [ΑΓ] [A2]

The defining predicates occurring in meaning definitions are lexical items of the object language. The defining predicates occurring in meaning definitions are not drawn from a universal, language-independent set of such predicates; this assumption is entailed by (ΑΓ).

Dik's first two corollaries state that only lexical items of the object language are regarded as predicates. Actual lexicographic practice can thus be taken as a model for the structure of meaning definitions. In line with this, the primary source of data that we have used for the elaboration of a lexical domain is information extracted from both alphabetically and conceptually organized English monolingual dictionaries (e.g. Roget's Thesaurus, Collins Cobuild, Oxford Advanced Learner's Dictionary, The Longman Lexicon of Contemporary English, etc.). The basic corpus of primary lexemes is that found in the Longman Lexicon of Contemporary English (LLCE). It is important to emphasize here that our inventory of lexical domains differs substantially from that found in the traditional thesauri, which group lexical items very roughly in terms of ideas expressed, and whose classification in thematic areas is based on extralinguistic criteria. For example, the sets of words in Roget's Thesaurus have no definitions, and thus it is not possible to differentiate one synonym from another. Although the organization is hierarchical, actual lexemes are provided only at the bottom levels. As a result, it is impossible to speak of hyponymy since there is no explicit ordering of specific terms in relation to more general ones. Bell (1991: 97) states that the lexical domain is an extension of the thesaurus model, but it is in reality much more than that. Our method of analysis is bottom-up (or data-driven), while the Thesaurus is based on a top-down (or concept-driven) type of processing.

FLM lexical description

61

Meaning components are extracted from various dictionaries, and on the basis of these, a new definition is elaborated. This new definition includes the most relevant features for the domain in question, or those that yield functional oppositions. [A3'] The defining predicates occurring in meaning definitions need not be primitives: they may themselves be predicates defined in other meaning definitions (Dik 1978b: 21). [A3*] The definiens of a meaning definition defining some predicate φ; may not contain a proper subconfiguration of predicates such that this subconfiguration in itself constitutes the definiens of a meaning definition defining some predicate φ (Dik 1978b: 24). The fact that the defining predicates must be hierarchically superordinate, and thus semantically more complex than the defined predicates, implies that definitions are organized hierarchically. This would avoid circularity in the definitions of lexemes. In fact, the paradigmatic organization of a lexical domain in subdomains solves this problem because lexemes belonging to the same subdomain share the same genus, and are differentiated from one another by at least one differentia. More specifically, in each lexical domain of the English verbal lexicon, there is a nuclear meaning common to all words, as well as different non-nuclear parameters of semantic differentiation which can be said to mark the distance of the hyponym from the more general action. In much the same way, the structure of the subdomains is hierarchical, each focusing on a particular conceptual area. [A4a] The structure of the definiens of meaning definitions is of the same formal type as the structures underlying sentences (Dik 1978b: 21). V

Corollary (A4a) guarantees that the combination of semantically simple predicates which form part of the definition are of the same type as the ones underlying sentences. This contrasts with formal theories such as GB and Lexical Functional Grammar (LFG), which postulate that the information specified in the argument structure is projected to theoryspecific initial syntactic structures. More functionally oriented theories

62

Simon Dik's Functional

Grammar

like FG or RRG propose one level of representation. This representation can be semantic (e.g. the FG predicate frame) or syntactic as in RRG, but always is in itself the underlying structure. Configurational representations are thus dispensed with altogether. [A4b'] The definiens of meaning definitions is not directly accessible to the operation of syntactic rules (Dik 1978b: 21).

In contrast to generative models, lexical structure is not directly built into the underlying structure of linguistic expressions. The reason for this is that given the autonomous status of the lexicon, meaning definitions may be used in the interpretation of underlying predications, but are not used in their formation. In the chapters that follow, it will become evident that semantic information codified in definitions can be used to predict the syntactic behavior of sets of semantically related lexemes. Consequently, information in meaning definitions should play a more salient role than it has until now. [A5] In every language there is a set of semantically simple lexical items the meaning of which cannot be defined by means of meaning definitions (Dik 1978b: 31). This is similar to the underlying premise of the Natural Semantic Metalanguage inventory of semantic primitives or the Meaning Text Theory hypothesis of a set of near primitives. Corollary (A5) allows for both the concentration of linguistic and extralinguistic information in superordinate terms, and at the same time permits the integration of their semantic content, which constitutes the informational core of the whole domain. Corollaries (A3), (A3*) and (A4a) are crucial in the establishment of meaning definitions because the information contained in each lexical item presupposes all the information in the genus. As a result, only the semantic characteristics that differentiate the defined lexeme from its generic term and from its co-hyponyms need be specified. In this type of procedure the relation of partwhole/class inclusion is vital, since the formation of hierarchies means

FLM lexical description

63

that certain lexemes will be treated as superordinate, and used to define the lower terms through explicitation. Although lexical interrelations may be of various types, the most favored one in domain structure is evidently that of hyponymy, a meaning postulate involving total inclusion, which is the basis of hierarchical relationships. The privileged status of hierarchical organization in cognition is in correlation with the importance that we have given it in our model of the lexicon. In this regard, Apresjan (1993: 82) underlines the hierarchical structure of the human psyche. Similarly, Waxman (1994: 234) states that hierarchical systems are exceptional for their efficiency in organizing existing information and for their power in generating new information. Hierarchies are also instrumental in the process of comprehension because through hierarchical organization, human beings can comprehend a schema containing a very large and complex number of discriminations (D'Andrade 1987: 112). In an onomasiologically organized FG lexicon, lexical entries with shared meaning components are located in the same lexical subdomain (Figure 3). Moreover, each lexeme is provided with its meaning definition as well as the grammatical information necessary for its use in different contexts. This type of lexical organization codifies the range of choices available to each speaker in the lexicalization of a given area of meaning. In this way the lexicon becomes a dynamic component where the choice of one lexeme instead of another is goal-directed. Although the lexicon has a primary hierarchical type of organization, nonhierarchical relations are also taken into account. The representation of the basic configurational pattern of a specific subdomain is the following:

64

Simon Dik's Functional Grammar SubdomX^ I \ Lex 1.1 / un J LeHim-1 · ILexeme-n

Figure 3.

Macrostructural organization of a lexical domain

FLM lexical description

65

As shall be seen, a lexicon with onomasiological structure permits the meaningful combination of linguistic information in the description of lexical entries. This means that the relevant meaning constituents contained in the primary vocabulary of a language are hierarchically stored. Through meaning definitions with a high degree of informativity and a maximum degree of economy, such a lexicon allows for the codification of semantic relations between different lexical units, and permits the development of a framework which classifies predicates (whether verbal, nominal, or adjectival) into semantic classes. It also clarifies the interrelation between syntax and semantics since the syntactic behavior of predicates is conceived as being motivated by the lexical subdomain to which these predicates belong.

2.4.

Summary

The organizational parameters of the FG lexicon have changed very little since the model was first formulated. Predicate frames are the basis of FG lexical representation, and contain information about predicates, predicate form, syntactic category, redundancy rules, quantitative valency, qualitative valency, selection restrictions, and meaning definition. Various issues arise that are relevant to the representation of lexical information. Firstly, it is not always clear how to distinguish between an argument and a satellite (argument obligatoriness). Secondly, there is the question of whether FG semantic functions are closer to conceptual categories or case roles, and if they can be said to form a complete inventory. Nevertheless, the most important issue for us is the question of meaning definitions and the extent to which they interact with predicate frames. Although at first Dik appears to rule out the possibility that meaning definitions have a role in determining syntactic expressions, certain new aspects of FG theory seem to leave the door open to this possibility. More specifically, the notion of paradigm is mentioned, though in relation to phonological form and syntactic variation. Presumably, it could be extended in the future to also include semantic information.

66

Simon Dik's Functional

Grammar

Even more promising is Dik's initial proposal for the inclusion of semantic information, which also influences syntactic form. In this regard, he suggests the addition of a feature [move] to the predicate frame of movement verbs since they take directional satellites. The existence of this type of feature indicates that there are other types of semantic information which would presumably have a direct influence on syntax as well. This proposal seems to point the way to the possible organization of predicates into coherent semantic classes, which encode both syntactic and semantic regularities. The goal of the Functional-Lexematic Model (FLM) is to develop Dik's proposal and expand the FG lexicon component to make its relational structure explicit. The FLM integrates the corollaries of Dik's Stepwise Lexical Decomposition and Coseriu's lexematic theory to organize the lexicon onomasiologically in semantic hierarchies which form lexical domains and subdomains. This type of organization is informative because the shared semantic information in these areas of meaning can be used to predict the syntactic behavior of sets of semantically related lexemes.

3.

Lexical organization and the FLM

3.1. Introduction: the concept of semantic field Semantic fields were first introduced by Trier (1934), and have a long tradition in European Structuralism in the work of Pottier (1964), Coseriu (1968, 1977) and Wotjak (1987, 1994abc, 1998abc). In America, semantic fields have been studied by Lehrer (1974, 1992), Kittay (1987, 1992) and Grandy (1987, 1992). Traditionally, a semantic field is regarded as a set of lexemes which cover a certain conceptual area and which bear certain specifiable semantic relations to one another. Undoubtedly, one of the main problems in establishing valid areas of meaning has been the vagueness of the concept itself. Although it is generally agreed that a semantic field is a collection of related words, how large a collection and how the items are related is in dispute (Grandy 1987: 260; Lyons 1977: 267). None of the solutions proposed up to the present has been generally accepted because agreement on this question would mean reaching a consensus of opinion on issues such as the possibility of semantic primitives and the structure of conceptual categories. Nevertheless, despite this lack of precise definition, the concept of semantic field dutifully makes at least a cameo appearance in the majority of studies in lexical semantics. Perhaps the popularity of semantic fields lies in the fact that people are intuitively aware that such a construct exists. Its psychological validity has been verified by experiments in neuroscience (H. Damasio et al. 1996) though its definition from a neurological perspective varies somewhat from its traditional linguistic conception. For example, in his study of the role of the right hemisphere in language comprehension, Beeman (1998: 256) defines semantic field as "the subset of semantic information activated by response to an input word". Although he obviously conceives meaning very traditionally in terms of a feature representation, he makes the distinction between a projective field, the set of internal representational units activated by an external input (a word), and a receptive field, the set of words which activate a given representation.

68

Lexical organization and the FLM

It is possibly because of this psychological relevance that the term semantic field is used by linguists on both continents with such frequency to describe any group of related words. This definition is usually implicit rather than explicit because the term often appears as though there were no need to further pinpoint its meaning, something that is hardly the case. Although a lexicon organized in terms of meaning would help to make explicit the interface between syntax and semantics, little or no effort has been made to specify the internal structure of semantic fields, their possible criteria for membership, or an inventory of relations which would contribute to the systematization of interfield connections. Semantic fields have never played a very important role in contemporary linguistic theory. In the most traditional representation of a semantic field, lexical items are positioned in a two-dimensional space. The items are differentiated by their situation in the field, which is divided on the basis of semantic dimensions with specific values. However, when most people think of semantic fields, they generally associate them with the kind of binary feature representation typical of componential analysis (CA), which Nida (1975) introduced in translation analysis in order to trace interlinguistic correspondences at the level of contrastive sets of lexemes. This type of meaning representation, popularized in linguistic theory by Katz—Fodor (1963), actually turns the two-dimensional analysis inside out by placing the distinctive dimensions that structure the field within the lexical items in the form of binary features. In CA, the meaning of a word is characterized by the sum of the binary features it possesses such as +/-human, +/-male, +/-concrete, +/-animate, etc. so as to render it distinct from that of all other nonsynonymous words. When CA was most popular, linguists hoped that it would enable them to ultimately arrive at a set of universal semantic features. However, it soon became clear that this type of analysis only works (if it can be said to work at all) in small highly structured sectors of vocabulary such as kinship. Although CA gives, at best, an incomplete description of word meaning, its supporters argue that this type of representation is not meant to be exhaustive, but only to account for systematic oppositions within a given lexical domain. To solve evident problems resulting from such an extreme oversimplification of word meaning, proponents of CA made an effort

Introduction: the concept of semantic field 69 to enrich it by distinguishing between diagnostic and supplementary components. Nida (1975) also proposed the possibility of using multivalued features, which, when all brought together, amount to a dictionary definition, but the question inevitably arises of whether such features are derived purely from observation or from a specific lexicographic source. If they have effectively been extracted from a source, then there is no escape from the fact that the analysis is circular. Another reason for the absence of semantic fields in contemporary linguistic theory is that such constructs have never been seriously contemplated within the mainstream syntax-oriented Chomskyan framework. Even in those theoretical schools that have included semantic fields, mention of them has been fleeting and further specification invariably left until some future date.

3.1.1. Cognitive Linguistics Cognitive Linguistics has accepted the concept of semantic field, but has adapted it to its theoretical principles, among other things, by eliminating CA. According to Croft (1993: 339), the concept of domain in Cognitive Semantics is distantly related to Trier's concept of field. However, it is hardly the same because in Cognitive Semantics, the standard semantic feature or criterial-attribute model of categorization has been vigorously rejected in favor of a model of word meaning based on prototype theory. In the same way that prototypes and the criterial-attribute model are two conflicting theories of word meaning, they are also conflicting models of the structure of conceptual categories. The classical view argues for a description of word meaning as a set of criterial properties. According to this theory, categories have clear boundaries as membership is limited to those entities possessing the conjunction of necessary and sufficient features particular to the category in question. Within the category itself, all members thus have equal status. In contrast, prototype theory holds that the meaning of a word should be described in terms of the clearest cases labeled by that word. Such prototypes serve as cognitive reference points, and so, membership within the category is graded (Rosch 1978; Taylor 1989: Chapters 2-4).

70

Lexical organization and the FLM

However, the indeterminacy of the notion of domain continues to be its most problematic aspect. Croft (1993: 339) admits that despite its centrality, this concept has not been described in detail, and goes on to define it rather fuzzily as "a semantic structure that functions as the base for at least one concept profile". This definition bears a certain resemblance to Fillmore's frames because in both approaches, a domain/frame can be almost anything from space to marriage to buying a pair of shoes. As a result, it is a bit like a wildcard which can be brought into the game when the player is in need of it, and whose meaning depends totally on the context in which it is used. When describing possible lexical domains, Langacker (1987) postulates those of time and space. However, time and space are not possible candidates for "domainship" because they are the first set of co-ordinates developed in the brain, and for this reason, they permeate the lexical distinctions that underlie the structure of all domains. Another area of indeterminacy in Cognitive Linguistics is the lack of consensus concerning the representational format to be used for the semasiological structures of lexical items, something that naturally affects a possible macrorepresentation of their relations with other lexemes. Geeraerts (1995) distinguishes three major types of formats: the radial set model popularized by Lakoff (1987), the schematic network as defined by Langacker (1987, 1991a), and the overlapping sets model introduced by Geeraerts (1989). Despite the existence of superficial differences, he argues that these three modes are in reality notational variants because all of them account for salience effects, metaphor and metonymy, hierarchical semantic links, and discrepancies between intuitive and analytical definitions of polysemy. Nevertheless, the problem lies in the analysis itself since none of the representational formats described explicitly provides a place for syntagmatic data. Nor do they solve the question of how to determine either the relevant prototypical center or the psychologically real schematic meaning within a concept. The specification of exactly how this should be done and the criteria such decisions should be based on have been left deliberately vague.

Introduction: the concept of semantic field 71 3.1.2. Generative Linguistics Lexical decomposition in generative linguistics has been largely syntax-oriented, and thus not greatly concerned with meaning, much less the structure of semantic fields. The priority in the generative framework has been to discover only those aspects of word-meaning that interface with grammar. One of the earliest approaches to semantics within the generative framework was componential (Katz—Fodor 1963; Katz—Postal 1964; Katz 1972). At the time, the principal objective was to find a finite set of universal semanticconceptual components into which all lexemes could be decomposed. Katz and his colleagues postulate a dictionary entry that includes the word being defined, its grammatical category, and a definition in the form of semantic markers and distinguishes. Markers are considered to be universale and as such, the most general classification components in the lexicon, while distinguishers take care of what is left and differentiate words from others which have the same marker. Meaning is thus divided in two parts: semantic meaning and extralinguistic meaning. Katz (1972) also includes selection restrictions, which are the main constraint in the amalgamation processes involved in projection rules (Saeed 1997: 236). However, as Melka—Schroten (1992: 38-39) point out, this type of analysis proves unsatisfactory because it is difficult, if not impossible, to define necessary and sufficient conditions for concepts as well as. to decide which features are linguistic and which are extralinguistic. To the extent that Katz makes an attempt to establish a semantic metalanguage, his theory can be said to be decompositional, but no reference is made to the possible existence of larger paradigmatic constructs such as semantic fields. In another context, Jackendoff (1990) explicitly mentions semantic fields. His version of semantic domains differs somewhat from that found in Cognitive Linguistics, but both he and Langacker subscribe to some form of localism, the hypothesis that motion provides the cognitive framework for more abstract domains of meaning such as POSSESSION, COMMUNICATION and CHANGE (Goddard 1998: 195). In Jackendoff s localistic approach, four semantic fields, spatial location, temporal location, property ascription, and possession cross-classify the basic ontological categories of EVENT and STATE

72

Lexical organization and the FLM

(Saeed 1997: 253). For example, the function BE is used to represent four categories of STATE, which Jackendoff calls semantic fields. By extending spatial conceptualizations into non-spatial domains, he distinguishes BELQC (location in space), BExemp (location in time), BEwent (ascription of a property in locational terms), and BEp0SS (possession as location). In the same way, he also distinguishes between GOloc, GOremp, GOid en t, and G 0 p 0 s s · However, the type of classification explicit in Jackendoff s model possesses the disadvantage of having very little internal structure. Dorr (1993: 125) points out that while Jackendoff s LCS framework provides a means for distinguishing between verbs across the classes that it establishes, it does not provide a principled account of constraints within its semantic classes. Although verbs within a particular class are frequently distinguishable by some feature corresponding to Manner (e.g. walk vs. run), Jackendoff says little or nothing about the differentiation between such verbs in his description of conceptual organization. If, as he claims, the business of conceptual structure is only to encode an appropriate argument structure, then there is no way of explaining lexical selection, given that all manner verbs have the same conceptual structure as their corresponding superordinate. According to Jackendoff (1990: 88), manner verbs would have to be differentiated by linking them to "a more detailed spatial structure encoding". Although he does not say anything about how such a representation could be constructed, the answer undoubtedly lies in a mixed model of conceptual structure. According to Pederson—Nuyts (1997: 3), the best-known work defending such image-based conceptualization can be found in Paivio (1972,1991), Kosslyn (1980), and Marr (1982). This is in consonance with recent speculation as to the possibility of there being multiple kinds of mental representation according to the different sensory modes of input (Levelt 1989; Jackendoff 1992). The description of the manner component in verbs would thus conceivably involve the specification of a corresponding set of relevant spatial parameters, which would constitute a language of visual distinctions.

Introduction: the concept of semantic field 73 3.1.3. Pustejovsky's Generative Lexicon Pustejovsky (1995: 57-58) presents a conservative approach to decomposition where lexical items are minimally decomposed into structured forms (or templates) rather than sets of features. He claims that the result is a generative framework for the composition of lexical meanings, thereby defining the well-formedness conditions for semantic expressions in a language. Verbs are classified into "aspectual types": ACTIVITIES (walk, run, swim, drink), ACCOMPLISHMENTS (build, destroy, break), ACHIEVEMENTS (die, find, arrive), and STATES (know, love, resemble, think, be) (Pustejovsky 1995: 16). In Pustejovsky's opinion, membership in an aspectual class determines much of the semantic behavior of a lexical item. His semantic classification for verbs is based on Aktionsarten. The basic premise is that verbs differ in the kinds of eventualities in the world they denote. In fact, one of the levels of representation of his generative theory of the lexicon is that of Event Structure, which represents information related to Aktionsarten and event type, in the sense of Vendler (1967), Dowty (1979), and related work. Unfortunately, he does not develop the idea of semantic domain any farther, but the concept is obviously contemplated in his theory, though more as a desideratum than anything else. He is in favor of the idea because he points out that the elaboration of taxonomies of entity types (or the semantic taxonomic tradition) should not be lightly dismissed, and that many of the generalizations it hopes to capture are legitimate goals for linguistic theory and cognitive science (Pustejovsky 1995: 19).

3.1.4. Fillmore's Frame Semantics Fillmore—Atkins (1992, 1998) propose frames as an alternative to semantic fields. This particular concept of frame initially resulted from a proposal made by Fillmore (1985), which in turn grew out of his work in case grammar (Fillmore 1968). Fillmore—Atkins (1992: 76-77) maintain that frame semantics and mainstream semantic field theory differ from each other in that field theorists posit systems of paradigmatic and syntagmatic relationships connecting members of

74

Lexical organization

and the FLM

selected sets of lexical items, whereas the frame semantics approach describes lexical meaning in terms of a structured background of experience, belief, or practices necessary for its understanding. Words are thus not related to each other directly, but only in virtue of their links to common background frames, which provide a conceptual foundation for their meaning (e.g. Schank—Abelson 1977). This approach has led to the development of the Berkeley/ICSI FrameNet project (Fillmore—Atkins 1998). Despite the authors' explicit rejection of the concept of semantic field (Fillmore—Atkins 1992), it is evident that FrameNet includes an inventory of semantic domains similar to those in Miller's WordNet. The categories established are communication, cognition, emotion, space, time, motion, body, interpersonal and institutional transaction, health and healthcare. Fillmore—Atkins (1998: 417) affirm that each entry is the result of the exploitation of corpus evidence as well as native speaker intuition, and thus, in principle is designed to provide a complete account of the semantic and syntactic combinatorial properties of one lexical unit. Words are collected in semantically related sets belonging to these domains, though the authors do not explain the criteria for domain membership or the internal structure of the domain. Part of the description of each word is the identification of the semantic frame underlying its analysis. This includes the vocabulary for describing the elements of the frame, information about the manner in which constituents realizing frame elements may or must be distributed within and around phrases headed by that word, together with a record of the other frames of which the target frame is an elaboration or a blend. Whereas frames are certainly not synonymous with semantic fields (especially if fields are regarded as purely paradigmatic constructs), frames are not really so different as to imply a totally new concept, particularly when field representation is enhanced with other types of information. In fact, Kittay—Lehrer (1992b: 6) suggest that frames and fields might be mutually derivable. In other words, by broadening the concept of field to include some systematic treatment of syntagmatic relations, they ask if it might be possible to build enough into a field so as to incorporate the pragmatic information that frames account for. Frames are certainly helpful as an interpretative device for the incorporation of pragmatic information, which clarifies the meaning

Introduction:

the concept of semantic

field

75

of a lexical item in a specific context. Nevertheless, they do not seem to be very susceptible to systematization and have the disadvantages of the "wild card" concept of domain in Cognitive Linguistics.

3.1.5. WordNet Fellbaum (1998c: 91) refers to a lexicon structured in terms of semantic fields as an alternative model to the one in WordNet. Her concept of semantic fields is based on Kittay—Lehrer (1992a), and seems to implicitly equate all work done in semantic fields with this collection of articles. Although there is a certain similarity between semantic fields and WordNet's field-like domains, she says that the two types of organization are different. The meaning of a lexeme in a given semantic field arises from its relations of similarity and contrast with other lexemes in the same area of meaning, whereas WordNet groups lexemes in sets of synonyms or synsets. Fellbaum remarks that in WordNet, syntagmatic information is not part of the network structure, while in semantic fields such information is taken into account. She apparently thinks that semantic fields are based on both paradigmatic and syntagmatic relations, when this is normally not the case. In fact, Kittay—Lehrer (1992b: 6) specifically mention the fact that semantic fields are essentially paradigmatic and as a future goal, suggest that they should be enriched by the incorporation of syntagmatic information. The elaboration of semantic domains in WordNet was carried out by using a pre-established inventory of semantic classes, and then fitting the lexemes into them. However, the domains themselves do not play an important role in this lexical database because the meaning of a lexeme is expressed primarily by its relation to other lexemes and synsets. Fellbaum (1998c: 70-74) goes so far as to say that another group of researchers faced with the same task would likely have carved up the lexicon in a different way. From this, it can be deduced that the classification in WordNet is heuristic.

76

Lexical organization and the FLM

3.1.6. Natural Semantic Metalanguage (NSM) In NSM Theory, the meaning of a word does not depend on the meaning of other words of similar meaning in the lexicon, but rather consists of a configuration of semantic primitives for each word (Wierzbicka 1996: 170). In this sense, Wierzbicka's conception of semantic fields differs somewhat from Trier's. Although semantic fields are included in her theory, the criteria for membership seem to be largely intuitive. However, a closer look at her inventory of semantic primitives shows this mini-lexicon has a tentative macrostructure of predicates grouped in various categories. Verbs are divided into the following sets: (i) (ii) (iii) (iv) (v)

Mental predicates (think, know, want, feel, see, hear) Speech (say, word) Actions, events and movement (do, happen, move) Existence (there is) Life (live, die)

These categories are indicative of a reduced form of lexical macrostructure, though the connections between different areas of meaning are not made explicit. Generally, the specification of lexical relatedness within the NSM seems to be based on definitional analysis. Each word is first defined in terms of semantic primitives or its most basic meaning components, and at the same time the word is compared with the meanings of other intuitively related words. By comparing a word to other words that intuitively are felt to be related to it, we can establish what each of these words really means; having done this, we can compare them again, this time more precisely, being able to identify the elements that are different. Proceeding in this way, we can often discover remarkable symmetries and regularities in the semantic structure of many words - as well as unexpected asymmetries and irregularities. We can discover self-contained fields of semantically related words with analogous semantic patterning (Wierzbicka 1996: 170).

Introduction: the concept of semantic field 77 Wierzbicka's claim seems to be that non-arbitrary semantic fields can be established through comparing configurations of semantic primitives in the definitions of words. By configurations, she means combinations of primitives or structured components such as "I want something" or "this is good". Thus, her version of semantic fields is apparently based on the recurrence of these primitives. The definitions through which she postulates semantic field relatedness take the form of a shared prototypical scenario describing a highly abstract cognitive structure. For example, emotion concepts such as sad, unhappy, distressed and upset, would all have the underlying semantic structure in (56): (56)

X feels something sometimes a person thinks something like this:

because of this, this person feels something X feels something like this (Wierzbicka 1996: 182). In this sense, Wierzbicka resembles Jackendoff in that there seems to be no way of differentiating troponyms or manner verbs. In the case of speech verbs, Wierzbicka (1992b: 217) observes that verbs that share semantic components also share certain syntactic frames or combinations of frames. And she very rightly points out that syntactic differences, which at first seem idiosyncratic, often are a sign of very real semantic differences, and thus confirm the reliability of syntactic clues in semantic analysis. She concludes by affirming that her approach to semantic primitives also offers necessary firm ground for the study of semantic fields, but so far very few studies have appeared which explore this possibility.

3.1.7. Meaning Text Theory and the Explanatory Combinatorial Dictionary In MTT greater emphasis has been placed on the microstructure of lexical entries, more concretely on the specification of semantic

78

Lexical organization and the FLM

relations of each lexical unit. Although semantic fields are mentioned, their role is not a salient one. Their inventory seems to be considered as pre-established, and only in Mel'cuk—Wanner (1996) is something said about their possible internal structure. In their study of Emotion Lexemes in German, Mel'cuk—Wanner divide this semantic field into eleven dimensions, each with a set of two or three mutually exclusive values. These definitions are arrived at through definitional analysis and represented in the familiar format of binary semantic features in a matrix. Working within this same framework, Escalier—Fournier (1997: 48) have a different idea of semantic field when they introduce "notion", a metalinguistic entity that is their name for a semantic paradigm. This paradigm is not the usual one, but seems more related to Fillmore and Atkins' frame. According to Escalier—Fournier (1997: 49), notion is a useful concept in that it "allows to gather under the same label different lexemes, syntactic constructions or grammatical elements that are related by a non-strict equivalence relation in a certain context. In other words, it constitutes an abstraction on lexemes which it refers to". However, notion differs from concept in that it is dependent on the structuring of meaning that is specific to a particular language. The example given is acheter 'buy', which is characterized by the notion trans, the name of the semantic content common to η expressions in a certain context. According to the authors, the concept of notion allows for a representation of lexemes difficult to describe in terms of semantemes, as well as a direct representation of the equivalence between lexical and non-lexical relations. In MTT, semantic field membership seems to be implicitly based on semantic decomposition in which absolute substitutability is required (Mel'cuk 1989: 81). The MTT (like the FLM) is in direct contrast to Wierzbicka's NSM in that intermediate semantic units are advocated in order to avoid awkward and unmanageable descriptions. Since a lexical unit should be defined in terms of lexical units that are semantically simpler than itself, Mel'cuk postulates that such consecutive semantic decompositions should lead to an inventory of semantic primitives, and underlines the importance of meaning definitions in semantic analysis. MTT semantic fields are identified by generic lexical units, which, in addition to a private entry, have another public entry with the

Introduction: the concept of semantic field 79 shared elements of the values of the lexical functions (LFs), associated with the more specific members of the field. In this sense, specific LF values, derived from the meaning shared by several lexemes of the field, are extracted from the individual entries of the lexemes in question and listed in the entry of the generic lexeme in accordance with the general principle of Lexical Inheritance (Mel'cuk 1996: 76): All lexicographic data shared by a family of semantically related lexical units should be stored just once - under one LU [lexical unit] of the corresponding vocable or under the generic LU of the corresponding semantic field, from where these data can be "inherited" in each particular case. Mel'cuk (1996: 97), nevertheless, underlines the high degree of idiosyncrasy in lexical co-occurrence, and warns against thinking that it is always possible to state the semantic conditions for the distribution of LF shared values. Lexical inheritance can also be both semantic and syntactic. Semantic inheritance is based on semantic classification of lexemes. For syntactic inheritance, lexemes with identical semantic and/or syntactic properties are grouped into classes such that the syntactic characteristics of those classes are formulated directly for each class. Although Mel'cuk—Wanner (1996: 210-211) note that syntactic and semantic inheritance classes overlap as well, they dismiss this as a rather constrained phenomenon. However, as shall be seen, it is exactly this type of phenomena that we wish to explore, and which lies at the base of the FLM model of the lexicon.

3.2. The FLM: paradigmatic axis The FLM uses the term lexical domain instead of semantic field. We have purposely avoided this label because as has been made obvious in the preceding sections, the concept of semantic field has been and still is the object of much imprecision. Our lexical domains are somewhat different from traditional semantic fields in that criteria for membership are specified and the internal structure is mapped out through the codification of both paradigmatic and syntagmatic information.

80

Lexical organization

and the FLM

The organization that we propose for the English verbal lexicon is based on the distinction between syntagmatic and paradigmatic relations, or the complementary principles of combination and selection (Saussure [1916] 1990; Lyons 1977: 241). Syntagmatic relations are those holding between elements that occur in combination with one another in linear sequences, while paradigmatic relations are those based on the potentiality of occurrence of elements in such combinations. We have found this distinction highly relevant because it is found in conceptual organization prior to and independently of the linguistic system (Nelson 1985: 179). As the semantic architecture proposed should have psychological validity, the basic structural principle involved must necessarily be that of our conceptual system. This distinction is also being used by others as a basis for lexical structure (e.g. in EuroWordNet). For example, Alonge (1998: 386) states that the conception of semantic fields in EuroWordNet is taken from Cruse (1986), but her description is evidently based in some way on the convergence of paradigmatic and syntagmatic relations: The point of view we have assumed is very similar to Cruse's approach to the meaning of a word (Cruse 1986). In his view the meaning of a word is conceived of as a kind of "semantic field", containing all of the possible (grammatical) sentential contexts of the word and all of the possible (grammatical) substitutes within those contexts. Although paradigmatic and syntagmatic relationships are complementary and inseparable in the description of meaning, they will be discussed separately purely for heuristic purposes. We shall begin with a description of the paradigmatic axis because in the FLM, a verb's semantic class, or the area of meaning it belongs to, is crucial in the delimitation of its syntactic properties as well as in its relations with other lexemes. The paradigmatic axis of the FLM Lexicon not only codifies how lexemes are arranged on the axis of selection, organizing them onomasiologically in a hierarchy of domains and subdomains, but is also a determining factor in their syntax. Traditional onomasiological dictionaries order lexical items in terms of conceptual meaning, while

The FLM: paradigmatic axis 81 semasiological dictionaries (of which alphabetical dictionaries are the most prominent subtype) classify lexemes according to form. Onomasiological organization has many advantages, not the least of which is its similarity to the organization of our mental lexicon. Kittay—Lehrer (1992b: 14) point out the multiple benefits of a lexicon organized in terms of meaning. For example, a structured lexicon would provide a way of looking at lexical universals, as well as principles of lexicalization and translation. It would also contribute to the formulation of hypotheses about the psychological reality of an internal lexicon and as a result, contribute positively to the construction of a machine-readable lexicon. The alphabetical organization prevalent in dictionaries has the drawback of not being psychologically relevant, and has little relation to the way words are actually stored within our mind. Psycholinguistic experiments such as word association and semantic priming have shown that semantic relations are a function of memory, and that words with related meanings are stored near each other in the mental lexicon (Aitchison 1994). A relational approach to the lexicon focuses on semantic domains and accepts the supposition that there are common properties that bind the items in a domain together, as well as properties that differentiate them from each other. Such a model represents human semantic memory as a network in which each node is a concept and concepts are linked together by a variety of semantic and lexical relations (Ms—Litowitz—Evens 1988: 263). In this respect, the verbal lexicon is different from the noun lexicon. Fellbaum (1998c: 70) points out that in the organization of verbs in WordNet, one of the factors to contend with was the fact that unlike noun-related concepts, verb concepts have no established set of lexical and semantic relations readily available. However, she found that the principal structural relationship for verbs is that of manner or troponymy. This relation can be represented by the formula, to W\ is to Υ2 in some particular way. The question is how to find the type of onomasiological organization most in consonance with that of our mental lexicon. Despite the fact that this was the modus operandi in WordNet, we do not believe that it is possible to find psychologically adequate areas of meaning by elaborating an inventory of semantic classes ad hoc in order to fit words into a lexical straitjacket. Since direct observation of our mental processes is impossible, one way to have an idea of the way

82

Lexical organization and the FLM

cognition works is to use the patterns in lexical structure as a window into the mind. In the traditional thesaurus, macro-areas of human experience are established a priori by the lexicographer, who then groups words accordingly. For example, Roget's Thesaurus (1987: xxxiv-xxxv) contains six main conceptual classes: (i) (ii) (iii) (iv) (v) (vi)

Abstract relations Space Matter Intellect: the exercise of the mind Volition: the exercise of the will Emotion, religion and morality

Such a thesaurus is based on a top-down (or concept-driven) type of processing. In this type of dictionary, we have no assurance that the descriptors elaborated by the lexicographer are really inherent cognitive reference points. This uncertainty can be resolved to a certain degree by using a bottom-up (or data-driven) type of analysis. In this way the areas of meaning or lexical domains established have a more solid basis and are not the product of intuition. The inventory of lexical domains that we have arrived at through this method differs somewhat from those normally found in traditional thesauri (Faber 1994b). For example, one semantic class that undergoes a radical change of location when classified according to the Functional-Lexematic Model is that of foretell, predict, forecast and prophesy. The Longman Lexicon of Contemporary English includes these words among those related to time, whereas Roget's International Thesaurus classifies them as mental attitudes. However, if we look closely at their definitions, we find that they do not really fall into either category.

The FLM: paradigmatic axis

83

(57) To say that something will happen '—foretell to say correctly that something will happen in the future. predict to foretell something accurately, based on ι previous knowledge/experience. I • forecast to predict something (of weather and events difficult to foretell). ->. prophesy to foretell something, especially through supernatural inspiration. The initial lexical organization in our model is based on definitional analysis. Verbs with definitions having the same genus belong to the same lexical domain or subdomain. The lexical hierarchy in (57), which encodes the possibilities of selection on the paradigmatic axis for this tiny area of meaning, shows these verbs to be either direct or indirect troponyms of say. As speech verbs, they are classified within the subdomain to say that something will happen (not implying responsibility of the speaker) as opposed to other subdomains such as to say that something will happen (implying responsibility of the speaker), which includes verbs such as promise, vow, swear and guarantee. The only possible justification for the thesaurus classification of these verbs can be found in the adverbial modification of the verbs, or most likely in the intuition of the lexicographer. The tracing and construction of lexical hierarchies through the analysis of dictionary definitions in itself is not new. Amsler (1980) did precisely that in order to derive hyponymic information about nouns. The basic premise is that the information in dictionaries can be thought of as constituting a huge relational network or grid which must by necessity have a non-trivial correspondence with the knowledge being expressed. These chains reveal the properties or knowledge specified in the differentiae of the more general words in these chains, so that they can be inherited by the more specific words. (Meijs—Vossen 1992: 144-145). Within the framework of the Acquilex Project, Vossen (1995) explains how such hierarchies can be extracted from machinereadable dictionaries (MRDs) in Dutch and English. In WordNet, lexical hierarchies have also been reconstructed by following the trail of hyperonymically related synsets:

84

(58)

Lexical organization and the FLM

{robin, redbreast} @ (Miller 1998: 25).

{bird} @

{animal, animate_being}

In (58) the parentheses indicate a synset, and @ —• is the semantic relation is-A, IS-A-KIND-OF. Both robin and redbreast are a kind of bird, and correspondingly, bird is a kind of animal or animate being. The question is whether this kind of hierarchical structure can also be applied to concepts of a more abstract nature such as verbs. Our study focuses on verbs because verbs provide the relational and semantic framework for their sentences, and thus can be regarded as the most important lexical and syntactic category of a language (Fellbaum 1990: 278). In a parallel way, we believe that verbs are also the most basic category in paradigmatic structure because the macrostructure implicit in the verbal lexicon also influences the organizational pattern of other categories as well. In the same way as others have done for the noun lexicon, we posit a lexical inheritance system for verbs in which meaning constituents are inherited as well as syntactic behavior. Our position, in this respect, is somewhat stronger than the one taken in WordNet in which syntactic inheritance is seen as a tendency rather than as a systematic organizational parameter: In the case of verbs, too, some knowledge about a verb concept is inherited from its superordinate. For example, the troponyms of (one sense of) the verb communicate inherit their argument structure from their superordinate (three arguments linked to nouns denoting a communicator (a source), a message, and a recipient) (Fellbaum 1998c: 91). Another important difference regarding other types of conceptual organization is that FLM lexical architecture is determined by working upward from words, not downwards from concepts. For this reason, we have given our set of generic terms the status of cognitive reference points. Givon (1984:13) affirms that the generic, culturally-shared knowledge stored in the lexicon pertains to relatively stable phenomena, concepts or points of reference, which constitute an intricate network that is part and parcel of our cognitive map of the phenomenological universe.

The FLM: paradigmatic axis 85 The structure of the lexicon is thus ideally based on conceptual invariants, though in practice, the status of these generic terms is somewhat more difficult to define. The capturing of the invariant has been the subject of much discussion in lexicography. Wierzbicka (1993: 51) has likened it to a search for truth, and has centered all her work on establishing a set for semantic primitives in terms of which all other words can be defined. In contrast, Lewandowska-Tomaszczyk (1987) takes a more cognitively oriented approach, arguing for the universality of a number of basic human conceptions and experiences, such as universal image schemata and sets of psycho-physiological preferences. In WordNet the generic terms are called unique beginners, and they are established previous to the actual classification. According to Miller (1998: 28), an important criterion in the choice of primitive semantic components is that collectively, they should provide a place for every lexical item, but he obviously makes no strong claims about their status as crosslinguistic generic terms. In this respect, our position is somewhere between Wierzbicka and Miller. We consider our inventory of superordinate terms to be neither conceptual invariants nor a heuristic set that varies with each researcher, but rather "near primitives" in the sense of Apresjan (1993: 86): In the majority of languages there are semantic primitives, or rather, near primitives, to describe the basic concept of each system: "to perceive" for perception, "to do" for bodily activity, "to think (that)" for intellect, "to want" for will, "to feel" for emotions, and "to say" for speech. However, these near-primitives are not necessarily the simplest semantic constituents imaginable. Apresjan maintains that such near primitives are really not cross-cultural, but this depends on which area of the lexicon one is talking about. Pinker (1997: 189), for example, claims that all human cultures ever documented have words for the elements of space, time, motion, speed, mental states, flora, fauna, weather, and logical connectors. This means that presumably certain domains are more basic than others, and if primitives can be said to exist, these would be the areas of meaning where they are most likely to be found. Such semantic near-primitives reflect conceptual

86

Lexical organization and the FLM

patterning (the naive picture of the world), an organization that does not necessarily coincide with the scientific model of reality. For example, in the domain of VISUAL PERCEPTION, the folk model of vision encoded in language is very different from what really occurs when we look at something. In the scientific model of visual perception, the eye is a passive receiver, whereas in the model encoded in language, the eye has a more active role (Faber—Perez 1997). Although this model of visual perception exists in English, Spanish, French and German (all SAE languages), obviously, more extensive research would have to be carried out before any type of crosscultural validity could be postulated. This type of semantic analysis at an intermediate level has the virtue of distinguishing "recurrent chunks of semantic molecules" in the sense of Goddard (1998: 254). We make no claims for universality, however, and admit to being increasingly skeptical about the feasibility of ever developing an ontology valid for all languages. If such a Herculean task could ever be accomplished, the result would be nothing less than a consensus on all concepts, relations and implications across all languages and cultures. It is for this reason that Vossen—Bloksma (1998: 400) take the position that wordnets are lexical ontologies rather than conceptual ontologies. The FLM offers an overview of the map of lexical meaning because it covers all of the domains within the English verbal lexicon, not only those that best lend themselves to such representation, and systematically relates categories with each other. With the excuse of fuzziness and lack of time or space, there have been few attempts to map out the macrostructure of lexical categories in detail. Without having a global idea of lexical design, it is very difficult to make any sort of significant statement about the lexicon. The lexicon ends up being described in very much the same way as the five blind men trying to describe an elephant. Depending on the part each happens to be touching, the elephant (or in this case, the lexicon) appears to be something totally different. In the FLM the lexeme is the central unit of linguistic description. The terms, lexical unit and lexeme, are used interchangeably in the sense of "a word taken in one well-specified sense and supplied with all the information specifying its behavior when it is used in this sense" (Mel'cuk 1988). Each lexical unit is conceived as a process in itself containing a variety of different types of

The FLM: paradigmatic axis 87 information at various levels. Hierarchies of lexemes, all of which share the same nuclear meaning and syntax, form lexical domains. This is in evident contrast to Ruhl (1989: 74), who makes the rather vague affirmation that a semantic domain is a superordinate with its hyponyms, but offers no formal criteria of how to determine domain membership. In our model, the genus in the definition of each lexeme marks the semantic territory covered by a specific domain or subdomain, and thus is the factor that determines lexical domain membership: (59)

Principle of Lexical Domain Membership: Lexical domain membership is determined by the genus, which constitutes the nucleus of the meaning of a lexeme.

Vossen (1995) organizes the noun lexicon in semantic fields by automatically generating hierarchies from MRDs and manually disambiguating the most frequent definition heads. He bases his model of lexical organization on the same classical structure of dictionary definitions as we do. Assuming that lexicographers use a substitution principle in writing definitions and assuming that definitions have a classical structure, we can use the definition heads to build up hierarchies of denotationally overlapping words. Applying inheritance to these relations ultimately leads to a fully semantic database (Vossen 1995: 227). Although we evidently agree with the substitution principle, we have serious reservations about the automatic extraction of information from only one dictionary. In our opinion, various dictionaries should be used and composite definitions formulated. In the case of verbs, syntactic behavior should be taken into account as well. Each FLM lexical domain has one or two generic terms or superordinates, in terms of which all the members of the domain are directly or indirectly defined. The fact that a lexical domain can have more than one superordinate is also seen in WordNet, which frequently has more than one unique beginner in each domain. Generic terms do not correspond to the entire semantic content of the domain, but rather

88

Lexical organization

and the FLM

are the nuclei of content modified by different lexemes. Examples of superordinate terms or multiword phrases in each domain are the following: (60) (i)

to be [EXISTENCE]

(ii) (iii)

to become different [CHANGE] to have/give [POSSESSION]

(iv)

to say

(v)

to feel [EMOTION] to do/make [ACTION] to know/think [COGNITION/MENTAL to move (go/come) [MOVEMENT]) to become aware (notice/perceive)

(vi)

(vii) (viii)

(ix)

[SPEECH]

PERCEPTION]

[GENERAL PERCEPTION] (x) (xi)

to see/hear/taste/smell/touch to be/stay/put [POSITION]

[SENSE PERCEPTION].

Generic terms are determined through Stepwise Lexical Decomposition (Dik 1978b), and a semantic hierarchy established within each lexical domain. This is also in consonance with Mel'cuk's Decomposition Principle, which stipulates that the definition of a lexical unit L must contain only terms that are semantically simpler than L (Mel'cuk 1988). Wierzbicka (1992a: 11) writes: Explicating involves reducing semantically complex words to semantically simple words, and hence the words used in an explication are not selected at random: there is a hierarchy among words, and a correct definition will reflect this hierarchy. Lexemes specify the nuclear meaning of the domain that they belong to through their differentiae, which in turn are indicative of semantic parameters codifying human perception and evaluation. These minimal differences in content are the basis upon which the lexemes are interrelated in domains and subdomains.

The FLM: paradigmatic

3.2.1.

axis

89

Definitional analysis and lexical decomposition

In contrast to Vossen (1995), we have chosen to elaborate our hierarchies manually, and incorporate information concerning complementation patterns as well. Wilks et al. (1996) claim that hand-building lexical resources is neither necessary nor desirable when they can be derived automatically from MRDs, but Fellbaum (1998d: 240) points out that MRDs are hardly theory-neutral or any less intuitive than traditional dictionaries. In fact hand-built lexicons have two important advantages: • •

Enriched entries can be created, which will be an improvement on those found in standard dictionaries. A consistent format will be more conducive to extracting relevant information (Fellbaum 1998b: 4).

The elaboration of lexical hierarchies through the analysis of similarities in the complementation patterns and argument structure of their members is another important aspect of our model. The necessity of using this type of information structure to verify such hierarchical structure has its explication in cognition. Palmer—Polguere (1995: 231) write: Criteria for building a conceptual description are linked to arguments borrowed from "real world" considerations or expressed in terms of denotation. What are the necessary conditions in the environment that allow the situation/event denoted by the verb to take place and what are the effects on the environment that are produced by this situation/event? Accordingly, a conceptual representation of a verb necessarily includes a specification of the number of arguments, their obligatoriness, and their semantic characteristics. An adequate meaning representation should evidently contain sufficient information to enable an accurate construction of the concept, but according to Miller—Gildea (1987), definitions found in most standard dictionaries often do not meet this admittedly difficult

90

Lexical organization

and the FLM

requirement. This is hardly the fault of the lexicographer, who must contend with practical restrictions of space and is, more often than not, forced to reduce an initially longer and more complete definition to a bare two lines. For this reason, Jackson (1988: 60) affirms that the meaning of a lexeme given in a dictionary entry must be regarded as "potential", a distillation of the essentials, awaiting actualization in a particular linguistic and situational context. Precisely because definitions are so succinct, one dictionary is not sufficient for the extraction of meaning. To obtain the meaning components, through which the definition of each lexical unit is elaborated, we found it necessary to consult several of the most widely used dictionaries. In this way, it was possible to obtain structured meaning definitions, which constitute a consensus of the information included in the most authoritative lexical resources. Hand-building definitions also facilitates the formation of hierarchies with a high degree of coherence, which are the basic structural relation in our lexical domains. According to Martin Mingorance (1984: 229), extracting semantic information from dictionaries is justified because they contain the body of knowledge gathered by lexicographic tradition, and thus are a treasure house of information for determining intermediate semantic constituents. Since their definitions have referential authority for users of the language in question, they provide a basis for extracting the stocks of more generic terms, which are intuitively felt by most speakers to be close to the status of conceptual reference points. B. Levin (1991: 20) also attributes dictionaries a fundamental role in the construction of a relational model of the lexicon: Monolingual collegiate-size dictionaries of a language can contribute information about a word's possible senses, and, for each sense, the relevant semantic class and the genus and differentiae that constitute its definition. Vossen (1995: 24) similarly emphasizes the usefulness of dictionaries as an empirical resource. He points out that since there is no theory of language that yields a complete descriptive apparatus, looking at dictionary data can, to some extent, help to fill in the blanks in such a theory. In fact, extracting information from

The FLM: paradigmatic axis 91 dictionaries is one of the methods used in projects such as EuroWordNet (Alonge 1998: 386). Although dictionaries contain all of the information necessary to create a relational network of the lexicon, the way the information is presented often does not facilitate its access. Some of the inconsistencies found in dictionary definitions can be attributed to the fact that there is a certain divorce between theory and practice in lexicography. Atkins—Kegl—Levin (1988: 100) underline the need for a theory of lexical organization to provide the context for building a good lexical entry, and attribute certain shortcomings in dictionaries to the lack of theory in the semantic-syntactic interdependency area. However, despite such criticism, dictionaries are an extremely valuable resource in any type of lexical research, and not using them because of certain understandable limitations would be a little like throwing the baby out with the bath water. They are undoubtedly the first place one must look in order to find information about meaning. The search for the most pertinent meaning components of each lexeme is a complex process since good defining practice dictates that a word cannot be defined in terms that are more complex than the word itself (Landau 1984; Zgusta 1971). The semantic constituents used in the FLM are natural language phrases found in definitions. Each verb is also associated with its inventory of complementation patterns or different activations of its argument structure, and thus has an underlying schema indicative of its quantitative and qualitative valency. In order to arrive at a verb's valency description, or the number and semantic characterization of its arguments, information in various monolingual dictionaries is analyzed. Definitions are segmented and the constituents compared and contrasted in order to find those that are the most appropriate, and semantically, the least complex (Faber—Mairal Uson 1997b). A practical example of this process is described in the following section.

3.2.1.1. Segmentation of dictionary information: embezzle Embezzle has a schema with very specific arguments, and also encodes a particular sociocultural context. As a troponym of steal, embezzle, belongs to the subdomain to cause to stop having [POSSESSION], which

92

Lexical organization and the FLM

like all negative subdomains in the lexicon, is highly lexicalized. Other verbs which share this same semantic space are rob, defraud, filch, pilfer, purloin, shoplift, lift, and swipe. The meaning of embezzle is the result of (i) semantic distinctions which differentiate it from its cotroponyms; (ii) its inventory of complementation pattern(s). In other words, the meaning of a lexical unit is a product of its lexical semantic structure (paradigmatic axis) and its combinatorial properties (syntagmatic axis). The dictionaries used to extract semantic and syntactic information are Collins Cobuild (CC), The American Heritage Dictionary (AHD), Longman Lexicon of Contemporary English (LLCE), Oxford Advanced Learners' Dictionary (OALD), Longman Dictionary of Contemporary English (LDCE), Webster's New World Dictionary for Young Readers (WDYR), and The Random House Dictionary of the English Language (RHD). These dictionaries give the following definitions of embezzle. (61)

embezzle

CC

If someone embezzles money, they take or use it illegally for their own purposes, when it belongs to a company or organization that they work for.

AHD

to take (money, for example) for one's own use in violation of a trust.

LLCE

to use (money placed by others in one's care) for one's own purposes, usu. unlawfully.

OALD

to use (money placed in one's care) in a wrong way to benefit oneself.

LDCE

to take and use for oneself in the wrong way (money that is placed in one's care).

WDYR

to steal money that has been placed in one's care.

RHD

to appropriate fraudulently to one's own use, as money or property entrusted to one's care.

The FLM: paradigmatic axis

93

In order to be able to ascertain the most appropriate semantic constituents, the definitions are segmented and put under the following headings: nuclear meaning, agent, goal, source, recipient, manner, and usage label!pragmatic information so that they can be contrasted, and thus facilitate the derivation of relevant information. Table 10.

Segmentation of the dictionary definitions of embezzle

Nuclear Meaning

Manner

take (AHD) use (LLCE, OALD) take or use (CC) take and use (LDCE) steal (WDYR) appropriate (RHD) illegally (CC) unlawfully (LLCE) fraudulently (RHD) in the/a wrong way (OALD, LDCE) in violation of trust/entrusted to one's care [LLCE, AHD]

Agent (A)

Somebody • works for company/organization [CC] • position of trust/direct responsibility

Goal (B)

Money • belonging to company/organization [CC] • placed/entrusted (by others) in/to one's care \placed in /entrusted to one's care: LLCE, LDCE, OALD, WDYR, RHD] Location • company/organization etc. where money is [CC]

Source (C) Recipient (D)

Beneficiary • for own purposes/benefit/oneself [CC, LLCE, OALD, RHD, LDCE]

Pragmatic information

Cultural context: modern society in which transactions are based on money

94

Lexical organization and the FLM

The definitions (61) and their corresponding segmentation in Table 10 show that embezzle has four argument slots, each with its corresponding semantic role: (62)

A [Agent] embezzles Β [Goal] from C [Source] for D [Beneficiary]

Three of these arguments are obligatory and typical of commercial transaction verbs whereas one is optional. Verbs of commercial transaction such as buy and sell are very popular as examples in lexical semantic analysis because of the number of arguments that they can take. They have been used as an example by Fillmore—Atkins (1992) in the description of their frame-based dictionary, and are also mentioned by Escalier—Fournier (1997) in their notional representation of meaning in MTM.

3.2.1.1.1.

Nuclear meaning and manner satellite

Within the category of nuclear meaning, four of the dictionaries analyzed have take and use, or various combinations of the two (take and use, take or use). This nucleus is modified by illegallyIunlawfully. On the other hand, the RHD uses appropriate modified by fraudulently, while the WYDR definition has steal as its nucleus. In the elaboration of a suitable definition, appropriate can be discarded as a possible genus since it violates our criterion of linguistic simplicity. Accordingly, words should only be defined in terms of other words whose meaning is included in their own. For example, if embezzle were to be defined in terms of appropriate, this would mean that its definition would include all the meaning components (stylistic register included) of appropriate, something which is manifestly untrue. Appropriate has a wider semantic scope for its second argument than embezzle because evidently one can appropriate other things besides money (e.g. land): (63)

The government appropriated the farmers' land in order to build the new freeway.

The FLM: paradigmatic axis 95 (64)

*The executive embezzled land from the government.

One of the meanings of appropriate is effectively to take something illegally. For example, when money is appropriated, it is either taken illegally or put aside for a special use. (65) (66)

He lost his job when he was found to have appropriated (= embezzled) some of the company's money. The government has decided to appropriate embezzle) millions of pounds for a new submarine.

Examples (63-66) show that the meaning of appropriate is more general than embezzle insofar as its semantic content goes. However, its pragmatic information is more specific because dictionaries such as Collins Cobuild classify it as a formal word whereas embezzle has no such usage label. Other possibilities are take, use, or a combination of the two. Of the two verbs, take seems to be the more basic action within embezzle, not only because intuition and common sense tell us so, but also because its complementation pattern is typical of those verbs within the general area of POSSESSION/TRANSACTION. (67)

a. b.

A (Agent) takes Β (Goal) from C (Source) for (Beneficiary). *A (Agent) uses Β (Goal) from C (Source) for (Beneficiary).

D D

The option of combining the two meanings together into one lump take and use would give embezzle a coordinated genus. We are fairly reticent about such an option because verbs with a coordinated genus are rare in the lexicon. Examples of this type of verbs are gloat (visual perception and feeling) and browse (visual perception and movement). In the meaning of both verbs, one conceptual area is equally important as the other. However, this is not true for embezzle. Despite the fact that money is embezzled for the purpose of using it in a certain way, this use is not part of the base meaning. It is certainly possible to embezzle money without using it or for no reason at all.

96 (68)

Lexical organization and the FLM a. b.

John embezzled the money from the company and left it under his mattress. John embezzled the money just for the hell of it.

If take is chosen as the genus for embezzle, it must be modified by illegally or unlawfully in order to differentiate it from other verbs whose definitions have the same genus. Illegally, though similar in meaning to unlawfully, is the better choice for reasons of simplicity. In the dictionaries consulted, unlawfully is defined in terms of illegally, and according to the OALD, has a more formal register. However, if to take illegally is selected as the nuclear meaning of embezzle, thus satisfying our simplicity requirement, we find that this definition basically coincides with the one already given to the verb steal. (69)

steal to take something away from somebody without their permission and not intending to return it.

The meaning constituents of steal in the same dictionaries are the following: Table 11.

Meaning components of steal

Nuclear Meaning

take

Manner

illegally/without the right to do so

Agent (A)

Somebody

Goal (B)

Something (possession)

Source (C)

Somewhere/somebody (location/possessor)

Pragmatic information

None

The argument structure for embezzle is essentially the same as that of steal. The basic difference between them is in their respective levels of specificity under all headings.

The FLM: paradigmatic axis 3.2.1.1.2.

97

Agent

The only possible specification for the agent in steal is the necessarily negative evaluation of anyone who would engage in such an activity. Possible fillers for this verb are agents such as a thief, robber, or burglar, whereas we have to bridge a considerable credibility gap before being able to accept slot fillers with positive values on the axiological scale, such as a nun, a two-year-old child or a volunteer worker in a leper colony. Embezzle, however, has a human agent with very definite specifications. Such an entity is human, prototypically someone responsible for somebody else's money, such as a bank employee, treasurer or company executive. Despite the fact that such people generally do not fit the typical description of thief, their actions classify them as one. The negative evaluation of the embezzler is accentuated not so much by the illegality of the action (which is part of the nuclear meaning of the verb anyway), but because of the resulting deviation from the norm of responsible behavior implicit in his/her position of authority and trust.

3.2.1.1.3.

Goal

The goal argument of steal has very few specifications. It could be any type of entity since it is possible to steal someone else's boyfriend, pet boa constrictor, new umbrella, or idea about how to get rich quick. The only selection restriction for this slot pertains to the ownership of the entity in question; it must be something that belongs to someone else, or is perceived as belonging to someone else. In contrast, the goal argument of embezzle is considerably more specific. The filler is invariably a sum of money which the agent is responsible for (because of his/her job) and which he/she takes without permission.

98

Lexical organization

3.2.1.1.4.

and the FLM

Source

Part of the argument structure of both verbs is the source from which they take the object in question. In steal, the source can be a place or person: (70)

a. b.

The thief stole the diamonds from the castle. The thief stole the diamonds from the duchess.

In embezzle, the source would typically be a company, bank, or organization which would function both as the location (where the money is) and group of people (whom the money belongs to). However, the source cannot be a single person: (71)

a. b.

The vice-president embezzled $25,000 from company. *The vice-president embezzled $25,000 from company treasurer.

the the

In fact, embezzling is often regarded as a higher or "cleaner" form of stealing not only because it involves money, but also because one does not embezzle money from a person, but from an organization. This somehow makes it seem less personal and thus less negative.

3.2.1.1.5.

Beneficiary/Recipient

Another possible argument slot is that of recipient/beneficiary. This is often left unspecified, because its default value is something that personally benefits the agent. (72)

a.

b.

The vice-president embezzled $25,000 from the company for Mother Theresa's Hospital in Calcutta [deviation from norm], The vice-president embezzled $25,000 from the company for a vacation spree in the Bahamas [specification of default value].

The FLM: paradigmatic axis

99

When the beneficiary is given, this would be either because it is a deviation from the norm (if it is not something that personally benefits the agent) (72a) or because it reaffirms the default value by giving it greater specificity (72b). After analyzing the dictionary definitions of embezzle and its argument potential, our definition is the following: (73)

embezzle [TRANSACTION]

to steal money placed in one's care for one's own purpose.

Argument structure: Agent

Goal Source Beneficiary

Somebody, working in a large organization/ company, with a position of responsibility, and in need of money. Money (usually a large sum) that the Agent is responsible for. Organization/company that the Agent works for. Self (own purpose, personal benefit).

The definition in (73) incorporates the information offered by the dictionaries consulted. In a final version, it would naturally include examples of the verb in possible contexts. Our discussion of embezzle shows why it is necessary not to only depend on one dictionary. The RHD uses a defining vocabulary that is semantically more complex than the word being defined. The AHD definition (to take (money, for example) for one's own use in violation of a trust) is too general to accurately describe the situation. Money is more than just an example of what one can embezzle; it is really the only possibility. Evidently, additional information also needs to be included about the semantic participants, as this lack of specification leaves the door open to erroneous interpretations. For example, if a 9year-old boy decides he wants to buy a bag of potato chips, and takes money out of his sister's piggy bank against her wishes after she has trusted him with the key, this would be taking money for one's own use in violation of a trust. It would thus fulfil all the conditions set in the AHD for embezzlement, yet it is obviously not embezzlement, because the other sentence elements do not fit into the verb frame. Furthermore,

100

Simon Dik's Functional Grammar

it is also important to remember that all language is embedded within a larger context. A verb such as embezzle obviously presupposes a certain type of society in which there are organizations where people place their money in the care of others with the ultimate purpose of keeping it safe or of acquiring more money. In a different type of society (e.g. Indian tribes in the Amazon rain forest) with an economy based on bartering, such a verb would obviously not exist, since such an activity would be inconceivable. The elaboration of definitions in this way, using meaning components and argument structure to relate verbs within the same area of meaning, is the framework for our lexical hierarchies. In this way, embezzle is pinpointed in an area of semantic space in terms of other words with similar meaning as well as complementation patterns. This leads to a representation similar to the one below: (74) To take something away from somebody without the right to do so steal to take something away from somebody illegally (without their permission and not intending to return it), rob to steal money/property from somebody/institution, defraud to rob somebody of their money by deception. embezzle to steal money placed in one's care for one's own purposes, purloin to steal something small [formal], shoplift to steal things from shops by taking them from the shelves and hiding them under clothes or in a bag pilfer to steal things that are small/of little value continuously over a period of time, filch to steal things secretly that are small/of little value [informal], lift to steal [informal]. swipe to steal something by removing it quickly. Embezzle is thus located in a subdomain of POSSESSION within the context of its co-troponyms, each of which encodes a different type of specification of the more generic term (74). The fact that each

The FLM: paradigmatic axis

101

lexeme has a full dictionary-type entry produces a richer lexical description than that offered in a traditional thesaurus because it is thus possible for any lexical domain or subdomain to compare the meanings of all the lexemes. Placing each verb within a paradigmatic context also makes it easier to view the lexicon as a whole.

3.2.2. Lexical hierarchies

Possession is not a basic concept, but one constructed from other more primitive ones. Whereas movement and action refer to observable processes, possession is derived by means of the perception and interpretation of abstract relations. According to Jackendoff (1994: 79), possession is an artificial relationship established between two entities, one of whom has the right or authority to use the other as he wishes, to control anyone else's use of the other, and to impose sanctions for uses other than those he/she permits. If we take this into account, each verb in this domain can be globally classified according to which of the two entities is the point of focus in its argument structure. More specifically, as we saw in the case of steal and its troponyms, take verbs focalize the loss of the transferred object as it leaves the sphere of control of its initial possessor, whereas get verbs focalize acquisition (the final possessor in control of his new possession). The following is a representation of part of the acquisition phase of the lexical domain of POSSESSION structured paradigmatically (See appendix 3 for the whole domain): As shown in Table 12 the genus is have, and all the other verbs are defined in terms of it. The adverbial modification or differentiae in each definition encodes the semantic parameters within each domain and subdomain. In this particular hierarchy, get is the most direct troponym of have, and thus becomes the genus of other words at more specific levels of the hierarchy. To come to have is the phrase that labels the lexical subdomain in question. Lexical subdomains are thus derived from the definitional structure of its members.

102

Lexical organization and the FLM

Table 12.

Acquisition phase of possession

HAVE TO HAVE SOMETHING

possess to have something [formal], TO COME TO HAVE SOMETHING

get to come to have something as a result of some previous activity, gain to get something wanted/beneficial through effort. win to gain something by work/struggle/skill, obtain to get something as a result of work/planning. procure to obtain something difficult to find [formal], acquire to obtain something with effort, adding it to previous possessions. TO GET MONEY FOR SOMETHING DONE

earn to get money as payment for work done. FROM SOMEBODY

extort to get money from somebody by force/threats, collect to get money from a number ofpeople/places. TO GET SOMETHING BY PAYING MONEY FOR ΓΓ

buy to get something by paying money for it. purchase to buy something [formal]. To come to have can also be put within the larger context of which is organized in terms of the following phasal distinctions:

POSSESSION,

(i) (ii) (iii)

to come to have [inception of an event] to continue to have [continuance of an event] to stop having [termination of an event]

In the verbal lexicon, this temporal relation is the only type of meronymic (PARTOF) relation possible. However in the FLM, it is conceived as a structural relation in the global context of an entire lexical domain, and not merely pertaining to individual lexemes. This kind of lexical organization is graphically represented in Figure 4.

104

Lexical organization and the FLM

The subdomains in Figure 4 also contain subdivisions that are the result of the semantic parameters of their members. However, a verb's place and membership in a hierarchy is not based on its meaning definition alone, but rather on the systematic interaction of its meaning and syntax. In some ways, this resembles Systemic Functional Grammar (SFG), in which lexemes are the result of a gradual selection of features over a systemic network. The representation of a lexical item is reduced to a set of choices which lead to other choices, thus narrowing down the scope of the network until one finally arrives at one lexical item (Butler 1985: 133-134). In the FLM lexicon, semantic structure and syntactic representation are intertwined because complementation patterns validate the membership of lexical items in a specific subdomain. In this regard, the mapping between the lexical representation of a predicate and its syntactic expression is predictable in that syntax is semantically motivated. Fellbaum (1998c: 95) affirms that an analysis of verbs in terms of their semantic and conceptual components can reveal many of the verbs' syntactic properties as shown in Table 13. The superordinate term of each subdomain tends to take a greater number of complementation patterns than its more specific troponyms. Table 13.

Possession: inventory of complementation patterns of get verbs get

0 NP NP]NP 2 NPreflNP2 NP PPlor NP PP,rom

-

He went to the corner to get some milk. Can I get you a drink? I went out to get myself a new computer. Can you get a newspaper for me? I got a letter from him. gain

0 NP NPiNP 2 NP PPtor NP PPfrom

-

The Nationalist Party has gained a lot of support lately. His newly acquired fortune gained him respectability in the eyes of society. The soldiers gained control of the government for their general. What do you hope to gain from the course?

The FLM: paradigmatic axis Table 13 (cont.)

105

Possession: inventory of complementation patterns of get verbs win

0

-

He easily won first prize in the chess competition. Her performance won her much critical acclaim. He easily won their support for his plan. He won respect from her through his honesty.

NP NP]NP2 N P PPFOR N P PP A O M

obtain 0 NP NPINP2 N P PPFOR N P PPFROM

Where can I obtain a copy of her latest book? -

He obtained this record for you with difficulty. She obtained her transcript from the university. procure

0 NP NPJNP2 N P PPFOR N P PPFROM

-

He was responsible for procuring supplies in the army. She procured some milk for the children. They procured fifty tons of hay fom the local farmers. acquire

0 NP

-

He acquired antique paintings/ a reputation for dishonesty/a taste for brandy.

NP,NP 2 N P PPLOR N P PPFROM

He managed to acquire a supply of caviar for the banquet. His aunt acquired a new pair of glasses from the optician.

106

Lexical organization and the FLM

Table 13 (cont.)

0 NP NP,NP 2 NP PPfor NP PPfrom

0 NP NP!NP2 NP PPfor NP PPfrom

Possession: inventory of complementation patterns of get verbs earn He earns quite well. Her husband earns $25,000 a year. His honesty earned great respect. Her uncle earned a good living for his family. He earned a lot of money from lecturing. collect He's collecting for famine relief. The government collects taxes every year. He is collecting money for famine relief. The government collected taxes from the people every year. extort

0 NP NP,NP 2 NP PPfor NP PPfrom

0 NP NP1NP2 NP PPfor (NP) PPfrom

0 NP NPINP2 NP PPfor NP PPfrom

The Mafia was accused of extorting money.

The police used torture to extort a confession from the prisoner. buy It's a good time to buy because the market prices are low. Where did you buy that coat? At Christmas he bought her an electric drill. They bought a new television set for their mother. My neighbor has decided to buy a new refrigerator from the department store. purchase Employees are encouraged to buy stock. He decided to purchase a tricycle for his son. The tricycle was purchased from the store.

The FLM: paradigmatic axis

107

In Table 13 the more specific verbs are indented. For example, if we compare the entries of buy and purchase, purchase obviously has fewer possible complementation patterns than buy because it is more specific. This is not only true insofar as the number of complementation patterns is concerned, but also includes the inheritance of the semantic characteristics of their arguments. As shall be seen, we take a fairly strong view of lexical inheritance. FLM lexical hierarchies thus become a source of information for the explanation of the syntactic behavior of lexemes. Both axes interact, and their convergence produces a set of synsem parameters (Chapter 4).

3.2.3. Differentiae: semantic parameters On the paradigmatic axis, lexemes are distinguished from each other by differentiae, which encode the divisions and distinctions that each language makes in the semantic continuum. Many of the parameters of semantic differentiation are recurrent, and can be classified at a higher level in more general categories (Martin Mingorance 1987: 377-378). In the FLM, there are three general types of higher-level meaning parameters, each of which belongs to a different layer of the structure of the lexicon: syntactic [e.g. transitivity], semantic [e.g. intensity], and pragmatic [e.g. formality]. Transitivity is instantiated in all of the causative subdomains, and is the schema of prototypical action. As such, it has three semantic features, each of which corresponds to one core aspect of the prototypical transitive event (Hopper—Thompson 1980; Givon 1995: 76): (i)

Agent: A controlling agent who is responsible for causing the event. (ii) Patient: A patient who is acted upon and who registers the event's change of state. (iii) Verbal modality. A verb that encodes an event that is perfective, sequential, and real (not hypothetical). Intensity is a semantic parameter that is present in a wide variety of domains. When applied to the domain of SOUND, it structures verbs

108

Lexical organization and the FLM

on a scale from murmur to shriek. In OLFACTORY PERCEPTION, the same parameter differentiates smell from stink. In other words, the value it encodes is a function of the meaning area where it operates. The pragmatic parameter of formality is also recurrent, and differentiates pairs of lexemes such as capture!apprehend or buy/purchase. The use of one lexeme or the other strictly depends on pragmatic features derived from the context of the speech act. Within the FLM framework, Jimenez Hurtado (1994) provides the basic classification used for the analysis of pragmatic information in the lexicon. In reference to our example of POSSESSION, semantic distinctions encoded in differentiae are of various types. In this case, they are manner specifications (through instrumentality or specification of intensity), optional argument focus (source), or pragmatic parameters such as formality. Table 14 illustrates how syntactic patterns become more and more specific as we move down the semantic scale. (See Appendix 4 for synsem interface of the entire domain). It also shows the semantic distinctions encoded in differentiae, which generate this specificity.

Table 14.

Interaction of semantic distinctions and syntactic complementation benef

goal 0

NP

TO COME TO HAVE

To come to have something [instr] by struggle /skill

[GOAL]

[instr] by work/ planning [prag] formality [manner] effort

GET gain win

obtain procure acquire

source

NP!

NP

(NP)

NP2

PPfor

PPfrom

-

+

+

+

+

-

+

-

+

+ +

+

—

+

+

—

+

—

+

+

-

+

-

+

-

+

-

+

+ +

The FLM: paradigmatic axis Table 14 (cont.)

109

Interaction of semantic distinctions and syntactic complementation

[GOAL] money [exchange] for sth. done [source] from sb. [source spec] many people [instr] threats/ violence [INSTR] by paying money [prag] formality

earn

+

+

+

+

+

collect

+

+

—

+

+

extort

—

+

—

—

+

buy

+

+

+

+

+

—

+

—

+

+

purchase

3.2.3.1. Manner-of-walking verbs The importance of the information found in recurrent semantic parameters cannot be overstressed, since the patterns implicit in them encode the way we process the sensory data received from the outside world. Such patterns are strongly linked to perception, and explain the abundance of manner verbs found in the lexicon (manner-of-moving, manner-of-speaking, manner-of-giving/taking, manner-of-laughing/ crying, manner-of-feeling, etc.). A case in point can be found in the verbs of manner-of-walking. The number of lexemes in this particular subdomain of MOVEMENT, as well as the complexity of its semantic distinctions are indicative of how influential perceptual information is in domain structure: (75)

To move using one's feet walk to move putting one foot in front of the other fairly slowly on the ground. stamp to walk quickly with loud, heavy steps, usually because of anger.

110

Lexical organization and the FLM stomp to stamp with very heavy steps, stump to walk with heavy steps in a clumsy way (as if with a wooden leg), pad to walk quickly and softly with fairly heavy steps, trip to walk quickly and lightly, plod to walk slowly with heavy steps, making an effort, usually because of tiredness, trudge to plod with very heavy steps because of tiredness/ unhappiness. amble to walk slowly in a relaxed way. ramble to walk slowly and without purpose for a long time, usuallyfor pleasure, stroll to walk slowly for pleasure [short distance]. promenade to walk slowlyfor pleasure, especially to be seen in a public place [formal] [oldfashioned], saunter to walk slowly in cm unhurried way. stride to walk quickly with long steps in a hurried way. waddle to walk with short, quick steps, moving slightly from side to side [like a duck], shuffle to walk without lifting one's feet from the ground, tiptoe to walk very quietly and carefully on one's toes, pace to walk with regular steps back andforth across a space. march to walk with very regular steps like a soldier, parade to march in a parade, usually on an important occasion /to display something, file to walk one behind the other in a line, swagger to walk proudly with a swinging movement, prance to walk with exaggerated movements to be admired. mince to walk with short, quick steps in a very affected/ effeminate way. strut to walk proudly so as to be noticed sidle to walk uncertainly and cautiously so as not to be noticed. stagger to walk very unsteadily as if about to fall.

The FLM: paradigmatic axis

111

totter to walk unsteadily usually because of being ill or drunk. limp to walk with difficulty in an uneven way, usually because one leg/foot is hurt, hobble to walk with difficulty in a clumsy way, usually because one'sfeet are hurt, hike to walk a great distance through rural areas, usuallyfor pleasure or exercise, tramp to walk heavily in a particular direction/along roads or streets, loiter to walk up and down without any real purpose. The hierarchy in the subdomain in (75) is flat, something typical of manner subdomains in general. As Fellbaum (1998c: 80) observes, troponymy is a kind of entailment since Vi as a troponym of a more generic term V2, also entails V2. Thus all manner-of-walking verbs entail walk. In contrast, although walking evidently entails stepping, walk cannot be considered a troponym of step. Fellbaum (1998c: 80) writes: In contrast with pairs like march-walk, a verb like walk entails and is included in step, but is not a troponym of step·, snore entails sleep, but is not a troponym of sleep. The verbs in these pairs are related only by entailment and proper temporal inclusion; verbs related by entailment and proper temporal inclusion cannot be related by troponymy. The similarity of meaning of the lexical items here is in direct relation to the number of shared semantic parameters. Manner-of-V-zwg verbs also signal the presence of a perceiver whose immediate sensory environment is affected in some way by this movement. Needless to say, what is lexicalized is what most catches the perceiver's attention. This is generally a deviation from the established norm. In fact, from the structure of the subdomain, we can deduce the default values for the action observed by analyzing the different semantic parameters: (i) (ii)

quick vs. slow movement long steps vs. short steps

112

Lexical organization and the FLM

(iii) (iv) (v)

heavy steps vs. light steps regular steps vs. irregular steps purposive vs. non-purposive movement

The oppositions explicit in the definitions of the lexemes indicate that walking is normally done in an upright position, and deviations from this position are negatively evaluated (stagger, totter). Walking is prototypically not a silent activity because as the definitions illustrate, there are ways of doing it softly as well as loudly. Consequently, steps can be said to have a normal level of sound, as well as a normal size/length. The normal speed for this type of movement is slow as specified in the definition of the generic term, walk. Walking is generally goal-directed, because when it is not, specification is necessary (ex. ramble, amble, loiter). In fact, walking without a specific destination can even be negatively evaluated in certain cultures. (Loitering is a legally punishable offense in certain states in the USA.) As people who walk move their feet without making excessive movements with other parts of their body, this type of movement, when it occurs, is perceived as significant in some way. Deviations from the norm become lexicalized because in a manner of speaking, they create a bigger splash in the perceiver's sensory environment. Once the perceiver's attention is caught, the movement is observed and categorized as one type or another. However, in this process, there are other factors being evaluated as well. These verbs also encode information about how this movement is different from the more generic type as well as information about the physical/emotional characteristics of the person who is moving. The meaning definitions of these verbs show us that the two senses involved are sight and hearing. There are three possible areas of focus: (i) (ii) (iii)

The agent (the animate entity who is moving) The motion itself (the degree to which it differs from the norm) The path followed/the reason for moving

Although most of the verbs in this set have sight-perceived differentiae as part of their definitions, a small group have aurallyperceived differentiae. The opposition heavy vs. light can be interpreted

The FLM: paradigmatic axis

113

in two different ways. Heavy steps can mean that the person walking is making an effort (plod, trudge) or a loud noise (stamp, stomp). There is also a difference between intentionally heavy steps (stamp) and steps that are perceived as heavy because of the size of the person/animal who is moving (pad). Such differentiae also combine to characterize both the motion itself (the degree to which the action in question is regarded as a good example of walking) as well as the person who is carrying out the specified action. This evaluation is generally negative since in most cases, it is the result of a deviation from the default values. In walking, such deviations translate into excessive noise and movement: Table 15.

Semantic and perceptual parameters in manner-of-walking verbs

Feeling/state attributed Deviation from norm to the agent

Motion perceived

annoyance

excessive loudness

stamp

anger

excessive loudness

stomp

tired

excessive slowness

plod

tired/unhappiness

excessive slowness

trudge

boastfulness

excessive body movement swagger

pride

excessive body movement strut

arrogance

excessive body movement prance

weakness/drunkenness

lack of uprightness

stagger

weakness/drunkenness

lack of uprightness

totter

Finally we can see how perception encodes cultural information because one reason for walking (though certainly a negatively valued one) is to be seen (swagger, strut, prance, parade). To want to be seen or to call attention to oneself through excessive body movement is not viewed very positively because it is considered a sign of excessive pride.

114

Lexical organization and the FLM

The parameters of these verbs are only semantic because they do not significantly affect syntax. As a result, in this type of semantic class, it is difficult to differentiate such verbs at a conceptual level. Despite JackendofFs (1990: 88) assertion that it is not the business of conceptual structure to encode different manner types, the question still remains as to how to distinguish between such verbs at a deeper level. Siskind (in Dorr 1993: 126) has elaborated a manner component for spatial verbs that is defined in terms of spatial notions, but offers no solution for more abstract predicates such as those of cognition or feeling. Dorr (1993: 126-127) proposes combining JackendofFs LCS with the use of macros such as BELIEVINGLY or WANTINGLY in order to avoid some of the well-known problems of extreme decomposition, such as the potential for deep recursion inherent in the Conceptual Dependency Framework of Schank (1973). As Schank (1973: 201) notes, deep recursion is particularly evident in instrumentality. He observes that if every ACT requires an instrumental case which itself contains an ACT, it is impossible to ever finish diagramming a given conceptualization. According to Dorr (1993: 127), the use of macros avoids this problem to a certain extent by leaving out the detailed mechanics underlying the modifying action. A speaker's choice in this subdomain depends on purely perceptual and evaluative information. Nevertheless, there are many other subdomains in which synsem parameters are activated because a verb's complementation can be determined through its semantics (Chapter 4). The paradigmatic axis is a rich source of meaning, and one that has not been sufficiently explored, particularly in regard to its interface with the syntagmatic axis. By analyzing lexemes in the context of their semantic class, lexical representations can be greatly enhanced.

3.3.

The FLM: syntagmatic axis

The syntagmatic axis specifies the complementation patterns for each verb. In this respect, complementation is a matter of matching the semantics of the higher predicate, as specified in the lexical subdomain in question, with the semantics of the complement phrase. Far from

The FLM: syntagmatic axis

115

being a mere inventory of syntactic patterns, the syntagmatic axis illustrates the extent to which semantic information on the paradigmatic axis is relevant to the form and function of verb complementation when it is analyzed within the larger framework of a lexicon. Meaning definitions are thus brought to the foreground and given the importance they deserve. Besides providing information about meaning relations, lexical organization leads to important insights regarding the internal complexity of complements and their relationship with the matrix predicate (Mairal Uson 1993).

3.3.1. The semantic characterization of complement selection The three basic issues in complement selection are the design of a lexical entry, the role of the lexicon as a source for predicting syntactic information, and the interface between syntax and semantics. In one way or another, complementation and how it is envisioned is at the root of all of these questions. At the risk of oversimplification, the dominant theoretical approaches and their view of complementation can be summarized by three assertions: (76)

a.

b. c.

Complementation and complement selection is exclusively syntactic (Chomsky 1965; Rosenbaum 1967; Bresnan 1970, 1979); Complementation is semantic, but actual complement selection is syntactic (Grimshaw 1979); Complementation and complement selection is exclusively semantic (i.e. Givon 1980, 1984, 1989, 1990; Dik 1997ab; Foley—Van Valin 1984; Van Valin—Wilkins 1993; Van Valin 1993ab; B. Levin 1995; B. Levin—Rappaport 1986).

The exclusively syntactic approach (76a) accounts for complement selection syntactically in the subcategorization frames of each lexical entry. Bresnan (1970, 1979), for example, encodes it in the subcategorization of complementizers, which is specified at the level of deep structure in terms of syntactic configurations:

116

Lexical organization and the FLM

(77)

remember

[—that] [—wh]

(78)

understand

[—that] [—wh]

In such models, each lexical entry contains a list of syntactic environments for the predicate, and no explanation is offered as to why a given set of complements appears within a lexical entry. The result is that the lexicon is more an "ad hoc tally sheet" (Van Valin—Wilkins 1993) than a source of information for the prediction of syntactic rules. Option (76b) views complementation from a slightly more semantic perspective. A typical example of this orientation can be found in Grimshaw (1979), who claims that the combination of higher predicates and their complements is dependent upon subcategorization and semantic selection, two types of co-occurrence restrictions that operate independently. Subcategorization reflects the restrictions that hold between the higher predicate and the syntactic form of its complement phrase, whereas semantic selection reflects the constraints of the higher predicate on the semantic type of its complement. Complements of the same semantic type are selected by similar predicates, and correspondingly, complements of different semantic types are selected by different (though not necessarily unrelated) predicates. Within such a framework, the lexical entry for know has the following format: (79)

know

[—S'] [—P] [—Q] [-E]

In (79) information related to the semantics of the different complement types (i.e. propositional (P), interrogative (Q) or exclamatory (E)) supplements the subcategorization frame. Although there is a timid effort to incorporate semantics, no attempt is made to relate the two frames to each other, and they appear as totally separate components. The third tendency (76c) has a purely semantic orientation, and is based on Kiparsky—Kiparsky (1970). It claims that the choice of

The FLM: syntagmatic axis

117

complement type is in large measure predictable from certain basic factors such as the semantics of the higher predicate or the semantics of the complement itself. Other lexical elements like tense, mood, and modality, are used as well to describe the semantics of the complement phrase (Langendoen 1970). For example, Ransom (1986) postulates that complement selection depends on the interaction of the modalities of the complement clause and those of the higher predicate. Despite evident differences at other levels, all of these approaches are based on the semantic motivation of complement selection. The advantage of explaining the syntactic properties of a sentence in terms of the semantics of the higher predicate is that subcategorization frames can be dispensed with altogether, given the fact that such information can be predicted from a set of linking rules (Carter 1988). In fact, studies have shown that the subcategorization frames of a verb are not the optimal source of information (Pinker 1994a). Even when such syntactic representations include the specification of argument structures and semantic roles, there is more to the picture. Grimshaw (1994: 421) goes so far as to say that without information about meaning it is impossible to know which subcategorization patterns should be grouped together, and which should be kept separate. Accordingly, lexical entries would contain semantic information relevant for the derivation of syntax. However, the principal difficulty in the exclusively semantic approach is devising the rules that link semantic information and syntactic expression. Although the form and function of linking rules vary, they fall into one of two types. They can either be based on the information encoded in a thematic grid in which each semantic argument (or function) is associated with a syntactic expression (Chomsky 1986), or they can be based on meaning definitions (Foley—Van Valin 1984; Van Valin 1993b; Jackendoff 1983; B. Levin—Rappaport 1996). The FLM also reflects an exclusively semantic orientation in which the combinatorial possibilities of predicates are described in terms of semantic restrictions following Dik's FG.

118

Lexical organization and the FLM

3.3.2. Complementation and FG Dik's FG envisions a multi-layered model of the clause, which also applies to complementation. Complements are thus characterized in terms of the highest layer they contain, as well the type of entity designated (Hengeveld 1988, 1989ab, 1990ab, 1992; Dik 1997ab; Dik—Hengeveld 1990; Bolkestein 1990, 1992). The structure underlying linguistic expressions consists of four hierarchically ordered layers, each of which designates a different type of entity, according to the classification proposed by Lyons (1977: 442^447). Each complement is classified according to whether it designates a first, second, third, or fourth-order entity. The type of layer involved in the complement is determined by the semantics of the higher predicate. The complement itself cannot contain a layer of a certain level without at the same time containing all subsequent layers. This means that in the same way as propositions can become the argument of a higher predicate, the other layers of the clause can also become the arguments of predicates. Thus, predicate frames have the format in (80): (80)

Predp ... (αη) ... (α = a variable ranging from x, e, X, and E)

In the multi-layered model of the clause, complements are of the types in examples (81-84): (81)

(82)

(83)

(84)

say [V] (x!)Ag (EI: clause (EI))GO "That woman was never in my office," said the president. [Direct speech] believe [V] (xi)p0(Xi: proposition (XI))GO The president believes that he will not be impeached. [Proposition] perceive [V] (xi)exp (ei: predication (ei))Go The president perceived a suspicious movement in the crowd. [Perceived event] build [V] (xi)Ag(x2: individual)Go The workers built a new house. [Physical object]

The FLM: syntagmatic axis

119

For example, a speech verb like say, when used in direct speech, has a fourth-order complement or quoted speech act (81). Cognition verbs {believe, consider, think) take a third-order complement or proposition (82). Physical perception verbs (see, hear, smelt) take a second-order complement or perceived event (83). The complement of creation verbs (build, make) is a physical object located in space and time (84). Regarding complement selection, an «-type of entity contains all η-1 subsequent layers. In other words, operators with a greater scope affect those with a lesser one. (85)

a. b. c.

(π2 e;: [πι say [V] (xi)Ag (Ej:[ π 4 ILL0 (π3Χ,: [π 2 e,: [πι Predb] (EJ))Go] (eO) ((π2 e,: [πι know [V] ( x ^ (π3Χ,:[π2 ej: e t c . ] ( X j ) ) G o ] (eO) ((π2 e;: [π ι see [V] (ΧΙ)ΡΠΚΕΧΡ faej: [πι Predb etc.] (ej))Go] (ei))

The predicate frames in (85) show how three matrix predicates take constructions of decreasing complexity as their goal argument. In (85a) say takes a clausal goal argument (E); in (85b) know takes a prepositional goal argument (X); and in (85c) see takes a predicational goal argument (e). All of these differences can be represented in the lexicon when the predicates are grouped into articulated semantic classes. In the same line, a complement of an «-type layer is only sensitive to the set of operators of that layer plus the set of operators of η-1 layers, but not to operators at a higher level: (86)

a. b.

?Sam saw a shark certainly coming towards him. ?Peter knows that honestly Ann will finish the painting.

Both the examples in (86) sound strange because in (86a) a secondorder complement or perceived event (a shark coming towards him) is modified by a third-order satellite (certainly). In contrast, in (86b) a third-order complement or proposition (Anne will finish the painting) is modified by a fourth-order satellite (honestly). The layered approach to complementation provides an explanation of many restrictions on the occurrence of certain satellites and operators within complements (Bolkestein 1990; Hengeveld 1996).

120

Lexical organization and the FLM

This relation between the complement and matrix predicate in terms of these operators can be seen in Table 16 (Dik—Hengeveld 1990: 237): Table 16.

Relations between complements and matrix predicates

Complement

Matrix predicate class

E, π3 X. CertXi Des Χι π 2 ei Sim ei Post ei

Utterance (Direct Speech) Pretense, Propositional attitude, Fearing Knowledge Desiderative (Hoping and Wishing) Commentative Immediate perception Desiderative (Wanting), Manipulative

Table 16 shows that utterance predicates subcategorize direct speech (fourth-order entities), while the predicates subcategorizing third-order entities are more heterogeneous. For example, a third-order complement can be modified by a boulomaic operator expressing desire (87) or an epistemological operator expressing certainty (88): (87) (88)

I wish 1 had won the lottery. [Des] I knew that she would be elected president. [Cert]

In contrast, other predicates in this category impose no restrictions whatsoever on their complements (π3 Χι):

(89)

Mondays are usually depressing. I believe that J her uncle will go bankrupt. Popeye ate a lot of spinach yesterday.

In the case of second-order entities (perceived events), there is also more than one group. (90) (91)

I saw smoke coming out of the house. [Sim] Jack ordered Peter to interrogate the prisoner. [Post]

The FLM: syntagmatic axis

121

One group of higher predicates (90) indicates that the time reference of the complement must be simultaneous with the time reference of the higher predicate (Sim), while the other group (91) indicates that the time reference of the complement must be subsequent to that of the higher predicate (Post).

3.3.3. Complementation and the FLM Although the Functional-Lexematic Model is based on FG, it places a greater emphasis on lexical representations as the key to the determination and explanation of the syntactic representations of predicates. Within FG, the predicate frame is used for the encoding of complement meanings, and the underlying clause structure can be derived from it by means of expression rules. Nevertheless, since no use has as yet been made in FG of the information contained within meaning definitions, one might well ask exactly what role (if any) meaning representations play in the construction of underlying clause structures. One possibility is for predicate frames to include information derived from a verb's paradigmatic context. However, for such information to be useful, it has to be clearly related to a verb's syntactic behavior. One of the basic premises of the FLM is that the English verbal lexicon is organized in sets of semantic classes and a verb's membership in one lexical set or another is a determining factor in its syntactic realizations). Each lexical subdomain thus becomes a repository of both syntactic and semantic information, and as such, can be regarded as indicative of the interface between syntax and semantics. We might even go so far as to say that a subdomain represents a kernel lexical grammar, which encapsulates all the linguistic properties of its members. The first step in the codification of the syntagmatic axis is an analysis of the semantic potential of the complementation pattern/s of a given verb. This is necessary in order to be able to place the semantics of the complement phrase within the wider context of the lexical subdomain to which the lexeme belongs. With this information, it is then possible to establish explicit connections between the semantic hierarchies and the syntactic typology of each lexical subdomain.

122

Lexical organization and the FLM

Externally, complements can be classified in terms of their relationship with the matrix predicate. In FG, this is the role assigned to the operators, whose scope is greater/lesser depending on the degree of fusion between the complement and its matrix predicate. (92) (93)

John cleaned the house. "Beware of vampire bats!" Peter exclaimed.

For example, the emphasized complement in (92) is obviously more tightly bound to the matrix predicate than in (93), where it has acquired the status of a sentence fully independent of the governing predicate. The problem is how to represent this information. Among the different factors to be considered are argument obligatoriness, complement form, and tense-aspect-modality (ΤΑΜ) operator distribution. An analysis of the way these elements are encoded in complementation patterns reveals a rich taxonomy of complement meanings since the codification of the grammatical features in complement structures is largely motivated by semantic parameters. The examination of the different syntactic realizations of complements within the wider context of the lexicon gives information regarding the restrictions on the possible set of operators, and provides the means to account for different complement forms of the same predicate.

3.3.3.1. Argument obligatoriness The argument structure of a verb can be seen as its minimal specification though by itself, it does not capture the semantic characterization of a lexical item. One question is how to distinguish between obligatory and non-obligatory arguments in the valency of a predicate. All of the elements of an utterance evidently do not have the same status. For instance, if we compare for $4 and on Friday in (94ab), we find that for $4 is more central than on Friday: (94)

a. b.

John bought a purple alarm clockfor Mary for $4. John bought a purple alarm clockfor Mary on Friday.

The FLM: syntagmatic axis

123

Although both (94a) and (94b) are grammatical without these elements, for $4 is definitely a candidate for argument status in verbs which typically codify a commercial transaction, whereas on Friday is not. If we base our answer on isolated examples, our judgement is largely the result of intuition, but if we consider buy in the context of its lexical subdomain, it is easier to offer a more systematic explanation. The following fragment from the lexical domain of POSSESSION (see Table 12) shows the paradigmatic location of buy in the FLM lexicon: (95)

to come to have get to come to have as a result of receiving, earning, buying, etc. obtain to get something [fairly formal]. procure to obtain something difficult to get [formal]. acquire to obtain something with effort, adding it to previous possessions. Φ buy to get something by paying money for it. purchase to buy something [formal]. POSSESSION:

All of the above verbs belong to the domain of POSSESSION, and share both semantic and syntactic constituents. The syntactic frame common to all of these verbs is an agent who comes to have something as result of giving its owner a certain amount of money. The agent gets the acquired entity from a source (in the case of buy, the recipient of the money), and afterwards can optionally give it to a beneficiary. Time is not a relevant factor in this process. In fact, this frame in itself is sufficient to transfer the meaning of the domain to other verbs whose meaning is generally quite different. (96)

a. b.

John tickled/smiled at/murdered Mary for $4. John tickled/smiled at/murdered Mary on Friday.

If we insert a verb from a different domain, such as tickle, smile or murder, in this type of frame, its meaning adjusts to that of a commercial transaction (96a), whereas no such change occurs when a time reference is added (96b):

124

Lexical organization and the FLM

Verbs that are incompatible with this type of frame or at the very least, sound extremely odd in it, are those which codify cognition and feeling. (97)

a. b.

*John thought about Mary for $4. * John hated Mary for $4.

No one can think or hate for money (97ab). The private nature of cognitive and affective processes make them less observable from the outside, and thus not really apt to be conceptualized as marketable goods. Curiously enough, love is a verb that can fit into this frame, but in the process, it completely loses its meaning as a verb of feeling. (98)

John loved (= had sex with) Mary for $4.

In (98) the argument structure and the semantic roles of the participants block affective meaning, and love becomes a verb of sexual intercourse or bodily action. The change in number and type of arguments is enough to project it into another lexical domain. Another problem regarding the determination of the number of arguments is when the structure of a verb calls for more arguments than those which actually appear. In example (99b) dupe has a threeargument valency, but in (99a) only two are profiled. The question arises whether the prepositional phrase in (99b), into thinking that he was a police officer, should be regarded as a satellite or an argument. (99)

a. b.

He duped me. He duped me into thinking that he was a police officer.

The answer can be found by relating these properties to the semantics of the lexical subdomain in which dupe is located. In this case, dupe belongs to the subdomain to cause somebody to believe that something is true when it is not. The verbs that share the same area of semantic space are those in example (100): (100) COGNITION: to cause somebody to believe that something is true when it is not

The FLM: syntagmatic axis

125

deceive to cause somebody to believe that something is true when it is not in order to take advantage of them, delude to deceive somebody by false promises, fool to deceive somebody, making them look foolish. trick to deceive somebody to get something from them as part of a plan. Φ dupe to trick somebody completely, usually to get them to do something, hoodwink to trick somebody [informal], bamboozle to trick somebody in order to gain an advantage [informal]. beguile to trick somebody into doing something by making it seem attractive. hoax to deceive somebody by playing a joke on them. con to deceive somebody by telling them things that are untrue [informal]. It is hardly accidental that the complementation pattern, NP wto-gerund happens to be common to all of the verbs in the subdomain. r

(101)

He

J

deceived deluded fooled tricked duped I hoodwinked [ bamboozled beguiled hoaxed conned

me into thinking that he was a police officer.

126

Lexical organization and the FLM

Prepositional phrases of the type in (101) usually co-occur with intransitive verbs which denote an atelic state of affairs. However, the fact that the prepositional phrase is a recurrent feature in this lexical set is indicative of its status as a defining characteristic of the semantic value of the lexical subdomain. This means that within this lexical subdomain, the prepositional phrase classifies as an argument, though an optional one, since it does not always have to be present in order for sentences to be considered syntactically acceptable. The complementation pattern thus has a three-argument potential and the number of arguments the speaker chooses to activate is indicative of the focus he/she ultimately wishes to give the SoA encoded in the lexical subdomain. This reveals the extent to which meaning definitions, and ultimately the relevant lexical subdomain, are determining factors in the prediction of the syntactic potential of a set of predicates. In other words, grammatical features, such as the (nonoccurrence of certain complements as well as their degree of obligatoriness, are motivated by specific semantic parameters characteristic of the verbs in question, as well as their corresponding lexical subdomain. This is the reason that no decision regarding argument structure can really be made without first examining a verb's meaning potential and its paradigmatic context.

3.3.3.2. Selection restrictions and prototypicality Each lexical subdomain gives the possible range of selection restrictions that affect a given area of meaning. Despite certain differences, we have found that there are regularities in the set of selection restrictions that define each lexical subdomain. A more detailed analysis of such restrictions is the first step towards the elaboration of an ontology of concepts. Insofar as their formalization is concerned, our proposal is for terms to be modified by a term operator called prototypical (prototyp), which signals the prototypical status of the full set of potential terms for that predicate. The introduction of such an operator is necessary because the encoding of selection restrictions is not a question of binary oppositions, but is rather a graded phenomenon.

The FLM: syntagmatic axis

127

For example, shatter is a predicate with a very simple syntactic frame. When it is used transitively in its causative sense, it has two arguments, both NPs. The first argument can be an Agent or a Force, and the second, that of Patient. The semantic characteristics of these arguments can vary, and depending on whether or not they are more or less prototypical, the base meaning of the verb varies accordingly. However, the degree of prototypicality is inevitably in direct relation to the concreteness of the arguments: (102)

shatter pieces.

Table 17.

to cause something to break into a lot of small

Shatter, inventory of selection restrictions

More Prototypical CONCRETE a. In a fit of rage, I shattered the vase against the wall.

Agent/Force

Patient

Hum Agent

Artifact

b. The explosion Force shattered the windows.

Artifact

c. His leg was shattered by the bullet.

Force

Hum body part

d. The noise shattered my ears.

Auditory percept

Hum body part

e. Noisy motorbikes shattered the peace.

Auditory percept

Mental percept

FIGURATIVE

f. The news of his death Abstract Event shattered us (=our ιr lives). g. The event shattered all Abstract Event Less of my ideas and hopes. Prototypical

Hum / Mental percept Mental percept

128

Lexical organization and the FLM

Table 17 shows the range of prototypicality in the inventory of selection restrictions typical of the arguments of shatter. The most prototypical subcategorization patterns are in accordance with the parameters characteristic of the prototypical transitive event (Hopper— Thompson 1980). They show a concrete, individuated agent that performs a forceful action upon a concrete, inanimate patient, which as a result, undergoes an important change of state. However, this same semantic frame can be instantiated through the insertion of other types of entities, thus producing a recategorization of entities along a progressively decreasing scale of prototypicality. The most prototypical examples are those under the heading of concrete. These are graded in the measure that they adjust to the canonical pattern of an animate agent acting upon a concrete, inanimate entity (Table 17a). Prototypicality begins to decrease when the agent becomes inanimate (Table 17a-g), and when the patient becomes animate (Table 17c-g). The figurative examples are also ordered on a scale of prototypicality. In the agent slot, the auditory percepts are obviously closer to animacy (and thus more prototypical) than abstract events. In the patient slot, human body parts are more prototypical than mental percepts such as ideas and hopes.

3.3.3.3. ΤΑΜ operator distribution Essential in the analysis of complement selection is the distribution of tense, aspect and modality (ΤΑΜ) operators. Such operators encode the degree of fusion existing between the complement and its matrix predicate. When this fusion is greater, the scope of the operators is reduced. (103)

a. b. c.

John baked a cake that was delicious. [Complement = noun phrase] I believe that he is too old to get marriedfor the seventh time. [Complement = that clause] "Have a nice wedding!" Peter exclaimed. [Complement = independent clause]

The FLM: syntagmatic axis

129

For example, the complement in (103 a) is more tightly bound to the matrix predicate than in (103bc). Both the complements in (103b) and (103c) are sentences, though (103b) is still dependent on the matrix predicate, whereas (103 c) is fully independent of the governing predicate.

3.3.3.3.1.

Time dependency

Certain higher predicates impose time restrictions on their complements. The operators Post and Sim act as filters which block complements designating events that fall outside the time frame signaled by the higher predicate. A third type of time dependency can be seen in the set of predicates whose complements require a future time reference (Ransom 1986: 34): (104)

a. b.

(105)

a. b.

Dorothy anticipates that the Wizard of Oz will help her return to Kansas. *Dorothy anticipates that the Wizard of Oz helped her return to Kansas. The scarecrow predicts that the Wizard of Oz will give him a brain. *The scarecrow predicts that the Wizard of Oz gave him a brain.

In the FLM lexicon, these verbs typically belong to the domains of COGNITION and SPEECH respectively. More specifically, they are in the subdomains to think that something will happen in the future and to say that something will happen in the future. The representation of these complements is characterized by a third-order entity which in turn, is modified by a tense operator designating future time: Pred [V] (χι) (X [Fut ei]).

3.3.3.3.2.

Aspectuality

Another important factor in complementation is aspectuality. The difference between aktionsart and aspect (the perfective/imperfective

130

Lexical organization and the FLM

distinction) is that the former is lexically expressed, whereas the latter is grammatically expressed (Dik 1997a: chapter 9). In reference to aktionsart, the state of affairs designated by certain complements shows no restrictions, while others are heavily constrained by the semantics of the higher predicate. For example, following Ransom (1986), predicates of a certain semantic class take complements which designate a future course of action instigated by the agent: (106)

a. b. c. d.

The bank robbers ordered everyone to lie down on the floor. Jack was still trying to persuade Jill to go up the hill. I promise to take the kids to the rock concert on Saturday. The pirates obliged him to keep on digging for the buried treasure.

In examples (106a-d), the agent causes someone else to do something (lie down, go up the hill, take the kids to the circus or dig for buried treasure). This action is to be carried out at some future point in time. The meanings of these verbs block the occurrence of permanent or already existing states in their complements: (107)

a. b. c. d.

*She ordered me to be ambidextrous. *Bill persuaded Susan to be wrinkled. */promised the kids to be crippled. * They obliged me to be a genius.

In this regard, constraints on the type of predication to be used in certain complements can be formulated in terms of selection restrictions: Pred[V] (xi) (ei: ). Another set of predicates, which also block complements designating states, are verbs of visual perception that subcategorize events: Pred [V] (xi) (ei: ). (108)

a. b.

I saw the students taking the exam (Event), *I saw the exam be difficult (State).

(109)

a.

I observed James playing the piano (Event).

The FLM: syntagmatic axis b.

131

*/ observed James be a male (State).

Examples (108b) and (109b) are ungrammatical because characteristics such as male and difficulty are an inherent part of exam and James respectively. Perceiving the exam or James logically means the perception of their characteristics, and so it is impossible to perceive an entity being what it is already. Quite another matter is the perception of a characteristic that does not normally form a part of the entity in question: (110)

I saw James be courteous (and I was very surprised).

Example (110) is possible because courteousness is not an inherent characteristic, and also because being courteous implies some sort of observable behavior. In some cases (11 labe), however, the complement clause exhibits no restrictions because it is a complete sentence in itself: (111)

a. b. c.

I noticed that she was speaking French (Action). I noticed that she was too tall for her age (Permanent state). I think that he is Chinese (Stat€)ihe is working very hard (Action).

The perfective/imperfective distinction provides the means to distinguish between complement patterns that usually occur with predicates of physical perception. When perfective (112a), the event designated in the complement clause is seen as a whole. When the complement is imperfective (112b), it is seen partially or as though it were in progress: (112)

a. b.

I saw you cross the street. I saw you crossing the street.

This distinction is marked in the predicate frame by the predicate operators Perf/Imperf. Pred [V] (xi) (Sim ei: Perf/Imperf). Due to the semantic parameters in their definition, certain predicates allow either the perfective or the imperfective, but not both.

132 (113)

Lexical organization and the FLM a. b. c.

I spotted him walking into the Louvre/*walk into the Louvre. Ipeeked at him washing the dishes/*wash the dishes. She glanced at him racing down the hall/*race down the hall.

For example, verbs that designate short duration (e.g. spy, glimpse, spot, sight, glance, peek, peep, scan, skim) only allow the imperfective reading. As we shall see, it is no accident that all of these verbs belong to the same lexical subdomain.

3.3.3.3.3. Modality Modality operators are also essential in the description of complement meanings (Ransom 1986). Second-order or third-order complements, designating a state of affairs (e) or a proposition (X), are modified by such operators. In line with this, certain higher predicates in the first person present also encode these distinctions. Following Hengeveld's (1989b) distinction between subjective and objective modality, we have distinguished various subtypes of meaning, which are also explicit in lexical structure. Subjective modality Subjective modality represents the speaker's commitment to the truth expressed within the complement (Hengeveld 1989b). Hengeveld also distinguishes boulomaic modality [Des] as a subtype of subjective modality. It is found in verbs like hope, wish and want, whose complements express the speaker's desire that something will occur in the future. The commitment to truth is represented by the parameter of factivity, which can be seen as a scale with the following values: Certain, Probable, Possible, Indeterminate (Ransom 1986). (a) Certain'. The speaker, who accepts the proposition as true, presents its content as being the case with no alternatives available. Such predicates act as lexical triggers because they presuppose the truth of their complement clause.

The FLM: syntagmatic axis (114)

133

Pred[V](xi)(CertXi) I regret that my student days are over a. *though in fact they aren't. b. *but it is not possible or the case that they are. c. *but it is not expected that they are. d. *but these may not be/*or not.

(b) Probable'. The speaker, slightly less committed to the truth of the proposition, presents the prepositional content as expected, but offers alternatives. (115)

Pred [V] (xi) (Prob Xi) I believe that there is life on Mars a * though in fact there isn't. b. *but it is not possible or the case that there is. c. but it is not expected that there is. d. but there may not be/or not.

(c) Possible: The speaker presents the information as possibly true with a preference for one alternative. This operator also explains the non-occurrence of satellites which negate the expectation of the SoA being the case: (116)

Pred [V] (xi) (Poss Xi) I guessed that the least expensive model was priced at $7,000 a. though in fact it wasn't. b. *but it is not possible or the case that it was. c. but it is not expected that it was. d. but it may not be/*or not.

(d) Indeterminate: The speaker presents the propositional content as an open question with two equal alternatives. (117)

Pred[V](xi)(IndetXi) I don't remember whether Keith came a. *but it is not possible or the case that he did. b. *but it is not expected that he did. c. but he may not have.

134

Lexical organization and the FLM

In the FLM, this scale is best exemplified in the semantic parameters in the domain of COGNITION, the subdomains of which are lexicalizations of degrees of probability. The semantic values present in the meaning definition of these verbs are projected onto the potential syntactic structure of these predicates: (118)

a. b. c. d. e. f.

To think something is true [know] To think something is probably true [believe] To think something is going to happen/be a certain way [expect] To think something is likely to be true [suppose] To think something without knowing if it is true [guess] To think something may not be true [doubt].

This can also be viewed in the context of Kiparsky—Kiparsky's (1970) semantic distinction between matrix predicates in terms of the speaker's commitment to the truth/falsity of the embedded proposition: (i) factive (the speaker's commitment to the truth of the embedded proposition); (ii) non-factive (the speaker has no commitment to either the truth or falsity of the embedded proposition); (iii) contra-factive (the speaker believe that the embedded proposition is false). Objective modality Objective modality refers to the speaker's evaluation of the actuality of the state of affairs. Two types can be distinguished: epistemic and deontic. In epistemic objective modality, the speaker assesses the state of affairs in terms of his/her knowledge, whereas in deontic objective modality, he/she evaluates the So A in terms of a moral or legal system. According to Hengeveld (1989b) and Dik (1997a: 296), each of these types of modality has its own particular range of values: (i) (ii)

Epistemic: Certain, Probable, Possible, Improbable, Impossible Deontic: Obligatory, Acceptable, Permissible, Unacceptable, Forbidden

The FLM: syntagmatic axis

135

Objective evaluation mainly affects those complements that designate a second-order entity such as an event or an action. The epistemic values, certain and indeterminate, are explicit within the architecture of domains, such as VISUAL PERCEPTION, and COGNITION. (Indeterminate does not appear in Dik's (1997ab) and Hengeveld's (1989b) epistemic scale because this evaluation has been taken from Ransom (1986). The lexicalization of these two values is also reflected in syntax: (a) Certain: All states of the world (SoWs) contain the SoA designated by the predication. (119)

He succeeded in cleaning the house a. * but it is not possible/the case that he did. b. * but it is not expected that he did. c. * but he may not have/*or not.

This value corresponds to Karttunen's (1971) implicatives and imply that the complement SoA was actually realized. Higher predicates with this type of operator would be represented as follows: Pred [V] (xi) (Cert ei). (b) Indeterminate'. The SoA is presented as an open question with two equal alternatives. (120)

They debated whether to leave home a. * but/and it is not possible/the case that they will. b. * but/and it is not expected that they will. c. * but/and they may or not.

(121)

He debated whether or not he would get a divorce.

Such complements are expressions negating two equal alternatives or the occurrence of a phrase of the type or not, which expresses an open alternative. Higher predicates with this type of operator would be represented as follows: Pred [V] (xi) (Indet X).

136

Lexical organization and the FLM

3.3.3.4. The codification of morphosyntax in a lexical domain The correlation of the different complementation patterns and the type of entity they designate can be analyzed within the wider framework of the semantic parameters of the meaning definitions of the higher predicate. In this regard, the different morphosyntactic realizations that reflect different types of meaning focalization are semantically motivated, because a single semantic parameter is capable of activating a rich inventory of syntactic features. An example of a verb in which this occurs can be found in plan (Faber—Mairal Uson 1998b): (122)

plan

to think about something (a method/way of doing something) carefiilly and deliberately in order to carry it out in the future.

(123)

Complementation patterns of plan a. She planned her departure. [NP] b. She planned to leave home. [Infinitive] c. She planned that they would leave home early. [Thatclause] d. She planned for them to leave home. [P (for) infinitive clause] e. She planned leaving home. [Gerund] f. She plannedfor the future. [PP (for)]

Plan [COGNITION] is part of the subdomain to think about something in order to make a decision in the future. The fact that plan is in COGNITION means that, at least in its most prototypical sense, it is a manner of thinking. However, the fact that think is the label of its basic concept does not necessarily imply that there is any explicit temporal relationship between the two. This predicate differs from consider (the other superordinate term within this subdomain) in that the decision to do something has already been taken, and the focus of the thought process is on the way to do it. Plan highlights how to carry out something in the future. This semantic parameter activates a rich syntactic scenario, thus presenting the future course of action from different perspectives. In the first case

The FLM: syntagmatic axis

137

(123 a), we are confronted with an NP, which is a concealed predication (to leave/to depart) . [(xi)Ag (ei: a course of action)^] Action. The second complementation pattern (123b), represents the prototypical syntactic structure for the encoding of a future action (i.e. the infinitive): [(ΧΙ)Α§ [(Prob ei: )]Go]Actioa The third pattern (123 c) introduces a radically different perspective. Although this course of action is the product of an intellectual exercise, what is presented is not so much a thought, but how to carry out the action in the future: [(xi)Ag [(Prob ei)]Go] Although the gerund (123e) in contrast to the infinitive presents this course of action as unbounded and therefore atemporal, the frame is identical to the infinitive. The two patterns introduced by a preposition (123 d) and (123f) both share an element of uncertainty or lack of confidence (Wierzbicka 1988). In other words, a future course of action is seen as a hypothetical state of affairs. Despite the diversity and apparent lack of systematicity in the range of syntactic patterns, these can be explained within the semantic context of the lexical subdomain. In this case, we have shown that the second argument in the meaning definition of plan activates different perspectives of presenting a future course of action. This way of highlighting a given semantic notion is complemented and validated by the rest of members that form part of the lexical domain. (124)

To think carefully about something

[COGNITION]

Γ consider to think carefully about something in order to make a future decision, contemplate to consider doing something in the future. meditate to consider something carefully and seriously as a possible course of action, ponder to consider something carefully, weighing it in one's mind debate to consider the argumentsfor and against doing something in the future. V

138

Lexical organization and the FLM

b.

^plan to think about something (a method/way of doing something) carefiilly and deliberately in order to carry it out in the future. design to plan something, making it a picture of it in one's mind, plot to plan something secretly. conspire to plot together secretly usually something illegal/harmful, connive to plot secretly and dishonestly for something to happen, scheme to plot something secretly in a devious way. intrigue to plot something secretly to gain somethingfor oneself/somebody else. V

In (124a), there are two subgroups: (124a) consider and its troponyms; (124b) plan and its troponyms. In the set of verbs governed by consider, the agent is thinking about a possible course of action, but has not as yet reached any firm decision. In contrast, plan and its troponyms have a somewhat different meaning because the agent has decided to do something and is thinking about how to do it. Table 18 illustrates the syntactic and semantic correspondences of each of the complements of plan. Table 18.

Semantic and morphosyntactic typology of complements of to think carefiilly about something. NP

1st order 2nd order 3rd order

Infinitive

Gerund

77iütf-clause

PP

—

—

—

—

-

+

+

+

+

+

-

-

-

-

All of the possible complements are second-order entities since they are events which take place. Plan establishes a possibly inventory of complementation patterns, some of which (but not necessarily all) are inherited by the more specific troponyms.

The FLM: syntagmatic axis Table 19.

139

Typology of syntactic frames: plan and troponyms 0

NP

Infinitive

Gerund

TTtßtf-clause

PPfor/against

plan

+

+

+

+

+

+

design

-

+

-

-

-

-

plot

+

+

+

-

-

+

scheme

+

+

+

-

-

+

conspire

+

-

+

-

-

+

connive

+

-

+

-

-

+

intrigue

+

-

+

-

-

+

Despite different morphosyntactic realizations, their meaning is compatible with the semantic value of the lexical subdomain, a course of action in the future. Even in those verbs in which where there is no explicit realization of the complement (plan, plot, scheme, conspire, connive, intrigue), the second argument is implicit in the meaning definition of the verb (a secret plan, plot, scheme, etc.). Consequently, according to the FG multilayered model, the candidate that best suits this meaning is the second layer. Given these correspondences, the next step would be the formulation of lexical rules, based on lexical semantic organization or articulated semantic classes, which account for the different morphosyntactic realizations (Mairal 1994). Despite the fact that the syntactic realizations of complements in the domain of COGNITION are often varied, their syntactic realizations are still compatible with the semantic content expressed within the lexical subdomain. This can be seen in the following partial representation of the complementation patterns for understand: (125)

understand a.

to know the nature or meaning of something, being fully aware of it. 77ia/-clause (xi: prototyp. human)Exp (Cert X)phen

140

Lexical organization and the FLM

Only today have I begun to understand that I do not love him. NP (xi: prototyp. human)Exp (Cert X)phen She understood him perfectly.

b.

The two complementation patterns for understand (125ab) are not really as different as they appear to be. In (125b) him is a concealed proposition because it refers to what he was saying or the way he was acting. Consequently, the meaning of both is compatible with that of the lexical subdomain. In this sense, different complementation patterns can be considered different instantiations of the same subdomain. Table 20.

Semantic and morphosyntactic typology of complements of understand

NP st

1 order 2nd order 3rd order

7% «/-clause

—

—

—

-

+

+

These correspondences can be stated in terms of rules whose format is similar to that of expression rules, but which are stored in the lexicon. Our hypothesis is that the productivity in a lexical domain gives rise to a set of lexical rules, which form the basis of what we have called lexical grammars (Faber—Mairal Uson, forthcoming). This presumably would lead to a simplification of the expression rule component as those expression rules which are concerned with the form of constituents can be derived from lexical structure.

3.4.

Summary

An overview of the concept of semantic field shows that it has never been seriously contemplated as a central element in linguistic theory. Even in those theoretical frameworks in which semantic fields/domains are mentioned, the term invariably lacks precise definition. This is partly due to the inherent fuzziness of meaning,

Summary

141

but also to the fact that traditional semantic fields do not give a very complete representation of meaning. Evidently, the concept is a useful one, but it needs to be enriched with other types of information and not restricted to mere binary feature representations that do not do justice to the complexities of word meaning. An organized lexicon is a goal to strive for because it would considerably facilitate the understanding of syntax. In this chapter, we explain how the concept of semantic field is enhanced in the FLM lexicon. We have used the term lexical domain instead of semantic field because we have substantially enriched the concept. A lexical domain is a map of paradigmatic and syntagmatic information, the distribution of which is motivated. Lexemes on the paradigmatic axis are organized hierarchically according to their definitional structure through Stepwise Lexical Decomposition. The information on which these hierarchies is based is extracted from various dictionaries and a composite definition elaborated for each lexeme. Lexical domain membership is determined by the genus, which constitutes the nucleus of the meaning of a lexeme. This result of this type of analysis is an inventory of basic lexical domains, each with one or two generic terms, which have the status of near primitives. Lexemes are thus the central units of linguistic description, and specify the basic meaning of their respective domains in different ways according to their differentiae. Recurrent differentiation parameters can be classified as syntactic, semantic or pragmatic. The analysis of differentiae also provides us with recurrent semantic information that can constrain syntactic representations on the syntagmatic axis. The syntagmatic axis of the FLM lexicon encodes the combinatorial possibilities of each lexeme. Our approach to complement selection has a semantic orientation. In other words, complementation is a matter of matching the semantics of the higher predicate, as specified in the lexical subdomain in question, with the semantics of the complement phrase. In accordance with the FG multilayered model of the clause, predicate frames are used for the representation of complement meanings, and complements are characterized in terms of the highest layer they contain, as well as the type of entity designated.

142

Lexical organization and the FLM

Although the FLM is based on FG, it places a greater emphasis on lexical representations as the key to the determination and explanation of the syntactic representations of predicates. It thus makes more explicit use of meaning definitions. The first step in the codification of the syntagmatic axis is an analysis of the semantic potential of the complementation pattern/s of a given verb in terms of argument obligatoriness, complement form, and ΤΑΜ operator distribution. An analysis of the way these elements are encoded in complementation patterns reveals a rich taxonomy of complement meanings since the codification of grammatical features in complement structures is largely motivated by semantic parameters. Our proposal is that information about a predicate's paradigmatic location in the lexicon be integrated into its predicate frame in order to represent the interface between syntax and semantics. A verb's subcategorization is semantically motivated, since the semantics of a lexical domain activates or triggers its complementation patterns. In this way the lexicon becomes a dynamic storehouse of words, which codifies and explains how we create and use language.

4.

Towards a semantic syntax

4.1.

Introduction

In the FLM lexicon, paradigmatic and syntagmatic information are closely interrelated to the extent that a verb's syntax depends on its location in semantic space. In other words, a verb's combinatorial possibilities and syntactic potential are semantically motivated. This interface between syntax and semantics in verbs has also been observed and studied by others (Dixon 1991; B. Levin 1993: 5; Van Valin 1993a). Nevertheless, in the majority of cases, syntactic coding has been used as the starting point to explain meaning, a procedure which has still to produce a satisfactory semantic classification. For example, words are often grouped in semantic classes on the basis of grammatical alternations such as the following: (126)

a. b.

(127)

a. b.

The harpoon plunged into the water. The sailor plunged the harpoon into the water. [Zeroderived causative form ofplunge] The hunter shot the tiger. The hunter shot at the tiger. [Conative grammatical alternation]

These criteria in themselves are insufficient as a basis for the determination of semantic classes because syntax alone does not provide a satisfactory explanation of semantic constraints. As Pustejovsky (1995: 10) points out: ... participation in one grammatical alternation does not sufficiently determine the semantic class of the verb. In fact, even once a complete cataloguing of participation in alternation classes is achieved, we must ask ourselves just what we have accomplished. Rappaport—Levin's (1998) lexical templates are an attempt to overcome this problem, though their closed inventory of lexical

144

Towards a semantic syntax

templates is very reduced, and makes what might seem an excessive use of template augmentation to account for individual verbs. The goal of any lexical semantic theory is to establish the premises upon which words can be organized into classes, which to a great extent predict their syntactic and semantic properties. Our assertion is that regularities in the lexicon can be mapped out by means of syntactic-semantic (synsem) parameters that operate throughout the lexicon in the various areas of meaning. Such parameters are the result of the convergence of paradigmatic and syntagmatic information, and are derived from the nuclear meaning of semantic sets, as well as the complementation patterns characteristic of each subdomain in the FLM lexicon. Though basically lexical semantic in nature, synsem parameters also influence a verb's syntax. The semantics of a lexical subdomain not only provides an explanation for questions such as argument obligatoriness, but also acts as a mechanism which either triggers or blocks certain types of complementation. This occurs because of the activation of synsem parameters such as duration, cessation, iteration, conation, achievement, etc., which operate in the subdomain in question. This means that for the prediction of complement form, the subdomain level is the most informative, and that it is not enough to know the general lexical domain to which a verb belongs. The fact that verbs in the same meaning area can show rather different syntactic behavior is evidence that a more fine-grained approach to lexical analysis is necessary.

4.2.

Meaning definitions and syntactic projections

Given the importance of semantics in syntactic representations, it is necessary to reconsider the ancillary status of meaning definitions in Dik's FG. Definitions, when considered in paradigmatic contexts, play an important role in determining the complementation of a given predicate as well as the occurrence of certain grammatical features such as the state of affairs of the predication and its cooccurrence with certain satellites. In this regard, the syntactic hierarchy of each FLM lexical domain can be mapped onto its corresponding semantic one. Synsem parameters are one of the products of this explicitation of the interface between syntax and

Meaning definitions and syntactic projections

145

semantics. Although such parameters play a role in the generation of the actual clause structure, some affect syntax more than others. We have classified these parameters in three categories, according to their scope of application: (i)

Lexically-realized grammatical parameters are obligatory and are always explicit in a verb's complementation structure. (ii) Lexically-realized optional parameters are semantically implicit, but are not always activated in the actual linguistic expression. (iii) Lexically-realized contextual parameters are not syntactically projected, but serve as clues for contextual setting.

4.2.1. Lexically-realized grammatical parameters Lexically-realized grammatical parameters have a direct effect on complementation, and act as a filter that blocks certain patterns while accepting others. Examples of such parameters are duration, temporal sequence, iteration, inception, achievement, cessation, causation, conation, and factivity. Though the inventory of parameters is by no means exhaustive, the ones cited are indicative of the interface between syntax and semantics. It is hardly accidental that these factors are basic to our perception of the event encoded by each verb. Experiencing an event means perceiving its shape; this signifies taking note of when it begins and/or ends, how long it lasts, if it is recurrent, its effect(s), and if it truly corresponds to the world or a state of the world. All of this information should logically be reflected in a verb's meaning as well as its syntax.15 Knowing a verb's semantic class is thus crucial to understanding its properties and to determining its relation to other words within the organizational scheme of English verbs (B. Levin 1991: 218). Such knowledge of meaning necessarily signifies the consideration of the lexicon as a whole.

146

Towards a semantic syntax

4.2.1.1. Duration Duration encodes how long an event lasts in our perception. As one of the most important synsem parameters, it has two subtypes, depending on whether the action or process takes place over a longer or shorter period of time. This distinction is important because whether or not an action is perceived as durative or momentaneous can affect complementation.

4.2.1.1.1.

Short duration

The semantic parameter of short duration is activated in the lexical domains of VISUAL PERCEPTION, CONSUMPTION and FEELING: ( a ) VISUAL PERCEPTION

(128)

VISUAL PERCEPTION: TO see somebody/something at a distance/briefly spot to see somebody/something [EFFORT + DIFFICULTY] with difficulty, making an effort to do so. sight to see somebody/some thing [EFFORT + SUDDENNESS] suddenly, making an effort to do so. glimpse to see somebody/something not very well/partially, [PARTIAL VISION] spy to see somebody/something, after looking for them for [PREVIOUS LOOKING] awhile.

The meaning of verbs such as spot, spy, sight and glimpse all contain the parameter of short duration (128). As troponyms of see, all of them specify the basic meaning of the subdomain in different ways.

Meaning definitions and syntactic projections (129)

a. b.

(130)

a. b.

147

I saw a movie actor walk across the street. [NP + bare inf.] */ glimpsed a movie actor walk across the street. [NP + bare inf.] I observed that the theater had become dark. [Thatclause] */ spotted that the theater had become dark. [Thatclause]

Much of the complementation acceptable for other perception verbs is filtered out here because of this parameter. Even though visual perception verbs such as see and observe can take an NP and a bare infinitive, as well as a that-oXdcasQ, glimpse and spot do not accept this sort of complementation. Examples (129b) and (130b) are ungrammatical because the act of perception encoded in visual perception verbs of short duration (glimpse, spot, sight, spy) does not last long enough to successfully detect the action in the clause. The complementation patterns that characterize this particular subdomain are an NP and an NP + /'«^-participle. However, there are instances when these complements also become unacceptable, and the reasons are purely semantic. (131)

a. b. c. d.

I spotted/spied/sighted/glimpsed a movie actor. [NP] */ spotted/spied/sighted/glimpsed a movie actor for a long time. [NP] I spotted/spied/sighted/glimpsed a movie actor walking down Hollywood Boulevard. [NP /«g-part.] */ spotted/spied/sighted/glimpsed a movie actor walking down Hollywood Boulevard for a long time. [NP ingpart.]

Despite the fact that (131b) and (13 Id) take the same complementation patterns as (131a) and (131c) respectively, they are anomalous because the adverbial phrase, for a long time, is in opposition to one of the basic semantic values of the subdomain. In contrast, adverbial modification that confirms this value (briefly/for a short time) is perfectly acceptable:

148

Towards a semantic syntax

(132)

I briefly spotted/spied/sighted/glimpsed walking down Hollywood Boulevard.

a movie

actor

In (132) briefly is a reiteration or confirmation of the semantic value of the subdomain. Its lexicalization within the expression has an intensifying effect. (b) CONSUMPTION

(133) consume to use (up) something until there is none left, drink to consume liquid, taking it into one's mouth and swallowing it. imbibe to drink alcohol [formal],

a.

QUICKLY IN LARGE { AMOUNTS

b.

SLOWLY IN SMALL AMOUNTS

gulp (down) to drink something very quickly, quaff to drink something quickly [old-fashioned], swig to drink something quickly in large amounts in a series of big swallows [informal]. swill to drink something quickly and greedily in large amounts [informal], guzzle to drink something (especially alcohol) very quickly, greedily and noisily in an unattractive way. tipple to drink something (especially alcohol) secretly and in small amounts [informal], sip to drink something slowly in very small amounts

Other predicates with the parameter of short duration can be found in the domain of CONSUMPTION. The semantic distinctions lexicalized in verbs of eating and drinking are largely motivated by social norms and pragmatic factors (Jimenez Hurtado 1994). Drinking, for example, is an activity that has very definite meaning specifications. One can drink a liquid quickly in large amounts (133a), or slowly in small amounts (133b). Curiously enough, there are no English verbs which lexicalize drinking a liquid quickly in small amounts, or conversely, drinking it slowly in large amounts.

Meaning definitions and syntactic projections

149

Verbs such as quaff, swig, swill, and guzzle thus contain the parameter of short duration. This means that in the same way as verbs of visual perception, consumption verbs cannot be modified by adverbial phrases that violate the semantic value characteristic of the subdomain: (134) a. b.

I gulped/quaffed/swigged/swilled/guzzled four martinis. *I gulped/quaffed/swigged/swilled/guzzled four martinis for a long time.

Example (134b) is ungrammatical precisely because this type of manner-of-drinking is regarded as a process that takes place quickly. Quaff, swig, swill and guzzle are also atelic, and focus on the categorization and evaluation of the activity (as well as of the person who is carrying it out), rather than on the achievement of any sort of goal. (135) a. b.

*He gulped/quaffed/swigged/swilled/guzzled himself to death. He drank himself to death.

Drink, the generic term of the subdomain, is the only one that can enter into a construction typical of a goal-directed action (135b). All of the verbs in this set must have an explicit goal argument with the exception of drink and imbibe, which can appear without one (136b). (136)

a. b.

*Just give me a glass of straight orange juice. I am afraid I don't gulp/quaff/swig/swill. Just give me a glass of straight orange juice. I am afraid I don't drink/imbibe (= consume alcohol).

Nevertheless, even when drink and imbibe appear with only an agent argument, the goal argument (an alcoholic beverage) is still present implicitly as a default value (136b). In contrast, verbs with the meaning parameter of short duration, such as gulp, quaff, swig and swill must take an NP because they have no goal argument of this type (136a) as part of their meaning.

150

Towards a semantic syntax

Although one can only gulp, quaff, swig and swill liquids, there are also certain cultural and contextual restrictions as to the kind of liquid that can be consumed in this way. (137)

a. b.

I gulped down/ quaffed,Ί ? ?swigged/? ?swilled/ ??guzzled several glasses of milk. I ??gulped down/??quaffed/??swigged/ ??swilled/ ??guzzled milk every day at lunch.

Manner-of-drinking verbs sound very strange in collocations with non-alcoholic beverages, especially healthy ones like milk (137a). And in (137b) where such verbs are used in the sense of a daily activity, the sentence would have to be strongly contextualized in order to be acceptable. Indeed, more than encoding the act of drinking, these verbs encode the perception of the act, as well as the evaluation of the agent. (c) FEELING

Short duration is not only a property of actions, but also of very strong feelings such as intense joy, fear and surprise. In this area of meaning, however, it takes the form of suddenness. (138)

a. b.

After the accident happened, he panicked. [Manifestation of fear] * After the accident happened, he panickedfor a long time/all afternoon.

For example, panic is the way fear can be manifested when one loses control (138a). It is something momentary, which cannot be maintained over a long period of time, and thus does not accept modification by adverbial time clauses (138b). This momentaneousness means that when it appears in participle form, it experiences a slight change of meaning. (139)

I saw him panicking (= on the verge of showing fear).

Meaning definitions and syntactic projections

151

Example (139) does not refer to the actual manifestation of fear (which is sudden), but to a prelude to the actual loss of control. In this case, the perceiver is the judge of the emotional state of the other person. However, as this type of evaluation cannot be done momentaneously, verbs of visual perception such as glimpse, spot, sight, and spy cannot be used in this context, since they are of short duration, and thus do not allow the perceiver enough time to make this type of judgement (140): (140)

*I glimpsed him panicking.

The syntactic properties of panic also apply to verbs that codify certain feelings of great pleasure {thrill, elate) and surprise {startle), which also have this element of suddenness: (141) (142) (143)

a. b. a. b. a. b.

She thrilled to his touch. [Manifestation of pleasure] *She thrilledfor a long time/all day to his touch. His success elated him. [Manifestation of pleasure] *His success elated him for a long time. The noise startled him. [Manifestation of surprise] *The noise startled him for a long time.

Examples (141-143) show that thrill, elate, and startle cannot take adverbial modification which violates their basic characteristic of suddenness. Moreover, as a momentaneous event, verbs with this semantic parameter cannot collocate with aspectual verbs that codify inception, continuance, and cessation. (144)

a. b.

*When they were together, he began/continued/ finished to thrill to her touch. *When they were together, he finished thrilling to her touch.

Another reason that examples (144ab) are ungrammatical is that as feelings, they are typically outside the control of the experiencer. As a result, it seems strange to speak of beginning or ending them. (145)

a. b.

* She spent the day being elated/star tied. *She was thrilling to his touch.

152 (146)

Towards a semantic syntax a. b.

*/ saw her thrilling/being elated/being startled. *I saw her thrill/be elated/be startled.

This type of event is over so quickly that it is virtually impossible to syntactically encode it as a process, and thus, it rarely appears in participle form (145a, 146a) or as a bare infinitive (145b, 146b).

4.2.1.1.2.

Long duration

In predicates affected by the parameter of long duration, the event/process takes place over a fairly long period of time. Such verbs cannot be modified by adverbs that violate this basic meaning (e.g. suddenly). Verbs of this type are present in a number of domains throughout the lexicon, however the examples below are from the domains of VISUAL PERCEPTION, PREPARATION, CHANGE and SPEECH. (a) VISUAL PERCEPTION

(147) look

to see somebody/something by intentionally directing one's eyes. watch to look at somebody/something paying attention to what is happening, observe to watch somebody/something carefully, often in order to learn something, regard to look at somebody/something with a particular feeling. contemplate to look at somebody/something steadily, in a quiet, thoughtful way. gaze to look at somebody/something steadily for a long time because they are attractive or surprising, stare to look at somebody/something for a long time with wide open eyes.

In the domain of VISUAL PERCEPTION, the temporal parameter of long duration is evident in the intentionality of the generic term, look. This parameter is inherited by its more specific troponyms, the majority of which share complementation patterns typical of this type

Meaning definitions and syntactic projections

153

of meaning. As a result, this parameter activates the wg-participle, which is typical of an unbounded event. (148)

a. b. c.

Ron saw/gazed at/watched/contemplated/regarded /stared at/observed Mary dressing. Ron saw/watched/observed Mary dress. *Ron /contemplated/regarded/gazed at/stared at Mary dress.

While the wg-participle is taken by all the verbs in the set, the bare infinitive is taken by only a more reduced inventory (148b). The fact that neither regard nor contemplate (148c) take this pattern is partly due to the fact that in another sense, they are also verbs of cognition. This signifies that despite their membership in VISUAL PERCEPTION, they are influenced by the syntactic behavior characteristic of the domain of COGNITION. This makes them less typical, and places them in a transition zone, in which meaning parameters from both areas are operative. Gaze and stare (148c) do not take the bare infinitive for purely semantic reasons. Both verbs lexicalize a long, highly intentional act of visual perception. It is thus impossible to gaze, stare, or visually focus for any length of time on something that only lasts a relatively short time. (b)

PREPARATION [of food]

The verbs of cooking have been studied extensively (Lehrer 1974). The tiny segment below shows those which are affected by the parameter of long duration. (149)

cook to be preparedfor eating by the use of heat (offood). stew to cook slowly in liquid in a covered container for long time, simmer to cook slowly in water at/just below boiling. poach to simmer (typically of eggs/fish).

The verbs of food preparation (149) encode a process which occurs over a period of time, the result of which is a change of state. In the

154

Towards a semantic syntax

same way as more prototypical verbs of movement and change, they have a zero-derived causative form: (150) (151)

I stewed the meat./The meat stewed. I poached the eggs./The eggs poached.

The fact that the action denoted by stew, simmer, and poach is durative filters out certain types of adverbial modification (152abc): (152)

a. b. c.

*He suddenly stewed the meat. * The meat stewedfor a second. * The eggs suddenly simmered/poached.

In (152ac), the adverbial modification suddenly is in direct opposition to the time parameter inherent in the semantics of these verbs. As a result, poach and stew sound rather bizarre in subordinate clauses with the bare infinitive (153 ab), which to some degree, encodes the meaning of immediacy, whereas they can be used with an wg-participle (153c): (153)

a. b. c.

?I saw the eggs poach/meat stew on the fire. *I glimpsed/spotted the meat stew. I saw the eggs poaching!meat stewing on the fire.

(c) CHANGE

(154)

To become different by becoming greater develop (fromAnto) to become larger or more complete gradually, mature to develop completely. ripen to mature, becoming ready to be eaten (offruit/crops), to be eaten/drunk (of cheese/wine), or complete (of qualities).

The verbs in this subdomain (154) encode processes that may also have an implicit or explicit time span in their definitions. For example, when mature and ripen refer to natural processes, no parameter of

Meaning definitions and syntactic projections

155

suddenness (either adverbial or syntactic) can be attached to them. Examples (155a-d) are weird enough to project us into the realm of science fiction. (155)

a. b. c. d.

* The apple tree ripened suddenly. *The wine matured suddenly. ?I saw the apple tree ripen/wine mature. ?I saw the apple tree ripening.

Nevertheless, when these processes describe changes in human beings as well as abstract entities, (156ab), rapid maturing/ripening becomes possible in certain contexts because of metaphorical extension. (156)

a. b.

He suddenly matured after he left home. I saw their plans suddenly ripen.

(d) SPEECH

(157)

To complain continuously nag to complain continuously so that somebody will do something. gripe to complain continuously andforcefully [informal], grouse to complain continuously and loudly [informal], bellyache to complain continuously and loudly for no reason [informal].

Within the domain of SPEECH, complaining is an area where examples of the parameter of long duration as well as iteration abound (157). In the same way as the previous examples, this parameter filters out certain types of adverbial modification, as well as syntactic constructions. As can be seen in Table 21, complain, as the generic term of the set, takes the greatest variety of complementation patterns:

156

Towards a semantic syntax

Table 21.

Typology of syntactic frames, to complain continuously 0

NP

Thaiclause.

PP (to)

PP (about) PP (of) Quote

complain

+

-

+

+

+

+

+

nag

+

+

-

-

+

-

+

gripe

+

+

-

-

+

-

+

grouse

+

-

-

-

+

-

+

bellyache

+

-

-

-

+

-

-

However, there is one pattern taken by the more specific members of the hierarchy, which is not included in the inventory of complain. Both nag and gripe can take NPs, whereas complain does not: (158)

a. b.

She was nagging annoying] me endlessly. His behavior really gripes [= annoys] me.

When nag (158a) and gripe (158b) take a human goal argument, they undergo a corresponding change of meaning. In other words, this change in syntax projects them into another lexical domain because in this context, the meaning of nag and gripe is closer to that of annoy in the domain of FEELING. Consequently, the fact that both can take this type of NP complement signals a meaning extension to another lexical domain because such a complement is typical of that particular semantic area. As verbs that encode repeated and continuous complaining, nag, gripe, grouse, and bellyache often appear in ingforms: (159)

a. b. c.

He kept nagging me endlessly. She was griping about the lack of heating. Stop your bellyaching!

It is considerably less frequent for them to appear in a bare infinitive construction such as the following:

Meaning definitions and syntactic projections

(160)

a. b. c.

157

*/ glimpsed him gripe. *He spotted her nag. ?I saw him gripe/nag.

Since these verbs encode a continuous action, they cannot appear with verbs of visual perception with the parameter of short duration. In a similar way, they do not accept adverbial modification such as suddenly or briefly, which violates their basic semantic parameters: (161)

*He nagged/bellyached griped suddenly.

Example (161) is ungrammatical because suddenly violates the meaning parameter characteristic of this set of verbs. Not surprisingly, these verbs are also atelic because they encode actions, which are not conducive to the achievement of any type of goal: (162)

a. b.

He griped/naggedfor an hour (*in an hour). *It took John three hours to bellyache/nag.

Griping and nagging are ongoing processes, which generally do not accomplish anything. Nag is the only one of these verbs that can enter into a resultative construction. (163)

a. b.

Her father-in-law nagged (= bothered) her to death. *Her father-in-law griped/bellyached her to death.

Nevertheless, in (163 a) this construction causes nag to change domain, and projects it into FEELING.

4.2.1.2. Temporal sequence The parameter of temporal sequence appears in verbs whose meaning depends on the temporal context of the encoded action, which is sequential to another previous related action/event. A parallel could be drawn to meronymy in nouns, though evidently, verbs cannot be taken apart in the same way. Miller—Fellbaum (1991: 218) point out that any acceptable statement about part relations among verbs always involves

158

Towards

a semantic

syntax

the temporal relation between the activities that the two verbs denote. One activity or event is part of another activity or event only when it is part of, or a stage in, its temporal realization.

(a) SPEECH

In the domain of speech verbs, this parameter is activated in the definitions of verbs like rephrase : (164)

rephrase to phrase something again in a different way.

For example, the act of rephrasing can only take place after another action of a similar type, something that makes the following example unacceptable: (165)

*He rephrased his question for the first time.

Since this type of meaning is signaled by the prefix re-, there are many other possible examples such as redo, repossess, rewrite, replay, etc., which belong to other domains as well as SPEECH.

(b) POSSESSION

(166)

To give something to somebody after one's death leave to give something to somebody for use after one's death by leaving written instructions to do so. Φ bequeath to leave something (money, properly) to somebody.

is a lexical domain imbued with pragmatic meaning. The act of giving is firmly rooted in a complex variety of social relations that dictate the context in which who can give what to whom and in exchange for what. Verbs such as leave and bequeath refer to the transfer of possessions (166). The specification in bequeath concerns the goal argument, which is restricted in this case to money and property (167a). In this sense, leave is more general (167b). POSSESSION

Meaning definitions and syntactic projections (167)

a. b.

159

He left/bequeathed me hisfortune/estate. He left/??bequeathed me a stack of old magazines/his cat/his debts.

However, both verbs also refer to the temporal context when such a transfer of possessions is appropriate. The act of bequeathing takes place in two phases, before and after the death of the giver. First written instructions are given, and secondly, possessions are transferred after the death or disappearance of the possessor: (168)

a. b.

I bequeath you my fortune. (= You will have my fortune after I am dead.) ??I bequeath you my fortune, and here it is.

For that reason, (168a) can be understood as a future transfer of possession, but (168b) is semantically anomalous because the possessor is still alive.

4.2.1.3. Iteration The parameter of iteration encodes a recurrent action, something that manifests itself in different ways depending on the particular area of meaning in which it is operating. Examples of this parameter can be found in CONTACT, USE and SOUND. (a)

CONTACT

(169) To hit somebody/something many times especially to hurt/damage them beat to hit somebody/something hard and continuously over a long period of time, batter to beat somebody/something with great force many times with one's fists/other object, clobber to beat somebody [informal], whip to beat somebody with a whip as punishment, lash to whip somebody (slave, sailor) as punishment [old-fashioned].

160

Towards a semantic syntax flog

to whip somebody (slave, sailor) with great force as punishment [old-fashioned], flagellate to whip somebody as a religious act of penance [formal], birch to whip somebody with a birch, cane to beat somebody with a cane as punishment. In CONTACT, the parameter of iteration is present in the set of verbs that encode physical punishment (169). All of these verbs basically share the same complementation and argument structure. Possible complementation patterns are: (170)

Complementation patterns of beat a. The overseer beat the slave. [NP] b. The overseer beat the slave to death. [NP PP {to death)] c. The overseer beat the slave senseless. [NP Comp] d. The overseer beat the slave with a stick. [NP PP (with)]

The selection restrictions for arguments mainly pertain to the entity affected by the action. One can beat and batter an inanimate object as well as an animate one, but in the other verbs, the scope of this argument narrows and only accepts living beings. The fact that whip, birch, and cane are denominal verbs derived from the instrument employed to carry out this action eliminates the PP (with) from their inventory (171a, 172a), unless the instrument specified in the phrase modifies the default value because it is of a different type or is more highly specified (171b, 172b). (171)

a. b.

(172)

a. b.

*The overseer whipped the runaway slave with a whip. The overseer whipped the runaway slave with the whip in the shed/a cat ο 'nine-tails. * The schoolmaster caned the boy with a cane. The schoolmaster caned the boy with a long, thin cane.

Meaning definitions and syntactic projections

161

Needless to say, the type of meaning encoded in the subdomain is that of hitting somebody repeatedly with something. The parameter of iteration and its corresponding intentionality are principally what connects the verb in question to sociocultural frames of institutionalized punishment. (173)

The overseer whipped him once (= hit him several times with a whip).

Iteration is always present in this particular set of verbs, and consequently, even when the action is specified as once, it does not mean one stroke of the whip or lash, but rather several (in sufficient number to constitute one punishment) (173). Another salient feature of these verbs is that the action they encode is goal-directed. Therefore the following constructions are possible: (174)

a. b.

The ship's captain beat/battered/clobbered/flogged the sailor to death. The ship's captain beat/battered/clobbered/flogged the sailor senseless.

Examples (174ab) are all resultative constructions in which the agent (in this case, the ship's captain) affects the patient {the sailor) in some way. The resulting state of the patient is codified by an adjectival complement {senseless) or a prepositional phrase {to death). This type of construction is very frequent with iterative verbs, in which the repetition of the action results in the attainment of a final state. (b)

(175)

CONSUMPTION

bite to use one's teeth to cut through something. chew to bite something repeatedly usually in order to taste/eat it. masticate to chew [formal], gnaw to bite something hard repeatedly.

162

Towards a semantic syntax

Other examples of the parameter of iteration can be found in verbs like chew, masticate and gnaw, all of which convey the meaning of iterative biting. Chew and masticate differ in degree of formality. The agent argument of both verbs is necessarily an animate being and the goal argument is solid food (soft enough to be eaten). Gnaw is slightly different in that the goal is something hard, which gets worn away little by little through continuous biting. All three of these verbs have teeth as an embedded instrument. (176)

a. b. c.

*He chewed/masticated the ham sandwich with his teeth. He chewed/masticated the ham sandwich with his newly repaired set of false teeth. The prisoner gnawed at the rope with his teeth.

However while the instrument can be lexicalized in gnaw (176c), this is impossible in chew and masticate, unless the argument is more highly specified (176ab). The parameter of iteration also makes (177) anomalous: (177)

* She chewed/gnawed the meat once.

This is another case of a verb being unable to accept adverbial modification that is contrary to its basic semantic parameters. (c)

SOUND

(178) laugh to make the sound expressing happiness or amusement. giggle to laugh softly and repeatedly, titter to giggle nervously, ITERATION J snigger to giggle softly in a secret, disagreeable way. snicker to snigger in a high-pitched way. Another example of iteration occurs in verbs of repeated laughter such as giggle, titter, snigger and snicker in the domain of SOUND. In

Meaning definitions and syntactic projections

163

the same way as chew and masticate, verbs of iterative laughter cannot appear in collocations with once/one time. (179)

*She giggled/tittered/sniggered/snickered

one time.

This parameter also filters out the collocation with at, which is possible with laugh. At, in this sense, indicates the specific target towards which the agent directs his/her action. (180)

a. b.

She laughed at the teacher. *The children giggled/tittered/sniggered/snickered the teacher.

at

It is to be noted that in (180a), laugh acquires a meaning which projects it into another domain. It becomes more than just sound, and in this sense, signifies mocking or making fun of someone/something. It thus becomes an event, something that is impossible for manner verbs in general.

4.2.1.4. Inception As its name indicates, inception refers to the beginning of an event. The most obvious examples of this parameter can be found in the domains of CHANGE and EXISTENCE. (a)

(181)

CHANGE

become to begin to be different in the way that is stated. get to become Ifairly informal]. turn (into) to become something different in nature/ quality/condition in a sudden way. grow to become something different in nature/ quality/condition in a gradual way.

Verbs of change are closely related to movement, though instead of movement from one place to another, they refer to movement from

164

Towards a semantic syntax

one state to another. Since the lexical domain of MOVEMENT encodes an event that occurs over a period of time, it is pervaded by a number of phasal distinctions. The same is true for CHANGE. The verbs in (181) contain the parameter of inception. In this subdomain, the process of change is slow because in this non-causative sense, time is often the implicit agent/force that causes the patient to become different in some way: (182) (183)

The bread became stale. Her parents grew old.

This is true in (182) and (183) in which the cause for bread becoming stale and parents growing old is evidently the passing of time. The parameter of inception also acts as a filter that blocks constructions such as the following: (184)

a. b.

*Her parents stopped becoming old. ?The prince stopped turning into a frog.

In fact, even when the change is not a natural temporal process (184b), constructions with stop still seem semantically anomalous because the process of change in this initial phase is conceived as something that once begun, really cannot be interrupted at will. The event in (184b) is prototypically one that cannot be controlled. The prince is not an agent, but rather an affected entity. However, if we conceive the prince turning into a frog as a repeated event, then it becomes more acceptable: (185)

The prince stopped turning into a frog every evening at sunset/every time the witch said the magic words.

In (185) the process has not yet begun, and so it is not a question of an implicit agent stopping an action which is happening.

Meaning definitions and syntactic projections

165

(b) EXISTENCE

(186)

appear to begin to exist in the perception of others. dawn to begin to appear (of light/mental phenomena) [formal], surface to appear, becoming obvious or known, especially after a period of remaining unseen, materialize to appear, taking bodily form/becoming reality. form to appear, beginning to be visible and having a clear shape/outline.

Lexical structure in the domain of EXISTENCE reflects our understanding of this concept. Life is conceptualized as an event, a temporal segment that can be broken down into a number of phasal distinctions such as beginning, continuing and stopping. The lexical set in (186) includes those verbs which belong to the phase of inception (Faber—Mairal Uson 1997a: 135-136). This subdomain is closely linked to that of PERCEPTION, since when something appears, it begins to exist, not in itself, but for the perceiver. Something can appear because it moves into somebody's field of vision, or it can appear because there is a change in contextual conditions, which facilitates visual perception (i.e. sufficient light, absence of barriers). Consequently, the lexical set in (186) refers to a process of change which takes place in the perceiver, who moves from a state of unawareness to awareness of the perceived entity. (187)

a. b.

* The idea stopped dawning/materializing. *The outline of the ship stoppedforming on the horizon.

In the phase of inception, this means that such verbs cannot appear in constructions which violate this basic synsem parameter such as (187ab).

166

Towards a semantic syntax

4.2.1.5. Cessation In comparison with inception, cessation is much more extensively lexicalized because ending seems to be more perceptually salient than beginning. Examples of this parameter can also be found in EXISTENCE. ( a ) EXISTENCE

(188)

disappear to stop existing in the perception of others. fade to disappear graduallyfrom sight/hearing/ memory, often remaining to some extent, vanish to disappear suddenly and completely, especially in a way that cannot be explained, dissipate to disappear completely, as if by scattering, evaporate to disappear like vapor.

Cessation is encoded in to stop existing in the perception of others, which includes the verbs in (188), all of which describe the SoA of not being able to see something any more. In vanish and dissipate, the disappearance is complete, so what is lexicalized is the manner of disappearance {suddenly, as if by scattering). Vanish also shows a secondary connection with COGNITION in that the suddenness of the process causes a state of perplexity in the perceiver. The fact that a sudden disappearance is something that needs an explanation signals a default value in this subdomain regarding the relative speed with which something can move outside the scope of our perception and be considered a normal state of affairs. In contrast, when something fades, the process is so gradual that there is usually some trace left over a period of time. All of these verbs are intransitive, and there are virtually no selection restrictions on the argument. Unlike the parameter of inception, cessation allows constructions with begin. In other words, one can talk about the beginning of the end much more easily than the end of the beginning. It is only semantically anomalous in vanish, primarily because the parameter of achievement blocks this type of meaning:

Meaning definitions and syntactic projections

(189)

a. b. c. d.

167

The bruise on her arm gradually began to fade. His enthusiasm for a new car began to dissipate when he saw what it would cost. Her desire for him began to evaporate as his real personality came to light. * The spot on the carpet began to vanish.

Alternatively, however, these verbs sound bizarre in constructions with stop because in examples (189a-d), cessation in fade, dissipate, evaporate and vanish is cancelled out rather than reinforced. (190)

a. b. c. d.

* The bruise on her arm gradually stopped fading. *His enthusiasm for a new car stopped dissipating when he saw what it would cost. *Her desire for him stopped evaporating as his real personality came to light. * The spot on the carpet stopped vanishing.

Cessation in this sense also has an important component of achievement. This parameter blocks an event/process that is not a sudden occurrence, which ends at the same time as it begins: (191)

a. b.

The two cars collided /*began to collide. The car exploded /*began to explode.

[CONTACT] [CHANGE OF

STATE]

c.

He mislaid/*began to mislay his car keys.

[POSSESSION]

Although examples (19labe) are anomalous when they refer to a single action, they become marginally more acceptable when every day is added, which makes the action iterative over a period of time (192ac). The only case in which this does not work is in (192b), since the explosion of the car means its total destruction, thus making it impossible for the car to explode again. (192)

a. b. c.

The two cars began to collide every day. * The car began to explode every day. He began to lose/mislay/misplace his car keys every day.

168

Towards

a semantic

syntax

Verbs in other subdomains are also affected by this parameter, which is instantiated in different ways, depending on the specific area of meaning in question. The semantic restriction pertaining to whether an event happens quickly or slowly explains the (nonoccurrence of certain syntactic expressions like the following: (193) (194)

(195)

a. b. a. b. a. b.

The mixture slowly decomposed. [CHANGE OF STATE] * The vase slowly shattered. He slowly trudged down the road. [MOVEMENT] *The Lone Ranger slowly galloped after the bank robbers. *The ghost slowly vanished. [EXISTENCE] His smile slowly faded.

4.2.1.6. Achievement Achievement is present in various areas of meaning and is most evident in resultative constructions which have unconscious, sick, flat, blind and silly as culminating predicates. (196)

a. b.

She beat him unconscious. [CONTACT] The thought of his going away worried her sick. [FEELING]

c. d.

They knocked him flat and stole his wallet. [CONTACT] The Minister of Defense robbed the government blind.

e.

She laughed herself silly.

[POSSESSION] [SOUND]

As Pustejovsky ( 1 9 9 5 : 15) observes, these can be classified as stagelevel predicates. These contrast with individual-level predicates, which are properties that an individual possesses more or less throughout a lifetime (Kratzer in Pustejovsky 1 9 9 5 ) . Such predicates cannot appear in resultative constructions: (197)

a. b.

*John ate himself handsome. *Paul worried himself tall.

Meaning definitions and syntactic projections

169

However, in certain lexical subdomains this type of construction is impossible no matter what type of predicate is used. For example, verbs of definitive change of state such as kill and destroy cannot participate in this kind of construction because the final destination has already been arrived at: (198)

a. b.

*John killed her blind, *Paul destroyed himself silly.

Another type of resultative construction can be seen in the following examples where death, insanity and exhaustion are conceptualized as the final destinations of the affected argument. (199)

a. b.

c.

They starved him to death. [EXISTENCE] She drank herself to death. [CONSUMPTION] The Lone Ranger galloped Silver to exhaustion. [MOVEMENT]

d. e.

She nagged him to death, [SPEECH] His indifference drove her to insanity, [MOVEMENT]

Certain of these verbs belong to the domain of MOVEMENT, but others such as starve, drink, and nag belong to other domains in which such movement is understood more figuratively. Evidently, achievement can be expressed in a variety of different ways. Table 22 shows examples of resultative constructions in verbs belonging to different lexical domains. Although all of the examples refer to completed action, in some the action affects the first argument, and in others the second.

170

Towards a semantic syntax

Table 22.

Affected arguments in resultative constructions 1st Argument

2nd Argument

a. CONTACT

*They knocked themselves flat. They knocked him flat. b. CONSUMPTION She ate/drank herself to death. *She ate/drank him to death. C. FEELING

She worried herselfsick

d. POSSESSION

*Peter robbed himself blind.

e. EXISTENCE f. MOVEMENT g. SPEECH h. SOUND

She worried him sick.

Peter robbed the government blind. She starved herself to death. He starved her to death. He ran himself into the He galloped his horse to ground. exhaustion. *She nagged herself to death. She nagged him to death. She laughed herselfsilly. *She laughed him silly.

4.2.1.7. Causation Wierzbicka (1980) relates causation to notions such as action, process, event, happening, state, and other aspectual classes of Vendler (1967). Paducheva (1997) makes a more detailed analysis of this relation, and suggests a correlation between the taxonomic category of a verb and its lexicographic definition. Though her criteria for establishing semantic classes are somewhat different from ours, she also affirms that verbs belonging to different taxonomic categories should possess different lexicographic formats, while those belonging to the same category should have the same format. Virtually all of the subdomains in MOVEMENT (or those in other domains closely related to MOVEMENT, like CHANGE or FEELING) are susceptible to the causative/inchoative alternation. B. Levin (1993: 2730) observes that when there are two verbs of identical form, one intransitive and the other transitive, and when the transitive form is the causative of the intransitive, such verbs are characteristic of "change of state" or "change of position". Pinker (1989) also points out the fact that there are large groups of motion and change verbs with identical

Meaning definitions and syntactic projections

171

causative and "anticausative" verb forms. A pertinent example can be found below in the subdomain of MOVEMENT IN/ON LIQUID: Table 23.

Movement in/on liquid

MOVEMENT

IN

LIQUID

MOVEMENT IN LIQUID (CAUSATIVE)

(NONCAUSATTVE)

To move/go downward sinkl to go slowly downward below the surface of a liquid.

To cause to move downward sink2 to cause somebody/something to go slowly downward below the surface of a liquid.

plungel to go into a liquid quickly andforcefully.

plunge2 to cause somebody/ something to go into a liquid quickly andforcefully.

divel to go into a liquid, jumping headfirst with one's arms straight above one's head.

dive2 to cause somebody/something to go into a liquid.

submergel to go under the surface of a liquid, especially at a planned speed. ^

submerge2 to cause somebody/ something go under the surface of a liquid, especially at a planned speed.

dipl to go down into a liquid for a short time.

immerse to cause somebody/ something to go down into liquid, covering them/it completely. dip2 to cause somebody/something to go down into liquidfor a short time. dunk to dip something into a liquidfor a particular purpose.

As can be seen in Table 23, the majority of these verbs have the same form for their transitive and intransitive use: (200)

a. b.

He plunged/sank/dove into the water. The gull dipped into the water.

172

Towards a semantic syntax

c.

He plunged/sank/dove/dipped his hand into the water.

Dive can be considered a special case in that its second argument is more semantically specified than that of the other verbs, and is necessarily a part of the body: (201)

a. b.

He plunged/sank/dipped the knife into the water. *He dove the knife into the water.

For that reason, dive cannot have knife as its second argument (201b). Some of these verbs, however, are only causative, and cannot be used intransitively, at least, in their sense of movement in/on liquid. (202)

* The ship immersed/dunked.

The more general lexemes in this set also have other meanings which refer to generic downward movement: (203)

a. b. c.

The seagull plunged into the ocean. The missile plunged downward. The drill plunged into the earth.

[WATER]

[AIR] [LAND]

With the exception of dunk, immerse, and submerge, the others (dive, plunge, sink, dip) can be used for general movement (203 c) as well as movement in air (203b). Sink and plunge can also be used to designate downward movement in more general and figurative contexts: (204)

a. b.

Ralph sank into unconsciousness/a deep sleep/ depression. Ralph plunged into thefire/hiswork/darkness.

However, the fact that its prototypical sense is movement in liquid can be seen in (205), where the medium is unspecified and the default value is obviously in liquid. (205)

He sank the ship/*plane/*drill.

Meaning definitions and syntactic projections

173

For plane and drill to be acceptable complements, they would have to be transformed into sea vessels floating on some liquid surface.

(b)

LIGHT

A second example of a domain marked by this alternation, though to a somewhat lesser degree, is that of LIGHT. In the subdomain of LIQUID MOVEMENT, causation is more evenly balanced with non-causation. However, the domain of LIGHT is different perhaps because of its generic semantic value in which light is emitted from a natural source. With the exception of shine and flash, verbs belonging to this domain have no zero-derived causative form. (206)

a. b.

The janitor shone/flashed the flashlight from the window. * The janitor twinkledJflickered/blazed/flaredJsparkled/ etc. theflashlightfrom the window.

In the causative subdomain, an agent causes a light source to give off light or alternatively, causes a space to be bright (206a). It follows that these verbs are those which have a transitive use, whereas the verbs in the parallel subdomain are one-place predicates which do not (206b).

4.2.1.8. Factivity The parameter of factivity emphasizes the different degrees of the thinker's commitment towards the truth or falsity of a proposition within the complement phrase. Chapter 3 describes the range of subjective, epistemological operators which act as the basic structuring parameters of an important sector of the domain of COGNITION, organizing its subdomains so that the verbs within them encode different degrees of certainty.16 This commitment on the part of the speaker has important syntactic repercussions. In our analysis we shall follow Hopper and Thompson (1973) and Ransom (1986), and will not discuss the formalization of root

174

Towards a semantic syntax 17

transformations within the FG framework. Our objective is only to show that syntactic information is semantically relevant. For example, a semantic value such as certain (the speaker's evaluation that the proposition holds in every possible world) is in consonance with topicalization, a syntactic construction involving the fronting of a non-subject NP interpreted as the topic in a sentenceinitial position. This topic fronting applies to definite NPs, and thus, is compatible with certain, an operator that offers no alternatives. (207)

a. b.

That she did a goodjob, he regrets. That he left, she knows.

Examples (207a) and (207b) show that topicalization is possible because the speaker's evaluation of the propositions (that she did a good job/that he left) is modified by the operator Certain. It follows from this that topic fronting is incompatible with operators that offer options. (208)

a. b.

* That she played well, she believes. * That he go to the dentist, he suggested.

In (208a) and (208b), the complement proposition does not have this operator because of the semantics of the higher predicate: A subjective epistemological operator like probable, which signals the speaker's evaluation that the proposition is expected to be the case, is in accordance with syntactic transformations known as "root transformations" (Hopper—Thompson 1973). Such transformations cannot operate on questions, reduced clauses, and presupposed clauses since it would not make any sense to emphasize constituents in a sentence whose proposition is presupposed or already known. Consequently, these syntactic constructions are compatible with a probable meaning and its possible alternatives. (209)

a. b.

He wants to leave the country, he says. [Complement preposing] The trade unions are obsolete, the president concluded. [Complement preposing]

Meaning definitions and syntactic projections (210)

a.

175

The group vows that disrupt the elections they will. [VP preposing] Barbie announced to Poindexter that marry him she would. [VP preposing]

b.

This type of meaning is therefore sensitive to constructions such as complement preposing (209ab) and VP preposing (210ab).

4.2.1.9. Conation The parameter of conation refers to the speaker's attempt to carry out an action. It is lexically marked by try, which appears as part of the genus of all of the verbs in (211). In the same way as start, stop, continue, it can either appear as the genus itself, or alternatively as part of the genus. (a)

(211)

GENERAL ACTION

try

to make an effort to do something. attempt to try to do something difficult. endeavor to try very hard to do something [formal]. strain to try very hard to do something, especially making a great physical effort, strive to try hard to do something, making a very great effort over a period of time.

The focus of this semantic value is thus on the initial phase of the activity without any mention of whether the action is actually carried out. It is explicitly marked by try, and appears in a variety of different areas of meaning. (b)

COGNITION

(212) coax

to try to persuade somebody to do something in a gentle, pleasant way.

176 (c)

(213)

Towards a semantic syntax POSSESSION

hawk to try to sell something by shouting in the street/taking it around to various people [informal].

The presence of fry as a semantic constituent of the definition has syntactic consequences insofar as it encodes an atelic state of affairs, and as a result, this sort of construction is compatible with the test for telicity (Dik 1997a: 108-111). (214) (215)

a. b. a. b.

I coaxed him for an hour (*in an hour). *It took me three hours to coax him. I hawked the jewelry for an hour (*in an hour). *It took me three hours to hawk the jewelry.

However, a change in the complementation of certain verbs brings about a change in the state of affairs, as can be observed in (216abc) (216)

a. b.

c.

I coaxed him into going to the store. (= He went to the store.) I coaxed him to go to the store (i) but he wouldn't. (ii) and he finally did. *It took me three hours to coax him to go to the store.

Here the inclusion of a prepositional phrase can cause the canonical meaning of coax to vary and acquire the meaning of achievement (216a), whereas the infinitive leaves the outcome uncertain (216b). Nevertheless, the same thing cannot be said for hawk, which belongs to another domain and thus does not accept this type of complementation.

4.2.2. Lexically-realized optional parameters

Lexically-realized optional parameters explain why certain arguments, though semantically present, are not syntactically prominent in the

Meaning definitions and syntactic projections

177

actual linguistic expression. The most obvious examples of this can be found in POSSESSION, in the subdomain of acquisition: (217)

a. b. c. d.

I bought/purchased a new car. I bought/purchased a new car from Tom. I bought/purchased a new car from Tom for Mary. I bought/purchased a new car from Tom for Mary for $5000.

Although all of the arguments are characteristic of verbs of possession, the only argument that is strictly necessary for the syntax of the sentence is a new car (217a). The others are all potential in the structure of this type of verb, but not necessarily activated all at once. In fact, when they all appear together, the result sounds awkward (217d). All of these arguments (are true arguments because they are syntactically realized in the verb in question. Pustejovsky (1995: 63-4) also distinguishes between default arguments and shadow arguments. Default arguments participate in the meaning, but are not necessarily expressed syntactically: (218)

a. b.

She made the wedding dress out of silk. The second little pig built his house out of sticks.

Shadow arguments, which are similar to Dik's implied satellites (Dik et al 1990), are semantically incorporated in the lexical item. As we have seen in previous examples from verbs of contact (17 lab), they can be expressed only by operations of subtyping or discourse specification. According to Pustejovsky (1995: 40-41), utterances with explicit shadow arguments are not ungrammatical, but certainly deviate from the notion semanticality. He argues that the licensed expression of certain types of verbal arguments is predictable from the semantics of the expression and semantic operations in the grammar.

4.2.3. Lexically-realized contextual parameters

Lexically-realized contextual parameters, though not syntactically projected, serve as clues for contextual setting. They are of different

178

Towards a semantic syntax

types and elaborate the typical schema of a lexical subdomain in contrasting ways, in terms of location, instrument, intensity, movement, etc. Examples of this parameter can be found in the domains of POSSESSION, VISUAL PERCEPTION, POSITION, and SPEECH:

4.2.3.1. Possession (219)

To give somebody something in exchange for money sell to give something in exchange for money. vend to sell [formal]. Φ peddle to sell small things by going from place to place. Φ hawk to try to sell something by shouting in the street/taking it around to various people [informal].

Peddle and hawk in the domain of POSSESSION (219) activate a schema of location and of movement because in both cases, the activity is carried out while moving from one place to another. (220)

a. b. c. d. e.

The traders hawked their wares on the street corners. The man peddled his wares from door to door. ?The man peddled his wares in the department store. ?The shop attendant hawked the perfume in the department store. ?He wentfrom door to door peddling Rolls Royces.

The respective meaning definitions of these two verbs make it anomalous for someone to peddle/hawk something in a department store. Nor is it possible, at least in the normal order of things, to peddle/hawk large and expensive objects such as a pair of matching Rolls Royces or a mansion in Malibu. Whereas (220cde) sound odd, they are remotely possible with a great deal of contextualization. However, the parameter that cannot be violated is that of sound in hawk.

Meaning definitions and syntactic projections (221)

a. b.

179

The traders loudly hawked their wares in the market place. *The traders silently hawked their wares in the market place.

The reason for this is that sound is an intrinsic part of the denotation of hawk, whereas the location of the action is not.

4.2.4. Position (222)

To stay somewhere, having it as one's home live to stay somewhere, having it as one's home. lodge to live somewhere (for a short period of time and paying money for it), board to live somewhere (in somebody's home and receiving meals) in return for payment, camp to live somewhere for a short time in a tent/caravan, bivouac to camp in a bivouac (a temporary shelter which is not a tent), squat to live somewhere (public land/unused building) illegally. room to live in a room/rooms in somebody else's house, cohabit to live together, having a sexual relationship (of unmarried people) [formal], shack (up) to start to live together, having a sexual relationship (of unmarried people) [informal].

In the domain of POSITION, verbs such as lodge, board, squat, camp, bivouac and cohabit with the base meaning of to stay in one place, all belong to the same subdomain. All of these verbs have live as their genus, and the descriptive parameters in their definitional structure refer to time, payment, location, and person(s) living with the positioner. The lexical contextual parameters within this subdomain are the following:

180

(223)

Towards a semantic syntax

Lexical contextual parameters a. Short period of time [during] , lodge b. Payment [in exchange for]: lodge, board c. Particular location [at/in]: camp, bivouac, room, squat d. Company [with] : cohabit, shack up

Although semantic parameters codifying sociocultural norms are different, depending on the domain and subdomain they are embedded in, it is relatively easy to deduce what such norms are because they are precisely what is not lexicalized. As was seen in manner-of-walking verbs (section 3.2.3.1), we tend to find words for what draws our attention, and what surprises us is generally what differs from our expectations or deviates from the socially accepted norm. Squat, for example, activates a whole sociocultural context of property rights, which includes those who make use of that property with or without the right to do so. From its original sense as a verb of body position, more specifically a rather unorthodox kind of sitting (on one's heels instead of on a chair/ground), squat has a meaning extension to another subdomain, to stay at a particular location. In this case, the location is an unused building/land where the positioner lives illegally (without permission and without paying rent). Cohabit and shack up are both negatively loaded, and lexicalize a relation between people of the opposite sex, who live together without being married. The fact that these verbs have become considerably less frequent is indicative of a gradual change in social norms.

4.2.4.1. Speech: codification of biological norms Biological norms are also codified in the lexicon. An example of this can be found in the following subdomain of speech: (224)

To speak with difficulty stutter to speak with difficulty (with short stops one cannot control). stammer to speak with difficulty (hesitating and repeating sounds and words).

Meaning definitions and syntactic projections

181

lisp to speak with difficulty (using "s " sounds which are not clear). bumble to speak with difficulty in an awkward, uncontrolled way. The subdomain in (224) lexicalizes deviations from the biological norm. If someone stutters, stammers or lisps, it is more difficult to communicate because of physical impediments. This kind of speech is negatively evaluated, not so much for its content, but for its sound quality. Consequently, example (225) is unacceptable because the speaker's lack of control is invariably regarded as negative: (225)

*He stuttered/stammered/lisped/bumbled well.

The definitions of these verbs tell us that speech should prototypically be fluid without stops or hesitations. Sounds should not be repeated, and phonemes (such as /s/) should not be unnecessarily emphasized. The norm that dominates all of the preceding ones is that of speaker control.

4.2.4.2. Speech: codification of power relations

Another important type of norm is codified in the power relation between arguments, as well as the context of the action, something that is clearly delimited within its meaning. A case in point can be seen in upbraid. (226)

To say that somebody has done something bad scold to say somebody has done something bad, talking to them angrily, chide to scold somebody gently [old-fashioned/. Φ upbraid to scold somebody for a long time [formal]. sermonize to scold somebody, giving them unwanted advice. moralize to sermonize somebody, telling them what is right/wrong.

182

Towards a semantic syntax

berate

to scold somebody loudly [formal][oldfashioned], reprimand to scold somebody severely and officially. Upbraid has an agent, a patient, and optionally a reason for scolding the patient in that way. Its argument slots can be filled by different types of entities: (227) Agent a. The teacher b. He c. ?The soldiers ά.?The children

upbraided

Patient Cause the students their poor marks his wife > for < her extravagance the general his bad temper their parents coming home late

Examples (227a-d) show us that one of the selection restrictions for the first two arguments is authority. The power relation between teacher and students (221 a.) and husband and wife (227b), gives the agent the certain right to upbraid the patient. In contrast, example (227d) is odd, and in need of contextualization because children do not prototypically have sufficient authority to upbraid their parents. The same is true for the inverted power relation between soldiers and the general in (227c). The third argument, the reason for being upbraided, is invariably a behavior/activity/attitude which the agent is not in agreement with. Furthermore, in contrast to other verbs in the same subdomain, the grounds for being upbraided are usually definite and justifiable (poor marks, extravagance, lack of discipline) etc. (228)

a. b.

The pharmacist upbraided his children for attending church. The pharmacist berated his children for attending church.

always always

Although both upbraid and berate unprototypically have a positively valued element for the third argument, (228a) is less acceptable than (228b) because upbraid has a negatively evaluated third argument, whereas berate does not. In consonance with the whole

Meaning definitions and syntactic projections

183

subdomain, the activity lexicalized in all these verbs is negatively valued, hence the unacceptability of example (229): (229)

*He upbraided/berated/chided

her well.

4.2.4.3. Contact (230)

To touch somebody/something quickly, using a lot of force hit to touch somebody/something quickly using a lot of force with one's hand/other object, usually to hurt them/push them into something else. strike to hit somebody/something with one's hand/stick/other object in anger/to make them go in a particular direction, knock to strike somebody/something a sharp blow, tap to hit somebody/something with a quick, light blow. rap to tap somebody/something, giving them a series of quick blows cuff to hit somebody lightly with one's hand, slap to hit somebody with the palm of one 's hand, smack to slap somebody sharply, spank to slap somebody on the buttocks as punishment, whack to hit somebody/something, making a sharp sound, swat to hit somebody/something with a quick, swinging movement with one's hand/folded newspaper, bash to hit somebody very hard, intentionally, bump to hit somebody/something with a dull sound, thump to hit somebody/something hard, usually with one's fist, making a dull, heavy sound, punch to hit somebody/something hard with one's fist, sock to hit somebody very hard with one's fist [Informal]. jab to hit somebody repeatedly with one's fist with quick, sharp movements, club to hit somebody with something blunt and heavy, clout to hit somebody hard with one's fist/ any heavy object, butt to hit somebody/something with one's head (like a goat), kick to hit somebody/something very hard with one's foot. dribble to kick a ball repeatedly, moving with it [soccer], hammer to hit somebody/something repeatedly (as i f ) with a hammer.

184

Towards a semantic syntax crown to hit somebody very hard on the top of the head IInformal]. brain to hit somebody very hard on the head [informal]. box to hit somebody on the ear with an open hand/fist, beat to hit somebody/something hard and continuously for a long time. batter to beat somebody/something with great force repeatedly times with one's fists/other object, clobber to beat somebody with great force [informal]. whip to beat somebody (slave, sailor) with a whip as punishment, lash to whip somebody (slave, sailor) as punishment [old-fashioned]. flog to whip somebody with great force as punishment [old-fashioned]. flagellate to whip somebody as a religious act of penance [formal]. birch to whip somebody with a birch, cane to beat somebody with a cane as punishment.

The domain of CONTACT is probably one of the most prototypical kinds of action. The generic term of CONTACT is hit, the central unit that defines the syntactic and semantic universe of discourse. The definitions of these verbs are indicative of different types of contextual parameters:

Meaning definitions and syntactic projections Table 24.

Verbs of contact: lexically-realized contextual parameters

PARAMETER

Type Of blow

185

SUBPARAMETER

Intensity

lightly (tap, cuff) hard (strike, punch, thump, beat, knock, clobber, clout, batter, cane, whip, lash, birch) very hard (sock, knock, clobber, bash, kick, flog, crown, brain)

Sound

sharp, punctual, loud (crack, whack, knock, smack, pb) dull, punctual (thump, bump) quick (tap, rap, swat, jab) swinging (swat) iterative (rap, jab, beat, batter, clobber, cane, whip, lash, flog, birch, hammer)

Movement

hand (hit, cuff, fist (punch, box, sock, thump, jab, strike, swat) batter, clout) palm (slap, spank, box, smack) head (butt) foot (kick, dribble) Instrument

object

club (club) hammer (hammer) cane (cane) whip (whip, lash) newspaper/fly swatter (swat) birch stick (birch) unspecified object (batter, clout)

Entity affected

head (crown, brain) buttocks (spank) ear (box) ball (dribble)

Reason

anger (strike) desire to hurt (spank, cane, whip, lash, flog, birch) desire to move something (hit, strike)

Either explicitly or implicitly in the context of the action, the elements lexicalized are: (1) an agent that effects the blow; (2) a blow measured

186

Towards a semantic syntax

in terms of force, sound and movement; (3) an instrument with which the blow is delivered (body part, object); (4) the entity affected by the blow (person, body part, object, ball); (5) the reason for the blow (anger, punishment, causation of movement); (6) contextual frames activated (sports, institutionalized punishment). Differentiating features and the information that they encode also determine the semantic characterization of the higher-level schema in which the lexeme in question is embedded. Verbs of contact are very much a part of their context because depending on the nature of the target, these verbs activate frames indicative of behavioral norms in our society. The most important contextual parameter in this lexical domain is that of institutionalized punishment for inappropriate social behavior within the context of family {spank), old-fashioned school (whip, cane, birch), conflictive male and female relationships {slap, beat, batter) and legally imposed power relations (e.g. servitude, imprisonment: lash, whip, flog). The second contextual parameter activated is that of competitive sports events, primarily boxing {punch, jab, belt, clobber), but also sports played with a ball such as football/soccer (kick, dribble).

4.3.

Lexical iconicity

In the FLM lexicon, verbs within the same subdomain have similar syntactic behavior. This is in consonance with research in child language acquisition in which researchers have observed a similar phenomenon, when trying to map syntax onto meaning. Grimshaw (1994: 418) writes that researchers working on syntactic mapping have investigated the idea that sets of subcategorization frames, and not just single frames, play a role in the acquisition of verbal meaning. Such frames, however, only acquire real significance when they are considered in the context of lexical macrostructure because syntax is not sufficiently fine-grained to constitute the basis for lexical organization. Gleitman—Gleitman (1997: 50) underline the fact that it is impossible to converge on construals of verbs from syntactic properties alone because most semantic distinctions are not formally expressed with this machinery. Nevertheless, the possibility of using semantic distinctions as well as syntactic properties for such construals

Lexical iconicity

187

offers a much richer panorama of lexical structure. The close interrelation of syntax and semantics in predicates leads us to postulate the following principle of iconicity: (231)

Principle of Lexical Iconicity The greater the semantic scope of a lexeme, the greater its syntactic variation.

There are two important corollaries which follow from this principle. Firstly, the genus marks the semantic scope of the rest of its hyponyms, and at the same time, it also introduces the syntactic universe relevant to the whole lexical subdomain. Secondly, the organization of both the semantic and the syntactic information is hierarchical, and the two resulting hierarchies are not independently motivated, but constrain and interact with one another.

4.3.1. The genus and its scope The genus of a subdomain is the most general or prototypical term. This means that the more prototypical a term is, the more prototypical effects it will show. Prototypical or generic terms are those whose definition is included in that of the other members of the lexical domain/subdomain. Conversely, the syntax of the more specific troponyms can be derived from that of the generic term. This is in consonance with what Vossen (1995: 251) has found in his study of nouns. He underlines the strong inverse correlation not only between frequency of use and restriction on syntactic context, but also between frequency of use and degree of polysemy. The examples offered in this section come from the domains of CONSUMPTION, VISUAL PERCEPTION, and COGNITION and are illustrative of a tendency which we have found to be generalized throughout the whole verbal lexicon.

188

Towards a semantic syntax

(a) CONSUMPTION

(232)

drink

to consume liquid, taking it into one's mouth and swallowing it. imbibe to drink alcohol [formal], gulp (down) to drink something very quickly, quaff to drink something quickly [old-fashioned], swig to drink something quickly in large amounts in a series of big swallows [informal], swill to drink something quickly and greedily in large amounts [informal], guzzle to drink something (especially alcohol) very quickly, greedily and noisily in an unattractive way. tipple to drink something (especially alcohol) secretly and in small amounts [informal], sip to drink something slowly in very small amounts.

The complementation patterns of the generic term drink (233a-e) form the syntax of the subdomain. (233)

Complementation patterns of drink a. He drinks heavily. [—] b. He drankfour martinis. [NP] c. He drank to the health of all of his relatives. [ PP (to)] d. He drank himself to death. [Refl PP (to)] e. He drank himself into a stupor. [Refl PP (into)]

The complementation of the other verbs in the subdomain can be derived from this same inventory:

Lexical iconicity

Table 25.

189

Typology of syntactic frames: liquid consumption

drink imbibe gulp (down) quaff swig swill guzzle tipple sip (at)

0 + + -

NP + + + + + + +

PP (to) +

REFL PP (toAnto) +

-

-

-

—

-

-

-

—

-

+

-

-

-

-

+

-

-

Table 25 shows how the troponyms of drink are much more restricted in their syntactic patterns. The only verb that can take more than one of the four possible syntactic patterns is imbibe, which basically differs from drink in its pragmatic feature of formality, as well as the evident restrictions on its goal argument: (234) (235)

Please give me a glass of ginger ale. I am afraid I don 7 drink/imbibe (= alcohol). a. He drinks milk/coffee/orange juice every day at lunch. b. *He imbibes milk/coffee/orange juice every day at lunch.

The default goal argument of both drink and imbibe is alcohol (234). However, when the goal argument is specified, imbibe restricts the possible fillers of this slot to alcoholic drinks (235b), whereas drink does not (235a). (b) VISUAL PERCEPTION

In the lexical domain of VISUAL PERCEPTION, see not only introduces the basic meaning, perceiving somebody/something with one's eyes, but also has the full set of complementation patterns relevant to the domain:

190 (236)

Towards a semantic syntax Complementation patterns of see a. Tom sees. (—) b. Tom saw the traffic light. [NP] c. Tom saw the traffic light turn red. [NP-bare infinitive] d. Tom saw the traffic light turning red. [NP iwg-participle] e. Tom saw that the traffic light had turned red. [77ia/-clause] f. Tom saw what was happening to the traffic light. [Wh-clause]

The full set of syntactic and semantic information is distributed throughout the domain in terms of inheritance mechanisms. Predicates at more specific levels of the hierarchy (and thus more constrained in their semantic scope) have meanings which pinpoint very specific areas of meaning, and this reduction in semantic scope brings with it a corresponding reduction in syntactic potential. As a result, visual perception verbs like watch, stare, glance, gaze and peek inherit the syntactic and semantic features of the governing predicate see : (237)

To see by intentionally directing one's eyes

look to see by intentionally directing one's eyes. glance to look at somebody/something quickly, peek to look at somebody/something quickly, secretly from a hiding place. peep to peek at somebody/something very quickly, watch to look at somebody/something, paying close attention to them. observe to watch somebody/something carefully, often in order to learn something, regard to look at somebody/something with a particular feeling, contemplate to look at somebody/something steadily, in a quiet, thoughtful way. gaze to look at somebody/something steadily for a long time because they seem attractive or surprising, stare to look at somebody/something for a long time with wide open eyes. goggle to stare at somebody/something in surprise.

Lexical iconicity

191

gape

to goggle at somebody/something with an open mouth. gawk to stare at somebody/something in a stupid, unthinking way [informal], glare to stare angrily at somebody/something, in an unfriendly way. glower to glare at somebody/something for a long time, ogle to stare at somebody with sexual interest.

Table 26.

Typology of syntactic frames: visual perception

see look glance peek peep watch observe regard contemplate gaze stare goggle gape gawk glare glower ogle

0

NP

+

+

+

-

+ +

-

-

-

-

+

+

-

+ + + + +

+ + +

+ + + -

-

NP NP-bare inf. j/ij?-part. + + + + + + + + + + + + + + + + +

Thatclause + -

-

-

-

-

-

-

Wh- PP {at) clause + + + -

+

+ +

—

+ + + + + + +

-

+

+

—

+

-

-

-

-

+ -

-

-

+

-

-

-

-

-

+

-

-

-

—

-

+

—

•

-

-

-

-

+

-

-

-

-

+

-

—

-

-

-

-

-

-

+ +

In Table 26 the only pattern which see does not take is the collocation with at. However, that is due to the fact that the other verbs are forms of intentional seeing, and at can be regarded as the lexicalization of that intentionality

192

Towards a semantic syntax

( c ) COGNITION

Exactly the same thing occurs in the lexical domain of COGNITION in which the superordinate term also introduces both the semantic and syntactic universe encoded in the lexical subdomain. (238) To think something is true believe to think something is true even when it cannot be proved. consider to believe that somebody/something is a certain way (after thinking it over), accept to consider something to be true/reasonable /satisfactory, often saying so or showing agreement, recognize to accept something as true, assume to accept something as true without proof. presume to assume something is true, taking it for granted. As the most general term, believe contains both the semantics and syntax of its more specific troponyms. Alternatively, Pustejovsky (1991) proposes a system of type coercion to explain how a verb like believe can accept arguments of varying type without requiring multiple verb entries in the lexicon. He defines type coercions as "a semantic operation that converts an argument to the type which is expected by a function, where it would otherwise lead to a type error". Table 27.

believe consider accept recognize assume presume

Typology of syntactic frames: to think something is true 0

NP

NP Inf

NP Comp

+

+ +

+ +

-

—

-

+ + + + +

-

-

-

+ -

+ + +

-

-

+

Thaiclause + + + + + +

PP (in) + —

—

-

—

-

Lexical iconicity

193

As the semantic hierarchy becomes increasingly specific in both VISUAL PERCEPTION and COGNITION, we move from a wide semantic conception of perceiving visually as well as mentally towards more specific ways of carrying out this same action. Given that this semantic correlation is in accordance with a reduction in the syntactic potential of the verbal predicates, semantics is the component that regulates the linguistic encoding of a lexical subdomain.

4.3.2. The interaction of semantic and syntactic hierarchies The second corollary of the Principle of Lexical Iconicity concerns the interaction of the semantic and syntactic hierarchies that define a given semantic subdomain. An analysis of the way these hierarchies are encoded reveals that both are mutually dependent and constrain one another. A pertinent example can be found in the lexical subdomain to say to somebody that something will be done in the future from the domain of SPEECH. The act of promising is a commissive speech act (Searle 1975) in which the speaker often commits himself/herself to a course of action. Wierzbicka ( 1 9 8 7 : 2 0 7 ) underlines the obligation inherent in promise: "In other words, by promising something, the speaker offers his personal credibility in general as a kind of guarantee that he really will perform the action in question." (239)

SPEECH:

promise

to say something will be done in the future to say that something will be done in the future or is certain. vow to promise something solemnly (before God). swear to promise solemnly to do something/that something is true. forswear to swear to give somebody/ something up [formal] [oldfashioned]. pledge to promise solemnly to do or provide something.

194

Towards a semantic syntax

commit

to pledge something (oneself/ one's properly) to a person/ cause. plight to pledge formally to marry somebody [old-fashioned], guarantee to promise that something will certainly be so, taking responsibility for it. vouch for to guarantee somebody/ something, giving personal assurance. This lexical subdomain (239) also has the following morphosyntactic information inherent in its members. As shall be seen, this hierarchy is not isolated, but is motivated to a certain extent by semantic parameters: Table 28.

Typology of syntactic frames: to say that something is certain 0

promise vow swear forswear pledge commit plight guarantee vouch (for)

+ •

-

—

-

—

-

NP + + + + + + + + +

NPI NP 2 + -

+

Toinfinitive + + +

That-

Quote

PP

clause + + +

+ + +

+ + +

-

-

•

+

-

-

—

•

-

+ + +

+

-

-

-

-

•

-

+

•

-

The interaction between the semantic and syntactic hierarchies becomes evident in the fact that semantic parameters act as constraints and filter the morphosyntactic realizations of the lexemes in the subdomain. A case in point is the synsem parameter of time, which indicates when the promised entity/action will actually happen. In this case, the parameter does not refer to the duration of the action encoded in the verb, but to the time reference of its complement. In (240) the complement encodes a future course of

Lexical iconicity

195

action, but is only acceptable in verbs at the first and second levels of the semantic hierarchy {promise, vow, swear, pledge, guarantee). (240)

promise vow swear pledge guarantee *vouch *forswear *commit *plight

that I will never smoke again, never to smoke again.

This parameter is also influential in the decodification of what at first appears to be an NP complement, but what in reality is a concealed proposition. In all of the extended complements (241 a-d), the action is to be understood in the future. (241)

a. b.

c.

d.

Paul promised her a chocolate cake (= that he would make/buy/steal/sell her a chocolate cake). Paul pledged fifteen no-cal chocolate cakes for the Weight Watchers party (= that he would make/buy/steal fifteen no-cal chocolate cakes for the Weight Watchers party). He vowed revenge for the insult that he had received (= that he would have/get revenge for the insult that he had received). The knight swore his allegiance to the king (= that he would give his allegiance to the king).

Another relevant parameter is that of factivity, which emphasizes the different degrees of the speaker's commitment towards the truth or falsity of a proposition within the complement phrase. Although these are not verbs of cognition, a relevant aspect of their meaning is the speaker's commitment and belief that what he promises/swears/vows/pledges, etc, will really come to pass. As such, many of these verbs are sensitive to topicalization:

196

(242)

Towards a semantic syntax

a. b. c. d. e.

That she did a good job, he promises. That John actually studied, he swears. That he would always be faithful, he vowed. That he would never drink again, he pledged. That they would receive the money, he guaranteed.

Lexically-realized contextual parameters are also important in this subdomain. To a certain extent, this is related to the parameter of speaker certainty, and the resulting presence of solemnly in the meaning definition. In the case of vow and swear, this means a projection into different cultural settings, something which has a direct effect on the semantic characteristics of the NP complement. In fact, the principal way vow is differentiated from the rest of the verbs is in respect to the place where this speech act takes place. One usually makes a vow in a place of worship, and the recipient of the vow is prototypically a deity. What one vows is generally a course of action which affects one's entire existence {revenge, celibacy, allegiance, one's life). (243)

a. b.

*/ vow a yellow toothbrush/a visit to the dentist/my butterfly collection. *I vow your allegiance/life/chastity.

For this reason, it is impossible to vow concrete entities even if they are possessions (243a), or something more abstract that does not belong to oneself (243b). Swear can take some of the same complements as promise and vow. (244)

He swore

revenge/allegiance/*poverty/*celibacy.

Notwithstanding, it is less versatile because of its other sense, to say rude words, which is the default value. As a result, greater specification is necessary {an oath of), and when this is added, practically all the complements of vow become possible: (245)

He swore an oath of

revenge/allegiance/poverty/celibacy.

Lexical iconicity

197

However, some of the possible complements of promise are still blocked for swear for obvious semantic reasons: (246)

*He swore an oath of a million dollars/making breakfast every Sunday/fewer wrinkles.

The scope of the complement is thus enlarged by specifying the precise type of oath, but in the same way as vow, the complement will always be something important which affects the course of the speaker's whole life: (247)

a. b. c.

*He swore an oath of courtesy/happiness. He swore an oath to be courteousfor the rest of his life. ?He swore an oath to be happyfor the rest of his life.

Although one can certainly make the resolution to be courteous forever after (247b), it sounds strange to do the same for happiness. The oddity of (247c) lies in the fact that being happy is something that the experiencer unfortunately cannot control. Another way that oath is restricted is in the specification of the scenario in which the act is taking place. Like vow, swear also prototypically evokes a specific setting in which a solemn promise is made, and this context is so essential because it is part of what confers meaning on the act. When one simply swears an oath without any further specification, the phrase can be ambiguous because an oath is also profane language. However, in a courtroom an oath would evidently be a promise that the speaker's words are true. (248)

The witness swore an oath [= promised to tell the truth] in the courtroom.

Curiously enough, the specification of the context in itself without an oath is not enough to override the primary meaning of swear. (249)

The witness swore [= used rude language] in the courtroom.

The meaning of swearing an oath likewise depends on the person receiving the oath as well as the location where the act of swearing

198

Towards a semantic syntax

takes place. For example, if the oath is sworn to a person of great authority (a king or bishop) in a castle or cathedral, then it could be a promise of loyalty or religious dedication. (250)

The knight/priest swore [= promised loyalty/religious dedication] an oath to the king/archbishop [in a castle/cathedral].

Forswear as a troponym of swear contains all the information in the superordinate plus the additional specifications. In this case, the semantic restrictions of the complement are significantly reduced. One can only forswear something bad for oneself (unless one is speaking humorously/ironically): (251)

a. b.

I shall forswear cigarettes/my sinful ways/drinking. ?I shall forswear helping others/a healthy lifestyle/ feeling happy.

Since forswear is to solemnly promise under oath to give something up, it is semantically odd to renounce something that is positively evaluated in our society. A further specification has to do with temporal variation, as this lexeme is now considered old-fashioned. Pledge is the speaker's promise to give something to someone. Here what is being given is as important as the obligation under which the speaker is putting himself/herself. The given entity can be either concrete or abstract: (252)

a. b. c. d.

I pledged my loyalty/friendship/support/help/a dollars to the president. *Ipledged revenge. *Ipledged the moon. */pledged the idea I had last night.

million

For this reason, in (252a) it is possible to pledge one's loyalty or a million dollars. Other things that can be pledged are positively evaluated entities that the speaker wishes to give to the receiver. Examples (252bcd) are semantically anomalous, but for different reasons. Revenge is not an acceptable complement because one

Lexical iconicity

199

cannot pledge negatively evaluated things. Neither can one pledge things that one does not possess or will never possess in the foreseeable future (e.g. the moon), and ideas cannot be pledged because they cannot be categorized as help or support for somebody else. The troponyms of pledge, commit and plight all take an NP and an NP PP (to). Commit and pledge are similar in that both accept a human argument in the sense of setting oneself on a course of action: (253)

I have pledged/committed myself to secrecy.

In a similar way, the agent can solemnly promise his possessions to help or support a cause. (254)

I have pledged/committed my fortune to finding a cure for cancer.

Plight, in much the same way as forswear, has a very reduced application. The only complement it can have is troth, and is now considered old-fashioned to the point of being archaic. The semantic parameters described, as well as the resulting inheritance or blocking of syntactic patterns can be seen in Figure 5. In the same way that there is an inheritance of meaning components in the hierarchy, there is also a similar inheritance of complementation patterns. Semantics constrains and interacts with syntax, and thus the configuration of the information encoded is not random.

200

Towards a semantic syntax

To say that something will be done

PROMISE

Speaker certainty Solemnly (formal act)

thatCL

forswear

pledge

to- N P , Inf n p 2

+ +

+ +

•

-

-

•

+

+

+ +

Speaker certainty Solemnly (formal act)

Negative (old-fashioned)

NP

Self/Possessions

commit

•

-

-

Marriage {troth)\ (old-fashioned)

plight

•

-

-

+

+

+

•

+

Speaker certainty

guarantee

Quote

—

-

—

—

-

+

vouch for

Figure 5.

4.4.

Synsem interface of promise and its troponyms

Summary

In a lexical subdomain, semantic parameters constrain and filter syntactic projections. Such lexically-realized parameters can be of three types: (i) grammatical; (ii) optional; and (iii) contextual. Synsem grammatical parameters such as duration, iteration, temporal reference, inception, cessation, causation, achievement, conation,

Summary

201

and /activity are a determining factor in the actual process of constructing an underlying clause structure. Optional parameters are primarily concerned with the obligatoriness and realization of arguments within the clause. They indicate which arguments are truly a part of a verb's argument structure as based on its position in the context of the entire lexical domain. Contextual parameters are not syntactically projected, but codify different types of pragmatic information such as biological, and social norms, as well as interpersonal relations. As a demonstration of the extent to which a verb's membership in a semantic set helps to explain its syntactic behavior, a wide range of verbs from different lexical domains have been analyzed within their respective paradigmatic sets. The fact that each set has been elaborated on the basis of shared semantic information obviously brings meaning definitions to the forefront because semantic information is projected onto syntax Another important aspect is that the semantics encoded in a lexical subdomain is in consonance with its syntactic potential. A verb's complementation is not random, but is systematically coherent with its semantics. The distribution of synsem information is also hierarchical insofar as the wider the semantic scope of the lexeme, the greater its syntactic variation. The configuration of both semantic and syntactic information is basically hierarchical, and the two resulting hierarchies are not independently motivated, but constrain and interact with each other.

5.

The cognitive axis

5.1.

Introduction

The third component in the FLM lexicon is the cognitive axis, the intersection point of language and thought. Meaning is conceived as an internal knowledge representation, or as a set of cognitive products and procedures, which elaborate and process information. In a manner of speaking, lexical meaning is in itself a cognitive model in which part of our knowledge about the world is symbolized by a linguistic expression. Thus, lexical units are translations of our interpretation of perceptual data because they and their combinatorial possibilities are the external representations of our model of the world. Such a language of internal meaning representation explains translation (among other things). To translate a text from one language to another, there must be a conceptual representation shared by both. This is the basis of the interlingua approach in machine translation in which a pivot language or non-language-specific representation is used to transfer meaning from the source language to the target language. In fact, underlying clause structures in FG have been proposed as an interlingual representation in machine translation (Dik 1986, 1990; Van derKorst 1990). This chapter explores how lexical structure on both the paradigmatic and syntagmatic axes can be said to reflect cognition through the codification of linguistic knowledge. Linguistic knowledge involves a systematization of the structural patterns in language behavior and a characterization of the mechanisms in this patterning. Patterns at the lexical level are the most important because they also affect those at the phonological, syntactic, and pragmatic levels. Lexical relations at different levels of the lexicon hint at a map of conceptual relations and give us a tantalizing glimpse of mentalese, a language of thought, which compared to any given language is richer in some ways and simpler in others (Pinker 1994b: chap. 3). Efforts to trace this map of conceptual relations have so far met with little success

204

The cognitive

axis

partly because of the complex interconnections present in any knowledge representation, and partly because of the many different types of behavior that must be examined in order to give an account of conceptual structure. Language is only one of the various kinds of behavior that must be studied, but it is a very privileged type because some of the clearest insights into how our mind works have been offered by linguistic analysis.

5.2.

Language and brain

According to Saint-Dizier—Viegas (1995b: 2), a mental lexicon is an inventory of lexemes, which are known and stored in our brain. Thanks to this inventory, we can (i) name and describe persons, objects, places, events and feelings; (ii) convey our thoughts through lexicalization and grammaticalization; (iii) recognize or produce words that we hear, read or pronounce. The two areas of the brain that are specifically related to language are Broca's Area (language production) and Wernicke's Area (language comprehension). As both areas are located in the same hemisphere, it is often believed that one hemisphere (generally the left) controls speech and thought processes, while the other hemisphere (generally the right) is less important. However, the idea that the left hemisphere is dominant for all aspects of language has been gradually replaced by the view that hemispheric dominance differs for different components of language. Heilige (1993: 68) observes that although injury to different areas of the left hemisphere selectively disrupts the ability to use syntactic or semantic information, injury to the right hemisphere disrupts the use of context and the ability to integrate information across sentences and larger units of language (i.e. the pragmatic aspects of language). Both our mental lexicon and a conventional dictionary contain information about lexical items, but the similarity evidently stops there. The obvious differences between the two stem from the fact that instead of printed pages, the mental lexicon is a collection of highly complex neural circuits in the brain. Meaning is the result of the activation of this circuitry because each time a certain word is encountered, the same set of neurons goes into action. Patterns of

Language and brain

205

neural activity are not static, but change with the result of acquired experience, and in function with the physical characteristics and the contexts in which they have occurred. However, certain components of these patterns remain constant when similar entities are encountered. Altmann (1997:122) writes: And naturally, objects that are similar to one another will tend to occur in similar contexts, to behave in similar ways, and to lead to similar experiences. Consequently, the neural activity associated with them will also be similar. Although the pattern of activity that is evoked in response to a shark will be different from the pattern evoked in response to a goldfish, components of those patterns will be similar, because the two fish share certain attributes. It is the similarity of their patterns of neural activity that corresponds to the concept of fish. Obviously this not only applies to fishlike entities, but also to abstract events such as promise, walk, and scold, and extends to all of the concepts that underlie our lexical representations. This amounts to quite a few if we consider that, depending on who happens to be doing the counting, the vocabulary of the average speaker has been estimated to be anywhere from 50,000 to 150,000 words. Even if we accept the lower estimate, the fact that we can accurately interpret and access very richly specified representations in a fraction of a second means that the organization in our mental lexicon is much too efficient to be the alphabetical type found in conventional paper dictionaries. In fact, alongside the mental lexicon, our pocket Webster resembles the class dunce because it cannot even come close to codifying all of the meaning relations between words, not to mention the mental storage system that permits us to access lexemes so quickly. Such efficiency argues for a streamlined system of organization that is heavily reliant on shared and related properties of meaning. Neurological evidence also supports a lexicon organized according to semantic categories. Studies carried out by A. Damasio— H. Damasio (1992) and H. Damasio et al. (1996) illustrate how braindamaged patients show category deficits depending on the area of the brain affected. In these studies, patients were asked to name pictures of

206

The cognitive axis

items belonging to three different categories: faces of well known people, animals and tools. Patients who had difficulty naming people had damage restricted to the left temporal lobe; those who could not name animals had left inferior temporal lobe lesions, and those with posterior inferior temporal lobe and temporo-occipito-parietal damage could not name tools. In a subsequent study of normal people, these brain areas were precisely the ones activated when subjects were asked to provide names for the same categories. In addition, previous studies had shown the use of different neural systems for the retrieval of words denoting action versus those denoting objects. All of this data evidently argues for lexical knowledge organized by category in distinct areas of the left temporal lobe (Caramazza 1996: 485). Instead of interpreting category deficits in terms of damage to shared semantic properties or damage to pathways leading from semantic properties to lexical representations, H. Damasio et al. (1996), posit that such deficits reflect the categorial organization of the mediating lexical representations. This means that there are very possibly three neural systems involved in the representation of lexical knowledge: (1) conceptual content (meaning); (2) phonological elements (sounds); and (3) modality-independent lexical knowledge which mediates between the other two systems. Such modalityindependent lexical representations also serve as the focal points about the grammatical class of words and their selection restrictions, and constitute part of the blueprint for a lexical entry or a set of lexical entries. Our proposal of predicate schemas is a step towards the formalization of the lexical knowledge postulated by these studies. Caramazza (1996: 486) effectively points out the need for further studies in this direction, which would tell us to what extent abstract entities are represented categorially, as well as how syntactic information is organized in the brain in relation to other aspects of world knowledge. In much the same line, Fromkin (1997: 23) also writes that more interdisciplinary studies are needed in order to understand how to relate the neuron to the noun phrase.

Language and memory

207

5.3. Language and memory The storage place for our knowledge of words, along with all our other knowledge, is our memory. Memory, however, is not a warehouse where items are randomly kept in static readiness to be used or referred to in some way. It is better described as a many-splendored thing of complex virtuosity. Memory refers to both the system of information processing (how memories become memories) as well as to the product of the system (memory itself). Memories are coded and stored by alternations in the patterns and excitability of the many synaptic connectors among the neurons in the brain. Many multistore models of information processing have been proposed, but probably the best known is that of Atkinson—Shiffrin (1968). In this model, information processing is done in three stages. First, sensory information enters a sensory register where it undergoes primary sensory processing and is given a perceptual representation. This sensory register retains information for a second or so, and acts as a filter, which screens incoming stimuli in order to decide what information will be discarded, and what information will be passed on to short-term memory (STM). Categorization mechanisms try to identify perceptual representations in the different sensory stores. As incoming stimuli become categorized, this new information enters the STM. Our consciousness of the information contained in the STM does not last very long, a few minutes at most if we keep rehearsing it. After processing, the information is either passed on to long-term memory (LTM) for permanent storage or simply forgotten. The fact that the STM has limited storage space and its stability is fragile leads Rupp (1998: 52) to compare it to an airlock in the brain, a holding zone where newly acquired information sits around, waiting to be sent to its final destination. However, the stability of this airlock is fragile because it can easily be punctured, and memory disrupted by any type of unexpected event such as a loud noise, bump on the head, or even a pleasant distraction. The information in the STM does not sit around in a deep freeze, but undergoes manipulation. Baddeley—Hitch (1974) extend the STM to include the concept of working memory, a kind of mental blackboard, which accounts for the operations that take place during

208

The cognitive

axis

information processing. Working memory consists of a set of mechanisms that all together perform strategic processing. These mechanisms are the behind-the-scenes machinery that helps us to make rapid associations between the input to the STM and information already stored in the LTM. In other words, working memory corresponds to the information activated in the LTM, new input in the STM, and the decision process that leads us to activate and retrieve certain bits of information instead of others. This type of memory permits us to consider and manipulate data to solve problems or to generate totally new ideas. In contrast to the STM, the LTM is our permanent store of memory and has unlimited storage capacity. Conscious retrieval of information from the LTM seems to depend strongly on two factors: reconstruction of the context and the degree of integration of knowledge in other knowledge (Vossen 1995: 96). For example, numerous experiments have shown that taxonomic organization greatly facilitates memory. Part of the reason for this is the fact that the more knowledge we have about an area of meaning, the easier it is for us to establish relations between concepts. The essentially hierarchical structure of the lexicon proposed is justified by the fact that conceptual knowledge is also structured in the same way because this type of economical storage facilitates the prediction of properties. Hierarchically structured conceptual categories reduce the infinite differences between stimuli to cognitively usable proportions. Of course, other types of knowledge are also used to organize information. Besides hierarchical organization, Barsalou (1992: 126— 127) mentions frames as another kind of organizational structure. In this sense, frames are representative of spatial, temporal, causal, and intentional relations between entities and events in familiar situations. The cognitive system stores newly encoded information in our LTM. This information can be divided into subtypes. Figure 6 shows the functional divisions initially proposed by Tulving (1972):

Language and memory

209

MEMORY

DECLARATIVE MEMORY

/X Episodic Memory

\ Personal Reminiscences of Life Events Figure 6.

PROCEDURAL MEMORY

i

Semantic Memory

How-tos

i Facts and Figures

Functional divisions of memory (in Rupp 1998: 26)

Memory can be divided into declarative memory (knowledge of what) and procedural memory (knowledge of how). It is the difference between knowing that bicycles are two-wheeled vehicles, and knowing how to ride one. Declarative knowledge is divided into episodic memory and semantic memory. Episodic memory is our knowledge of what has happened to us in the past, whereas semantic memory refers to conceptual knowledge, This includes knowledge related to language as well as other activities such as categorization, imagery or problem solving. Before being transformed into semantic (or conceptual) memory, newly acquired knowledge is often initially episodic as it is tied to the learning experience itself. For example, the knowledge that Madrid is the capital of Spain (semantic memory) gradually becomes detached from the fourth-grade classroom and Wednesday morning when this piece of information was initially learned (episodic memory).

210

5.4.

The cognitive

axis

Lexical structure and conceptual structure

Conceptualization sits at the core of the blackbox problem of the human mind (Nuyts 1992). The difficulty in studying conceptualization lies in the fact that it never reveals itself directly, but only in disguise, either coded in or filtered through different types of human behavioral systems such as language (Pederson—Nuyts 1997: 3). Both linguists and psychologists have long been aware of the significance of lexical structure as a means of ascertaining and exploring the organization of concepts in the mind. This awareness has logically coincided with a growing interest in issues such as knowledge engineering, artificial intelligence, and psycholinguistic research in categorization, areas in which the study of conceptual structure is vitally important (Connolly 1990). Nevertheless, the extent to which semantic representation and conceptual representation coincide is an open question. For example, linguists such as Langacker (1987), Jackendoff (1983), and Dik (1990) equate semantic structure with conceptual structure, whereas others such as S. Levin (1997) argue against it. Whatever the exact relation between language and conceptual structure, whether it is the whole story or only part of it, the fact remains that in an adequate psychological account of human mental life, linguistic competence must play a central role (Miller 1990: 321). At all levels, language structure is proof that we are not passive recipients of information, but active experiencers and interpreters of a world upon which we impose a certain organization. Human beings are uniquely endowed with the capacity to build flexible and creative linguistic and conceptual systems. Of the small set of basic conceptual structures which relate senses by connecting them to people's shared knowledge of the world, the most important is categorization. In fact one of our first tasks in life is to form categories that capture the different traits and properties that entities share, and to learn words that describe these categories. In a parallel way, categorization is the major psychological principle governing lexical structure. The two conflicting models of conceptual categories are the classical (or the criterial-attribute) model and the prototype model. The classical view of word meaning argues for a description of word meaning as a set of criterial properties.

Lexical structure and conceptual structure

211

According to this theory, categories have clear boundaries as membership is limited to those entities possessing the conjunction of necessary and sufficient features particular to the category in question. Within the category itself, all members have equal status since all of them possess the same set of properties required for category membership. This model is in consonance with the semantic categories posited by Katz (1972) in which category membership is specified in terms of a set of binary properties. In contrast, the prototype model of categorization, based on Rosch (1978), is more cognitively realistic, but has the disadvantage of being difficult to pin down. Categories are represented in terms of a single, centralized category representation. Depending on the theorist, this representation may take the form of a specific exemplar, judged to be the best member of the category, or it may take the form of a set of abstract properties representative of the category's exemplars. In this model, entities are categorized as one type of thing or another in terms of their similarity to the prototype. Such prototypes serve as cognitive reference points, and membership within the category is consequently graded. (Rosch 1978; Taylor 1989: Chapters 2-4). Barsalou (1992: 29) points out that one of the problematic aspects of this model is that it fails to specify constraints on the abstraction process, and that an infinite number of properties can be abstracted for any given category. He states that it is necessary to place principled constraints on abstraction so that only those properties relevant to goals of the cognitive system are abstracted. In the FLM, categorization is approached by means of an abstract prototype or schema because such constructs are reflected in lexical organization, which in turn is the symbolic representation of cognitive processes. Broadly speaking, a schema is an underlying organizational pattern of cognitive perception, which encodes both mental and physical experience. Schemas were first discussed by Bartlett (1932), who posited that our interactions with the world are mediated by largescale memory schemas, each of which encapsulates our knowledge about everything connected with an object or event. Schemas have also been used to refer to the plans we have for coping with recurring sequences of basic actions. These plans or schemas are involved in the storing of information concerning stereotyped situations, global

212

The cognitive

axis

patterns of knowledge or generalized events (Schank—Abelson 1977; Schank 1973, 1982). Our notion of schema is evidently somewhat more restricted in that it is reduced to the common properties of a set of predicates. Nevertheless, all schemas, both great and small, are formed as the result of a process of decontextualization by which linguistic elements achieve varying degrees of emancipation and context independence. Langacker (1997: 236) underlines the fact that consistently recurrent features encountered on multiple occasions become progressively "entrenched", whereas others that do not recur simply "cancel out" and fail to achieve conventional symbolic association with the form. He defines schema in the following way: ... an abstract characterization that is fully compatible with all the members of the category it defines (so membership is not a matter of degree); it is an integrated structure that embodies the commonality of its members, which are conceptions of greater specificity and detail that elaborate the schema in contrasting ways (Langacker 1987: 371). If linguistic categories are the symbolic representations of conceptual categories, and conceptual categories have schemas, then it stands to reason that linguistic categories should also have such schemas reflected in their structure.

5.5.

Predicate schemas in the FLM lexicon

In the FLM lexicon, the information codified in both axes of the lexicon combines to form a representation called a predicate schema. This type of lexical representation is in consonance with the type of modality-independent lexical knowledge that A. Damasio—H. Damasio (1992) postulate in their intermediate representation, as well as the categorization by prototype described by Barsalou (1992: 2829). Such representations would presumably form part of our semantic memory in the LTM and considerably facilitate lexical retrieval. This type of construct helps to overcome the inadequacies of sense

Predicate schemas in the FLM lexicon

213

enumeration lexicons. Pustejovksy (1995: 46) points out that word sense enumeration cannot characterize all the possible meanings of the lexical item in the lexicon, and that somehow lexical semantics must be able to account for the creative use of words in different contexts without allowing for complete, unrestricted interpretations. A lexicon component organized in a hierarchical network of predicate schemas necessarily entails certain modifications in Dik's Functional Grammar because FLM lexical domains are conceived as lexical microgrammars in which pragmatic, semantic and syntactic regularities converge. In this sense, the FG lexicon structured according to FLM principles can be regarded as a core lexical grammar formed by rules and procedures the realization of which is dependent on the lexical domain. This concentration of information in the lexicon evidently has a strong impact on the FG linking algorithm. If a set of syntactic, semantic and pragmatic factors can be abstracted from a group of predicates with similar meaning, information previously classified as unpredictable is no longer so when viewed within the wider context of a lexical domain and subdomain. This means that expression rules can be greatly simplified and reduced in number. The set of domainspecific expression rules thus become lexical rules. Predicate schemas are cognitive constructs obtained by the convergence of both paradigmatic and syntagmatic features. Based on Langacker (1987: 371), a predicate schema can be defined as a modular, dynamic characterization that subsumes linguistic symbolic units obtained through the activation of lower-level schemas. These schemas are linguistically motivated and reflect our perceptions of reality. Predicate schemas are modular in that they are organized hierarchically within lexical domains. This allows for the elaboration of grammars of individual languages within a more general or specific area of meaning. Underlying predications are constructed on the basis of predicate schemas at different levels within the hierarchy of lexical subdomains. Predicate schemas are linguistic in that the units which embody them are obtained through lexical structure. As they are linguistic items, they do not belong to any type of metalanguage (Dik 1986: 3-4). This contrasts with other approaches such as Jackendoffs Conceptual

214

The cognitive axis

Semantics in which semantic functions (GO, CAUSE, PATH, etc.) are assigned the role of conceptual primitives. Other examples are the lexical templates of Rappaport—Levin or the logical structures in Role and Reference Grammar, where the use of a metalanguage in meaning definitions has become a methodological tenet. This is not to say that we are not ultimately in favor of using such a language, but there are still not enough crosslinguistic studies to be able to formulate one with any assurance of psychological validity. Predicate schemas are also dynamic in that they are not frozen structures, but often undergo mutations, which establish new connections with other schemas, a process which is the basis of metaphor and metonymy. Table 29 represents the dynamic nature of some lexemes belonging to COGNITION: Table 29.

Domain of cognition: interrelations with other domains

COGNITION

RELATED DOMAIN

foresee to know something will happen. view to believe that somebody/ something is a certain way (by looking at them in one's mind). enlighten to cause somebody to understand (see) something better (as if by shining light on it). plan to think about something (a method/way of doing something) carefully and deliberately in order to carry it out in the future. perplex to confuse somebody by making them feel slightly worried because they do not understand. persuade to cause somebody to believe that something is true by saying something acting in a certain way. swindle to deceive somebody in order to obtain something valuable from them.

EXISTENCE [happen = to begin to exist in time] VISUAL PERCEPTION [look at =

to see by intentionally directing one's eyes] LIGHT [mental light = knowledge] ACTION [do]

FEELING [feel\

SPEECH [say] DO [act)

POSSESSION [obtain = to come to ave something]

Predicate schemas in the FLM lexicon

215

Table 29 shows examples of the some of the relations that verbs of cognition have with other lexical domains. Although the genus of foresee, view, enlighten, plan, perplex, persuade, and swindle places these verbs squarely in the domain of COGNITION, their respective differentiae contain verbs which signal connections to other domains (Faber—Mairal Uson 1994). For example, swindle is a troponym or a manner-of-deceiving (to cause somebody to think something that is not true). Despite the fact that it is a verb of cognition, its location in a causative subdomain makes it more concrete or action-like. In its differentiae, we see that the reason for swindling somebody else is to come to have something. This signals the fact that in this case the domain of COGNITION is directly related to POSSESSION. Much the same is also true of perplex, persuade, plan, foresee, view and enlighten, whose adverbial modification encodes links to other domains. Evidence of the central position that the domain of COGNITION occupies in the lexicon is precisely the multiplicity of its interconnections with other domains. Another example of interdomain connections can be found in the lexical domain of FEELING, which also shows a wide range of connections with other areas of meaning such as MOVEMENT, CHANGE, PERCEPTION, and COGNITION. Movement-based extensions are represented by the subdomain schemas, to cause somebody to feel, as well as more specific ones like to cause attraction/aversion in somebody. Within to cause somebody to feel, the verb move means to cause somebody to experience a fairly strong emotion. The movement involved is from a normal state of feeling to a different one, which necessarily constitutes a deviation from the norm. The fact that the type of emotion involved is invariably one of sadness or sympathy for someone implies that our default emotional state is not neutral, but rather positive. (255)

a. b.

His words moved me (to tears). *His words moved me (to laughter).

While (255a) is perfectly acceptable, (255b) is not. This same phenomenon can be observed in troponyms of MOVEMENT such as stir and agitate, which in one of their senses also belong to FEELING. These

216

The cognitive axis

verbs refer to both quick, sudden movement and strong, often violent feelings. (256)

stirl

to cause something (a liquid) to move by mixing it with a spoon, agitaiel [MOVEMENT] to stir a liquid, shaking it violently, stir2 [FEELING] to cause a strongfeeling in somebody. agitate2 [FEELING] to cause somebody to feel so worried that they cannot think clearly or act calmly. [MOVEMENT]

In the domain of MOVEMENT, agitate is a troponym of stir, while in FEELING, the connection between the two verbs is somewhat less direct. Like move, stir is a more general verb and the strong emotion it causes can be either excitement or anxiety. In contrast, agitate is a troponym of worry, and falls into the extensively lexicalized anxiety group. The definitions in (256) also show how feelings are conceptualized as liquids. The strength of the feeling is in function with the corresponding violence of the movement. Every domain related to MOVEMENT is also related to CHANGE. In FEELING this is evident in the fact that many verbs of the domain are de-adjectival since they lexicalize a change in which the experiencer goes from one emotional state to another (e.g. sadden, gladden, sicken, madden, etc.). It is also significant that one of the generic terms in FEELING is a generic term in TACTILE PERCEPTION as well. This is a clear instance of the close relationship between the physical and the non-physical dimensions of experience, as it is used to refer to the inner perception of a non-physical stimulus. The reason that our emotions are linked to the sense of touch instead of one of the other senses lies in the fact that when we touch something (or something touches us), we do not necessarily understand what it is, but we do sense whether it gives us pleasure or not. This is in direct contrast to see, which belongs to both VISUAL PERCEPTION and COGNITION, and thus implies understanding. A link between VISUAL PERCEPTION and FEELING occurs in the subdomain to feel attraction or interest. For example, when we love someone, they also become physically attractive to us. Thus, to love someone is also to love to look at them. However, love does not necessarily imply knowledge, just as looking does not necessarily

Predicate schemas in the FLM lexicon

217

imply seeing or understanding. As a result, it is not a contradiction that most of these verbs can be modified by blindly. (257)

Despite the way he treated her, she loved/adored/ worshipped him blindly.

These are just a few of the examples that point to the fact that the lexicon is an intricate web of meaning relations, which link different areas of meaning in a principled way, and which have a direct effect on all levels of linguistic representation.

5.5.1. Towards a typology of predicate schemas in FG Predicate schemas can be found at different levels of the lexicon, more specifically at the levels of lexeme, domain and subdomain. At each of these levels, we have postulated a corresponding type of schema (Faber—Mairal Uson 1998a). The most basic is that found in each lexeme in which syntactic, semantic and pragmatic units combine to form a pattern representing our knowledge of the lexical item in question and the area of meaning to which it belongs. The second type is a subdomain-level schema made up of those prototypical syntactic, semantic and pragmatic units obtained through the factorization of the lower-level (or lexeme-level) schemas. Finally, the most general type is a domain-level schema, which is obtained in a similar way through factorization from lowerlevel schemas. A lexical domain has the following macrostructural pattern.

218

The cognitive axis

Table 30.

Domain schema

Subdomain-level schema 1

Subdomain-level schema 2

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Subdomain-level schema 3

Subdomain-level schema 4

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Lexeme Schema

Accordingly, the selection of one lexeme over another is a set of paradigmatically oriented instructions, which will eventually lead us to the actual speaker's lexical choice. When this type of hierarchical structure is applied to a specific lexical domain, in this case, EXISTENCE, the instantiation would be the following:

Predicate schemas in the FLM lexicon

220

The cognitive

axis

In Figure 7, the paradigmatic choices open to the speaker at various levels are not random. The principal distinctions in this domain are firstly phasal, and secondly, temporal and perceptual. Needless to say, there are more lexemes in each box than the one or two that have been offered as examples. The complete representation would also take into account the differentiation parameters between individual lexemes, and that would mean single paths that lead to each lexical item. On the basis of our lexical domain structure, it is also possible to elaborate a generation dictionary similar to the kind used in certain artificial intelligence approaches to meaning extraction. This type of dictionary is useful in that it makes explicit a speaker's syntagmatic choices and how they influence lexical selection. Such dictionaries take the form of discrimination nets with choice points marked by particular units of semantic knowledge, usually in the form of semantic roles. Nirenburg (1987a: 6) gives an example of such a discrimination net, but it is extremely basic, and evidently does not take into account any type of paradigmatic context. However, if we enrich it with the information specific of lexical domains and subdomains, it becomes considerably more informative. Figure 8 is a discrimination net based on our configuration of the lexical domain of verbs of CONTACT:

Form and representation Start

Figure 8.

)

Discrimination net for verbs of contact

221

222

The cognitive

axis

In Figure 8, the first question asked is if the Actor (the entity that carries out the action) is human. If the answer is yes, then the next decision point concerns the intentionality of the action. The fact that the first argument is human does not necessarily indicate that it has the semantic function of Agent: (258)

a. b.

The soccer player hit the wall (= hit the wall with his hand/other object), The soccer player hit the wall (= was shoved/pushed against the wall by somebody/something else).

In the case of (258b), the path then goes directly to the superordinate term. If the decision is (258a), then the next decision point concerns the nature of the Patient. If the Patient is inanimate, then the troponyms are discriminated in terms of instrument {hammer) or sound, which may be iterative (tap) or not (whack). Animate Patients are divided into two categories, human and animal. For example, an action affecting a human Patient can have two possible routes; the first focuses on the affected entity and the second, on the action itself. When the focus is on the affected body part, then it is a question of whether this body part is the head (crown, brain) or the ear (box). In English, there are no other lexicalizations for specific body parts affected by this type of action. The second route focuses on the action and the person who carries out the action. Here there are many possible discrimination points, which ask about possible lexicalizations concerning the instrument. The instrument can be non-specified, but discriminated in terms of sound (thump), force (knock), or transmitted emotion (pat). If the instrument is specified, it can either be an object or a body part. Specified objects can be stick-like (birch) or whip-like (lash). If the instrument is a body part, it can either be a hand (open or closed), foot, head, elbow, or knee. All of the possible paths have default options that lead back to the generic terms, and which transmit the basic conceptual content of the domain. The questions asked and the lexical choices available are based on the paradigmatic structure of verbs of CONTACT in the FLM lexicon and the predicate schema characteristic of each lexical subdomain. The progressively specific distinctions are based on the result of the

Form and representation 223 analysis of the valency of these verbs and the semantic characteristics of their arguments.

5.5.1.1. Lexeme-level predicate schemas A lexeme-level predicate schema is the meaningful configuration of definitional components in a lexical entry. This type of schema encodes the parameters of the speaker and hearer's communicative contract, the nature of which touches upon the three major components of language, i.e. semantics, syntax, and pragmatics. The source for this information can be found in the meaning definition of each lexeme, which is in one sense a translation of our perceptions of reality. This coincides with Miller—Johnson-Laird's (1976) affirmation that perception is a structured activity, the result of which enables us to extract distinctive characteristics from an entity (e.g. form, color, dimension, orientation,, etc.). Such perceptual information is necessarily encoded in lexicographic definitions, which can be regarded as a nutshell description of the concept in question. In this sense, each dictionary definition is a micro-knowledge representation of how we categorize objects and events on the basis of sensory data. The various kinds of parameters of semantic differentiation show the divisions and distinctions that each language makes in the semantic continuum, and indicate how different senses of a predicate are related (Faber 1997). For example, dictionaries enumerate the following senses of bawl. (259)

a.

bawl

b.

bawl

c.

bawl

d.

bawl

to shout in a loud, rough voice; to cry noisily [Longman Dictionary of English Language and Culture], to shout or cry loudly [Oxford Advanced Learner's Dictionary], If you bawl, you shout or sing something very loudly and rather harshly. If a child is bawling, it is crying loudly [Collins Cobuild], to shout or sing in a very loud voice; to cry loudly [Cambridge International Dictionary of English],

224

The cognitive axis

e.

bawl

to cry or sob loudly; wail; to cry out loudly and vehemently; shout [American Heritage].

An analysis of this information shows that bawl is in reality not one lexeme, but three, and its sense components can be factored into different types: Table 31.

Sense components of bawl bawll

bawl2

bawl3

Genus

shout

cry

sing

Manner 1

loudly

very loudly

loudly

Instrument

voice

voice

voice

Manner2

harshly

unpleasantly

badly

of children

negatively evaluated by the perceiver]

Pragmatic [speaker authority information and/or strong emotion]

The components in Table 31, which have been extracted from dictionaries, provide the basis upon which we can construct definitions for each sense, which are a consensus of the information. (260)

bawll bawl2 bawl3

to shout loudly and harshly. to cry very loudly in an unpleasant way (of children). to sing badly in a very loud voice.

The definitions in (260) show that bawl is not one type of sound, but three. This information places bawl in three different subdomains within SOUND, depending on the quality of the sound, its intensity, and the emotion that it expresses.

Form and representation

225

(261)

To make a loud sound by speaking shout to speak loudly yell to shout loudly (because of excitement, anger, pain), whoop to shout loudly (because of excitement, happiness), cry to shout (because of excitement, surprise, fear). scream to cry out loudly in a high-pitched tone (because of great excitement, anger, pain, fear). shriek to scream very sharply and loudly. screech to shriek disagreeably and continuously. + bawll to shout loudly and harshly. roar to shout loudly and continuously in a very deep voice, bellow to shout in a loud deep voice.

(262)

To make a sound expressing unhappiness whine to make a long, high-pitched sound (because of unhappiness). whimper to make short, high-pitched sounds (because of unhappiness). moan to make a long, low-pitched sound (because of pain, grief suffering). groan to make a long, low-pitched, rather loud moan (because of pain, worry, disapproval), cry I to make a loud sound (because of unhappiness, fear, pain). wail to make long, loud, high-pitched cries (because of sorrow, pain). howl to make a long, loud cry (because of pain, fear, unhappiness). yowl to howl loudly. cry2 to make a sound while producing tears, which expresses strong emotions. weep to cry very softly, sob to cry, breathing in short breaths, blubber to cry very loudly in an unpleasant, childish way. Φ bawl2 to cry very loudly in an unpleasant way.

226 (263)

The cognitive axis To produce musical sounds sing to produce musical sounds with/without words. hum to sing with closed lips, without pronouncing words. croon to sing in sweet, low voice. warble to sing pleasantly in a high-pitched voice with trills (like a bird), yodel to sing, changing quickly and continuously from a normal voice to a very high one (as in Switzerland), chant to sing a religious prayer/song, carol to sing happily (esp. Christmas songs), troll to sing loudly to celebrate something [oldfashioned], Φ bawl3 to sing badly in a very loud voice.

The hierarchies show that bawl belongs to three different areas of meaning. It can be a manner of shouting, crying or singing. All of these subdomains fall in a transition zone between SPEECH and SOUND, and thus share certain characteristics such as intensity of sound and axiological evaluation. These hierarchies of meaning constitute the basic structural pattern of each area of semantic space. (264) (265)

bawl to sing badly in a very loud voice. warble to sing pleasantly in a high-pitched voice with trills (like a bird).

Examples (264) and (265) show that both bawl (in one of its senses) and warble are ways of singing. The fact that they share the same genus (sing) means that they are located in a subdomain of SOUND. In both cases, their definitions can be factored in the following way:

Form and representation Table 32.

227

Definitional constituents of bawl and warble

Definitional constituents Genus (indicating domain membership) Manner 1 Instrument Manner2 Pragmatic information

bawl

warble

SING (to make a musical sound)

SING (to make a musical sound)

badly

pleasantly

loud voice

high-pitched —

[negatively evaluated by perceiver]

voice

with trills like a bird

Other troponyms of sing which fall in the same area of semantic space are hum, croon, yodel, chant, carol, and bawl. Most of the verbs of this domain are predicates which subcategorize one argument that is prototypically a human source of sound emission, and optionally a second argument in the form of a cognate object inherent in the meaning of the verb {song, hymn, anthem, etc.). (266) The soprano sang a(n) anthem/aria/hymn/*speech/ *bachelor/ *penguin/*cup. The second argument for all of these verbs has very specific selection restrictions imposed by the meaning of each verb as well as the characteristics of the first argument. Although sing admits all types of song, its more specific hyponyms do not: (267) The monks chanted a hymn/?anthem/??aria. (268) The alpine guide yodeled a Swiss song/*hymn/*anthem/ *aria. In this particular subdomain, the source of sound is human because the sound is harmonious and thus controlled. The different types of information within the meaning definitions and the basic similarities

228

The cognitive

axis

that can be derived from them are the components of a lexeme-level predicate schema.

5.5.1.2. Subdomain-level predicate schema Subdomain-level schemas possess both a micro- and macrostructure. Microstructurally these schemas act as a repository of lexical regularities (shared synsem parameters). Within the lexicon, subdomain-level predicate schemas fall halfway between domainlevel and lexeme-level predicate schemas, acting as a link between the two. As a result, the operational scope of these schemas is much broader than any other unit of analysis in the lexicon, extending upwards towards the more general level of domain, as well as downwards towards the more basic level of lexeme. In this sense, subdomain predicate schemas are conceptualized as dynamic patterns that establish links and connections with other domains. These connections often represent metaphoric and metonymic extensions, and are formalized in what we have called a semantic macronet. In this macronet, subdomain predicate schemas are the means by which a semantic value is mapped from a source domain onto a target domain, and thus are the basic units of the cognitive axis of the lexicon. The fact that a lexical subdomain is the source of lexical productivity signifies that certain phenomena are, to a large extent, motivated, and provide the major structuring parameters that define a domain. According to LakofF (1987), a given construction in grammar is motivated to the extent that its structure and properties are inherited from other items in language. Therefore, if a lexical item X is defined in terms of (or based on) a lexical item Y, then X inherits all the properties of Y which are not incompatible with its own defining parameters. In this regard, Lakoff claims that the more redundant the property of a given lexical item, the greater its motivation, and thus the better it fits into the grammar. Most of our exemplification is confined to subdomain predicate schemas since they are the basic units in the construction of a lexicallybound domain grammar. A case in point is the subdomain to think carefully about something [COGNITION], whose microstructure and

Form and representation 229 macrostructure show the systematic organization of semantic and syntactic information. (269)

To think carefully about something consider to think carefully about something in order to make a future decision, meditate to consider something seriously as a possible course of action, contemplate to consider doing something in the future. ponder to consider something, weighing it in one's mind. debate to consider the arguments for and against doing something in the future.

The semantic parameter of time (long duration) is activated throughout the subdomain (269). All of the predicates designate an atelic state of affairs, as shown in the following examples: (270)

a. b. c.

He meditated/considered/contemplated/pondered/ debated the question for an hour. *He meditated/considered/contemplated/pondered/ debated the question in an hour. *It took him three hours to meditate/consider/ contemplate/ponder/debate the question.

These verbs can also occur with adverbial modification that intensifies the duration of the activity already encoded, but not with adverbs such as carelessly, which violate the basic semantic parameter present in the definitions of all of the members of the semantic set: (271)

a. b.

He carefully meditated/considered/contemplated/ pondered/debated the issue for an hour. *He carelessly meditated/considered/contemplated/ pondered/debated the issue for an hour.

All of these verbs can take the types of complementation shown in Table 33:

230

The cognitive axis

Table 33.

Typology of syntactic frames: to think carefully something

consider meditate contemplate ponder debate

(272)

a. b. c. d.

NP [Agent] + + + + +

W7i-clause [Goal] +

•

Gerund [Goal] + + +

about

PP [Goall + —

—

+

+

-

John considered/contemplated/meditated/pondered/ debated the issue for a long time. [NP] Ruth considered/pondered what she had heard at school. [Wh-clause] Leo considered/contemplated/meditated/debated staying at home. [Gerund] Frank meditated/pondered on the meaning of life. [PP]

Examples (272a-d) are similar in that they all refer to unbounded events. In fact, their syntactic forms are in many respects iconic with the semantic value encoded in their meaning definition, especially if we consider that a gerund is the canonical form for the representation of unboundedness. Nevertheless, even though these verbs belong to the domain of COGNITION, they are not prototypical verbs of this domain because their behavior is more similar to that of verbs of VISUAL PERCEPTION. Not surprisingly, consider, contemplate, meditate and ponder are predicates with double domain membership. All of them have at least one sense that designates a type of visual perception when their goal argument is a first-order entity: (273)

He pondered/considered/contemplated the museum.

the Greek statue in

In contrast, in their more cognitive senses, the second argument of these verbs is always a second-order entity (in any of its various morphosyntactic realizations). When issues or questions (270-271)

Form and representation

231

appear in nominal form, they become closer to objects (which the thinker looks at in his/her mind). For this reason, we have called them 2nd-order entities though perhaps they are fair candidates for a gray area somewhere in-between. The semantic value of duration presupposes the existence of an event, which must last over a period of time. Table 34.

Typology of arguments: to think carefully about something first-order entity

NP ^-clause Gerund PP (about, over, etc)

-

-

second-order entity + + + +

third-order entity -

-

-

The information in Table 34 illustrates the fact that these verbs are not strictly typical of COGNITION because among other things, they do not admit third-order entities: (274)

*He pondered/contemplated/meditated

that it was late.

Consider is the only one of these verbs that can take a proposition as its complement. However, when it enters into that type of construction, its meaning changes and it is projected into a different subdomain. Evidence of the double nature of this set of cognition verbs can be seen in the fact that there are clear semantic and syntactic traces of their connection with VISUAL PERCEPTION. For example, contemplate is basically a visual perception verb, whose meaning has been mapped onto COGNITION, where it signifies to look at something in one's mind. Table 35 shows the two different meanings of contemplate and their corresponding argument structure:

232

The cognitive axis

Table 35.

Contemplate: argument structure and meaning correlations

contemplate

[VISUAL PERCEPTION]

to look at somebody/something quietly and steadily for a long time, while thinking about them in an often favorable way.

contemplate [COGNITION]

to think about something carefully and deeply for a long time (a future state of affairs).

Argument structure Exp = prototypically human Goal = concrete (usu. somebody/something attractive)

Agent = prototypically human Goal = future course of action

Meaning correlation (i) (ii) (iii) (iv)

look somebody/something [concrete] steadily quietly

(i) (ii) (iii) (iv)

think future course of action [abstract] carefully deeply

In Table 35 we can see that contemplate contains the meaning components of looking and thinking. When it refers to visual perception, the vision component is foregrounded in the genus of the definition, while the thought component appears in the adverbial modification. In contrast, when perception is mental rather than physical, the experiencer argument becomes an agent argument. Although there is obviously no direct mention of visual perception, the vision component is implicit in the activity of thinking since the basic schema of visual perception is inherent in cognition. The only difference is that instead of our eyes looking outwards towards the world and perceiving concrete entities, they are looking inwards towards our mind and perceiving abstract ones. The iconic distribution of the set of semantic and syntactic parameters within a lexical subdomain is the abstract characterization of all the synsem parameters common to its members. Each of the

Form and representation

233

members in turn provides different ways of specifying the lexical content codified in a lexical domain.

5.5.1.3. Domain-level predicate schemas A domain is a complex network of subdomains, which contains the information typical of an area of meaning. The question is to what extent this information within such an extensive area can be systematized. In this regard, we are skeptical about the possibility of discovering any set of absolute rules valid for entire lexical domains. What we have found, however, is compelling evidence of certain recurrent organizational patterns and regularities, which justifies the kind of lexical productivity we postulate. What cannot be overstressed is that such productivity is semantically motivated. Domain-level predicate schemas are the macro-organizational patterns that encapsulate the inventory of linguistic units relevant to the lexemes in a domain. However, not all the domains in the lexicon have the same prototypical weight. Prototypicality in this case is measured according to the notion of verb schema (Givon 1984; Langacker 1987). In this connection we believe that the linguistic features in a given lexical domain are not arbitrarily coded, but are a reflection of the prototypical status of the domain along the time-stability scale displayed in a verb schema. In such a scale, the category verb prototypically denotes rapid changes, codified in terms of visible, kinetic events or actions. Members of this category are graded as to the number of prototypical features they display. For example, a domain-level schema for cognition verbs (e.g. know, believe, hypothesize, conjecture, understand etc.) which encodes long-lasting events exhibits fewer prototypical properties for verbs than the domain-level schemas for ACTION (e.g. hit, break) and MOVEMENT (e.g. go, come, take, bring) which primarily encode instantaneous events. This justifies the presence of an SoA variable as well as a lexical domain variable in the predicate schema. As a result, domains can be graded on a concrete-abstract scale. MOVEMENT and GENERAL ACTON are conceptualized as basic activities

because they are more concrete than verbs belonging to the domains of

234

The cognitive

axis

and FEELING, which are more abstract and thus less observable. This scale illustrates connections between different domains, and traces their path throughout the lexicon. From a purely syntactic point of view, it is unrealistic to assume that the lexemes in a domain can be reduced to a finite number of constructions, but we firmly believe that the paradigmatic organization in the lexicon reveals an important correlation between syntax and semantics.

EXISTENCE, COGNITION

5.5.1.3.1.

Domain-level semantic patterns

A domain is primarily composed of a set of semantic parameters, which are in turn responsible for its lexical architecture. These parameters are not only indicative of the linguistic encoding of a domain, but also reflect the categorization of that area of meaning. Accordingly, the configuration in lexical domains mirrors basic structures of perception, which are possible candidates for near primitives with cross-cultural validity. This has led us to postulate the following principle: (275)

Lexical Principle of Conceptual Salience The conceptual salience of a given semantic domain is proportional to its degree of lexicalization.

It follows from (275) that a domain with both temporal and spatial macrostructural patterns such as MOVEMENT with over a thousand lexemes and multiple subdomains, is more salient than LIGHT, a domain with fewer verbs and with a much simpler structure. Conceptual salience can be seen as well in recurrent patterns of domain organization. Some of these parameters are specific to the particular domain or subdomain, while others such as spatiality, axiology, and time appear across a wide range of domains. In our analysis of domain structure we have found four recurrent macro-organizational patterns: (i) (ii) (iii) (iv)

Space (environmental and biological space) Time (phasal distinctions) Sociocultural context Axiological evaluation (positive/negative).

Form and representation

23 5

Space Space can refer to the space outside or inside ourselves. Our map of outside space is derived from our immediate environmental context. Inner space refers to the biological space of our bodies, which necessarily affects the way we perceive our environment. (a)

Environmental space

Our perceptions of the world (and the order that we as perceiving entities impose on it) appear as a macro-pattern in certain lexical domains. The most obvious example can be found in MOVEMENT, one of the largest and most complex semantic areas. The degree of complexity in this domain is evidence of the above because the movement of our bodies and of entities in general is one of the most prominent events in our daily experience. Miller (1972: 338-339) writes: Nothing is more common in our environment than the movement of people and things: in order to characterize that environment, a language must have a rich supply of words for indicating how an object that is at place Pi and at time Ti comes to be at place P2 at some subsequent time T2. The words that serve this purpose directly are the verbs of motion. Movement, as perceptual experience, is structured in terms of both space and time. However, the primary macrostructural pattern in this particular domain is space, more particularly the medium in which movement is perceived to take place: (276)

MOVEMENT

a. b. c. d.

movement movement movement movement

in general (go, come) in/of liquid (flow, swim) in/of air (fly, blow) on land (walk, run)

236

The cognitive

axis

Because movement in space is easier to perceive than movement in time, this domain is primarily organized in terms of the possible mediums of our physical environment in which this movement can occur. The fact that lexical structure is anthropomorphic means that general movement to a great extent coincides with movement on land, which is evidently the default pattern for the domain. Proof that MOVEMENT is a very basic area of meaning is the fact that it is the source domain for countless types of metaphoric extensions at all levels of lexical organization. (b)

Biological space

Biological space is the organizational pattern of verbs of physical perception. This domain is divided into five groups, one for each of the five senses: (277)

PERCEPTION

(i) (ii) (iii) (iv) (v)

visual perception (see, look) auditory perception (hear, listen) olfactory perception (smell, scent) tactile perception (touch, feet) taste perception (taste)

This organization (277) reflects the biological structure of our perception. The complexity of each subdomain (the number of verbs and the number of lexical distinctions made) is in accordance with its cognitive salience. In the domain of PERCEPTION, the default value is visual perception (just as in the domain of MOVEMENT, the default value is movement on land). If we compare the structure of VISUAL PERCEPTION with that of the other senses we find that it is the most complex. Our prototypical way of perceiving is with our eyes (most of our information about the world is perceived visually), and this is mirrored in the complexity of the distinctions lexicalized in VISUAL PERCEPTION.

Form and representation Table 36.

237

Lexical distinctions in perception

SENSE PERCEPTION

LEXICAL DISTINCTIONS

To become • aware of somebody/something by using one's eyes (see, look) • VISUAL PERCEPTION:

•

• • • • • AUDITORY PERCEPTION:

To become aware of somebody/something through one's ears (hear, listen) OLFACTORY PERCEPTION:

To become aware of somebody/something through one's nose (smell, scent) TACTILE PERCEPTION:

To become aware of somebody/something by using one's hands (touch, feel) TASTE PERCEPTION:

To become aware of something through one's mouth (taste)

To see somebody/something (notice, observe) To see somebody/something intentionally (distinguish, discern) To see somebody/something by intentionally directing one's eyes towards them (gaze, stare) To be seen (show) To cause somebody/something to be seen (show) To not see (be blind) To cause somebody to be unable to see (blind, dazzle) To cause somebody/something not to be seen any more (hide, conceal)

• •

To hear somebody/something (hear) To hear somebody/something intentionally (listen)

•

To cause somebody to become aware of somebody/something through their nose (smell stink)

238

The cognitive axis

The eight distinctions in VISUAL PERCEPTION contrast eloquently with the considerably lesser number or even absence of distinctions in the other four senses. We have found this to be true for other European languages as well, such as Spanish, German, and French. Taking into account that vision is our most central perceptual experience and the one that we most depend on for our knowledge about the world, it is not surprising to find that this lexical domain is larger and more complex than those of the other senses. It is also the one that has the most extensive range of extensions into more abstract domains.

Time: phasal

distinctions

The second macro-organizational pattern at domain-level is temporal. This means that actions are prototypically perceived as temporal segments that can be broken down into phasal distinctions: beginning, continuing and ending. Such distinctions are an important factor in the organization of certain lexical domains. Phasal distinctions coincide with the syntagmatic axis, and are also clearly linked with perception since our perception of movement in time or space is what classifies something as an event. We realize that such movement has occurred because we notice a change either in space, time or resulting state. For example, the fact that existence is perceived in a sequence of phasal distinctions can be observed in the structure of this domain. Table 37.

Phasal distinctions in existence

INCEPTION Ψ CONTINUATION Ψ CESSATION

To begin to exist To begin to exist in the perception of others To exist To exist as the representation of something To exist in time To stop existing To stop existing in the perception of others

Part of the concept of existence is the progression of life over a well defined period of time. In this respect, it is related to movement in that an event necessarily involves some kind of dynamism. This is in

Form and representation

239

accordance with the idea that our life from birth to death is a progression through time, a progression involving duration as well as our perception of duration. Of course, as living creatures we generally do not have an active perception that we are moving in time. If we realize that time moves, it is only because we experience the effects of this movement when our own bodies and other entities in our immediate environment undergo changes in state (static effect). These changes, which we experience in ourselves and perceive in others around us, are codified in the lexical domain of CHANGE, which along with MOVEMENT is one of the largest in the lexicon. Although we have considered CHANGE and MOVEMENT to be two different domains, they are, of course, very closely related because CHANGE (either in terms of state or position) codifies the end result of MOVEMENT. Beginning, continuing and ending are phasal distinctions which reappear in other domains as well. It can take the form of movement through time in the sense of past, present, and future. Spatial meanings are clearly fundamental to cognition and the system for encoding them provides the core structuring principles for many meanings that are not fundamentally spatial (Choi—Bowerman 1991: 84). Table 38. PAST

V PRESENT

V FUTURE

Phasal distinctions in cognition

To think about something, bringing it back into one's mind from the past [remember]. To think about something that has happened in the past [reflect].

To think about something over a period of time to understand it better [meditate]. To think about something in order to make a decision in the future [consider].

A case in point is COGNITION in which certain subdomains are structured in terms of temporal phases or movement in time. Reflect and remember take the thinker back into the past, whereas the process of meditation usually concerns something that is a part of the thinker's

240

The cognitive axis

immediate context. Consider and its troponyms project the thinker into the real of possible future action. Sociocultural context Sociocultural context is codified in the different types of pragmatic information recurrent throughout the lexicon. Biological/ social/ aesthetic norms, and variational features of different types are codified in word meaning, and often appear as the differentiation feature in words whose basic meaning otherwise coincides. The fact that such information consistently appears across different domains confers greater systematicity on lexical information. Social norms refer to the values imposed by a given society. As these are functions of meaning, there are different types, depending on the semantic value of the domain. In CONSUMPTION, gobble encodes the violation of a socially-set aesthetic norm since the semantic parameters, quickly and greedily, are negatively evaluated with respect to our conceptualization of how people should eat. Verbs, like wolf or gorge, which lexicalize the consumption of large quantities of food, are negatively charged. Variational features also appear throughout all lexical domains. These can be geographical, social or temporal. In the domain of SPEECH, we can find many examples of this type of pragmatic information: Table 39.

Variational features

Pragmatic information

Generic meaning

Troponyms

Geographical variation

talk

yammer [American English] witter [British English]

Social variation

complain

bitch [vulgar] gripe [informal] cavil [formal]

Temporal variation

promise

yorswear[old-fashioned] plight (one's troth) [archaic]

Form and representation

241

Finally, pragmatic features also activate institutionalized cultural models whereby certain acts prototypically occur as part of a given procedure in a certain place: Table 40.

Institutionalized cultural models Lexemes

Social context activated

acquit, sentence, annul, plead, summon, [Court of law] petition, sue, cross-examine, indict, testify, attest, contest beatify, canonize, consecrate

[Church]

interrogate, grill, charge

[Police station]

command, requisition

[Military]

All of the verbs in Table 40 evoke a specific cultural frame or scene in the sense of Fillmore— Atkins (1992), which is embedded in social structure. For this reason, they do not easily accept arguments that do not belong to one of these social contexts. (278)

a. b.

(279)

a. b. a. b. a.

(280) (281)

b.

The jury acquitted the accused of the crime. ??The students acquitted the teacher of giving them bad grades. The bishop consecrated the church. *The chancellor consecrated the university. The policeman interrogated the suspects. ??The mother interrogated the children. The supply officer requisitioned all of the vehicles available. *The secretary requisitioned pencils from the factory.

Although strictly speaking, examples (278b-281b) are grammatical, their strangeness lies in the fact that the entities in the argument slot do not fit the prerequisites for typical participants in the situation evoked by the verb in question. As verbs of speech, they codify an activity, as well as the sociocultural context in which this activity takes place.

242

The cognitive

axis

Axiological evaluation Lexical domains are also structured at some level in terms of positive and negative oppositions either as their primary or secondary structural pattern. The problem of the connotative/emotive meaning of lexical items is one that has yet to be satisfactorily dealt with. Although this aspect of word meaning has been relegated to the background, affective meaning is difficult to ignore since there are very few (if any) totally neutral words in our language (Krzeszowski 1990; Felices 1991; Faber 1992). Escalier—Fournier (1997: 53) underline the existence of "subjective" terms in the sense of Kerbrat-Orecchioni (1980). They claim that such terms express an emotional commitment of the speaker (e.g. emotional words such as funny or pathetic). Alternatively, subjective terms can imply the existence of a norm internal to the speaker and are related to his/her systems of ethical/aesthetic evaluation (e.g. evaluative terms such as good, genius, and hate). Nevertheless, Escalier—Fournier give no indication of how "emotional" words are to be differentiated from "evaluative" terms. The classification seems to be the result of largely intuitive criteria. In fact, in the majority of their examples (e.g. hate, pathetic) there seems to be a certain overlap between the two categories. Krzeszowski (1990) has also pointed to the dominant function that values perform in the structure of concepts, and underlines the fact that emotions determine information processing rather than merely modifying it (Osgood—Suci—Tannenbaum 1957; Osgood 1980). In fact, he affirms that all lexical items are assessable on an axiological scale. Although such a broad generalization may be difficult to prove, it is certainly true that many lexical items carry heavy axiological weight, and that this weight is semantically relevant. In consonance with Krzeszowski's assertion, we have observed that the opposition of good and bad consistently appears in the lexical semantic structure of English verbs. This opposition forms a part of the world of the average person who, as a conceptual point of reference, typically functions in an upright position, looks ahead, moves forward, performs actions, and views himself/herself as being good (LakofF— Johnson 1980: 132). Generally speaking, lexemes that carry heavier axiological weight tend to be more abstract, and refer to mental

Form and representation 243 activities and processes. According to Krzeszowski (1990: 150), words have a tendency to be axiologically loaded with positive or negative connotations in proportion to the degree of the human factor associated with them. This human factor is vitally important since we are constantly evaluating our surroundings in terms of what is more or less agreeable for us. The polarity of good and bad constitutes a subjective scale upon which we weigh and assess lexical items, and is a built-in feature of many areas of the lexicon. The relevance of this scale is evident in the fact that it is basic to the organization of domains such as CHANGE, SOUND, POSSESSION, ACTION, THOUGHT, a n d FEELING.

When domain structure is polarized, positiveness and negativity are instantiated in various ways depending on the domain in question. These instantiations can be structured in three levels. This is in accordance with Tischner (in Kreszowski 1990: 142) who postulates a three-level hierarchy of values. At the lowest level are those values that are associated with direct sensory experience, and which produce pleasure/repulsion. At the middle level are those values having to do with life and health. At the third level of the hierarchy are spiritual values such as truth, beauty, and goodness. Prototypical polarizations at the first level can be found in the domain of SOUND {to make a loud sound/to make a soft sound) in which sounds are classified as pleasant/harmonious or unpleasant/discordant. These oppositions are based on prototypical default values. Obviously, it is possible, though less frequent, to have loud, pleasant sounds as well as soft, unpleasant ones. Although the macrostructural pattern of the domain is primarily based on the categorization of the sound source (human/nature/objects), there is a secondary pattern, which is a definite polarization between good and bad.

244

The cognitive axis

Table 41.

First-level axiological polarizations in sound SOUND

POLARIZATION

1. Sound produced by living creatures LOUD VS. SOFT

1.1. To make a sound by speaking 1.1.1. To make a loud sound {shout, screech, shriek) 1.1.2. To make a soft sound (murmur, whisper) 1.1.3. To make a sound indicating an emotion 1.1.3.1. sadness (cry, whine) 1.1.3.2. happiness (laugh) 1.1.3.3. disapproval (hoot, boo)

1.2. To make a sound related to body conditions 1.2.1. To make a sound by breathing (sigh) 1.2.1.1. To make a sound by breathing quickly and audibly (gasp, puff) 1.2.2. To make a sound by expelling air 1.2.2.1. To make a sound by expelling air from one's throat (cough, burp) 1.2.2.2. To make a sound by expelling air from one's nose (sneeze, snort) 1.2.2.3. To make a sound by expelling air through one's anus (fart) 1.2.3. To make a sound by inhaling air (sniff)

SADNESS VS. HAPPINESS

NO POLARIZATION

Form and representation Table 41 (cont.)

1.3.

First-level axiological polarizations in sound

To make a sound like an animal 1.3.1. To make a sound like a wild/angry animal (growl, howl) 1.3.2. To make a sound like a domestic animal (bark, meow) 1.3.3. To make a sound like a bird (chirp, squawk) 1.3.4. To make a sound like an insect (buzz)

2.

245

Sounds produced by Nature (bubble, rustle)

WILD vs. TAME (DANGEROUS VS. SAFE)

LOUD VS. SOFT

2 .1. To make a loud sound (thunder) 2.2. To make a soft sound (rustle)

3.

Sounds produced by objects

HARMONIOUS VS. DISCORDANT

3.1. To make a musical sound (ring, chime) 3.2. To make a metallic sound (clink, clank) 3.3. To make a frictional sound (rasp, grate) 3.4. To make a punctual, explosive sound (crash, bang) 3 .5 . To make a dull, punctual sound 3.5.1. Something hitting against something else (thud, thump) 3.5.2. Something heavy hitting against a liquid (splash) 3.6. To make a durative sound (roar, boom) 3.7. To make an iterative, sibilant sound (rustle, swish)

LOUD VS. SOFT

246

The cognitive

axis

The categorization of SOUND is based on the various types of soundproducing entities ranging from animals to natural forces to inanimate objects. This is a reflection of the importance that sound perception has had in evolution. We are presently here because our ancestors were capable of accurately processing sensory data. Species which heard a lion roar and interpreted it as a coconut falling from a tree are no longer around to tell us what happened to them. Curiously enough, the structure in the domain of SOUND is a prototypical example of the schema of ME (my body inwards) against the WORLD (my body outwards). Sounds are used to interpret events, and the polarizations in Table 41 inform us of the type of event that the sound represents and exactly how it will affect us. The one exception to this can be found in the sounds that come from inside our body. This type of sound is not polarized because we consider that everything that comes from inside ourselves is necessarily good. As a result, this type of noise cannot pose a threat to our well being, and any negative meaning is a question of social and cultural norms. This is the case of burp and fart, which designate sounds that are more or less socially acceptable according to the culture. At the second level of values, we have FEELING (to feel happiness/to feel sadness). FEELING is the conceptual label we have given to the lexical domain which symbolizes how we perceive and generally talk about affective states of consciousness (as opposed to cognitive and volitional states). Feelings can express (i) an awareness of a phenomenon in the outside world; (ii) the physical/mental state of the experiencer; (iii) an idea based on intuition but not on evidence. In Table 42 polarization is the primary organizational pattern because we experience these states in terms of what is more or less pleasurable to ourselves, something that is obviously important to our life and health. Polarizations in this domain are various instantions of the basic opposition between bad and good.

Form and representation

Table 42.

Second-level axiological polarizations infeeling

FEELING

1.

247

Feeling (to become aware of something other than sight, having a sensation) 1.1. To feel something bad 1.2. To feel something good 1.3. To feel sadness 1.4. To feel happiness 1.5. To feel aversion producing: 1.5.1. disgust 1.5.2. anger 1.6. To feel attraction 1.6 .1. To feel a loss of attraction 1.7. To feel something bad in one's body 1.7.1. To cause somebody to feel less physical/mental suffering/pain 1.8. To feel fear/anxiety (fear about bad things happening in the future) 1.8.1. To feel less fear/worry/anger

POLARIZATION

BAD VS. GOOD

SADNESS VS. HAPPINESS AVERSION VS. ATTRACTION

FEAR VS. LESS FEAR

1.9. To feel surprise 1.10. To feel shame 1.11. To feel a need to do something or to have/get something 1.12. To cause somebody to feel hope/courage 1.13 . To cause somebody to lose (not to feel) hope/courage/confidence

HOPE VS. LOSS OF HOPE

Domains such as COGNITION and SPEECH are at the third level because their structure is polarized in terms of truth and falsehood. In fact, the similarity between the two domains shows their relatedness:

248

The cognitive axis

Table 43. Third-level axiological polarizations in cognition and speech COGNITION

SPEECH

To think and arrive at a conclusion (reason) To think something without knowing if it is true (guess)

To say something that you have come to know as a result of thinking (reason) To say something without knowing if it is true (guess)

To think something is true (accept)

To say something is true (accept)

To cause somebody to think something is true (persuade) To cause somebody to think something is true when it is not (deceive) To think something will happen (foresee) To cause somebody to remember something (remind)

To say something causing somebody to think that something is true (persuade) To say something knowing it is untrue (lie) To say something will happen (foretell) To say something to somebody, causing them to remember it (remind)

In Table 43 the only difference between reason, persuade, and deceive as cognition verbs and speech verbs is that in one domain there is only thought, whereas in the other, the content of the thought is voiced. This is in accordance with the conceptualization of speech as the transmission of thought. Although such a representation may not always be the case, it is part of our naive theory of the world and thus the default value of the subdomain.

5.6.

Form and representation

In introducing the notion of schema in the FG lexicon, our goal is to establish a format flexible enough to allow for the systematic inclusion of the recurrent sets of syntactic and semantic properties. Obviously, our proposal implies certain changes in the FG framework, such as the establishment of an inventory of lexical redundancy rules which will

Form and representation

249

capture such regularities. Example (282) is the representational format proposed for this construct: (282)

Format of a predicate schema A. Intra-domain analysis Lexical domain Lexical subdomain 1. Typology of paradigmatic prototypes 1.1. Meaning definitions 1.1.1. Genus 1.1.2. Semantic parameter 1 1.1.3. Semantic parameter 2 1.1.4. Semantic parameter Ν 1.1.5. Pragmatic information 1.2. Domain variable 2. Typology of syntagmatic prototypes 2.1. Predicate frame 2.1.1. Quantitative valence (xi x„). 2.1.2. Qualitative valence 2.1.2.1. Selection restrictions or lexical co-occurrence 2.1.2.2. State of Affairs variable B. Interdomain connections 1. Semantic macronet 1.1. Type of movement 1.1.1. Primary vs. secondary connections 1.2. Scope of movement. 1.3. Taxonomy of lexical metaphors

A predicate schema operates from both an inter- and an intra-domain perspective. A complete representation of the meaning of a lexical item necessarily implies the consideration of both its internal and external structure. From an intra-domain perspective, a predicate schema accounts for those recurrent semantic and syntactic parameters which play a determining role in the organization of a lexical subdomain. A schema contains semantic parameters which constrain syntactic projections, and this information triggers other types of information in

250

The cognitive axis

terms of a set of inheritance mechanisms. This linguistic information is represented in the predicate frame by means of the variable Σ. From an inter-domain perspective, a predicate schema also signals the connections established with other lexemes in the lexicon. The result is a semantic macronet which shows how domains are related, and which provides information concerning the movement involved as well as its scope. This set of systematic mappings between domains constitutes the basis for a principled explanation of lexical metaphor. Since predicate schemas are obtained via the factorization of predicate frames, their codification will be in a similar format: (283) I

[Σ: (predicate) (Σ) ] df= α

SoA + Lexical domain I (Metaphorical connections)

PREDICATE FRAME

I

df= meaning definition of subdomain predicate schema

|

A predicate schema contains a slot for the information relevant to both the type of SoA and the lexical domain. The encoding of an SoA variable is informative about predicate argument structure as well as other relevant linguistic information. Although the SoA is not encoded in every predicate, we postulate a new variable that contains the information typical of each area of meaning, and which is derived from the type of lexical structure proposed. This schema has the advantage of capturing generalizations across different types of predicates. For example, the lexical subdomain to think about something in order to make a future decision has a set of regularities that concern its arguments, selection restrictions, quantitative valence, SoA, type of entity designated, as well as metaphorical projections. (284)

Predicate schema: consider, meditate, ponder, debate, contemplate [(xi: human)Ag (ei) Go] Activity + Cognition (VISION) df= [think [V] (xi: prototyp. human)Ag (ei)Go (σι: carefully)Μωη/M^>rime]purpose]

Form and representation

251

The representation in (284) informs us that the verbs in this subdomain are transitive, and belong to the domain of COGNITION. It also specifies a two-argument valency in which there is an agent who is carrying out an atelic cognitive activity. This basic frame is instantiated differently in function to the different contextual restrictions, semantic parameters, perspective, and focus that the speaker wishes to convey. This particular transitive schema is closely related to VISUAL PERCEPTION.

5.7. A semantic macronet A lexicon which meets standards of textual adequacy should be conceptualized as a network of elements interconnected by cohesive, associative, lexical and encyclopedic functions. These are best represented by links between lexemes in a network of cross-references. The two important issues in the macrostructural design of the lexicon are the mapping of its internal structure, as well as a description of the types of link established.

5.7.1. Connections and dependencies A semantic network reflecting the internal structure of the lexicon is meaningful because the structures formed by semantic connections are representations of category relationships. On the basis of our lexical semantic analysis, we have established a hierarchy of domains in terms of primary connections and dependencies. However, since domain interrelationships are multiple and various, each domain can be said to have a set of secondary connections as well. These secondary connections are evident, for example, in the double domain membership of certain verbs in which one set of meaning components or another is highlighted, depending on its location (see Appendix 2). The names of domains in their primary locations are written in uppercase letters. The domains in lowercase letters indicate the existence of secondary relationships or where certain lexemes have double membership. To a certain extent, this lexical dependence also reflects a conceptual dependence. A specific example of the

252

The cognitive

axis

interrelations between domains can be found in the section of the macronet where VISUAL PERCEPTION is located: (285) 1. CHANGE: To become different/alike 1.1. PHYSICAL PERCEPTION to become aware of something through one's senses or experience [perceive] 1.1.1. speech to become aware of something through one's senses or experience and say it [note] 1.1.2. + VISUAL PERCEPTION to become aware of something with one's eyes [see] 1.1.2.1. LIGHT stimulus domain, necessary for sight [shine] 1.1.3. cognition [ VISUAL PERCEPTION). This establishes a mapping relationship between the two definitions: the first argument, which in glarel (LIGHT) is prototypically a force (light source), becomes an experiencer in glarel (VISUAL PERCEPTION), and the disagreeable light becomes a disagreeable emotion, in this case, anger.

A semantic macronet Table 48.

263

Glare: argument structure and meaning correlations

glarel [LIGHT] glare2 [VISUAL PERCEPTION] to shine with a very bright light that is to stare angrily difficult to look at argument structure Exp = human Force = light source

(i) shine (ii) light source (iii) disagreeable light

5.7.2.3.2.

meaning correlation (i) look (ii) human (iii) negative emotion (anger)

Visual perception

Image schemas derived from our analysis of lexical structure underline different types of relationships between the perceived object and the perceiver, depending on the different meanings conveyed by see/look and their troponyms. Interestingly enough, from an historical perspective, the schemas implicit in lexical domains reflect the theory of visual perception proposed by the Pythagoreans twenty-five centuries ago. Pythagoras and his followers affirmed that vision was the result of an invisible fire (or a collection of rays, according to Euclid) emitted from the eye itself. This fire touched the object(s) of perception, thus enabling the perceiver to distinguish forms and colors. They believed that the eye acted on the object in the following way by emitting invisible light: (294)

Eye

-

• Object

Of the various explanations of vision proposed at that time, this was the one that achieved the greatest popularity. The following arguments were put forward in favor of eyes as light sources:

264

The cognitive axis

1. The eyes of certain animals glow in the dark. 2. The eyes of magicians can mesmerize/hypnotize people. 3. The eyes have a convex shape appropriate for emission in contrast to other sense organs, like ears, which have a concave shape more apt for reception. 4. We can look for a needle at our feet for a long time without finding it. (To be able to see such a small object, it was believed that one of the luminous rays from the collection of those emitted from our eyes had to fall on it.) 5. Someone else's gaze on us can produce a burning/tingling sensation on the back of our neck. Some traces of these pre-Socratic explanations can still be found in expressions, such as to look daggers at somebody or simply to glower at someone. In this type of folk model, the perceived entity is affected in some way by the perceiver. Instead of the role of experiencer, the first argument assumes the role of an agent, who actually does something to the object, in this case, cutting or burning the entity in question with the intensity of his/her gaze. Needless to say, the scientific model of visual perception is very different from that encoded in our language, since in reality our eye is only a passive receiver. The light shining from a light source such as the sun or a reading lamp is reflected off of the perceived entity. The features of the entity are transmitted to our eye by light rays. The image of the object is then formed on the retina as a result of the refraction of the light rays on the lens. object

Figure 11.

Scientific model of vision

As science teachers are aware, the scientifically correct model of vision is not spontaneously acquired. The fact that an object reflects light is far from obvious. We only conceive of an object in relation to light when it is a light source. Of course, no one believes any more

A semantic macronet

265

that we can see because our eyes emit luminous rays, but this folk theory is still present in language. Everyday language, which may reflect and reinforce the common ways of thinking, conveys the same idea: it attributes an active role to the eye, while the object "looked at" has only a passive role; an eye examines, probes, scrutinizes; in romantic literature, eyes flash fire, one looks daggers at someone. Indeed, when looking at an object, there is more a feeling of being an active subject than a passive receptor (Guesne 1985: 26). The active role of the perceiver evident in the vision model of the Pythagorean School can be seen in the way the activity of vision is lexicalized. Within the hierarchical structure of the lexical domains under analysis we have found that the basic image schema that emerges from our physical interactions with the physical world around us can be represented as shown in Figure 12. Figure 12 does not correspond to the scientific model of visual perception, but rather to how we conceptualize seeing/looking at something. This type of perception can be lexicalized both as an event (see and its troponyms), and as a process (look and its troponyms).

concrete entity

Figure 12. Visual perception (physical) Vision as a process is derived from vision as an event, and is representative of the synsem parameter of duration. A long event at some stage ceases to be an event and becomes a process, and accordingly, look is a troponym of see as shown in Table 49:

266

The cognitive axis

Table 49.

Vision as an event and process

VISION AS AN EVENT

VISION AS A PROCESS

see to become aware of somebody/something by using one's eyes. distinguish to see somebody/ something making an effort to do so, because there is difficulty, discern to distinguish somebody/ something but not clearly.

look at to see somebody/something by intentionally directing one's eyes. watch to look at somebody/ something, paying attention to what is happening. observe to watch somebody/something carefully, often in order to learn something, regard to look at somebody/ something with a particular feeling. contemplate to look at somebody/something steadily, in a quiet, thoughtful way.

[short duration]

[short duration]

spot to see somebody/something making an effort, spy to see somebody/something, catching sight of them, sight to see somebody/ something making an effort, briefly and suddenly, glimpse to see somebody/ something very briefly and not very clearly.

glance to look at somebody something quickly, peek to look at somebody/ something quickly, secretly from a hiding place. peep to peek at somebody/ something quickly. [long duration]

gaze to look at somebody /something steadily for a long time because they are attractive or surprising, stare to look at somebody/ something for a long time with wide open eyes.

A semantic macronet

267

However, the image schema for vision (Figure 12) is not a fixed or static image, but rather a dynamic pattern. It can thus take on different specific instantiations depending on the various situations lexicalized within the hierarchical structure of verbs of visual perception. An elaboration of the basic schema can be found in manner-of-staring verbs. (295)

stare

to look at somebody/something for a long time with wideopen eyes. goggle to stare at somebody/something in surprise. gape to goggle at somebody/something with an open mouth. gawk to stare at somebody/something in a stupid, unthinking way [informal], glare to stare angrily at somebody/something, in an unfriendly way. glower to glare at somebody/something for a long time. ogle to stare at somebody with sexual interest.

This group of verbs implies the presence of two perceivers, one that focuses on the perceived entity, and the other whose focus is on the first perceiver as he/she stares in a certain way. The presence of the second perceiver, though implicit, is very real because if someone is gaping (staring at somebody/something in surprise with an open mouth), they cannot see their own face and realize that they are gaping unless they happen to be looking in a mirror. Therefore, someone else is obviously doing the evaluation.

Figure 13.

Manner-of-staring verbs

268

The cognitive axis

In manner-of-staring verbs, there is also information as to how the physical activity is perceived as well as an evaluation of physical/emotional characteristics of the perceiver. The way a person stares causes others to judge him/her as surprised {goggle), angry (glare), or stupid {gawk). Inherent in the action are also facial characteristics such as having an open mouth (gape) or wide-open eyes (stare). Another group of verbs which represent another instantiation of the basic schema are those that convey a mental process parallel to the visual perception, and in most instances, as a consequence of it. (296)

watch to look at somebody/something, paying attention to what is happening. observe to watch somebody/something carefully in order to learn something. contemplate to look steadily at somebody/something, in a quiet, thoughtful way. examine to look carefully and closely at somebody/something. scrutinize to examine all of something (esp. to find information from or about it), scan to examine the whole of something to find one thing in particular. survey to look carefully at something, considering it as a whole, inspect to look carefully at somebody/somebody to check that it is all right.

Verbs such as watch, denoting both attention and intention, or contemplate, denoting evaluation, focus on the steadiness of the perceiver, who is carrying out some sort of parallel mental process as he looks at the concrete entity.

Figure 14. Visual perception with a parallel mental process

A semantic

macronet

269

In other verbs of the same subdomain, the parallel mental process is caused by the carefulness of the perceptual act, made explicit in its purpose. For example, survey implies looking at the total extension of an entity to ascertain its condition, whereas in examine, the perceiver examines something in order to reach a conclusion about it. This same basic schema can also be extended to abstract entities. Vision is thus mapped onto cognition.

Figure 15. Mental perception In its transfer to COGNITION, both the descriptive parameters and the aspectual distinctions found in the lexical structure of VISUAL PERCEPTION remain essentially unchanged. Consequently, the distinction between visual acts or events and visual activities or processes depending on the role of the subject, becomes now a distinction between mental acts or events on the one hand, and mental activities or processes on the other. In the first case seeing becomes understanding, and correspondingly, looking becomes thinking. In both cases the first argument is an experiencer, but the second argument becomes an abstract goal argument. The fourth instantiation of the basic schema appears in verbs in which actual vision is involved. Nevertheless, the perceived entity is an image reproduced within the minds, which may or may not have a physical match in the outside world. (297)

To see somebody/something in one's mind imagine to see somebody/something in one's mind, forming a picture of them/it. dream to imagine somebody/something one wants to happen.

270

The cognitive

axis

daydream to imagine pleasant things and lose touch with reality, fantasize to imagine something very pleasant that is unlikely to happen, picture to imagine somebody/something in order to have a clearer idea of it. visualize to imagine somebody/something very clearly. For example, when we imagine something, the entity perceived in our minds can be either concrete or abstract. (298)

a. b.

Imagine a cat on the table. Imagine a good reason to stop smoking.

On the other hand, its tropoynms, visualize, picture and fantasize, give the perceived mental entity both shape and size. (299)

He visualized a good reason to stop smoking (= his neighbor who had died of lung cancer the month before).

In Figure 16 there are two arrows extending out from the eye. The unbroken line signifies that the perceived entity is concrete and as such, belongs to the outside world, while the dotted line indicates the perception of an abstract entity. Two arrows have been used because these verbs entail both types of perception.

Figure 16.

To see something in one's mind

A semantic macronet

271

In the mapping from the physical to abstract domains, we move from the outer to the inner world of the experiencer. As a result, his/her physical perceptual space becomes a mental perceptual space. However, an analysis of such metaphoric projections shows that despite the change of experiential domain, the relation between the perceiver and the perceived entity remains constant. The recurrence of such perceptual-evaluative parameters is normal if we consider that as human beings, we have common biological characteristics, and develop through interaction with roughly comparable environments. Langacker (1997: 233) writes: Far from being detached and autonomous, the mind is identified with aspects of the functioning of the human body, which is fundamentally alike for all individuals, and thus creates and delimits a common range of potential experience. Individuals also function in a real world likewise the same for everyone in basic respects - which shapes and constrains experience and cognitive development. All of these types of domain interrelations make the lexicon cohere, and lead us to postulate the existence of a semantic network, which is a reflection and subset of conceptual structure. Although the most basic structural relation of this network is hierarchical, there are other nonhierarchical relationships which enhance lexical structure, and are indicators of the multidimensional nature of concepts.

5.8.

Summary

Lexical structure on both the paradigmatic and syntagmatic axes reflects cognition. The link between language and thought is very real. There are traces of conceptual structure at all levels of language, but the problem is that when one tries to arrive at an actual representation, there is no escape from words. Nevertheless, the correlation between the conventionalized meaning in paper dictionaries at the level of microstructure and the meaning representations of our mental lexicon can be regarded as a partial

272

The cognitive axis

indication of how concepts are organized in our mind, and in this sense, lexical structure is a subset of conceptual structure. Evidently, language is certainly not the whole story when it comes to analyzing conceptual structure. Other types of behavior presumably have to be examined as well, but so far, it is language that offers the best window into the mind. Neurological evidence argues for a lexicon organized in areas of meaning and based on the shared and related properties of lexemes. Meaning in the brain is the activation of highly complex neural circuits. The fact that synaptic connections vary as new experience is acquired gives the mental lexicon an eminently dynamic character that is difficult, if not impossible to reproduce outside the brain. The difference between a lexeme in a dictionary and its corresponding representation in semantic memory is something like the difference between reading a tourist brochure about Majorca, and actually going to a luxury resort there, and getting a suntan on the beach. While it is unrealistic to assume that a lexical entry will ever be like having a holiday at an island paradise, the representation in the tourist brochure can be greatly enhanced. The configurational parameters of the FLM lexicon are based on shared properties of meaning as well as combinatorial behavior. This form of lexical categorization is the result of the convergence of both types of meaning, something that facilitates the abstraction of predicate schemas for each level of the lexicon. Predicate schemas are derived from the common properties of a set of predicates, and are thus compatible with all members of the category. These schemas are possible candidates for the representation of modality-independent lexical knowledge. Our model of lexical organization is configured in hierarchies of predicate schemas at the levels of lexeme, subdomain, and domain. This kind of representation, which is derived from the factorization of lexical units, provides the set of paradigmatically and syntagmatically oriented instructions as to a speaker's lexical choice. Lexeme-level predicate schemas are explicit in the definitional structure of each lexeme. They are represented by natural language phrases, and are derived from the configuration and analysis of meaning constituents. Of the three types of predicate schemas, those at subdomain-level are the most productive because they codify links between different lexical

Summary

273

sets throughout the lexicon, and are the basis for metaphorical projections. Domain-level schemas are the macro-organizational patterns of entire domains, and can be of four types: (i) space; (ii) time; (iii) sociocultural context; and (iv) axiological evaluation. As an expansion of the predicate frame, predicate schemas capture generalizations and encode lexical productivity. The dynamic nature of these constructs is evident in the information encoded in them. Both inter- and intra-domain relations are taken into account, as well as systematic mappings between domains, which account for lexical metaphor. To describe the links established in the lexicon, we postulate a semantic macronet in which lexical dependency acts as a pointer to conceptual dependency. Links between domains are traced linguistically. The most salient links between domains take the form of transition zones, causation, and image schemas. Transition zones are areas of overlap between domains. The lexemes located in this area have double domain membership. The second type of link is causation, which brings lexical sets in more abstract domains such as CHANGE closer to concrete action because the action is carried out relatively quickly and the agent is prototypically human. The third type of domain interconnection is that of the image schema. Image schemas, as preconceptual structures for organizing experience, are instantiated in paradigmatic structure in the lexicon, and encode connections between the domains of LIGHT, FEELING, VISUAL PERCEPTION, and COGNITION.

6.

Conclusion

The lexicalist revival that has taken place over the last two decades has centered linguistic theory increasingly on the lexicon. Lexemes and the information they contain are the basis for language at all levels, and accordingly, a lexical representation should evidently convey the internal and external relations of each lexeme, as well as the knowledge inherent in these relations. Much of the interest in the design of a lexical entry stems from the fact that lexical representations are the key as well as the source for predicting and explaining syntactic properties. Nevertheless, the various proposals offered so far for the elaboration of fully specified lexical entries have been very programmatic, and do not contain enough information about meaning. Hitherto, the importance of the interface between syntax and semantics has been generally acknowledged, but more emphasis has been placed on syntax to the detriment of the semantic information within word definitions. As a result, the principal approach to lexical representation has often been to use syntax as a starting point and to elaborate representations in terms of the syntactically relevant meaning components. As has become clear in the previous chapters, we believe it is more productive to begin with semantics and use meaning parameters as a starting point to better understand and represent syntax. This is in consonance with Simon Dik's FG, which operates on the basic premise that the properties of clause structure are predictable from the semantics of predicates. However, after Dik's (1978b) description of Stepwise Lexical Decomposition, its principles and corollaries were not developed any further in the FG theoretical framework, and the information conveyed in meaning definitions was relegated temporarily to the background. In this book, we have described how the FG lexicon component can be expanded by organizing it according to FLM principles. This means using definitional analysis and semantic information in meaning definitions to give the lexicon the onomasiological type of organization

276

Conclusion

described in Chapter Two. Lexemes are structured hierarchically, along with their complementation patterns. In this respect, the degree of specificity encoded in a predicate's semantic content should have a parallel degree of syntactic specificity in relation to that of the other members of the meaning hierarchy. The lexicon is thus conceived in terms of a set of articulated semantic classes called lexical domains. These domains are hierarchically configured in subdomains, whose members share a set of semantic, syntactic and pragmatic regularities typical of that particular area of meaning. This provides a format to account for syntactic and semantic generalizations across different types of predicates which share a given area of meaning. The resulting inventory of lexical domains is not the ad hoc set of categories found in traditional thesauri, but rather the product of a datadriven analysis in which the conceptual categories are derived principally from the configuration of the meaning constituents found in a variety of different monolingual dictionaries. Semantic and syntactic typologies are deeply enmeshed because the semantic architecture of each domain and subdomain is in consonance with by the number and characteristics of the complementation patterns of its predicates. This type of onomasiological structure reveals that the lexicon contains information that is more productive than has hitherto been supposed. In fact, each lexical subdomain can be said to constitute a source of lexical productivity in the sense that all of its members share a set of properties. These properties, both semantic and syntactic, interact with each other in such a way that the greater the semantic scope of a lexeme, the greater its syntactic variation. In a lexical subdomain, we have postulated that semantic information constrains and ultimately explains the syntactic projections of its members. We have consequently extracted meaning components that are grammatically relevant within each subdomain. These take the form of synsem parameters such as time, iteration, causation, achievement, conation, /activity, duration, inception, cessation, etc. which play a determining role in the complementation patterns of the different lexemes in a specific area of meaning. The activation of these parameters is in function to the shared meaning inherent in the semantics of the lexical subdomain. The information codified in both the paradigmatic and syntagmatic axes of the lexicon converges to form a type of

Conclusion

277

representation that we have termed predicate schema, which contains the recurrent semantic, syntactic and pragmatic information at a particular level of meaning and is an enriched predicate frame. In this regard, we also believe that meaning should be formulated relative to some background, scene, or frame, similar to certain proposals in Cognitive Linguistics. Although schemas can be found at all levels of the lexicon, from the level of a specific lexeme to that of an entire lexical domain, the most productive type for lexical representations is found at the level of the subdomain, which constitutes the link between the other two levels. This fact runs parallel to our assertion that each lexical subdomain can be regarded as a kernel lexical grammar. The formalization of a predicate schema contains a variety of different types of information that is both inter and intra-domainspecific. Whereas the recurrent organizational patterns in lexical domains are indicative of perception and categorization on a more extensive scale, those at the level of subdomain are indicative of synsem parameters which can trigger or activate other types of information in terms of inheritance mechanisms. Predicate schemas at this level also encode the relational structure of the lexicon, and indicate how lexical representations are mapped from one lexical domain/subdomain to another. Furthermore, this notion also accounts for multiple verb senses, which can be derived from one meta-entry by means of the application of lexical rules. We have formalized lexical connections in the form of a semantic macronet in which domain and subdomain relations are represented through definitional structure. Systematic mappings between different domains help to explain lexical metaphor. Such schemas are the basis upon which the specific lexemes of each domain can be generated. This generation would be done by means of lexical rules, which would accordingly enhance the FG expression rule component. These rules would translate the regularities encoded in each predicate schema into the actual predicates of a language. In this way, the lexicon becomes truly meaningful not only because of the lexical items it contains, but also because of how they are organized. The systematization of meaning opens a new perspective on the configuration of a lexical entry, and allows a better understanding of the interface between syntax and semantics.

Appendix 1: Lexical domains

1.

Existence 1.1. 1.2. 1.3.

1.4.

1.5.

1.6.

1.7. 1.8.

General: To exist / to continue to exist To begin to exist [be, live] 1.2.1. To cause something to exist [create, make] To exist in the perception of others [appear] 1.3.1. To cause something to exist in the perception of others [show] To exist in time (becoming real) [happen] 1.4.1. To cause something to exist in time [induce, provoke] 1.4.1.1. To cause something to exist in time in a particular way [precipitate, hasten] 1.4.1.2. To cause something to happen, making it possible [allow, permit] 1.4.1.3. To cause something not to happen [prevent, avoid, stifle, smother] To exist as something 1.5.1. To exist as the representation of something else [represent, express] 1.5.1.1. To cause something to exist as a representation of something [copy, reproduce] 1.5.2. To exist as a part of something [comprise, constitute] To begin to exist [start, commence, be born] 1.6.1. To cause to begin to exist [start, commence] 1.6.1.1. To cause to be born [abort] 1.6.2. To begin to exist in the perception of others [arise, form] 1.6.3. To begin to exist in time (becoming real) [start, originate] 1.6.3.1. To cause something to begin to exist in time [start, initiate] To continue to exist [last, endure] 1.7.1. To stop something from continuing [interrupt] To stop existing [die] 1.8.1. To cause somebody/something to stop existing [kill, murder] 1.8.2. To stop existing in the perception of others [disappear, vanish]

280

Appendix 1 1.8.2.1. 1.8.2.2. 1.8.2.3.

2.

To cause something to stop existing in the perception of others [erase, delete] To stop existing in time [end,finish,cease] To cause something to stop existing in time [end, finish, cease]

Movement 2.1.

General [move, go, come] 2.1.1. To move in a particular way 2.1.1.1. To move quickly [speed, race, hurry] 2.1.1.1.1. To cause somebody/something to move quickly [speed, race, hurry] 2.1.1.2. To move slowly [slow, lumber, trundle] 2.1.1.2.1. To cause something to move slowly [slow, brake, rein] 2.1.1.3. To move slightly [stir] 2.1.1.4. To move smoothly, easily [glide, slide] 2.1.1.4.1. To cause something to move smoothly, easily Iglide, slide] 2.1.1.5. To move (forward) suddenly [leap, lunge] 2.1.1.6. To move in a circular manner [circle, tum, spin] 2.1.1.6.1. To cause something to move in a circular manner [circle, turn, spin] 2.1.1.7. To move from side to side/back and forth/ up and down repeatedly [swing, rock, shake] 2.1.1.7.1. To cause something to move from side to side/back and forth/ up and down repeatedly [swz'wg, rock, shake] 2.1.1.8. To move from an upright position [lean, bend] 2.1.1.8.1. To cause something to move from an upright position [lean, bend] 2.1.2. To move ofi/away 2.1.2.1. To move offi'away from a place/ thing/ person [separate, leave] 2.1.2.1.1. To cause somebody/something to move away from a place/ thing/person [send, separate, remove] 2.1.2.1.1.1. To cause somebody/something to go away because it is not wanted any more [discard, scrap] 2.1.3. To move towards a place [go, travel, advance]

Lexical domains 2.1.3.1.

281

To cause somebody/something to move towards a place [advance] 2.1.3.2. To move towards a common point, coming together [meet, join, gather] 2.1.3.2.1. To cause people/things to come together at a common point [join, gather, assemble] 2.1.3.3. To move backward [back, reverse] 2.1.3.3.1. To cause something to move backward [back, reverse] 2.1.3.4. To move upwards [rise, ascend] 2.1.3.4.1. To cause somebody/something to move upwards [raise, lift] 2.1.3.5. To move downwards [fall, drop] 2.1.3.5.1. To cause something to move downwards [lower, drop] 2.1.4. To move across [cross] 2.1.5. To move over/through [pass, clear] 2.1.5.1. To cause something to move over/through something [pass] 2.1.6. To move in a different direction [turn, twist, bend] 2.1.6.1. To cause something to move in a different direction [turn, twist, bend] 2.1.7. To move about in no particular direction [wander, drift] 2.1.8. To move in relation to somebody/something 2.1.8.1. To move together [accompany] 2.1.8.1.1. To cause people/animals to move together [herd] 2.1.8.1.2. To cause somebody/something to go with you [take, bring] 2.1.8.1.3. To move with somebody, going before [lead] 2.1.8.1.4. To move with somebody, going after/behind [follow] 2.1.8.2. To come together [join, connect] 2.1.8.2.1. To cause to come together [join, connect] 2.1.8.3. To move slower in relation to somebody/ something [lag, trail] 2.1.8.4. To move towards and beyond somebody/ something [pass] 2.1.8.5. To move round in order to be on all sides of [surround, circle] 2.1.8.6. To move out in all directions [spread, scatter]

282

Appendix 1 2.1.8.6.1.

To cause to move out in all directions [spread, scatter] 2.1.8.6.2. To move apart [separate, divide] 2.1.8.6.2.1. To cause something to move apart from something else [separate, divide] 2.1.8.6.2.2. To come apart [break, disintegrate] 2.1.8.6.2.2.1. To cause something to come apart [break, disintegrate] 2.1.8.7. To move back and forth between places 2.1.8.7.1. To cause somebody/something to move back and forth between places 2.1.8.8. To move into a place [enter] 2.1.8.8.1. To cause somebody/something to move into a place [enter] 2.1.8.8.2. To move into a building by force [burgle] 2.1.8.9. To move out of a place [emerge] 2.1.8.10. To move to a different place/position [change, switch, transfer] 2.1.8.10.1. To cause somebody/something to move to a different place/position [change, switch, transfer] 2.1.8.10.2. To move somebody/something to a different place/position by holding and drawing them along with force [pull, drag] 2.1.8.10.3. To move somebody/something to a different place/position by holding by walking behind them and exerting force on them, esp. with one's hands [push, shove] 2.1.9. To not move any more, after having moved [stop, halt] 2.1.9.1. To cause somebody/something to not move any more, after having moved [stop, halt] 2.2. Liquid 2.2.1. To move as liquid in a particular way [flow, ripple] 2.2.1.1. To move slowly in small quantities [drip, dribble] 2.2.1.2. To move quickly in large quantities [pour, gush, flood] 2.2.1.2.1. To cause a liquid to move quickly in large quantities [pour, flood] 2.2.1.3. To move through an opening [squirt, spurt] 2.2.1.3.1. To cause a liquid to move through an opening [squirt, spurt]

Lexical domains

283

2.2.1.3.2. To cause (liquid) to go in [inject] To move in/downward below the surface of a liquid [sink, plunge] 2.2.2.1. To cause somebody/something to move in/downwards below the surface of a liquid [sink, plunge, dip] 2.2.2.2. To move in liquid using parts of the body [swim, paddle] 2.2.3. To move over liquid [sail, cruise] 2.2.3.1. To cause something to move over liquid [launch, sail, paddle] 2.2.3.2. To move over liquid in no particular direction [float, drift] 2.2.4. To move upwards to the surface of a liquid [surface] 2.2.4.1. To cause something to move upwards to the surface of a liquid [surface] 2.3. Atmosphere 2.3.1. To move as air [blow] 2.3.2. To move through the air [fly,flutter,float] 2.3.2.1. To cause something to move through the air [fly, float] 2.3.3. To move upwards in the air [rise, soar] 2.3.4. To move downwards in the air [swoop, dive] 2.4. Land 2.4.1. To move in a particular way [skulk, creep, scamper] 2.4.1.1. To move using one's feet [walk] 2.4.1.1.1. To move quickly using one's feet [run] 2.4.1.1.2. To move up and down using one's feet [jump] 2.4.1.2. To move downwards to the ground [fall] 2.4.2. To cause somebody/something to move downwards to the ground [trip, tumble] 2.4.3. To move one's body [gesture, exercise] 2.4.3.1. To move one's body from side to side [squirm, wriggle] 2.4.3.2. To move one's body by raising it [stand] 2.4.3.3. To move one's body by lowering it [ί/ί, bend] 2.4.3.4. To move a part of one's body [duck, lick, wave] 2.2.2.

Appendix 1 POSITION (TO be in a particular/state/condition/position without moving/ changing) 3.1.

To be in a particular/state/condition/position without moving/changing [stay, lie] 3.1.1. To cause somebody/something to stay in a particular state/ condition/ position [keep, maintain] 3.1.1.1. To cause somebody to be in a particular place/position [put, place] 3.1.1.1.1. To put things together [join, attach] 3.1.1.1.2. To put many things together [pool] 3.1.1.1.3. To put something around something else [wrap] 3.1.1.1.4. To put something on (the surface of) something else [cover, spread] 3.1.1.1.4.1. To cover something with something else to protect it/make it more attractive [paint, coat] 3.1.1.1.4.2. To cover somebody/something so that it cannot be seen [enshroud, cloak] 3.1.1.1.5. To put something on top of something else [pile] 3.1.1.1.6. To put something in/onto something else [insert, fill] 3.1.1.1.7. To put something in a particular position/order [arrange, classify] 3.1.1.1.8. To put somebody/something in the place of somebody/ something else [replace, substitute] 3.1.1.1.9. To put somebody/something in a bad situation [risk, endanger]

Contact 4.1.

To come into contact with somebody/something [hit] 4.1.1. To hit somebody/something hard and continuously for a long time [beat, batter] 4.1.2. To hit a ball [slice, hook] 4.1.3. To hit against something [collide] 4.1.3.1. To cause something to hit against something [crash, ground] 4.1.4. To not hit [miss]

Lexical domains 5.

285

Change (To begin to be different) 5.1.

To become [change] 5.1.1. To change somebody/something [vary, transform] 5.1.2. To change by increasing (becoming more) [grow, increase] 5.1.2.1. To cause something to change by increasing it [increase, augment] 5.1.3. To change by decreasing (becoming less) [lessen, decrease, diminish] 5.1.3.1. To cause something to change by decreasing it (making it less) [lessen, decrease, reduce, diminish] 5.1.4. To change becoming better [improve] 5.1.4.1. To cause something to change by making it better [improve, enhance] 5.1.4.2. To cause something to change by making physically more attractive [decorate, adorn] 5.1.4.3. To cause something to change by making it better/right again [correct,fix,amend] 5.1.4.4. To cause something to change by making it better/free of something bad [clean, purify, cure] 5.1.5. To change by becoming worse [worsen, deteriorate, spoil] 5.1.5.1. To cause something to change by making it worse [worsen, deteriorate, spoil] 5.1.5.2. To cause something to change by making it less good/ attractive/satisfactory [harm, damage] 5.1.5.3. To cause something to change by making it impure/dangerous/ dirty [dirty, defile, pollute] 5.1.6. To change in time [age] 5.1.6.1. To cause something to change in time [age] 5.1.7. To change in appearance 5.1.7.1. To change color [color, redden] 5.1.7.1.1. To cause something change color [color, redden] 5.1.7.2. To change in texture [coarsen] 5.1.7.2.1. To cause something to change in texture [smooth, crumple] 5.1.7.3. To change in state [solidify, liquefy] 5.1.7.3.1. To cause something to change in state [solidify, liquefy] 5.1.8. To change, becoming like something else [normalize]

286

Appendix 1 5.1.8.1.

6.

Perception (To become aware of the existence of somebody/ something) 6.1. 6.2.

6.3. 6.4.

6.5.

6.6. 1.

To cause something to change, making it like something else [standardize, normalize]

General perception (all senses): to become aware [perceive, find, discover] Visual perception (To become aware by using one's eyes) [see, look] 6.2.1. To see (notice, observe) 6.2.2. To see intentionally [distinguish, discern] 6.2.2.1. To see intentionally, directing one's eyes [gaze, stare] 6.2.3. To be seen [show] 6.2.3.1. To cause somebody/something to be seen [show] 6.2.4. To not see [be blind] 6.2.4.1. To cause somebody to be unable to see [blind, dazzle] 6.2.4.2. To cause something not to be seen any more [hide, conceal] Tactile perception (To become aware by using one's hands) [touch, feel] Olfactory perception (To become aware through one's nose) [smell, scent] 6.4.1. To cause somebody to become aware of something through one's nose [smell, stink] Auditory perception (To become aware through one' s ears) 6.5.1. To perceive something with one' s ears [hear] 6.5.2. To hear intentionally [listen] Taste perception (To become aware through one's mouth) [taste]

Cognition (to become aware through one's mind) [know] 7.1.

To become aware of something, (having it) in one's mind [know] 7.1.1. To come to know something [learn] 7.1.1.1. To cause somebody to learn [teach] 7.1.1.2. To cause something to be known [show] 7.1.2. To know the nature/meaning of something [understand] 7.1.2.1. To cause somebody to understand something [enlighten, illuminate]

Lexical domains

287

7.1.2.2.

To cause something to be understood better [clarify] 7.1.2.3. To understand with difficulty (grasp] 7.1.2.4. To not understand [mistake] 7.1.2.4.1. To cause somebody not to understand/understand with difficulty 7.2. To use one's mind to become (more) aware of something in a certain way [think about] 7.2.1. To think about something bringing it back into one's mind from the past [remember] 7.2.2. To think about something that has happened in the past [reflect] 7.2.3. To think about something (usu. in order to understand it better [meditate] 7.2.4. To think about something in order to make a decision (in the future) [consider] 7.3. To use one's mind to form an opinion/idea [think (of)] 7.3.1. To think something, having formed an opinion/come to a decision about it [decide] 7.3.2. To think something is true [believe] 7.3.3. To think something is going to happen [expect] 7.3.4. To think something is likely to be true [suppose] 7.3.5. To think without knowing if it is true [guess] 7.3.6. To think something may not be true [doubt] 7.3.7. To think (of) something, forming it in one's mind as an idea/picture [imagine] 8.

Feeling (to become aware of something other than by sight, having a sensation) 8.1. 8.2. 8.3.

8.4.

8.5.

To feel something bad [suffer] 8.1.1. To cause somebody to feel something bad [punish] To feel something good [enjoy] To feel sadness [grieve] 8.3.1. To cause somebody to feel sadness [sadden, grieve, distress] To feel happiness [delight in, thrill, rejoice] 8.4.1. To cause somebody to feel happiness [please, gladden, thrill] To feel aversion [dislike, hate, detest] 8.5.1. To cause somebody to feel physical aversion [disgust]

288

Appendix 1 8.5.2.

To cause somebody to feel emotional aversion [anger, enrage] 8.6. To feel attraction [like, love, admire] 8.6.1. To cause somebody to feel attraction [attract, fascinate, interest] 8.6.2. To feel a loss of attraction [tire, weary] 8.6.2.1. To cause somebody to feel a loss of attraction [tire, weary] 8.7. To feel something bad in one's body [hurt, ache] 8.7.1. To cause somebody to feel something bad in their body [hurt, wound] 8.7.2. To cause somebody to feel less pain [alleviate, assuage] 8.8. To feel fear [fear, dread, worry] 8.8.1. To cause somebody to feel fear [frighten, scare, terrify] 8.8.1.1. to cause somebody feel less fear/anger [calm, soothe] 8.9. To feel surprise [wonder, marvel] 8.9.1. To cause somebody to feel surprise [surprise, astonish, amaze] 8.10. To feel shame 8.10.1. To cause somebody to feel shame [shame, disgrace, embarrass] 8.11. To feel a need to do something or to have/get something [want, wish, desire] 9.

Speech [say, speak, talk] 9.1.

To say something in a particular way 9.1.1. To say something formally [address, state, declare] 9.1.2. To say something informally [gossip, chat] 9.1.3. To say something firmly [insist, emphasize] 9.1.4. To say something precisely [specify] 9.1.5. To say something briefly [mention] 9.1.6. To say something again [repeat] 9.1.7. To say something with difficulty [stutter, stammer] 9.1.8. To say something quickly/continuously [chatter, babble] 9.1.9. To say something suddenly/loudly [exclaim, shout] 9.1.10. To say something in a soft way [whisper] 9.1.11. To say something angrily [snarl] 9.1.12. To say something unhappily in a dissatisfied way [complain, lament] 9.1.13. To say something in a proud way [boast, brag]

Lexical domains

289

9.1.14. To say something in a rude way [insult] 9.2. To say something 9.2.1. To say something is the case (positive things) 9.2.1.1. To say yes to somebody/something [accept, agree] 9.2.1.2. To say that something is true [acknowledge, admit] 9.2.1.2.1. To say that something will happen [foretell, predict] 9.2.1.2.2. To say that something is certain [promise, guarantee] 9.2.1.3. To say positive things about somebody/ something 9.2.1.3.1. To say positive things, speaking favorably Ipraise, extol] 9.2.1.3.2. To say positive things, saying that something should be considered [suggest, advise] 9.2.2. To say something is not the case (negative things) 9.2.2.1. To say no to somebody/something [refuse, reject] 9.2.2.2. To say negative things about somebody/ omething [criticize, slander] 9.2.2.2.1. To say negative things so that others will laugh [ridicule] 9.2.2.3. To say something bad may happen [warn, threaten] 9.2.3. To say something without knowing if it is the case [guess] 9.3. To say something for a particular purpose/ with a specific result 9.3.1. To say something to somebody so that they will do it [direct, order] 9.3.1.1. To say something to somebody else to put an idea in their mind [suggest] 9.3.2. To say something in order to get something else [ask, request] 9.3.2.1. To say something in question form in order to get an answer [ask, question] 9.3.2.1.1. To say something in return to something [answer] 9.3.2.1.2. To say something not in return to something [remark, comment] 9.3.3. To say something to somebody to tell them about it 9.3.3.1. To say something expressing an opinion or judgement one has arrived at [reason] 9.3.3.2. To say something to somebody so that they will remember it [remind]

290

Appendix 1 9.3.3.3.

9.4. 9.5.

To say something to somebody so that they will know it [inform, notify] 9.3.3.4. To say something to somebody giving an account of it [describe] 9.3.3.5. To say the main points of [outline] 9.3.3.6. To say something to somebody else, talking it over from several points of view [i/wcwii] To say something by a particular means [telephone] To say something in a different language [translate]

10. Sound 10.1. Sound produced by living creatures 10.1.1. To make a sound by speaking 10.1.1.1. To make a loud sound [shout] 10.1.1.2. To make a soft sound [murmur] 10.1.1.3. To make a sound indicating an emotion 10.1.1.3.1. unhappiness [cry, whine] 10.1.1.3.2. happiness [laugh] 10.1.1.3.3. disapproval [hoot, boo] 10.1.2. To make a sound related to body conditions 10.1.2.1. To make a sound by breathing [sigh] 10.1.2.1.1. To make a sound by breathing quickly and audibly |gasp, puff] 10.1.2.2. To make a sound by expelling air 10.1.2.2.1. To make a sound by expelling air from one's throat [cough, burp] 10.1.2.2.2. To make a sound by expelling air from one's nose [sneeze, snort] 10.1.2.2.3. To make a sound by expelling air through one's anus [fart] 10.1.2.3. To make a sound by inhaling air [sniff] 10.1.3. To make a sound like an animal 10.1.3.1. To make a sound like a wild/angry animal [growl, howl] 10.1.3.2. To make a sound like a domestic animal [bark, meow] 10.1.3.3. To make a sound like a bird [chirp, squawk] 10.1.3.4. To make a sound like an insect [buzz] 10.2. Sounds produced by Nature [bubble, rustle] 10.3. Sounds produced by objects 10.3.1. To make a musical sound [ring, chime]

Lexical domains

291

10.3.2. To make a metallic sound [clink, clank] 10.3.3. To make a factional sound [rasp, grate] 10.3.4. To make a punctual, explosive sound [crash, bang] 10.3.5. To make a dull, punctual sound 10.3.5.1. Something hitting against something else [thud, thump] 10.3.5.2. Something heavy hitting against liquid /splash] 10.3.6. To make a durative sound [roar, boom] 10.3.7. To make an iterative, sibilant sound [rustle, swish] 11. Light 11.1. To give off light/ be bright [shine] 11.1.1. To shine brightly |glare, blaze] 11.1.2. To shine softly/faintly [shimmer, glimmer] 11.1.3. To shine steadily [beam] 11.1.4. To shine unsteadily [flicker] 11.1.5. To cause something to give off light/be bright [shine, illuminate] 11.2. To be^ecome without light [darken, dim] 12. Possession 12.1. To have something [possess, own, hold] 12.1.1. To come to have something [get, obtain] 12.1.1.1. To get something as a result of force/skill [take, capture] 12.1.1.2. To get something through effort/as a reward [gain, earn] 12.1.1.3. To get something after it has been given/sent to you [receive] 12.1.1.4. To get a large number of things over a period of time [collect, accumulate] 12.1.1.5. To get something back after it has been lost/stolen [recover] 12.1.2. To continue to have something [keep, save] 12.1.2.1. To have something within as a part [contain, include] 12.1.2.2. To cause something to have something as a part [include, incorporate] 12.1.2.2.1. To not include [omit, exclude] 12.1.3. To stop having [lose]

292

Appendix 1 12.1.3.1. To cause to stop having [take, steal] 12.1.4. To cause somebody/something to have [give] 12.1.4.1. To give something formally [present, confer] 12.1.4.2. To give something, having the power/authority to do so [concede] 12.1.4.3. To give something in return for something else [exchange, trade] 12.1.4.4. To give something in return for something good done [reward, honor, bestow] 12.1.4.5. To give money in return for something [pay] 12.1.4.6. To give somebody/something what is needed [provide, supply] 12.1.4.7. To give something for use after death [leave, bequeath] 12.1.4.8. To give something in shares/units to many people [distribute] 12.1.4.9. To cause somebody/something to have more money [enrich] 12.2. To not have [lack]

13.

Action 13.1. To do something [perform] 13.1.1. To start to do something [undertake] 13.1.2. To continue doing something [proceed] 13.1.3. To do something bringing it to a finish [accomplish, fulfil] 13.1.4. To do something again [repeat] 13.1.5. To do something well [succeed] 13.1.6. To do something badly [botch] 13.1.7. To do something in particular [work] 13.1.7.1. To stop working [retire] 13.1.7.1.1. To cause somebody to stop working [retire] 13.1.8. To do something for somebody [help] 13.2. To not do something [fail, neglect] 13.2.1.1. To cause somebody not to do something [prevent] 13.2.2. To stop doing something [end, finish] 13.2.3. To make an effort in order to be able to do something [try, attempt] 13.3. To do something for a particular purpose [act] 13.3.1. To cause somebody to do something [influence] 13.3.2. To do something to cause enjoyment [amuse] 13.3.3. To do something to cause suffering [punish]

Lexical domains

293

13.3.4. To do something to get a certain result [deal with, use] 13.3.4.1. To use something without the right to do so [pirate] 13.3.4.2. To use a writing instrument [write, scribble] 13.3.4.3. To use something until it is finished [consume, exhaust] 13.3.4.3.1. To cause somebody to consume something [drug, dope] 13.3.4.4. To use one's teeth to cut through something [bite] 13.3.4.5. To use violence against somebody/something to hurt/damage [fight, attack] 13.4. To make something [create] 13.4.1. To make something as a result of a natural process [sweat] 13.4.2. To make something by doing something [produce] 13.4.3. To make something in a particular way [fashion] 13.4.4. To make something in a particular place [manufacture] 13.4.5. To make something as a representation of something else [copy]

Appendix 2: Semantic network

1.

EXISTENCE: To be in space or time 1.1. POSITION: To be in a particular state/condition/position without moving/hanging. 1.2. MOVEMENT: To begin/continue/stop existing 1.3. POSSESSION: To have (for something/somebody to be with you/as a part of you). 1.4. action [make): To cause something to be [create] by making it. 1.5. action [do] : To cause something not to be by doing something (carrying out a specific action)[end, cancel], 1.6. perception [physical & mental]: (i) To be in the perception of others [appear] (ii) To begin to be in the perception of others [arise] 1.6.1. causative perception: (i) To begin to develop, showing the first signs of something [dawn], (ii) To cause to exist in the perceptionof others [show]. 1.6.2. visual perception: To be in the perception of others, becoming visible [surface, materialize, form], 1.7. change: (i) To be in the perception of others, becoming visible [surface, materialize, form], (ii) To be in time, becoming real [happen], 1.7.1. causative change: to cause something to be/become [make, render] 1.8. feeling: to be regularly in a certain place, usually causing fear, problems, worry [haunt],

2.

CHANGE: To become different/alike 2.1. PHYSICAL PERCEPTION: To become aware of something through one's senses or experience [perceive], 2.1.1. speech: To become aware of something through one's senses or experience and say it [note]. 2.1.2. VISUAL PERCEPTION: To become aware of something with one's eyes [see], 2.1.2.1. LIGHT: stimulus domain, necessary for sight [shine].

Semantic network 2.1.3.

295

mental perception (cognition): [contemplate, regard]

2.1.4.

speech: To become aware of something with one's eyes and say it [observe], 2.1.5. position: To (cause something) not to be seen because of putting it in a place [hide, secrete] or putting something over it [cloak, veil], 2.1.6. AUDITORY PERCEPTION: To become aware of something with one's ears [hear], 2.1.6.1. SOUND, SPEECH: stimulus domains for hearing. 2.1.7. 2.1.8. 2.1.9. 2.2.

TACTILE PERCEPTION: [touch] OLFACTORY PERCEPTION: [smell]

TASTE PERCEPTION: [taste] MENTAL PERCEPTION: To become aware

of something in one's

mind [know]. 2.2.1.

speech.

(i) (ii) (iii) 2.2.2.

To become aware of something in one's mind and say it [reason], To (try to) cause somebody to think something is true by saying things to them [persuade] To cause somebody to think something is true by telling them things that are untrue [con].

visual perception:

(i) (ii) 2.2.2.1.

To become aware of something in one's mind though one is seeing it. [see] To cause somebody to be aware of something their mind by causing it to be seen [show]. light: To cause somebody to be aware something in their mind by causing it to be seen though by shining a light on it. [illuminate]

as in of as

2.2.3.

possession/tactile perception: To become aware of

2.2 A.

something in one's mind as though by getting hold of it in one's hand [grasp], possession: To cause somebody to think something is true when it is not in order to take something away from them [defraud, swindle],

2.2.5.

action:

(i) (ii)

To cause somebody to become aware of something by doing something [publicize], To cause somebody to think that something is true when it is not to make them do something you want them to. [dupe].

296

Appendix 2 2.2.6.

change/existence: To think that something is going to become real (happen) [expect], 2.3. FEELING: To experience (become aware of a feeling/experience/ sensation 2.3.1. visual perception: (i) To feel attracted to somebody/something, perceiving them visually (looking on them) with interest/approval [admire], (ii) To feel pleasure at somebody else's misfortune and show it by one's facial expression [gloat], 2.3.2. speech: (i) To feel something, expressing that feeling in words [/im, fret]· (ii) To cause somebody to feel a certain way by saying words to them. 2.3.3. action: To cause somebody to feel pain /pleasure by doing something, [hurt, punish] 2.4. movement: To move (To change from one place/position to another). 2.5. action: To become/cause to become different by doing something, [straighten, widen, etc.] ACTION: To do/make something 3.1. USE: To do something with something else to get a certain result. 3.1.1. speech: To use one's mouth to speak/say words. To use something to speak/say words [telephone] 3.2. feeling: To do something causing somebody/something to feel something (pain, pleasure, etc.) [torture, punish, spank], 3.3. possession: To do something to have something else. 3.4. change: To do something that causes something to become different [clean, damage], 3.5. movement: (i) To do something by coming together/apart. (ii) To do something that causes something to come together/apart [join, separate], (iii) To do something in which something moves /causing something to move (with something else) [rake, plow, ski, etc.], 3.6. position: To do something (i) to cause somebody/something to be in a place/position [hold, save].

Semantic network (ii)

3.7. 3.8. 3.9. 3.10.

297

to put something in/on a particular place (focus on instrument [tape, nail, screw], location [bottle, can, box, etc.], force/difficulty [wedge, jam], substance [paint, wax]. existence: To make something, causing it to be [create], mental perception: To do something to make something known [publicize], speech: To do something by saying something [annul, consecrate, acquit, etc.] SLEEP: To not do anything active for a period of time.

Appendix 3: Possession

To have something possess to have something [formal], to come to have something get to come to have something as a result of some previous activity, gain to get something wanted/beneficial through effort. win to gain something by work/struggle/skill, obtain to get something as a result of work/planning, procure to obtain something difficult to find [formal], acquire to obtain something with effort, adding it to previous possessions. to get money for something done earn to get money as payment for work done. To get money from somebody extort to get money from somebody by force/ threats. collect to get money from a number of people/places. to get something by paying money for it buy to get something by paying money for it. purchase to buy something [formal], to get somebody/something as one's own from somebody/ something else take to get somebody/something as one's own from something/ somebody else (esp. by force/skill). to take somebody/something with a certain movement grab to take something esp. picking it up suddenly/ roughly, snatch to take something quickly with a fest often violent movement. to take goods/property confiscate to take goods/private property from somebody with the right to do so, esp. as punishment.

Possession

299

impound

to take goods/property from somebody officially and keep it until it is claimed, appropriate to take something for oneself without having the right to do so. expropriate to take something from somebody for public use often without payment to the owner, commandeer to take private property (vehicles, buildings, etc.) officially/by force for military use

take something by force seize to take somebody/something by official order/by force, annex to seize something (an area of land by force, taking control of it), wrest to take something from somebody by force [oldfashioned], capture take somebody/something by force/skill to control them. catch to capture somebody/something esp. after chasing them. to catch a person apprehend to catch somebody in order to arrest them [formal], nab to catch somebody [informal], to catch an animal bag to catch an animal hunting, fish to (try to) catch fish, hook to catch a fish with a hook, trap to catch an animal in a trap (a device for catching animals/ birds), snare to trap an animal with a snare (a noose that jerks tight around its body).

take something away from somebody by deception swindle to take money from somebody by cheating or deceiving them, con to swindle somebody, telling them things that are untrue.

300

Appendix 3 to take something away from somebody without the right to do so steal to take something away from somebody without their permission and not intending to return it. rob to steal money/property from somebody/ institution. defraud to rob somebody of their money by deception. embezzle to steal money placed in one's care for one's own purposes, purloin to steal something small [formal], shoplift to steal things from shops by taking them from the shelves and hiding them under clothes or in a bag pilfer to steal things that are small/of little value continuously over a period of time, filch to steal things secretly that are small/of little value [informal], lift to steal [informal]. swipe to steal something by removing it quickly. to get something back again recover to get something back again that has been lost/stolen. recuperate to recover something (one's health after being ill/ money spent or lost), recoup to recover something (a sum of money), regain to recover a desirable condition/state, retrieve to get something back from a difficult place [formal], redeem to get something back by paying money for it.

Appendix 4: Synsem interface of to come to have

0

TO COME TO HAVE

To come to have [GOAL]

sth beneficial [manner] straggle/skill [instr] work/planning [prag] formality [manner] effort

Go

Ben. So

Go

NP NPI NP2

NP PP

NP NP PP PP

for

from

of

GET

-

+

+

+

+

-

gain

-

+

+

+

-

-

-

+

+

+

-

-

-

+

-

+

+

-

procure

-

+

-

+

+

-

acquire

-

+

-

+

+

-

+

+

+

+

-

-

win

obtain

[GOAL] money

[exchange] [source] firomsb [source spec] many people [instr] violence [INSTR] by paying money [prag] formality

earn

-

collect

extort

buy purchase

+

+

-

-

+

-

+

+

+

+

-

+

-

+

+

-

+

+

-

+

+ -

302

Appendix 4

Synsem interface of to come to have (cont.) [SOURCE]

from sb else [manner] quick, violent movement [manner spec] rudely [manner spec] suddenly [goal] goods/property as punishment officially for public use for military use without permission force [manner] officially [goal] area of land [prag] old-fashioned [purpose] control [temporal seq] after a chase [goal] person [prag] formal [prag] informal [goal] animal [goal spec] fish [instr spec]

-

+

-

+

+

-

grab

-

+

-

-

+

-

snatch

-

+

-

-

+

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

TAKE

confiscate impound expropriate commandeer

-

+ + + +

appropriate

-

+

-

-

-

-

seize

-

+

-

-

+

-

-

+

-

-

-

-

wrest

-

-

-

-

+

-

capture

-

+

-

-

+

-

-

+

-

-

-

-

-

+

-

-

-

-

-

+

-

-

-

-

-

-

-

-

-

-

-

-

-

[INSTR] by

annex

catch

apprehend nab fish bag, hook, trap

+ -

+ +

Synsem interface Synsem interface of to come to have [INSTR] by deception [goal] money [instr spec] with lies [MANNER] unlawfully [goal] money/property [instr] deception

(cont.)

-

+ +

+

-

-

-

+

-

-

-

+

+

+

-

+

-

-

+

-

-

-

+

-

+

-

-

-

+

embezzle

+

+

purloin pilfer

+

+ +

-

lift swipe

swindle con

-

steal rob defraud

[goal spec] money in one's care [goal spec] small items [prag] formality [duration] continuously [manner] secretly

303

filch

[manner] movement [movment spec] silently [movement spec] quickly [location] shop

shoplift

+

-

-

-

-

+ +

+

-

-

+

-

-

+

-

-

+

-

-

+

-

-

+

-

+

+

-

-

-

-

+

+

-

+

+

-

-

+

-

+

-

-

-

-

[TEMPORAL

SEQ] again [goal] sth lost/taken [goal spec] desirable state

recover regain

304

Appendix 4

Synsem interface of to come to have (cont.) [goal spec] health/money [goal spec] money

recuperate

+

+

-

+

+

-

recoup

-

+

-

-

-

-

-

+

-

+

+

-

-

+

-

+

+

-

[location]froma retrieve difficult place [instr] by paying redeem money

Notes 1.

2.

3.

4.

5. 6.

7. 8.

Paradoxically enough, this simplification of syntactic rales seems to be in opposition to one of the basic tenets of the Chomskyan paradigm, the Cartesian uniqueness of the Faculte de langage (Gonzalez Escribano 1991: 211-212). In GB several notational devices have been proposed to mark the distinction between the external argument and the internal arguments, e.g. underlining the external argument, the use of angle brackets around the internal argument, and finally others that only designate the direct argument (Grimshaw 1990: 2): (i) give: [+V, -N] THETAGRID: [Actor, Theme, Goal] (Radford 1988) (ii) see (A, Th) (Di Sciullo—Williams 1987: 29) (iii) put: χ (Rappaport—Levin 1986: 9) (iv) give (theme, goal) (v) put (theme, location) (Marantz 1984: 18). RRG makes a distinction between Aktionsart (the lexical aspect encoded within the lexical meaning of a verb) and the state of affairs, which refers to the lexical meaning of the verb or predicating element within the entire clause. The former is encoded within the lexical entry in the lexicon, whereas the latter determines the semantic representation of the core of a clause. Dik (1995a: 55) proposes a fifth layer (f). At this level reference is made to properties and relations and underlines the fact that predicates not only have a predicating function but also a referring function. Keizer (1992) and Hengeveld (1992) provide empirical evidence for such a proposal. For a further discussion of these two functionally oriented approaches to the structure of the clause, see Butler (1990). For the typology of operators, see Dik (1997a) and Hengeveld (1988; 1989a; 1990a). A classification of satellites can also be found in Dik— Hengeveld (1990). Term operators (Ω) (Rijkhoff 1990) have not been included here. There are some exceptions. Studies of the FG lexicon component

306

9.

10.

11.

12.

13.

14.

15.

16. 17.

Notes have been carried out by Meijs (1990), Schack Rasmussen (1994), Vossen (1995), and Weigand (1990). This is obviously in direct contrast to other theoretical frameworks that map the structures defined by predicates onto theory-specific initial syntactic structures, which can be based on constituency (e.g. GB, LFG), dependency (e.g. Hudson's Dependency Grammar), or relations (e.g. Perlmutter's Relational Grammar). De Groot (1989: 56) formulates a Demoted Agent, which is meant to account for the differences between the agent of causative constructions and the agent in other constructions. In line with this, Martin Mingorance (1985) also proposes a semantic function Adscriptum, which acts as a complement of the Zero semantic function expressed in the FG inventory. This is useful in the description of the adjectival lexicon because the term inserted into this slot always entails the assignment of a quality or feature. Schack Rasmussen (1994: 53-62) adopts a model of lexical decomposition similar to JackendofFs, and analyzes the interaction of semantic functions and the typology of SoAs with a special emphasis on the first argument functions. This claim has been discussed and rejected by Schack Rasmussen (1994), Garcia Velasco (1998), Faber—Mairal (1997), who claim that that meaning definitions should play a more active role in the explanation of syntactic and semantic regularities in the lexicon. This has been one of the major concerns of Schack Rasmussen (1994), who elaborates a system of meaning definitions in terms of a set of primitives or near primitives. She claims that this is not in contradiction with Dik's claim that meaning definitions are a language-internal affair. This methodological position is not compatible with the orthodox position in FG, which reduces morphology to an inventory of predicate formation rules. See Martin Mingorance (1985b) for an alternative proposal. Coseriu (1967, 1977, 1978, 1981; Geckeler, 1977, 1981) developed this initial theoretical framework for a paradigmatic organization of the lexicon. For a general discussion of the role of semantic parameters in the lexicon, see Martin Mingorance (1998b). More specifically, see Felices (1991) and Portero (1997) for an analysis of parameters, such as axiology and intensification. The following discussion is based on Hopper—Thompson (1973) and Ransom (1986). Note that transformations as such are dispensed with in FG (Dik

Notes 307 1997a: Chapter 1). This is a modified version of Faber—Perez Hernandez (1993, 1997).

References

Aitchison, Jean 1994

Words in the mind: An introduction to the mental lexicon. Oxford: Basil Blackwell. Alonso Ramos, Margarita—Agnes Tutin—Guy Lapalme 1995 "Lexical functions in the explanatory combinatorial dictionary for lexicalization in text generation.", in: P. Saint-Dizier—E. Viegas (eds.), 351-366. Altmann, Gerry 1997 The ascent ofBabel. Oxford: Oxford University Press. Amsler. R. A. 1980 The structure of the Merriam-Webster pocket dictionary. Technical Report TR- 164. University of Texas at Austin. Anderson, Stephen—Kiparsky, Paul (eds.) 1973 A festschrift for Morris Halle. New York: Holt, Rinehart and Winston. Alonge, Antonietta 1998 "Data on verb semantics in the EuroWordNet Database", in: A. Rubio —N. Gallardo—R. Castro—A. Tejada (eds.). Vol. 1, 385-389. Apresjan, Juri D. 1993 "Systemic lexicography as a basis of dictionary-making", Journal of the Dictionary Society of North America 14: 79-87. Aronoff, Mark 1976 Wordformation in generative grammar. Cambridge, MA: MET Press. Atkins, Beryl—J. Kegl—Beth Levin 1988 "Anatomy of a verb entry: From linguistic theory to lexicographic practice", International Journal of Lexicography 1: 84-126. Atkinson, R.C.—Shiffrin, R.M. 1968 "Human memory: A proposed system and its control processes", in: K.W. Spence—J.T. Spence (eds.), Vol. 2. Austin, John L. 1962 How to do things with words. Oxford: Clarendon Press.

References

309

Bach, Emmon—Robert T. Harms (eds.) 1968 Universals in linguistic theory. New York: Holt, Rinehart and Winston. Baddeley, A.D.—G. Hitch 1974 "Working memory", in: G. H. Bower (ed.) 47-90. Barsalou, Lawrence W. 1992 Cognitive psychology: An overview for cognitive scientists. Hillsdale NJ: Erlbaum. Bartlett, Frederic C. 1932 Remembering: A study in experimental and social psychology. Cambridge: Cambridge University Press. Beeman, Mark 1998 "Coarse semantic coding and discourse comprehension", in: M. Beeman— C. Chiarello (eds.), 255-284. Beeman, Mark—Christine Chiarello (eds.) 1998 Right hemisphere language comprehension. Mahway, NJ: Erlbaum. Bell, Roger 1991 Translation and translating. London: Longman. Bierwisch, Manfred—Karl Frich Heidolph 1970 Progress in linguistics. The Hague: Mouton. Bloomfield, Leonard 1933 Language. New York: Holt, Rinehart and Winston Bolkestein, Machtelt 1990 "Sentential complements in Functional Grammar; embedded predications, propositions, utterances in Latin", in: J. Nuyts—M. Bolkestein—C. Vet (eds.), 71100.

Bolkestein, Machtelt 1992 "Limits to layering: locability and other problems", in: M. Fortescue—P. Harder—L. Kristoffersen (eds.), 387407. Bower, G. H. 1974 The psychology of learning and motivation. Vol. 8. New York: Academic Press. Bransford, J. D.—J. J. Franks 1971 "The abstraction of linguistic ideas", Cognitive Psychology!·. 331-350. Bransford, J. D.—J. R. Barclay—J. J. Franks 1972 "Sentence memory: A constructive versus interpretive approach", Cognitive Psychology 3: 193-209.

310

References

Bresnan, Joan 1970

Bresnan, Joan 1979 Bresnan, Joan 1982

"On complementisers: Toward a syntactic theory of complement types", Foundations of Language 6: 297321. Theory of complementation in English syntax. New York: Garland. The mental representation of grammatical Cambridge, ΜΑ: ΜΓΓ Press.

relations.

Bühler, Karl 1934 Sprachtheorie [Language theory], Jena: Fischer. Butler, Christopher 1990 "Functional Grammar and Systemic Grammar: A preliminary comparison", Working papers in Functional Grammar 39. University of Amsterdam. Butt, Miriam—Wilhelm Geuder (eds.) 1998 The projection of arguments: Lexical and compositional factors. Stanford: CSU. Calzolari, Nicoletta 1994 "Issues for lexicon building", in: A. Zampolli—N. Calzolari—M. Palmer (eds.). Caramazza, Alfonso 1996 "The brain's dictionary", Nature 380: 485^86. Carter, Ronald 1988 "Compositionality and polysemy", in: B. Levin—C. Tenny (eds.), 167-204. Choi, Soonja—Melissa Bowerman 1991 "Learning to express motion events in patterns", in: B. Levin—S. Pinker (eds.), 83-122. Chomsky, Noam 1957 Syntactic structures. The Hague: Mouton. Chomsky, Noam 1965 Aspects of the theory of syntax. Cambridge, ΜΑ: ΜΓΓ. Press. Chomsky, Noam 1970 "Remarks on nominalization", in: R. Jacobs—P. S. Rosenbaum (eds ), 11-61. Chomsky, Noam 1973 "Conditions on transformations", in: S. Anderson—P. Kiparsky (eds.), 232-286.

References

311

Chomsky, Noam 1981a "On the representation of form and function", The Linguistic Review. 3-40. Chomsky, Noam 1981b Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam 1986 Knowledge of language: Its nature, origin and use. New York: Praeger. Cole, Peter—Jerry Morgan (eds.) 1975 Speech acts. New York: Academic Press. Connolly, John 1990 "Functional grammar and artificial intelligence", in: J. Connolly—S. C. Dik (eds.), 217-228. Connolly, John—Simon. C. Dik. (eds.) 1990 Functional Grammar and the computer. Dordrecht: Foris. Coseriu, Eugenio 1967 "Lexicalische Solidaritäten" [Lexical solidarities], Poetica 1: 293-303. Coseriu, Eugenio 1977 Principios de semantica estructural [Principles of structural semantics]. Madrid: Gredos. Coseriu, Eugenio 1978 Gramätica, semantica, universales [Grammar, semantics, universale]. Madrid: Gredos. Coseriu, Eugenio 1981 Lecciones de lingiiistica general [Lectures in general linguistics], Madrid: Gredos. Croft, William 1993 "The role of domains in the interpretation of metaphors and metonymies", Cognitive Linguistics 4 (4): 335-370. Cruse, Donald 1986 Lexical semantics. Cambridge: Cambridge University Press. Cunchillos Jaime, Carmelo—Francisco Ruiz de Mendoza (eds.) 1993 Adas del XV Congreso de AEDEAN [Proceedings of the XV Conference of AEDEAN], Logrono: University of Logrofio. Damasio, Antonio—Hanna Damasio 1992 "Brain and language", Scientific American 267 (3): 6371.

312

References

Damasio, Hanna—Τ. Grabowski—Daniel Tranel—Richard Hichwa— Antonio Damasio 1996 "A neural basis for lexical retrieval", Nature 380: 499505. D'Andrade, Roy 1987 "A folk model of the mind", in: D. Holland—N. Quinn (eds.), 112-148. Devriendt, B.—Louis Goossens—Johan van der Auwera (eds.) 1996 Complex structures: A functional perspective. Berlin: Mouton. Dik, Simon C. 1978a Functional Grammar. Dordrecht: Foris Publications. Dik, Simon C. 1978b Stepwise lexical decomposition. Lisse: de Ridder. Dik, Simon C. 1986 "Linguistically motivated knowledge representation", Working papers in Functional Grammar 9, University of Amsterdam. Dik, Simon C. 1990 "Relational reasoning in functional logic", in: J.C. Connolly—S.C. Dik (eds.), 273-288. Dik, Simon C. 1997a The theory of Functional Grammar : The structure of the clause. Berlin: Mouton de Gruyter. Dik, Simon C. 1997b The theory of Functional Grammar: Complex and derived constructions. Berlin: Mouton de Gruyter. Dik, Simon C.—Kees Hengeveld 1990 "The hierarchical structure of the clause and the typology of perception verb complements", Working papers in Functional Grammar 37, University of Amsterdam. Dik, Simon C.—Kees Hengeveld—E. Vester—Co Vet 1990 "The hierarchical structure of the clause and the typology of adverbial satellites", in: J. Nuyts—M. Bolkestein—C. Vet (eds.), 25-70. Di Sciullo, Anna-Maria—Edwin Williams 1987 On the definition of word. Cambridge, ΜΑ: ΜΓΓ Press. Dixon, Robert M. W. 1991 A new approach to English grammar on semantic principles. Oxford: Clarendon Press.

References

313

Dorr, Bonnie Jean 1993 Machine translation: A view from the lexicon. Cambridge ΜΑ: ΜΓΓ Press. Dowty, David 1979 Word meaning and Montague Grammar. Dordrecht: Riedel. Driver, Rosalind (ed.) 1985 Children's ideas in science. Milton Keynes: Open University Press. Droste, Flip G.—John Joseph (eds.) 1991 Linguistic theory and grammatical description. Amsterdam: Benjamins. Eikmeyer, Hans-Jürgen—Hannes Rieser (eds.) 1981 Words, worlds and contexts: New approaches in word semantics. Berlin: Walter de Gruyter. Emmorey, Karen—Victoria Fromkin 1987 "The mental lexicon", in: F. Newmeyer (ed.), Vol. Ill, 127-149. Endrushat, A.—Mario Vilela—Gerd Wotjak (eds.) 1994 Verbo e estruturas frasicas [Verb and phrase structures]. Oporto: Universidade. Engberg-Pedersen, Elisabeth—Lisbeth Falster Jakobsen—Lone Schack Rasmussen (eds.) 1994 Function and expression in Functional Grammar. Berlin. Mouton de Gruyter. Escalier, Marie Christine—Corinne Fournier 1997 'Towards a notional representation of meaning in the Meaning-Text Model: The case of the French SI", in L. Wanner (ed.), 25-55. Evens, Martha (ed.) 1988 Relational models of the lexicon: Representing knowledge in semantic networks. Cambridge: Cambridge University Press. Faber, Pamela 1991 "The relational structure of the lexicon: Its application to the field of sound verbs", in: J. Carmelo Cunchillos—F. Ruiz de Mendoza (eds.), 351-361. Faber, Pamela 1992 "The salience of negativity in semantic domains", in: J. Ruiz—P. Abad—J. Barrio (eds ), 291-297.

314

References

Faber, Pamela 1994a

Faber, Pamela 1994b

"Rasgos perceptuales en la estructuracion del lexico" [Perceptual features in lexical organization]. Verbo e Estructuras Frasicas: Rev. Fac. Letras, Linguas e Literaturas Anexo VI: 111-120.

"The semantic architecture of the lexicon", Lexicographica. Series Maior 57: 37-50. Faber, Pamela—Ricardo Mairal Uson 1994 "Methodological underpinnings for the construction of a functional lexicological model", Miscelanea. A Journal of English and American Studies 15: 193-217. Faber, Pamela—Ricardo Mairal Uson 1997a "The paradigmatic and syntagmatic structure of the lexical field of EXISTENCE in the elaboration of a semantic macronet", Studies in Language 21 (1): 119-154. Faber, Pamela—Ricardo Mairal Uson 1997b "Definitional analysis in the Functional-Lexematic Lexicographic Model", Alfinge 9: 219-232. Faber, Pamela—Ricardo Mairal Uson 1998a 'Towards a typology of predicate schemata in a Functional-Lexematic Model", in: G. Wotjak (ed.), 1137. Faber, Pamela—Ricardo Mairal Uson 1998b "Methodological criteria for the elaboration of a functional lexicon-based grammar of the semantic domain of cognitive verbs", in: H. Olbertz—K. Hengeveld—J. Sanchez Garcia (eds.), 3-24. Faber, Pamela—Ricardo Mairal Uson forthcoming Lexical rules. Faber, Pamela—Chantal Perez Hernandez 1993 "Image schemata and light: A study in contrastive lexical domains in English and Spanish", Atlantis XV 1-2: 117— 135. Faber, Pamela—Chantal Perez Hernandez 1997 "Image schemata and lexical oppositions in functional lexicography", Acta Universitatis Lodziensis: Folia Linguistica 36: 63-107.

References

315

Faber, Pamela—Jesüs Sanchez Garcia 1990 "Semäntica de prototipos: El campo semantico de los verbos que expresan la manera de habla frente al de los verbos de sonido en ingles y espanol" [Prototype semantics: A contrastive study of verbs of speech and sound in English and Spanish], Revista Espanola de Lingüistica Aplicada 6: 19-29. Farkas, Donka—Wesley Jacobsen—Karol Todrys (eds.) 1978 Papers from the parasession on the lexicon. Chicago: Chicago University Press. Felices, Angel 1991 El componente axiologico en el lenguaje. Su configuration en los adjetivos que expresan emotiones y conducta en la lengua inglesa [The axiological component in language and its configuration in English adjectives of emotions and behavior], Granada: University of Granada. Fellbaum, Christiane 1990 "English verbs as a semantic net", International Journal of Lexicography 3 (4): 279-301. Fellbaum, Christiane (ed.) 1998a WordNet: An electronic lexical database. Cambridge, ΜΑ: ΜΓΓ Press. Fellbaum, Christiane 1998b Introduction, in: C. Fellbaum (ed.), 1-20. Fellbaum, Christiane 1998c "A semantic network of English verbs", in: C. Fellbaum (ed.), 69-104. Fellbaum, Christiane 1998d Review of Y. Wilks—B. Slator—L. Guthrie 1996, International Journal of Lexicography 11 (3): 238-241. Fillmore, Charles 1968 "The case for case", in: E. Bach—R. Harms (eds ), 1-68. Fillmore, Charles 1985 "Frames and the semantics of understanding", Quademi di semantica VI: 222-254. Fillmore, Charles—Beryl Atkins 1991 'Towards a frame-based lexicon: the semantics of RISK and its neighbors", in: E. Kittay—A. Lehrer (eds.), 75102.

316

References

Fillmore, Charles—Beryl Atkins 1998 "FrameNet and lexicographic relevance", in: A. Rubio— N. Gallardo—R. Castro—A. Tejada (eds), Vol. 1, 41723. Fodor, Jerry 1970 "Three reasons for not deriving 'kill' from 'cause to die'", Linguistic Inquiry 1: 429^438. Fodor, Jerry 1977 Semantics: Theories of meaning in generative grammar. Sussex: Harvester Press. Foley, William—Robert Van Valin Jr. 1984 Functional syntax and universal grammar. Cambridge: Cambridge University Press. Fortescue, Michael—Peter Harder—Lars Kristoffersen (eds.) 1992 Layered structure and reference in a functional perspective. Amsterdam: Benjamins. Fromkin, Victoria 1997 "Some thoughts about the brain/mind/language interface", Lingua 100: 3-27. Garcia Velasco, Daniel 1998 La predication nuclear en la gramatica funcional de S C. Dik: una perspectiva conceptual. Unpublished Doctoral Dissertation. University of Oviedo. Geckeier, Horst 1977 Sermntica estructural y teoria del campo Ιέχίοο [Structural semantics and lexical field theory], Madrid: Gredos. Geckeier, Horst 1981 "Structural semantics", in: H. Eikmeyer—H. Rieser (eds.). Geeraerts, Dirk 1989 Wat er in een woord zit. Facetten van de lexicale semantick. [What is in a word. Aspects of lexical semantics]. Leuven: Peeters. Geeraerts, Dirk 1995 "Representation formats in cognitive semantics", Folia Linguistica XXJXJl-2: 21-41. Givön, Talmy 1980 "The binding hierarchy and the typology of complements", Studies in Language 4(1).

References Givön, Talmy 1984 Givön, Talmy 1989 Givön, Talmy 1990

317

Syntax: A functional-typological introduction. Vol. I. Amsterdam: Benjamins. Mind, code and context: Essays in pragmatics. London: Erlbaum. Syntax: A functional-typological introduction. Vol. II. Amsterdam: Benjamins.

Givön, Talmy 1995 Functionalism and grammar. Amsterdam: Benjamins. Gleitman, Lila—Henry Gleitman 1997 "What is language made out of?", Lingua 100: 29-55. Goddard, Cliff 1998 Semantic analysis: A practical introduction. Oxford: Oxford University Press. Goldberg, Adele 1995 A Construction Grammar approach to argument structure. Chicago: University of Chicago Press. Goldberg, Adele 1996 "Jackendoff and construction-based grammar", Cognitive Linguistics 7-1: 3-19. Gonzalez Escribano, Jose Luis 1991 Una teoria de la oracion [A theory of the sentence]. Oviedo: University of Oviedo. Grandy, Richard 1987 "In defense of semantic fields", in: E. LePore (ed.), 259280.

Grandy, Richard 1992 Grimshaw, Jane 1979

"Semantic fields, prototypes and the lexicon", in: E. Kittay—A. Lehrer (eds.), 103-122. "Complement selection and the lexicon", Linguistic Inquiry 10.2: 279-326.

Grimshaw, Jane 1990 Argument structure. Cambridge, ΜΑ: ΜΓΓ Press. Grimshaw, Jane 1994 "Lexical reconciliation", Lingua 92: 411-430. Groot, Casper de 1987 "Predicate Formation in Functional Grammar" Working papers in Functional Grammar 20.

318

References

Groot, Casper de 1989 Predicate structure in a functional grammar of Hungarian. Dordrecht: Foris. Gruber, Jeffrey 1965 Lexical structures in syntax and semantics. New York: North-Holland. Guesne, Edith 1985 "Light", in: R. Driver (ed.), 10-32. Halliday, Michael A. K. 1994 An introduction to Functional Grammar. London: Arnold. Hartmann, Reinhart (ed.) 1984 LEX'eter' 83: Proceedings of the International Conference on Lexicography. Tübingen: Niemeyer. Hellige, Joseph 1993 Hemispheric asymmetry: What's right and what's left. Cambridge MA: Harvard University Press. Hengeveld, Kees 1987 "Clause structure and modality in Functional Grammar", in: J. van der Auwera—L. Goossens (eds.), 53-66. Hengeveld, Kees 1988 "Elocution, mood and modality in a functional grammar of Spanish", Journal of Semantics 6: 227-269. Hengeveld, Kees 1989a "Layers and operators in Functional Grammar", Journal of Linguistics 25: 127-157. Hengeveld, Kees 1989b "Illocution, mood and modality in a functional grammar of Spanish", Journal of Semantics 6: 227-269. Hengeveld, Kees 1990a "The hierarchical structure of utterances", in: J. Nuyts— M. Bolkestein—C. Vet (eds.), 1-23. Hengeveld, Kees 1990b "Semantic relations in non-verbal predications", in: J. Nuyts—M. Bolkestein—C. Vet (eds.), 101-122. Hengeveld, Kees 1992 "Parts of speech", in: M. Fortescue—P. Harder—L. Kristoffersen (eds.), 29-55. Hengeveld, Kees 1996 "The internal structure of adverbial clauses", in: B. Devriendt.—L. Goossens—J. van der Auwera (eds.), 119-147.

References

319

Hoinkes, Ulrich (ed.) 1995 Panorama der Lexikalischen Semantik [Panorama of lexical semantics] Tübingen: Gunter Narr. Holland, Dorothy—Naomi Quinn (eds.) 1987 Cultural models in language and thought. Cambridge: Cambridge University Press. Honnech, Richard P.—Robert R. Hoffmann (eds.) 1980 Cognition and figurative language. Hillsdale, N.J.: Erlbaum. Hopper, Paul J.—Sandra Thompson 1973 "On the applicability of root transformations", Linguistic Inquiry 4 (4): 465-498. Hopper, Paul J.—Sandra Thompson 1980 'Transitivity in grammar and discourse", Language 56.2: 251-299. Hudson, Richard 1976 Arguments for a non-transformational grammar. Chicago: University of Chicago Press. Ingria, R—B. Boguraev—James Pustejovsky 1992 "Dictionary/Lexicon", in: S C. Shapiro (ed.), 341-365. Iris, Madelyn—Bonnie Litowitz—Martha Evens 1988 "Problems of the part-whole relation", in: M. Evens (ed.), 261-287. JackendofF, Ray 1972 Semantic interpretation in generative grammar. Cambridge ΜΑ: ΜΓΓ Press. JackendofF, Ray 1975 "Morphological and semantic regularities in the lexicon", Language 51: 639-671. JackendofF, Ray 1983 Semantics and cognition. Cambridge, ΜΑ: ΜΓΓ Press. JackendofF, Ray 1987 "The status of thematic relations in linguistic theory", Linguistic Inquiry 18: 369-411. JackendofF, Ray 1990 Semantic structures. Cambridge, ΜΑ: ΜΓΓ Press. JackendofF, Ray 1992 "Parts and boundaries", in: B. Levin—S. Pinker (eds.), 9 46.

320

References

Jackendoff, Ray 1992 Jackendoff, Ray 1993

Languages of the mind: Essays on mental representation. Cambridge, ΜΑ: ΜΓΓ Press. Patterns in the mind: Language and human nature. New York: Basic Books.

Jackson, Howard 1988 Words and their meaning. London: Longman. Jacobs, Roderick—Peter Rosenbaum (eds.) 1970 English transformational grammar. Waltham, MA: Ginn. Jimenez Hurtado, Catalina 1994 El componente pragrmtico en el lexicon verbal del espanol, aleman e ingles [The pragmatic component in the Spanish-German-English verb lexicon], [Unpublished PhD thesis]. University of Granada. Johnson, Mark 1987 The body in the mind. Chicago: University of Chicago Press. Karttunen, Lauri 1971 "Implicative verbs", Language 47 (2): 340-358. Katz, Jerrold 1972 Semantic theory. New York: Harper and Row. Katz, Jerrold—Jerry Fodor 1963 "The structure of a semantic theory", Language 39: 170— 210.

Katz, Jerrold—Paul Postal 1964 An integrated theory of linguistic description. Cambridge, ΜΑ: ΜΓΓ Press. Kay, Paul 1997 Words and the grammar of context. Stanford: CSLI. Keenan, Edward L. 1970 "A logical base for a transformational grammar of English", in: Transformation and discourse analysis papers 82. Philadelphia: University of Pennsylvania. Keizer, M. Evelien 1992 "Predicates as referring expressions", in: M. Fortescue— P. Harder—L. Kristoffersen (eds.), 1-27. Kerbrat-Orecchioni, C. 1980 L'enonciation—De la subjectivite dans le langage [The utterance—On subjectivity in language], Paris: Colin.

References

321

Kiparsky, Paul—Carol Kiparsky 1970 "Fact", in: M. Bierwisch—K.E. Heidolph (eds.), 143-173. Kittay, Eva 1987 Metaphor: Its cognitive force and linguistic structure. Oxford: Oxford University Press. Kittay, Eva 1993 "Semantic fields and the individuation of content", in: E. Kittay—A. Lehrer (eds.), 229-252. Kittay, Eva—Adrienne Lehrer (eds.) 1992a Frames, fields and contrasts. Hillsdale, NJ: Erlbaum. Kittay, Eva—Adrienne Lehrer 1992b Introduction, in: E. Kittay—A. Lehrer (eds.), 1-20. Kosslyn, Stephen 1980 Image and mind. Cambridge, MA: Harvard University Press. Kreszowski, Tomasz P. 1990 "The axiological aspect of idealized cognitive models", in: J. Tomaszczyk—B. Lewandowska-Tomaszczyk (eds ), 135-166. Lakoff, George 1987 Women, fire and dangerous things. Chicago: University of Chicago Press. Lakoff, George— Johnson, Mark 1980 Metaphors we live by. Chicago: University of Chicago Press. Landau, Sidney 1984 Dictionaries: The art and craft of lexicography. Cambridge: Cambridge University Press. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar. Stanford: Stanford University Press. Langacker, Ronald W. 1991a Concept, image and symbol: The cognitive basis of grammar. Berlin: Mouton. Langacker, Ronald W. 1991b "Cognitive grammar", in: F. Droste and J. Joseph (eds ), 275-306. Langacker, Ronald W. 1997 The contextual basis of cognitive semantics, in: J. Nuyts— E. Pederson (eds.), 227-252.

322

References

Langendoen, D. Terence 1970 "The 'can't seem to' construction", Linguistic Inquiry 1 (1): 25-36. Lappin, Shalom (ed.) 1996 The handbook of contemporary semantic theory. Oxford: Blackwell. Lehrer, Adrienne 1974 Semantic fields and lexical structures. Amsterdam: North-Holland. Lehrer, Adrienne 1992 "Names and naming: Why we need fields and frames", in: E. Kittay—A. Lehrer (eds.), 123-142. LePore, Ernest (ed.) 1987 New directions in semantics. London: Academic Press. Levelt, Willem 1989 Speaking: From intention to articulation. Cambridge MA: ΜΓΓ Press. Levin, Beth 1991 "Building a lexicon: The contribution of linguistics", International Journal of Lexicography A (3): 205-226. Levin, Beth 1993 English verb classes and alternations: A preliminary investigation. Chicago: University of Chicago Press. Levin, Beth 1995 "Approaches to lexical semantic representation", in: D. Walker—A. Zampolli—N. Calzolari (eds ), 53-91. Levin, Beth—Malka Rappaport 1986 "The formation of adjectival passives", Linguistic Inquiry 17,4:623-661. Levin, Beth—Malka Rappaport 1995 Unaccusatives: At the syntax-lexical semantics interface. Cambridge, ΜΑ: ΜΓΓ Press. Levin, Beth—Malka Rappaport 1996 "Lexical semantics and syntactic structure", in: S. Lappin (ed.), 487-507. Levin, Beth—Steven Pinker (eds.) 1992 Lexical and conceptual semantics. Oxford: Blackwell. Levin, Beth—C. Tenny (eds.) 1988 On linking. Papers by Richard Carter. ΜΓΓ Lexicon Project Working Paper 25. Department of Linguistics and Philosophy. Cambridge ΜΑ: ΜΓΓ Press.

References Levin, Stephen 1997

323

"From outer to inner space: Linguistic categories and nonlinguistic thinking", in: J. Nuyts—E. Pederson (eds ), 1345. Lewandowska-Tomaszczyk, Barbara 1987 Conceptual analysis, linguistic meaning, and verbal interaction Ledz: Acta Universitatis Lodziensis. Lyons, John 1977 Semantics. Vols. 1-2. London: Cambridge University Press. Mairai, Ricardo 1993 Complementation patterns of cognitive, perception and speech act verbs in the English language. A functionalcognitive approach. [Unpublished PhD thesis]. University of Zaragoza, Spain. Mairai, Ricardo 1994 "Parämetros para la organization de una Sintaxis Lexica Funcional" [Parameters for the organization of a lexical functional syntax], in: J. Martin Arista (ed.), 23-73. Marr, David 1982 Vision. San Francisco: Freeman. Martin Arista, Javier (ed.) 1994 Estudios de Grarmtica Funcional [Studies in Functional Grammar]. Zaragoza: Mira. Martin Mingorance, Leocadio 1984 "Lexical fields and stepwise lexical decomposition in a contrastive English-Spanish verb valency dictionary", in: R. Hartmann (ed.), 226-236. Martin Mingorance, Leocadio 1985a "La semantica smtagmatica del adjetivo. Parämetros para la organizaciön de un lexicon ingles/espanol de Valencias adjetivales" [The syntagmatic semantics of adjectives. Parameters for the organization of an English-Spanish lexicon of adjective valencies], in: R. Monroy-Casas (ed.), 329-340. Martin Mingorance, Leocadio 1985b "Bases metodologicas para un estudio contrastivo del lexico derivado" [Methodological underpinnings for a contrastive study of word derivations], Revista Espanola de Lingüistica Aplicada 1: 37-54.

324

References

Martin Mingorance, Leocadio 1987 "Classematics in a Functional-Lexematic grammar of English", in: C. Olivares (ed.), 377-382. Martin Mingorance, Leocadio 1989 "Functional Grammar and Lexematics", in: J. Tomaszczyk—B. Lewandowska-Tomaszczyk (eds.), 227-253. Martin Mingorance, Leocadio 1995 "Lexical logic and structural semantics: methodological underpinnings in the structuring of a lexical database for natural language processing", in: U. Hoinkes (ed.), 461474. Martin Mingorance, Leocadio 1998a "Semes, semantic classemes, and dimensions: The lexicological and lexicographic perspectives", [paper presented at the XIV International Congress of Linguists, Berlin 1987], in: Martin Mingorance (1998c), 209-224. Martin Mingorance, Leocadio 1998b "Pragmatic features in the lexicon of a Functional Grammar" [paper presented at the International Pragmatics Conference, Amberes 1987], in: Martin Mingorance (1998c), 91-100], Martin Mingorance, Leocadio 1998c El modelo lexermtico-funcional [The FunctionalLexematic Model], Marin Rubiales, Amalia (ed.), Granada: University of Granada McCawley, James 1968 "The role of semantics in grammar", in: E. Bach—R. Harms (eds.), 124-169. Meijs, Willem 1990 "Spreading the word: knowledge-activation in a functional perspective", in: J. Connolly—S.C. Dik (eds ), 201-217. Meijs, Willem—Piek Vossen 1992 "In so many words: Knowledge as a lexical phenomenon", in: J. Pustejovsky—S. Bergler (eds.), 137— 153. Mel'cuk, Igor 1981 "Meaning-Text Models: A recent trend in Soviet linguistics", Annual Review ofAnthropology 10, 27-62.

References Mel'cuk, Igor 1988

Mel'cuk, Igor 1989

Mel'cuk, Igor 1996

325

"Semantic description or lexical units in an explanatory combinatorial dictionary: Basic principles and heuristic criteria", International Journal of Lexicography 1 (3): 165-188. "Semantic primitives from the viewpoint of the MeaningText Linguistic Theory", Quademi di Semantica 10 (1): 65-102.

"Lexical functions: A tool for the description of lexical relations in the lexicon", in L. Wanner (ed.), 37-71. Mel'cuk, Igor—A. K. Zholkovskij 1970 'Towards a functioning Meaning-Text Model of language", Linguistics 57: 10-47. Mel'cuk Igor—N. Arbatchewsky-Jumarie—L. Elnitsky—L. Iordanskaja— A. Lessard 1984 Dictionnaire explicatif et combinatoire du franqais contemporain. Recherches lexico-semantiques I. [An explanatory combinatorial dictionary of modern French: Lexical semantic studies]. Montreal: University of Montreal Press. Mel'cuk, Igor—Leo Wanner 1996 "Lexical functions and lexical inheritance for Emotion lexemes in German", in: L. Wanner (ed.), 209-227. Melka, Francine—Jan Schroten 1992 "On semantic representation", Cahiers de Lexicologie 60 (1): 37-50. Melton, A.W.—E. Martin (eds.) 1972 Coding processes in human memory. Washington D C.: Winston. Miller, George A. 1972 "English verbs of motion: A case study in semantics and lexical memory", in: A. W. Melton—E. Martin (eds.). Miller, George, A. 1986 "Dictionaries in the mind", Language and cognitive processes 1 (3): 171-185. Miller, George A. 1990 "Linguists, psychologists and the cognitive sciences", Language 66 (2): 317-322.

326

References

Miller, George A. 1998 "Nouns in WordNet", in: Christiane Fellbaum (ed.), 2346. Miller, George A.—Phillip Johnson-Laird 1976 Language and perception. Cambridge, MA: Harvard University Press. Miller, George A — R. Beckwith—C. Fellbaum—D. Gross—K. Miller—R. Tengi (eds.). 1993 Five papers on WordNet. CSL Report 43, July 1990, revised March 1993. Miller, George A.—P. M. Gildea 1987 "How children learn words", Scientific American 257: 9499. Miller, George A.—Christiane Fellbaum 1991 "Semantic networks of English", Cognition 41: 197-229. Monroy-Casas, Rafael (ed.) 1985 Actas del II Congreso Nacional de la Asociacion Espanola de Lingüistica Aplicada [Proceedings of the II Nacional Conference of the Spanish Association of Applied Linguistics]. Madrid: Sociedad General Espanola de Libreria. Montague, Richard 1970a "Universal Grammar", in: R. Thomason (ed.), 222-246. Montague, Richard 1970b "English as a formal language", in: R. Thomason (ed.), 188-221.

Montague, Richard 1973 "The proper treatment of quantification in ordinary English", in: R. Thomason (ed ), 247-270. Moreno Ortiz, Antonio 1998 Disefio e implementation de un lexicon computacional para lexicografia y traduction automätica [The design and implementation of a computational lexicon for lexicography and machine translation], [Unpublished PhD thesis]. University of Cordoba, Spain. Mufwene, Salikoko S. 1978 "English manner-of-speaking verbs revisited", in: D. Farkas—W. Jacobsen—K. Todrys (eds.): 287-288. Nelson, Katherine 1985 Making sense: The acquisition of shared meaning. Orlando: Academic Press.

References

327

Newmeyer, Frederick 1980 Linguistic theory in America. New York: Academic Press. Newmeyer, Frederick (ed.). 1988 Linguistics: The Cambridge survey. Vols I-IV. Cambridge: Cambridge University Press. Nida, Eugene A. 1975 Componential analysis. The Hague: Mouton. Nirenburg, Sergei 1987a "Knowledge and choices in machine translation", in: S. Nirenburg (ed.): 1-15. Nirenburg, Sergei (ed.) 1987b Machine translation: Theoretical and methodological issues. Cambridge: Cambridge University Press. Nowakowski, Miroslaw. 1990 "Metaphysics of the dictionary versus the lexicon", in: J. Tomaszczyk—B. Lewandowska-Tomaszczyk (eds.), 320.

Nuyts, Jan 1990

Nuyts, Jan 1992

"Linguistic representation and conceptual knowledge representation", in: J. Nuyts—M. Bolkestein—C. Vet (eds.), 263-293.

Aspects of a cognitive-pragmatic theory of language. Amsterdam: Benjamins. Nuyts, Jan—Machtelt Bolkestein—Co Vet (eds.) 1990 Layers and levels of representation. Amsterdam: Benjamins. Nuyts, Jan—Eric Pederson (eds.) 1997 Language and conceptualization. Cambridge: Cambridge University Press. Olbertz, Hella—Kees Hengeveld—Jesus Sanchez Garcia (eds.) 1998 The structure of the lexicon in Functional Grammar. Amsterdam: Benjamins. Olivares, Carmen (ed.) 1987 Actas del X Congreso de la Asociacion Espanola de Estudios Anglo-Norteamericanos [Proceedings of the X Conference of the Spanish Association of AngloAmerican Studies], Zaragoza: University of Zaragoza. Osgood, Charles 1980 "The cognitive dynamics of synaesthesia and metaphor", in: R. P. Honnech—R. R. Hoffinann (eds ), 203-238.

328

References

Osgood, Charles—George Suci—Percy Tannenbaum 1957 The measurement of meaning. Urbana, Illinois: Illinois University Press. Paducheva, Elena 1997 "Verb categorization and the format of a lexicographic definition (semantic types of causative relations)", in: L. Wanner (ed.), 61-74. Paivio, Allan Imagery and verbal processes. New York: Holt, Rinehart 1972 and Winston. Paivio, Allan 1991 Images in mind. New York: Harvester. Palmer, Martha—Alain Polguere 1995 "A preliminary lexical and conceptual analysis of BREAK: A computational perspective", in: P. Saint-Dizer—E. Viegas (eds.), 2 3 1 - 2 5 0 . Pederson, Eric—Jan Nuyts 1997 "Overview: On the relationship between language and conceptualization", in: J. Nuyts—E. Pederson (eds.), 112.

Perlmutter, David Deep and surface constraints in syntax. New York: Holt, 1971 Rinehart and Winston. Pinker, Steven 1989 Leamability and cognition: The acquisition of argument structure. Cambridge MA: Harvard University Press. Pinker, Steven 1994a "How could a child use verb syntax to leam verb semantics?" Lingua 92: 377-410. Pinker, Steven The language instinct: How the mind creates language. 1994b New York: William Morrow. Pinker, Steven 1997 How the mind works. New York: Norton. Portero, Maria del Carmen 1997 Intensiflcacion: Estudio de un clasema [Intensification: Study of a classeme] [Unpublished PhD Thesis] University of Cordoba, Spain.

References Pottier, Bernard 1963

329

"Vers une semantique moderne" [Towards a modern semantics], Travaux de linguistique et de litterature 2 (1): 107-137. Pustejovsky, James 1991 "The generative lexicon", Computational Linguistics 17 (4): 409-441. Pustejovsky, James 1995 The generative lexicon. Cambridge, ΜΑ: ΜΓΓ Press. Pustejovsky, James—Sabine Bergler (eds.) 1992 Lexical semantics and knowledge representation. Berlin: Springer. Radford, Andrew 1988 Transformational grammar. A first course. Cambridge: Cambridge University Press. Ransom, Evelyn 1986 Complementation: Its meanings and forms. Amsterdam: Benjamins. Rappaport Hovac, Malka—Beth Levin 1988 "What to do with theta-roles", in: W. Wilkins (ed.), 7-36. Rappaport Hovac, Malka—Beth Levin 1998 "Building Verb Meanings", in: M. Butt—W.Geuder (eds.), 97-134. Ravin, Yael 1992 Lexical semantics without thematic roles. Oxford: Oxford University Press. Rijkhoff, Jan 1989 'Toward a unified analysis of terms and predications", in: J. Nuyts—M. Bolkestein—C. Vet (eds.), 165-192. Rosch, Eleanor 1978 "Principles of categorization", in: E. Rosch—B.B. Lloyd (eds.). Rosch, Eleanor and B.B. Lloyd (eds.) 1978 Cognition and categorization. Hillsdale, NJ: Erlbaum. Rosenbaum, P. S. 1967 The grammar of English predicate complement constructions. Cambridge, ΜΑ: ΜΓΓ Press.

330

References

Rubio, Antonio—Natividad Gallardo—Rosa Castro—Antonio Tejada (eds.) 1998 Proceedings of the First National International Conference on Language Resources and Evaluation. Vols. 1-2. Paris: Elra. Ruhl, Charles 1989 On monosemy: A study in linguistic semantics. Albany: State of University of New York Press. Ruiz, Jose—Pilar Abad— Jose Barrio (eds.) 1994 Actas del XVI Congreso de AEDEAN (Proceedings of the XVI Conference of AEDEAN). Valladolid: University of Valladolid. Rupp, Rebecca 1998 Committed to memory: How we remember and why we forget. New York: Crown. Saeed, John 1997 Semantics. Oxford: Blackwell. Saint-Dizier, Patrick—Evelyne Viegas (eds.) 1995a Computational lexical semantics. Cambridge: Cambridge University Press. Saint-Dizier, Patrick—Evelyne Viegas 1995b "An introduction to lexical semantics from a linguistic and psycholinguistic perspective", in: P. Saint-Dizier—E. Viegas (eds.), 1-30. Sanchez de Zavala, V. 1976 Semantica y sintaxis en la lingüistica tranformatoria [Semantics and syntaxis in transformational linguistics], Madrid. Alianza Universidad. Saussure, Ferdinand de 1916 Cours de linguistique gemrale [Course in general linguistics]. Paris: Payot. 1990 [reprinted Paris: Payot] Schack Rasmussen, Lone 1994 "Semantic functions in perspective - reconsidering meaning definitions", in: E. Engberg-Pedersen, L. Falster Jakobsen, L. Schack Rasmussen (eds.), 41-63. Schank, Roger 1973 "Identification of conceptualizations underlying Natural Language", in: R. Schank—Κ. M. Colby (eds.), 187-247.

References

331

Schänk, Roger 1982

Dynamic memory: A theory of reminding and learning in computers and people, Cambridge: Cambridge University Press. Schank, Roger—Robert Abelson 1977 Scripts, plans, goals and understanding. Hillsdale, NJ: Erlbaum. Schank, Roger—Κ. M. Colby (eds.) 1973 Computer models of thought and language. San Francisco, CA: Freeman. Searle, John 1975 "Indirect speech acts", in: P. Cole—J. Morgan (eds.): 5982.

Sells, P. 1985

Lectures on contemporary syntactic theories: An introduction to Government and Binding Theory, Generalized Phrase Structure Grammar and Lexical Functional Grammar. Stanford: CSLI.

Shapiro, Stuart C. 1992 Encyclopedia of artificial intelligence. New York: John Wiley. Shopen, Thomas (ed.) 1986 Language typology and syntactic description. Vols. 1-3. Cambridge: Cambridge University Press. Siewierska, Anna 1991 Functional Grammar. London: Routledge. Spence, K.W.—J. T. Spence (eds.) 1968 Advances in the psychology of learning and motivation. New York: Academic Press. Talmy, Leonard 1985 "Lexicalization patterns: Semantic structure in lexical forms", in: T. Shopen (ed.), Vol. 3, 57-149. Taylor, John R. 1989 Linguistic categorization: Prototypes in linguistic theory. Oxford: Clarendon Press. Thomason, R. (ed.) 1974 Formal philosophy. Selected papers of Richard Montague. New Haven: Yale University Press. Tomaszczyk, Jerzy—Barabara Lewandowska-Tomaszczyk (eds.) 1990 Meaning and lexicography. Amsterdam: Benjamins.

332

References

Trier, J. 1934 Tulving, E. 1972

"Das sprachliche Feld", in: Neue Jahrbücher für Wissenschaft und Jungenbildung 10: 428-449. "Episodic and semantic memory", in: Ε. Tulving—W. Donaldson (eds.), 381^03.

Van der Korst, B. 1991 "Functional Grammar and machine translation", in J. Connolly—Simon C. Dik (eds.), 289-316. Van Valin, Robert D. 1990 "Layered syntax in Role and Reference Grammar", in: J. Nuyts—Machtelt Bolkestein—Co Vet (eds.), 193-232. Van Valin, Robert D. (ed.) 1993a Advances in Role and Reference Grammar. Amsterdam: Benjamins. Van Valin, Robert D. 1993b "Synopsis of Role and Reference Grammar", in: R. D. Van Valin (ed.), 1-164. Van Valin, Robert D — Randy LaPolla 1997 Syntax: Structure, meaning and function. Cambridge: Cambridge University Press. Van Valin, Robert D — D. Wilkins 1994 "Predicting syntax from semantics", in: R. D. Van Valin (ed.), 499-534. Vendler, Zeno 1967 Linguistics in philosophy. Cornell: Cornell University Press. Vossen, Piek 1995 Grammatical and conceptual individuation in the lexicon. Amsterdam: IFOTT. Vossen, Piek—Laura Bloksma 1998 "Categories and classifications in EuroWordNet", in: A. Rubio—N. Gallardo—R. Castro—A. Tejada (eds ). Vol. 1: 399-407. Walker, Don—Antonio Zampolli—Nicoletta Calzolari (eds.) 1995 Automating the lexicon I: Research and practice in a multilingual environment. Oxford: Oxford University Press. Wanner, Leo (ed.) 1996a Lexical functions in lexicography and natural language processing. Amsterdam: Benjamins.

References Wanner, Leo 1996b Wanner, Leo 1997

333

"Introduction", in: L. Wanner (ed.), 1-36.

Recent trends in meaning-text theory. Amsterdam: Benjamins. Waxman, Sandra R. 1994 "The development of an appreciation of specific linkages between linguistic and conceptual organization", Lingua 92:229-257. Wiegand, Hans 1990 Linguistically motivated principles of knowledge base systems. Dordrecht: Foris. Wierzbicka, Anna 1980 Lingua mentalis. Sydney/New York: Academic Press. Wierzbicka, Anna 1987 English speech act verbs: A semantic dictionary. New York: Academic Press. Wierzbicka, Anna 1988 The semantics ofgrammar. Amsterdam: Benjamins. Wierzbicka, Anna 1992a "In search of tradition: The semantic ideas of Leibniz", Lexicographical: 10-25. Wierzbicka, Anna 1992b "Semantic primitives and semantic fields", in: E. Kittay— A. Lehrer (eds.), 209-228. Wierzbicka, Anna 1993 "What are the uses of theoretical lexicography?", Journal of the Dictionary Society of North America 14: 44-78. Wierzbicka, Anna 1995 "Universal semantic primitives as a basis for lexical semantics", Folia Lingüistica XKXJ1-2: 149-169. Wierzbicka, Anna 1996 Semantics: Primes and universals. Oxford: Oxford University Press. Wilkins, W. (ed.) 1988 Syntax and semantics 21. Thematic relations. New York: Academic Press. Wilks, Yorick A.—Brian Slator—Lousie Guthrie 1996 Electric words. Cambridge ΜΑ: ΜΓΓ Press.

334

References

Williams, Ε. S. 1981

"On the notions 'lexically related' and head of a word'". Linguistic Inquiry 12: 245-274. Wingfield, Arthur— Debra Titone 1998 "Sentence processing", in: J. Gleason—N. Ratner (eds.), 227-276. Wotjak, Gerd 1987 "La semantique lexicale: etat actuel et perspectives" [Lexical semantics: present and future perspectives]. Linguistische Studien 166: 45-85. Wotjak, Gerd 1994a "Actantes y circunstantes y sus correspondencias semanticas en verbos" [Actants and circumstantials and their semantic correspondences in verbs] in: A. Endrushat—M. Vilela—G. Wotjak (eds ), 71-91. Wotjak, Gerd 1994b "El potencial comunicativo de las unidades lexicas" [The communicative potential of lexical units]. Vozy letra V/l: 155-173. Wotjak, Gerd 1994c "Escenas cognitivas y significado lexico sistemico y contextual" [Cognitive scenes and systemic and contextual lexical meaning], Les cahiers du C.RI.A.R 14: 235-254. Wotjak, Gerd (ed.) 1998a Towards a functional lexicology. Frankfurt: Peter Lang. Wotjak, Gerd 1998b "El cuadro predicativo nuclear sememico como punto de intersection entre eje paradigmätico y eje sintagmätico asi como entre semantica lexica y combinatoria sintactica y cognition" [The nuclear sememic predicate frame as an intersection point for the paradigmatic and syntagmatic axes as well as for lexical semantics, syntactic potential, and cognition] in: G. Wotjak (ed.) (1998a), 57-92. Wotjak, Gerd (ed.) 1998c Teoria del campo y semantica lexica [Semantic field theory and lexical semantics], Frankfurt: Peter Lang. Zampolli, Antonio—Nicoletta Calzolari—Martha Palmer (eds.). 1994 Current issues in computational linguistics: In honour of 7 Don Walker. Linguistica Computazionale Vol. ΓΧ-Χ. Pisa: Giardini.

References Zgusta, Ladislav 1971

335

Manual of lexicography. The Hague: Mouton.

List of Dictionaries

The American Heritage Dictionary of the English Language. New York: Houghton Mifflin, 1992. Cambridge International Dictionary of English. Cambridge: Cambridge University Press, 1995. Collins CobuildEnglish Language Dictionary. London: Collins, 1987. Longman Dictionary of Contemporary English. Burnt Mill, Harlow, Essex: Longman, 1987. Longman Dictionary of English Language and Culture. Burnt Mill, Harlow, Essex: Longman, 1992. Longman Lexicon of Contemporary English. Burnt Mill, Harlow, Essex: Longman, 1981. Oxford Advanced Learners' Dictionary. Oxford: Oxford University Press, 1989. The Random House Dictionary of the English Language (RHD). New York: Random House, 1987. Roget's International Thesaurus. Harper & Row: New York, 1984. Webster's New World Dictionary for Young Readers. New York: Simon & Schuster, 1983.

Author index

Abelson, 74, 212 Aitchison, 81 Alonge, 80, 91 Alonso Ramos, 13, 14 Altmann, 205 Amsler, 83 Apresjan, 13, 65, 85, 254 Aronoff, 7 Atkins, 73, 74, 78, 91, 94, 241 Atkinson, 207 Austin, 40

Damasio, Α., 205,212 Damasio, H., 57, 205, 206, 212 D'Andrade, 63 Di Sciullo, 305 Dik, 1-3, 16, 22, 23, 35-66, 88, 115, 117, 118, 120, 130, 134, 135, 144, 176, 177, 203, 210, 213, 275, 305-307 Dixon, 143 Dorr, 72, 114 Dowty, 73

Baddeley, 207 Barclay, 16 Barsalou, 208,211,212 Bartlett, 212 Beeman, 67 Bell, 60 Bloksma, 86 Bloomfield, 4, 5 Bolkestein, 43, 1118, 119 Bowerman, 239 Bransford, 16 Bresnan, 6, 115 Bühler, 40 Butler, 48, 104, 305

Emmorey, 15 Escalier, 77, 94, 242 Evens, 81

Calzolari, 19 Caramazza, 206 Carter, 117 Choi, 239 Chomsky, 5-7, 54, 69, 115, 117 Connolly, 210 Coseriu, 3, 57-59, 66, 67, 306 Croft, 69, 70 Cruse, 80

Faber, 82, 86, 91, 136, 140, 165, 215, 217, 223, 242, 252, 255, 306, 307 Felices, 366 Fellbaum, 1, 16, 75, 81, 84, 89, 104,111,157 Fillmore, 17, 32, 70, 73, 74 78, 94, 241 Fodor, 6, 54, 68,71 Foley, 40, 77, 94, 115, 117, 242 Fournier, 77, 94, 242 Franks, 6 Fromkin, 15, 206 Garcia Velasco, 306 Geckeler, 306 Geeraerts, 70 Gildea, 89 Givon, 84, 107, 115,233 Gleitman, H., 195

Author index Gleitman, L., 195 Goddard, 8-10, 71, 86 Goldberg, 2, 23, 32, 34 Gonzalez Escribano, 15, 305 Grandy, 67 Grimshaw, 6, 24, 115-117, 186, 305 Groot, 39, 46, 306 Guesne, 265 Halliday, 40 Heilige, 204 Hengeveld, 39-47, 118, 119, 132-135, 305 Hitch, 207 Hopper, 107, 128, 173, 174, 306 Hudson, 5, 306 Ingria, 19 Iris, 81 Jackendoff, 4, 7, 10, 16, 17, 25, 71,72, 77, 101, 114, 117,210, 213,306 Jackson, 90 Jimenez Hurtado, 108, 148 Johnson, 17, 242, 258,259 Johnson-Laird, 54, 223 Karttunen, 135 Katz, 6, 7, 54, 68,71,211 Kay, 2, 23 Keenan, 54 Kegl, 91 Keizer, 44, 47, 305 Kerbrat-Orecchioni, 242 Kiparsky, C., 7, 116, 134 Kiparsky, P., 7, 116, 134 Kittay, 67, 74, 75, 81 Kosslyn, 72 Kreszowski, 243

337

Lakoff, 17, 70, 228, 242, 258, 259 Landau, 91 Langacker, 1, 17, 70, 71, 210, 212,213,233,271 Langendoen, 117 Lapalme, 13, 14 LaPolla, 31, 31 Lehrer, 67, 74, 75, 81, 153 Levelt, 72 Levin, B. 20, 23-30, 35, 46, 51, 90, 91, 115, 117, 143, 145, 170 Levin, S , 210 Lewandowska-Tomaszczyk, 85 Litowitz, 81 Lyons, 40, 57„ 67, 80, 1118 Mairal Uson, 91, 115, 136, 139, 140. 165,215,217, 306 Marantz, 305 Marr, 72 Martin Mingorance, 2, 3, 56-59, 90, 107 McCawley, 6 Meijs, 83 Mel'cuk, 8, 11-13, 78, 79, 86, 88 Melka, 71 Miller, 16, 18, 20, 21, 74, 84, 85, 157, 210, 223, 235 Montague, 6 Moreno Ortiz, 19 Mufwene, 255 Nelson, 80 Newmeyer, 7 Nida, 68, 69 Nirenburg, 220 Nowakowski, 7 Nuyts, 16, 17, 52, 72, 210

338

Author index

Osgood, 242 Paducheva, 170 Paivio, 72 Palmer, 89 Pederson, 16, 72, 210 Perez Hernandez, 86, 307 Perlmutter, 306 Pinker, 85, 117, 170, 203 Polguere, 89 Portero, 306 Postal, 7, 71 Pottier, 67 Pustejovsky, 2, 73, 143, 168, 177, 192,213 Radford, 305 Ransom, 117, 129, 130, 132, 135, 173 Rappaport, 23-20, 35, 46, 51, 115, 117, 143,214, 305 Ravin, 7 Rijkhoff, 305 Rosch, 17, 18, 69, 211 Rosenbaum, 115 Ruhl, 87 Rupp, 207, 209 Saeed, 71, 72, 258 Saint-Dizier, 204 Sanchez Garcia, 255 Sanchez Zavala de, 7 Saussure, 12, 57, 80 Schack Rasmussen, 52, 306 Schank, 74, 114,212 Schroten, 71

Searle, 193 Shiffrin, 207 Siewierska, 54 Suci, 242 Talmy, 7 Tannenbaum, 242 Taylor, 69, 211 Thompson, 107, 128, 173, 174, 306 Titone, 16 Trier, 67, 69, 76 Tulving, 208 Tutin, 13, 14 Van der Korst, 203 Van Valin, 7, 30, 31, 39, 40, 115-117, 143 Vendler, 28, 73, 170 Viegas, 204 Vossen, 56, 83, 86-90, 187, 208, 306 Wanner, 12-14, 78, 79 Waxman, 63 Wiegand, 306 Wierzbicka, 5, 8-10, 76-78, 85, 88, 137, 170, 193 Wilkins, 115, 116 Wilks, 89 Williams, 51, 305 Wingfield, 16 Wotjak, 67 Zgusta, 91 Zholkovskij, 8, 11, 12

Subject index

accomplishment, 26-30, 51, 73 achievement, 27-30, 73, 144, 145, 149, 157, 166-169, 176,200, 276 Acquilex, 20, 83 acquisition, 15, 16, 19, 101, 102, 177,186 actant, 12 semantic actant, 12 deep-syntactic actant, 12 adequacy, 4, 15 descriptive adequacy, 4 explanatory adequacy, 4 observational adequacy, 4 psychological adequacy, 15 textual adequacy, 251 agent, 23, 24, 31-34, 39, 45, 50-52,55, 93-99, 107, 112, 123, 127, 128, 130, 138, 149, 150, 161-164, 173, 182, 185, 199, 222, 230, 232, 251,256, 264, 273 aktionsart, 28, 29, 73, 130, 305 antonymy, 13, 22 argument, 4, 11, 23-26, 29, 31-34,38-41,50-55,65, 84, 89, 91, 94, 97, 99, 107, 108, 118, 119, 123, 124, 126-128, 137, 139, 149, 156, 158, 160, 162, 169, 170, 172, 176, 177, 181, 182, 189, 192, 199, 201, 222, 223, 227, 230-232, 250, 260, 264, 269 default argument, 177,189 shadow argument, 177

true argument, 177 argument obligatoriness, 65, 89, 122, 142, 144 argument role, 32 argument slot, 10, 23, 54, 94, 98, 182,241 artificial intelligence, 210, 220 aspect, 28, 42, 73, 122, 128, 129, 151, 170, 269 axiological evaluation, 31, 57, 97, 226, 234, 242-248, 273 axiology, 234, 306 behavioral system, 212 binary properties, 211 biological space, 234-236 brain, 70, 204-207, 272 categorial component, 6, 7 categorial selection, 7 categorization, 17, 18, 69, 149, 207-212, 234, 244, 246, 272, 277 category case category, 52 conceptual category, 65, 67, 79, 208-212, 222, 242, 272, 276 lexical category, 86, 272 linguistic category, 212 syntactic category, 3, 21, 39, 47, 48, 65, 84 taxonomic category, 170 category deficit, 205, 206 category membership, 211

340

Subject index

causation, 17, 145, 170, 173, 186, 200, 254-256, 273, 276 certainty, 120, 173, 196, 200 cessation, 144, 145, 151, 166, 167, 200, 238, 276 circularity, 9, 61 clause structure, 32, 35-41, 45, 46, 56, 121, 145,210, 203, 275 cognition, 11, 17, 63, 82, 89, 203, 232, 23941,262, 269, 271 cognitive axis, 35, 203-274 Cognitive Linguistics, 69-71, 75, 258,277 cognitive processes, 16, 124, 203, 211 cognitive reference point, 69, 82, 84, 90,211,242 cognitive science, 18, 73 cognitive structure, 77 cognitive system, 2, 15, 56, 208, 211

Cognitive Semantics, 17, 18, 69, 258 collocation, 14, 150, 163, 191 communicative strategy, 42, 43 competence, communicative competence, 37 lexical competence, 4, 15, 57 linguistic competence, 210 complement, 6, 42, 43, 53, 114-122, 132, 135-147, 156, 160, 173-176, 194-199,210, 229, 231,252 adjectival complement, 161 fourth-order complement, 119 second-order complement, 119,138

third-order complement, 119-120 complementation pattern, 43, 58, 89-92, 95, 100, 104-109, 114, 121, 122, 125, 126, 131, 136-144, 147, 152, 155, 160, 188, 190, 199, 258, 276 complement selection, 115, 117, 119, 128, 141 componential analysis (CA), 68, 69 computational linguistics, 6, 14, 18-22, 34 computer applications, 14, 19, 22 computer science, 19 conation, 144, 145, 175, 200, 276 concealed proposition, 140, 195 concept, 9, 12, 16, 17, 20-22, 56, 58, 60, 70, 71, 77, 78, 81-86, 89, 101, 126, 136, 165, 205-210, 223, 238, 242, 271, 272 nominal concept, 17, 18 verb concept, 81, 84 conceptual component, 2, 71, 104 conceptual dependency, 251, 273 Conceptual Dependency Framework, 114 conceptual invariant, 85 conceptual level, 114 conceptual organization, 16, 21, 22, 72, 80, 84 conceptual relations, 203 conceptual role, 52 conceptual salience, 234 Conceptual Semantics, 213, 214 conceptual system, 80, 210 conceptual structure, 2, 15-17, 31,56, 67, 69, 72, 114,204, 210,242, 271-273

Subject index conceptualization, 16, 17, 114, 210, 240, 248 image-based conceptualization, 72 spatial conceptualization, 72 connection, 68, 76, 121, 204, 214-216, 228, 231,234, 249-254, 258, 262, 272, 273, 277 primary connection, 249, 251,253 secondary connection, 166, 251,253 constant, 25-28 constraint, 20, 23, 54, 71, 72, 116, 130, 143, 194,211 construction, 1,2, 23, 32-35 ditransitive construction, 32-34 grammatical construction, 54, 228 resultative construction, 157, 161,168-170 syntactic construction, 78, 119, 149, 155, 156, 164-167, 174-176, 231, 234 Construction Grammar (CG), 1, 23,32, 34-37,51 context canonical context, 10 grammatical context, 80 linguistic context, 90 paradigmatic context, 101, 121, 126, 144, 220 semantic context, 137 situational context, 90 sociocultural context, 91, 93, 180, 234, 240, 241,273 syntactic context, 187 temporal context, 157, 159

341

contextual restrictions, 5, 150, 251 contextualization, 178, 182 co-occurrence, 10, 13, 79, 144, 249 criterial-attribute model, 69, 212 decomposition, 3, 6, 7, 13, 25, 26, 29, 32, 46,71,73, 78, 89, 114, 256 decontextualization, 45, 212 deep structure, 6, 115 default value, 98, 99, 111, 113, 149, 160, 166, 172, 196, 215, 222, 236, 243,248, 261 definiens, 61, 62 deviation, 97-99, 111-113, 181, 215,261,272 dictionary, 1, 15, 21, 82, 87, 90, 94, 99, 204 generation dictionary, 220 mental dictionary, 18 dictionary definition, 58, 69, 83, 87, 91, 99, 223 dictionary entry, 57, 71, 90, 93, 101 differentiae, 59, 61, 83, 88, 90, 101, 107, 108, 112, 113, 141,215 dimension, 225 semantic dimension, 78, 223 discrimination net, 220, 221 domain conceptual domain, 59, 61, 67, 95, 254 lexical domain, 3, 35, 58-70, 79-83, 87-90, 101, 102, 124, 129, 134-146, 152-160, 164, 165, 169, 170, 173, 176-180,

342

Subject index

184-193,213-222, 227-263, 272, 273, 276,277 semantic domain, 56, 71-75, 81, 87, 234 domain membership, 87, 227, 230, 251,252, 273 duration, 144-146, 194, 200, 229, 231,239, 265, 276 long duration, 152-155, 229,266 short duration, 132, 146-151, 157, 266 dynamism, 15, 51, 238 emotion, 13, 74, 77, 78, 82, 85, 112, 151,215,216, 222, 224, 242, 259-263, 268 entailment, 21, 112 entity, 16, 23, 24, 38, 40, 44, 51, 52, 69, 73,78, 97, 101, 112, 118, 119, 123, 128, 131, 136, 141, 155, 160, 164, 165, 182, 185-188, 196, 198, 205-208,211,222, 223, 232, 235, 239, 241, 246, 256, 265-271 first-order entity, 40, 43, 118, 230, 231 fourth-order entity, 40, 118, 120 second-order entity, 40, 42, 118, 120, 135, 138, 230, 231, 153 third-order entity, 40, 43, 118, 120, 129, 231 epistemic scale, 135 Euclid, 263 EuroWordNet, 80, 91

event, 9, 16 24-29, 40, 71, 73, 76, 89, 102, 107, 118-120, 127-131, 135, 138, 145, 146, 151-153, 157, 158, 163-170, 204, 205, 208, 211,212, 223,230-235, 238, 246240, 248, 258, 268-270 event structure, 23-28 exemplar, 17, 211 experiencer, 24, 31, 51, 52, 151, 197, 210, 216, 232, 246, 262, 264, 269, 271 Explanatory Combinatorial Dictionary (ECD), 12, 77 factorization, 58, 217, 250, 272 feature, 18, 32, 49, 50, 54, 61, 66, 69-73, 104, 126, 186,211, 212, 233,240, 243 binary feature, 68 grammatical feature, 122, 126, 142, 144 multi-valued feature, 69 pragmatic feature, 57, 108, 189,241 prosodic feature, 46 semantic feature, 69, 78, 107, 190 syntactic feature, 136, 190 variational feature, 240 focalization, 136 focus, 112, 126, 149,175,222, 251,267, 268 formality, 107-109, 162, 189 frame, 2, 17, 32, 70-74, 77, 78, 99, 129, 208, 251 illocutionary frame, 40, 41 semantic frame, 128

Subject index sociocultural frame, 161, 186,241 subcategorization frame, 6, 7, 53, 115-117, 186 syntactic frame, 26, 123, 124, 127, 139, 156, 189-194, 230, 277 Frame Semantics, 32, 73, 74 FrameNet, 74 function, 57, 72, 192, 240, 251 grammatical function, 24, 33 ideational function, 40 interpersonal function, 40 lexical function (LF), 12-14, 79 pragmatic function, 37, 45 semantic function, 24, 39, 48, 50-53, 65, 117, 214, 222 syntactic function, 37, 45 textual function, 40 Functional Grammar (FG), 1-3, 23,35, 37-66, 117, 118, 121, 122, 139-144, 174, 204,213,217, 248, 275,276 Functional-Lexematic Model (FLM), 2,3,8,11, 14, 34, 35, 56-144, 186, 203, 211-213,222, 272, 275 functional opposition, 57, 61 fund, 37, 38 generative linguistics, 7, 62, 71 Generative Semantics, 6, 7 genus, 59-63, 87, 90, 94-96, 101, 141, 175, 179, 187, 215, 224, 226, 227, 232, 249, 253, 255, 262 gerund, 125, 136-139, 230, 231

343

Government and Binding Theory (GB), 1,23,24,50, 55,61 government pattern, 12 grammatical alternation, 143 grammaticalization, 204 hemisphere, 204 left hemisphere, 204 right hemisphere, 67, 204 hierarchy, 18, 24, 58, 63, 80, 84, 88, 89, 208,213,218, 243, 251,265,267 lexical hierarchy, 83, 101, 104, 107, 111, 141, 156, 190, 194, 199, 226, 256, 265, 276, 271, 276 semantic hierarchy, 88, 121, 193-195, 201 syntactic hierarchy, 144, 187, 193, 194, 201 three-level hierarchy of values, 243 human factor, 243 hyponymy, 12, 22, 60-63, 83, 87, 187, 227 illocutionary force, 42-45 image schema, 17, 85, 254, 258, 259, 263, 265, 267, 273 implicatives, 135 inception, 102, 145, 151, 165-166, 200, 238, 276 indeterminacy, 70 infinitive, 136-139, 147, 152-154, 157, 176, 190, 194 information processing, 207, 208, 242 inheritance, 87, 107, 190, 199, 250,277

344

Subject index

lexical inheritance, 79, 84 semantic inheritance, 79, 199 syntactic inheritance, 79, 84, 199 instrument, 14, 23, 27, 43, 114, 160, 162, 178, 185, 186, 222,224, 227 instrumentality, 108, 114 intensity, 107, 108, 148, 178, 185, 226, 257, 260-264 intentionality, 55, 113, 152, 153, 161, 109, 191,208,214, 222,224,237, 268 interface, 2, 3, 15, 25, 46, 68, 71, 108, 115, 121, 142-145, 200,275, 276 Interpretative Semantics, 7 iteration, 144, 145, 155, 159-162,200, 276 knowledge, conceptual knowledge, 14, 56, 208 declarative knowledge, 209 lexical knowledge, 1, 14, 15, 19,206,207, 212,217, 272 linguistic knowledge, 12, 19, 57, 203 semantic knowledge, 220 language acquisition, 19, 186 language comprehension, 5, 16, 19, 20, 67, 204 language generation, 20 language processing, 19 language production, 16, 19, 204 lexematic structure, 58 lexematic theory, 3, 57-59, 66 lexical chain, 22

lexical combinatorial zone, 12 lexical-conceptual structure (LCS), 17, 72, 114 lexical database, 20, 22, 75 lexical entry, 3-7, 12, 22, 27, 28, 34,35, 47, 56, 59,91, 115, 116, 206, 223,272, 275,276 Lexical Functional Grammar (LFG), 55, 61 lexical insertion, 5 lexical knowledge base (LKB), 20 lexical organization, 3, 19, 56, 58, 63, 67, 83, 87, 91, 102, 115, 186,211,236, 254, 272 lexical productivity, 229, 233, 273, 276 lexical revival, 4, 6, 19 lexical rules, 139, 140, 213, 277 lexical selection, 72, 220 lexical semantics, 4, 7, 8, 56, 67, 213 lexical universal, 9, 81 lexicography, 6, 8, 34, 56, 85, 91 lexicon, 1-10, 15-22, 32-39, 46-92, 101, 107-109, 115, 116, 119, 122, 140-145, 152, 180, 192, 203-205, 208,213,215,228, 233, 234, 239,240,243,250, 251,271-277 FG lexicon, 3, 35, 38, 48, 50, 56-59, 63-66,213, 248,275 FLM lexicon, 35, 57, 80, 104, 123, 129, 141-144, 186, 203, 212,223, 272 Generative Lexicon, 73 mental lexicon, 15, 16, 19, 81,204, 205, 271,272

Subject index

345

noun lexicon, 81, 84, 87 sense enumeration lexicon, 213 verb lexicon, 2, 18, 53, 56, 61, 80,81,84, 86, 102, 121,187, 254 lingua mentalis, 11 linking algorithm, 46, 213 linking rule, 4, 117 localism, 71 logical structure, 28-31

semantic memory, 81,211, 214, 275 short-term memory (STM), 209 working memory, 209 meronymy, 22, 157 metaphor, 17, 30, 54, 57, 70, 155, 214, 228, 236, 249, 250-253, 258, 271, 273 metonymy, 70, 214 microstructure, 3, 56, 77, 228, 271

macrostructure, 3, 35, 56, 57, 76, 84, 86, 186, 228, 229 mapping, 2, 11, 24, 25, 32, 35, 56, 104, 186, 250, 251,262, 271, 273, 277 meaning component, 23-26, 58, 61,63, 76,91,94, 96, 100, 199, 232, 251,255, 275,276 meaning constituent, 58, 65, 96, 272,276 meaning definition, 3, 35, 38, 46, 47, 52, 55,56, 59-65, 78, 90, 104, 112, 115,117, 121, 126, 134-139, 142, 144, 178, 196, 201,223, 227, 230, 249, 275 meaning extraction, 222 meaning postulate, 63 Meaning Text Model (MTM), 11, 94 Meaning Text Theory (MTT), 8, 11-14, 77, 78 memory, 15, 16,31,207, 209 declarative memory, 209 episodic memory, 209 lexical memory, 16, 20 long-term memory (LTM), 207, 208, 212 procedural memory, 211

modality, 42,43, 107, 117, 122, 128,132 epistemological modality, 42, 43 objective modality, 42, 132, 134 subjective modality, 42, 132 morphology, 7 multi-layered model of the clause, 118, 141 Natural Semantic Metalanguage (NSM), 8-11, 14, 76-78 network, 3, 9, 10, 57, 58, 75, 81-84,91, 104,213, 233, 251,271 conceptual network, 3 schematic network, 70 semantic network, 12, 19, 251,258, 271 neural activity, 205 neural circuit, 204, 272 neural system, 208 neuron, 19, 204, 206, 207 neutralization, 255 norm, 98, 99, 111-113,215,242, 261 aesthetic norm, 240, 242 behavioral norm, 186, 246

346

Subject index

biological norm, 180, 181, 201 social norm, 148, 180, 201, 240 notion, 78 nuclear meaning, 61, 87, 88, 93-97, 144 onomasiological structure, 56, 58, 65, 276 ontology, 86, 128 operator, 29, 30, 38, 40-44, 48, 49, 119, 120, 122, 126, 135, 137,176 boulomaic operator, 120 epistemological operator, 120, 173, 174 illocution operator, 41 predicate operator, 41, 131 predication operator, 41 proposition operator, 41 ΤΑΜ operator, 122, 128, 129, 132, 142 term operator, 41, 126 paradigmatic axis, 35, 59, 79, 80, 83, 92, 107, 114, 115, 141 paradigmatic relations, 57, 80 paradigmatic structure, 2, 17, 58, 84, 222, 258, 273 parameter differentiation parameter, 141, 220, 223, 259, 260 lexically-realized contextual parameter, 145, 177-179, 184-186, 196, 200, 201 lexically-realized grammatical parameter, 145,200

lexically-realized optional parameter, 145, 176, 200,201 pragmatic parameter, 108, 107 semantic parameter, 88, 101, 104, 107, 109-114, 122, 126, 131, 134, 136, 142, 146, 151, 180, 194, 195, 199, 200, 229, 234, 240, 249,251 spatial parameter, 72 synsem parameter, 35, 59, 107, 114, 144-176, 195,228, 265 performance, 14, 31 phasal distinction, 102, 164, 165, 234, 238, 239 phrase structure rules, 6, 7 polarization, 243-248 polysemy, 70, 187 positioner, 39, 51, 52, 179, 180 pragmatic component, 46, 223 pragmatic factor, 7, 23, 148, 213 pragmatic information, 20, 57, 74, 93-96, 108, 201,224, 227, 240, 249, 277 pragmatics, 37, 223 predicate activity predicate, 29, 31 individual-level predicate, 168 matrix, predicate, 115,119, 120, 122, 128, 129, 134 stage-level predicate, 168 state predicate, 29 predicate frame, 38, 40, 41, 44-66, 118-121, 131, 141, 142, 249, 250, 273, 277

Subject index predicate schema, 206, 212-214, 217, 222, 223, 249, 250, 273,277 domain-level, 233 lexeme-level, 223, 228, 272 subdomain-level, 228, 272 predicate-argument structure, 1, 23,35,46 predicate-centered approach, 23-25, 52 predicate formation rule, 38, 39, 48 predicate variable, 43, 47 presupposition, 7, 20 primitive, 24-26, 29-31, 52, 61, 77, 85, 141, 234 conceptual primitive, 9, 214 semantic primitive, 25, 62, 67, 76-78, 85 probability, 43, 134 problem solving, 209 projection argument projection, 26 metaphoric projection, 250, 258, 271, 273 syntactic projection, 35, 144, 200,249, 276 projection rule, 71 proposition, 39, 41, 45, 118, 119, 132-134, 140, 173, 174, 195,231 propositional attitude, 120 propositional content, 42-44, 135 propositional layer, 40 proto-argument structure, 24 prototype, 19,211,212 paradigmatic prototype, 249 semantic prototype, 58 syntactic prototype, 58 syntagmatic prototype, 249 prototype theory, 69, 210, 211

347

psycholinguistics, 15, 16, 19, 20, 81,210 Pythagoras, 263 radial set, 70 reductive paraphrase, 8 redundancy rules, 47, 49, 65, 248 representation conceptual representation, 16, 22, 72, 89, 203, 205, 210-212 feature representation, 67, 68, 141 FG representation, 37, 40, 41,44, 46-49,51,55, 62, 65 knowledge representation, 4, 14, 57, 203, 204, 223 lexical representation, 2, 3, 8, 9, 14-16, 23, 30-37, 46, 68, 73, 74, 78, 100, 101, 104, 114, 121, 142, 205,206,212, 220, 248, 251,253, 272 meaning representation, 11, 23,68, 89, 121, 141, 203, 249, 271 mental representation, 17, 18, 72 morphological representation, 11 phonetic representation, 11, 12 propositional representation, 16 semantic representation, 4-7, 11, 15,26,210 syntactic representation, 3, 11, 104, 117, 121, 129, 139-144

348

Subject index

resultative construction, 159, 163, 170-172 representational format, 70, 249 Role and Reference Grammar (RRG), 1, 23, 24, 28-32, 46, 46,51,62 role-centered approach, 23, 24, 35 root transformation, 174 satellite, 40-45, 49, 51, 55, 65, 66, 94, 119, 124, 133, 144 clause satellite, 41, 43 illocutionary satellite, 41, 43 implied satellite, 177 predicate satellite, 41, 43 predication satellite, 41, 43 proposition satellite, 41, 43 schema, 21, 34, 35, 63, 91, 107, 178, 186,211,212, 246,251 scope, 43,94, 104, 119, 122, 128, 145, 160, 166, 187, 190, 197, 201, 228, 249, 250, 276 selection restriction, 6, 13, 23, 29,38, 47,53,54, 65,71, 126-128, 130, 160, 166, 182, 206, 227, 249, 250 semantic class, 3, 25, 26, 46, 49, 56, 65, 66, 72-75, 79-82, 90, 114, 119, 121, 130, 139, 143, 145, 170, 276 semantic macronet, 228, 249-253, 273, 277 semantic marker, 71 semantic priming, 81 semantic relations, 13, 14, 16, 19, 22, 65, 67, 78, 81, 84 semantic restriction, 117, 198 semantic role, 23, 38, 51, 53, 94, 117,124,220 semantic zone, 12

semanticality, 177 semasiological structure, 70 sensory data, 109, 223, 246 sensory register, 207 situation, 2, 29, 89, 90, 211, 241, 254, 258, 267 speech act, 39-44,108, 119, 193, 196, 254, 255 state of affairs (SoA), 23, 29, 42, 43, 49,51,69-71, 144, 151-153, 184, 251,268 Stepwise Lexical Decomposition, 3, 55, 58-60, 66, 88, 141, 275 Structuralism, 67 subdomain, 35, 58-66, 80, 83, 87, 88,91,92, 100-104, 107-111, 114, 121-126, 129, 132-141, 144, 154, 161,164,168, 169-173, 177-183, 186, 187, 192-196, 200, 201, 213-236, 239, 248-251, 254, 255, 262, 269, 272, 276, 277 subject, 24, 33, 34, 45, 53, 174, 265, 269 substitutability, 78 substitution, 12 substitution principle, 87 superordinate term, 18, 59, 62, 72, 84-88, 104, 136, 192, 198, 222,252 synaptic connector, 207 synset, 20, 22, 75, 83, 84 syntactic alternation, 25, 39 syntactic potential, 35, 126, 143, 190, 193,201 syntactic relations, 12, 56

Subject index syntactic structure, 5, 61, 134, 137,306 syntactic zone, 12 syntagmatic axis, 35, 59, 92, 114, 115, 121, 141, 142, 238 syntagmatic choice, 220 syntagmatic structure, 15, 258 taxonomic organization, 226 taxonomy, 122, 142, 249 telicity, 176 template, 24-30, 35, 46, 73, 143, 144,214 Template Augmentation, 28, 144 temporal lobe, 206 temporal sequence, 145, 157 tense, 42, 117, 122, 128, 129 term, 8,38,41,52-54, 126 tertium comparationis, 9 thematic role, 7, 23-25 thesaurus, 61, 82, 83, 101 time reference, 121, 123, 129, 194 time-stability scale, 233 top-down, 60, 82 topicalization, 174, 195 Transformational Grammar (TG),5,6 transition zone, 153, 226, 254, 255, 273 transitivity, 107 translation, 68, 81, 203, 223 troponym, 77, 83, 84, 91, 92, 100, 101, 104, 138, 139, 146, 152, 187, 189, 192, 198-200,215,216, 222, 227, 240, 256, 257, 263, 265 troponymy, 81, 111 truth, 40, 132-134, 173, 195 typicality, 17, 18

349

valency, 10, 39, 91, 122. 124, 223,251 qualitative valency, 3, 47, 65,91 quantitative valency, 3,39, 47, 50, 65, 91 verb action verb, 9, 26-29, 50-52, 76, 88, 101, 128-131, 135, 137, 150, 154, 157, 175, 214,215, 222, 233, 238, 243, 254-257, 296 change verb, 26, 71, 88, 152, 154, 163, 164, 167, 168, 170,215,216, 239, 243, 252-257, 273, 285 cognition verb, 74, 88, 114, 114, 124, 129, 134-139, 153, 166, 173, 175, 187, 192-195,214-216, 227-234, 239, 247-255, 258, 262, 269, 273, 286 consumption verb, 146-149, 161, 169, 170, 187, 187-189, 240 contact verb, 161, 162, 169, 170, 172, 179, 186-188,222-224,284 existence verb, 9, 76, 88, 163-170,214,218, 219, 234, 238, 253-255, 279 feeling verb, 85, 95, 109, 113, 114, 124, 146, 150, 156, 157, 168, 170,214-216, 234,

350

Subject index

243, 246, 247, 258-260, 273, 287 light emission verb, 173, 214, 252, 258-264, 291 manner-of-deceiving verb, 215 manner-of-drinking verb, 149, 150 manner-of-speech verb, 254, 255 manner-of-staring verb, 267, 268 manner-of-walking verb, 109, 111, 113, 180 movement verb, 49, 66, 76, 88, 95, 101, 109, 154, 164, 168-173,215, 216, 233-236, 239, 253, 280 physical perception, 29, 31, 85, 88, 119, 120, 131, 147,215, 236, 237, 252, 258, 286 position verb, 88,178-180, 252-254, 284 possession verb, 31, 71, 72, 88, 91, 95, 100-108,

123, 158, 167-70, 176-178,214,215, 243, 291 preparation verb, 152, 153 sound emission verb, 25, 26, 107, 159, 162, 168, 170, 224-227, 243-246, 254, 255, 290 speech verb, 9, 30, 31, 76, 77, 83, 85,88, 119, 129, 152, 155, 158, 169, 170, 178-181, 193, 214, 226, 240, 241, 247, 248, 252-255, 288 visual perception verb, 86, 95, 113, 130,135, 146-153, 157, 178, 187-193,214,216, 230-232, 236-238, 251-253,258, 262, 263, 267,269, 273, 286 vocabulary matrix, 21, 22 VP preposing, 175 word form, 21, 22 word-formation, 7, 57 WordNet, 20, 22, 74, 75, 81, 83-87